Raymond S. T.

Lee

Quantum
Finance
Intelligent Forecast and Trading Systems
Quantum Finance
Raymond S. T. Lee

Quantum Finance
Intelligent Forecast and Trading Systems

123
Raymond S. T. Lee
Division of Science and Technology
Beijing Normal University-Hong Kong
Baptist University United International
College (UIC)
Zhuhai, Guangdong, China

ISBN 978-981-32-9795-1 ISBN 978-981-32-9796-8 (eBook)

https://doi.org/10.1007/978-981-32-9796-8
© Springer Nature Singapore Pte Ltd. 2020
This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar
methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publication does not imply, even in the absence of a specific statement, that such names are exempt from
the relevant protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publisher remains neutral with regard
to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
This book is dedicated to all readers and
students in UIC taking Quantum Finance
course; your enthusiasm for learning new
concepts and seeking knowledge prompts me
to write this book.
Preface

Motivation of This Book

Everything happens for a reason. Anything that might be happened with a cause.
These two sentences seem to be very simple, but they are major motivations to
drive the advances in science and technology of human history.
However, modern cosmology tells us that we live in a multiverse, the world we
all experience is just one of the many realities. Quantum theory tells us that all
fundamental particles in the subatomic world coexist in a particle–wave duality
state. Heisenberg’s uncertainty principle even tells us that we can never 100%
correctly measure both the position and momentum of any quantum particle
simultaneously. Chaos theory shapes the world as complex systems with all pre-
dictions are highly sensitive to its initial conditions; only a tiny observation error
will lead to a totally different future. Fuzzy logic tells us to accept this as nature of
reality; all we can do is to model the world as a collection of fuzzy linguistics and
events.
Does it mean that we cannot predict the future?
The answer is definitely … NO.
The motivation of this book—quantum finance, a cross-discipline of financial
engineering with the integration of quantum field theory, classical finance theory,
computer science, and AI technology—is to find a scientific path to resolve this
grant puzzle.

Quest for the Knowledge of Financial Prediction

This book is also about an over 30-year journey of knowledge seeking on financial
prediction.
The idea of using quantum field theory and AI to solve financial problems (now
called quantum finance) emerged when I was an undergraduate physics student at

vii
viii Preface

the Hong Kong University (HKU) in 1986 studying quantum mechanics and
general relativity courses. At that time, I was deeply impressed by all great minds
and innovative ideas for the interpretation of the world in two totally different
perspectives: the subatomic world of quantum mechanics versus the cosmic world
of general relativity. Since then, I strongly believe that both the Schrödinger
equation and Feynman’s path integrals hold the keys to unlock the dynamics of
everything—to truly observe the past and predict the future. They also hold the keys
to resolve two major forecasting challenges in human history: weather forecast in
meteorology and worldwide financial prediction in finance.
After graduation from HKU in 1989, I joined the Hong Kong Observatory of the
Government of Hong Kong Special Administrative Region and worked as exper-
imental office on weather prediction and in the building of computational weather
prediction model known as numerical weather prediction (NWP) system using
Fortran. Although at that time it was still based on Newton’s laws of motion, fluid
dynamics, and thermodynamics equations to predict weather, it gave me an
opportunity to realize two important facts: (1) In order to ultimately unlock the
mystery of prediction using Schrödinger equation by computer system, we need to
convert (model) the Schrödinger equation wisely into other formats such that it can
be easily computed by digital computers using numerical computation techniques;
(2) Only Schrödinger equation and Feynman’s path integral are not adequate to
predict the future. In Chaps. 3–5, these two mathematical models are the keys to
model the quantum dynamics of quantum financial particles (QFPs) and to evaluate
the quantum price levels (QPLs), but they are not the ideal tools to predict future. In
order to predict the financial market wisely and effectively, we need some other
intelligent tools. Such motivation precipitated me to pursue M.Sc. and Ph.D. studies
at Hong Kong Polytechnic University in 1994 and 1997, respectively.
During this period, I have learned and explored all different kinds of artificial
intelligence (AI) technologies including artificial neural network (ANN), fuzzy
logic (FL), genetic algorithms (GAs), and chaos and fractals theories. Among them,
I realized in order to solve this grant puzzle, one must look for four puzzle pieces
first:
(1) All these AI tools and technologies are not only closely related but also each
of them has its own advantages and shortcomings. If we can integrate these AI
tools and techniques wisely, we might have an opportunity to solve more
complex and fundamental problems, worldwide financial prediction—for
instance;
(2) All these AI tools and techniques are in fact different faces of the same coin;
how to integrate them wisely and effectively should not only focus on the
methodology but rather the fundamental structure of the AI tools themselves;
(3) How to model quantum finance effectively is another important issue. In
Chap. 3, we will see that traditional Feynman’s path integral although can
provide a feasible solution to model quantum dynamics of financial markets, the
computational complexity of this method hinders us to apply it for real-time
financial market prediction. In other words, we need a totally new idea to model
Preface ix

quantum finance effectively, not only at the analytical level but also at the
numerical computation level;
(4) Even though if we can model financial markets using quantum finance effec-
tively, how to integrate such model with AI is another problem. In order to
integrate quantum finance model and AI technologies nicely to implement
worldwide financial prediction and intelligent quantum trading system, we need
a common ground to link up these two worlds. Again, not at the methodology
level, but rather at the core model structure itself.
Not until 2004, I proposed the Lee-oscillator—a chaotic time-discrete neural
oscillator with profound transient-chaotic property, which gave us some new hopes
to resolve part of this puzzle. The interesting thing is that, when the Lee-oscillators
are used to replace the neurons in any time series neural networks, such as tradi-
tional feedforward backpropagation networks (FFBPNs), they would convert the
neural networks into powerful chaotic neural oscillatory networks (CNONs) with
promising prediction and machine learning capabilities. In Chap. 9, such chaotic
neural oscillators can be served as the missing link to integrate all different kinds of
AI technologies effortlessly, which sheds light on the resolution of the first two
puzzle pieces to model complex time series prediction problems such as real-time
worldwide financial forecast. After that, it leaves us with the last two puzzle pieces.
In 2007, Dasgupta et al. published their paper titled as Simple systematics in the
energy eigenvalues of quantum anharmonic oscillators in the Journal of Physics A
(Chap. 3) which explained how to model Schrödinger equation as quantum
anharmonic oscillators (QAHOs). More importantly, this paper nicely described
how to resolve a special type of QAHO so-called kx2m AHO in which at the same
time I had derived the quantum finance Schrödinger equation (QFSE) as an AHO of
order 4 (i.e., m = 2) quantum dynamic model (Chap. 4). By the integration of this
kx4 AHO mathematical model and the finite difference method (FDM) learnt for the
implementation of the NWP model in Hong Kong Observatory years ago, I can
finally construct a numerical computation model with the integration of AHO (to
model quantum financial particles) and CNON (to model chaotic time series pre-
diction) into the so-called quantum finance forecast system (QFFS).
In order to prove the applicability of this totally new AI-fintech invention in
real-world financial markets, during 2012–2017, I joined Leanda Investment Group
(one of the major commodity investment and trading companies in China) as the
chief analyst and group CTO for the implementation and national promotion of
QFFS in China. With the success of the implementation of QFFS in commodity
product forecast for over 5 years in China and Hong Kong, in Mar 2017, I set up the
quantum finance forecast center (aka QFFC, website QFFC.org), a nonprofit
making, AI-fintech R&D, and worldwide financial forecast center aims at the R&D
and the provision of an open platform for worldwide traders and individual
investors to acquire free worldwide real-time financial prediction. QFFS now
provides daily (weekly) financial forecasts for over 120 worldwide financial
products including 9 major cryptocurrencies, 17 worldwide financial indices, 84
forex, and 19 major commodities. As of March 2019, QFFC has over 10,000+
x Preface

registered worldwide members which consists of worldwide professional traders,

quants, and independent investors from major fund houses and financial institutions
using the free quantum finance daily forecast services.

Organization of This Book

This book is the collection of my past 20 years (1999–2019) of R&D works and
practical implementation of quantum finance and related AI technology including
chaotic neural networks, fuzzy logics and genetic algorithms (GAs) in financial
market modeling, time series financial prediction, and intelligent agent-based
quantum trading techniques and methodologies.
This book consists of two parts: Part I describes the basic concept and theory of
quantum finance and related AI technology. Part II describes the current and
ongoing R&D projects for the application of quantum finance technology on
intelligent real-time financial prediction and intelligent agent-based quantum trad-
ing systems.
This book is organized as follows:
• Chapter 1—Introduction to Quantum Finance
This chapter introduces the origin of quantum finance and the two different
perspectives to see the world. It investigates the wave–particle duality and
Heisenberg’s uncertainty principle in quantum mechanics and how these phe-
nomena are related to quantum finance, which leads to our study on the phi-
losophy of quantum finance. After that, it introduces the major components of
quantum finance and the quantum finance four-tier concentric sphere model.
• Chapter 2—Quantum Field Theory for Quantum Finance
To avoid complex mathematical derivation, the chapter mainly focuses on the
basic concept and physical meaning of how quantum field theory can be applied
to financial engineering for the modeling of worldwide financial markets.
• Chapter 3—An Overview of Quantum Finance Models
This chapter investigates two main branches of quantum finance models—the
Feynman’s path integral model and the quantum anharmonic oscillator model. It
also explores the future of quantum finance which relates to intelligent finance
systems’ development.
• Chapter 4—Quantum Finance Theory
This chapter starts with a basic concept of quantum finance theory—quantum
anharmonic oscillator model. It focuses on the complete mathematical deriva-
tion of the author’s original work of quantum finance Schrödinger equation
(QFSE). It also investigates the quantum dynamics in financial markets, the
notion of quantum price levels (QPLs) and their relationship with quantum
finance energy levels (QFELs).
• Chapter 5—Quantum Price Levels—Basic Theory and Numerical
Computational Technique
Preface xi

This chapter focuses on detailed mathematical and numerical derivations of

quantum price levels to solve the quantum finance Schrödinger equation (QFSE)
effectively using numerical computational method—the core of quantum
finance in financial market modeling. First, it presents the basic concept of
quantum price levels (QPLs) and its relationship with quantum finance energy
levels (QFELs) in QFSE. Second, it shows how to interpret the quantum price
return wave function in terms of probability density function (pdf) using finite
difference method (FDM). Third, it explores how to solve QFSE using
numerical computational technique and describes the computer algorithm to
determine all QFEL and QPL. This chapter will be ended with the first quantum
finance computing workshop for QPL evaluation on worldwide financial
products using metatrader query language (MQL) in metatrader (MT) platform.
• Chapter 6—Quantum Trading and Hedging Strategies
• This chapter focuses on the practical use of quantum finance in quantum trading
and hedging operations. First, it starts with the discourse of seven major trading
and hedging techniques—a collection of the author 20+ years of trading
experience in stock, commodity, and forex trading with numerous technical
trading courses conducted in China and Hong Kong for past 10 years. Second, it
presents the main concept on how to make use of quantum finance, especially on
quantum price levels (QPLs), quantum forecasts to design quantum trading and
hedging strategies.
• Chapter 7—AI Powerful Tools in Quantum Finance
This chapter introduces three major AI tools in quantum finance: artificial neural
networks (ANNs) on machine learning and time series prediction, fuzzy logics
(FL) on fuzzy and inexact financial modeling, and genetic algorithms (GAs) on
trading strategy optimization. More importantly, it shows how these AI tools are
integrated with quantum finance technology to implement quantum finance
forecast and intelligent agent-based program trading systems.
• Chapter 8—Chaos and Fractals in Quantum Finance
This chapter introduces two vital contemporary finance engineering theories:
chaos and fractals. It explains the duality behavior of financial markets modeled
by these two theories, and more importantly, on how they are related to quantum
finance for financial engineering in contemporary financial institutions.
• Chapter 9—Chaotic Neural Networks in Quantum Finance
This chapter presents basic theory of chaotic neural oscillators and neural net-
works with the author’s original work on Lee-oscillator, how it can be adopted
with quantum field signals (QFSs) into quantum finance oscillators (QFOs) and
converted into chaotic neural networks for financial prediction.
• Chapter 10—Quantum Price Levels for Worldwide Financial Products
As Part II’s first chapter on quantum finance applications, this chapter discusses
R&D project to apply quantum finance technology to calculate the quantum
price levels (QPLs) for over 129 worldwide financial products including: forex
(84), cryptocurrency (9), major commodities (17), and worldwide financial
indices (17).
xii Preface

• Chapter 11—Time Series Chaotic Neural Oscillatory Networks for Financial

Prediction
This chapter discusses R&D project on how quantum price levels (QPLs) can be
integrated with chaotic neural oscillatory network learnt in Chap. 9 to time series
chaotic neural oscillatory networks (TSCNONs) for financial prediction.
• Chapter 12—Chaotic Type-2 Transient-Fuzzy Deep Neuro-Oscillatory Network
(CT2TFDNN) for Worldwide Financial Prediction
This chapter discusses the author’s latest R&D for the integration of chaotic
type-2 transient-fuzzy logic with chaotic neural networks for the implementation
of chaotic type-2 transient-fuzzy deep neuro-oscillatory network (CT2TFDNN)
for worldwide financial prediction.
• Chapter 13—Quantum Trader—A Multiagent-based Quantum Financial
Forecast and Trading System
This chapter discusses the latest research project to apply multiagent-based
real-time financial prediction and trading system using quantum finance
technology.
• Chapter 14—Future Trends in Quantum Finance
This chapter concludes all concepts and theories in quantum finance, their
relations to contemporary AI technology on intelligent financial systems. It also
discusses the latest research on financial engineering and the trend of quantum
finance.

Readers of This Book

This book is both a textbook and professional book tailored for

• Undergraduate and master-degree students for various courses on quantum
finance/computational finance/AI/fintech/machine learning;
• Researchers and scientists working in the field of financial engineering and
intelligent systems, AI, computational finance, econophysics, data science, and
computational intelligence;
• Professional traders/quants/independent investors who would like to learn the
basic concept and theory of quantum finance, more importantly on how to use
this innovative AI-fintech technology to implement intelligent financial forecast
and program trading systems;
• Lecturers and tutors who would like to teach students quantum finance with
related AI technologies, and to conduct labs to teach students on how to design
and implement intelligent financial forecast and trading systems.
Preface xiii

How to Use This Book?

This book can be served as both textbook and research monograph for various
undergraduate and postgraduate courses on quantum finance and related
AI/AI-fintech courses.
Basically, Part I (Chaps. 1–9) covers all basic concepts and theory about
quantum finance and related AI technology can be conducted as a self-contained
undergraduate/postgraduate course on quantum finance.
Part II (Chaps. 10–14) covers all advance studies and applications of quantum
finance, which can be adopted as the selected topics on various advanced AI/fintech
courses for postgraduate and research studies.
In UIC, starting from early 2019, the author launched a new free elective course,
namely, “Quantum Finance” for Year 3/4 undergraduate students of different dis-
ciplines including computer science, financial mathematics, data science, and
statistics. In terms of course syllabus, this course covers Chaps. 1–9 in the 14-week
study, and Chaps. 10 and 11 as workshops and term projects for the implementation
of QPL-based quantum finance forecast systems. Besides, students and course
instructors can use MQL workshops provided by QFFC.org enabling students to
learn MQL programming and how to develop simple program trading programs.
For advanced studies such as postgraduate AI courses or M.Phil./Ph.D. studies,
Chaps. 11–13 provide advanced topics on how to combine quantum finance theory
with contemporary type-2 fuzzy logic, chaotic neural oscillators, and reinforcement
learning agent theories for further R&D on quantum finance and related studies.
Last but not least, for quantum finance application developers such as quants and
financial analysts make use of the QFSDK provided by QFFC.org to develop
quantum finance applications. They can directly read Chaps. 10 and 11 first as
technical reference for QPLs’ calculation to implement quantum finance forecast
system for their own interesting financial products and revert to Part I to learn the
basic concepts and theories.

Companion Website—Quantum Finance Forecast Center

(QFFC.org)

Quantum finance forecast center (QFFC.org) is a nonprofit making, AI-fintech

R&D and worldwide financial forecast center that aims at R&D and the provision of
an open platform for worldwide traders and individual investors to acquire free
worldwide financial products’ quantum finance forecast.
As a companion website, QFFC provides a web portal for readers and worldwide
quantum finance system developers to learn and develop quantum finance systems
with the following:
xiv Preface

• Interactive quantum finance computing workshops (QFCWs) provide

step-by-step online tutorials and programming workshops for readers to apply
the knowledge learnt from this book.
• Quantum finance software development kit (QFSDK) is a C-library fully inte-
grated with MT platform which consists of all related quantum finance and AI
tools and functions for quantum finance developers to build their own quantum
finance forecast and intelligent agent-based trading and hedging systems. In
other words, this book can be also used as the guidebook and main reference to
learn the basic concept and theory of quantum finance technology.

Zhuhai, China Raymond S. T. Lee

Acknowledgements

While it only required around 6 months to write this book, the whole journey of
quantum finance—from seeking concepts and knowledge to the actual implemen-
tation—took me almost 30 years, started when I was still an undergraduate at HKU
since 1986 studying quantum mechanics and general relativity courses to currently
teach quantum finance to my fellow students at UIC.
I would like to express my gratitude to the following people for their support and
assistance.
To my wife Iris for her patience, encouragement, and understanding, especially
during my time spent on research and writing in the past 30 years.
To Ms. Celine Cheng, executive editor of Springer Nature and her editorial team
members including Jane Li and Haiying Li, for their support, valuable comments,
and advice.
To Prof. James Liu, my M.Sc. and Ph.D. mentor, for his support to led me to the
sacred land of artificial intelligence.
To Jim He, my TA, and all my lovely research students in UIC including Turing
Qiu, Henry Liang, Alex Yuan, and Peter Chen, for their help and assistance for the
development of Quantum Finance Software Development Kit and online tutorials,
labs, and workshops.
To Prof. Tang Tao, President of UIC, for the provision of excellent environment
for the research, teaching, and the writing of this book.
To Prof. Hua Xiong Huang, Dean of Division of Science and Technology of
UIC, and Prof. Weifeng Su, Program Director of Computer Science and
Technology of UIC, for their support for the opening of new course in quantum
finance in UIC.
To all fellow colleagues in UIC, for their support and fruitful advice in course design.
To UIC, for the prominent support on research grant # R201948; and for the
provision of an excellent environment, facilities for system development, testing,
and evaluations.

Zhuhai, China Raymond S. T. Lee

June 2019

xv
Contents

Part I Quantum Finance Theory

1 Introduction to Quantum Finance . . . . . . . . . . . . . . . . . . . . . . . ... 3
1.1 A Tale of Two Worlds—From the World of General
Relativity to the World of Quantum Mechanics . . . . . . . . . ... 4
1.1.1 Newton’s Physical World of Classical Mechanics . ... 5
1.1.2 Einstein’s Cosmic World of General Relativity . . ... 5
1.1.3 Heisenberg’s Subatomic World of Quantum
Mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 6
1.2 Two Sides of the Same Coin—Wave–Particle Duality
in Quantum Finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 7
1.2.1 Wave–Particle Duality and Double-Slit
Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ... 7
1.2.2 Wave–Particle Duality in Finance . . . . . . . . . . . . ... 9
1.3 Quantum Field Theory and the Birth of Quantum Finance . ... 10
1.4 Heisenberg’s Uncertainty Principle in Finance and Modern
AI Technology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.1 Schrödinger’s Cat Paradox . . . . . . . . . . . . . . . . . . . . 11
1.4.2 Uncertainty Principle in Finance . . . . . . . . . . . . . . . . 12
1.5 Basic Components of Quantum Finance . . . . . . . . . . . . . . . . . . 13
1.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2 Quantum Field Theory for Quantum Finance . . . . . . . . . . ...... 17
2.1 Quantum Mechanics—The Basics . . . . . . . . . . . . . . . . ...... 18
2.1.1 A Brief History Quantum Mechanics . . . . . . . ...... 19
2.1.2 Quantum Mechanics—Wave–Particle Duality
Phenomena . . . . . . . . . . . . . . . . . . . . . . . . . ...... 19
2.1.3 Heisenberg’s Uncertainty Principle . . . . . . . . ...... 23

xvii
xviii Contents

2.1.4 Quantization of Energy . . . . . . . . . . . . . . . . . . . . . . . 24

2.1.5 Quantum Entanglement . . . . . . . . . . . . . . . . . . . . . . . 25
2.2 From Quantum Mechanics to Quantum Field Theory . . . . . . . . 27
2.3 Classical Field Versus Quantum Field . . . . . . . . . . . . . . . . . . . 27
2.4 Feynman’s Path Integral . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.4.1 A Thought Experiment . . . . . . . . . . . . . . . . . . . . . . . 29
2.4.2 Feynman’s Path Integral Formulation . . . . . . . . . . . . . 32
2.4.3 Path Integral Formulation in Quantum Finance . . . . . . 32
2.5 Quantum Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.5.1 Quantum Harmonic Oscillator—The Basics . . . . . . . . 33
2.5.2 N-Dimensional Quantum Harmonic Oscillators . . . . . 34
2.5.3 Quantum Anharmonic Oscillators . . . . . . . . . . . . . . . 36
2.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3 An Overview of Quantum Finance Models . . . . . . . . . . . . . . . . . . . 41
3.1 A Brief History of Quantum Finance . . . . . . . . . . . . . . . . . . . . 42
3.2 Major Works and R&D in Quantum Finance . . . . . . . . . . . . . . 43
3.3 Two Pillars of Quantum Finance Models . . . . . . . . . . . . . . . . . 43
3.4 Feynman’s Path Integral Approach in Quantum Finance . . . . . . 44
3.4.1 Forward Interest Rate in a Nutshell . . . . . . . . . . . . . . 44
3.4.2 Zero Coupon Bond (Treasury Bond) . . . . . . . . . . . . . 46
3.4.3 Quantum Field Theory of Interest Rate Model . . . . . . 46
3.4.4 Feynman’s Path Integral Formulation of Forward
Interest Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 48
3.5 Quantum Anharmonic Oscillator Approach in Quantum
Finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.5.2 QAHO Modeling—Prices Versus Returns . . . . . . . . . 49
3.5.3 Quantum Anharmonic Oscillations of Returns . . . . . . 50
3.5.4 A Quantum Anharmonic Oscillator Model
of Financial Markets . . . . . . . . . . . . . . . . . . . . . . . .. 52
3.5.5 Physical Meaning of Hamiltonian in Quantum
Finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.5.6 Probability Density Function in Quantum Finance . . . 53
3.5.7 Quantum Anharmonic Oscillator Formulation . . . . . . . 54
3.5.8 Quantum Finance Schrödinger Equation—QFSE . . . . 55
3.5.9 Significance of QFSE in Quantum Finance . . . . . . . . 56
3.6 Quantum Finance—Which Way to Go . . . . . . . . . . . . . . . . . . . 56
3.6.1 Feynman Feynman’s Path Integral Approach Versus
Quantum Anharmonic Oscillator Approach . . . . . . .. 56
3.6.2 Quantum Finance—Which Way to Go . . . . . . . . . . .. 58
Contents xix

3.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4 Quantum Finance Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.1 Quantum Finance Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.1.1 The Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.1.2 The Notion of Quantum Price . . . . . . . . . . . . . . . . . . 67
4.1.3 Revisit of Schrödinger Equation . . . . . . . . . . . . . . . . 69
4.2 Quantum Finance Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.1 The Mathematical Model . . . . . . . . . . . . . . . . . . . . . 70
4.2.2 Physical Meaning of Wave Function w . . . . . . . . . . . 71
4.3 Financial Dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.1 Key Players in Secondary Financial Markets . . . . . . . 72
4.3.2 Market Maker (MM) . . . . . . . . . . . . . . . . . . . . . . . . 72
4.3.3 Arbitrageurs (ARs) . . . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3.4 Speculators (SPs) . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
4.3.5 Hedgers (HGs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.3.6 Investors (IVs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
4.4 Financial Dynamics and the Notion of Excess Demand . . . . . . . 77
4.5 Quantum Dynamics in Financial Markets . . . . . . . . . . . . . . . . . 79
4.5.1 Market Makers (MMs) . . . . . . . . . . . . . . . . . . . . . . . 79
4.5.2 Arbitrageurs (ARs) . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.5.3 Speculators (SPs) . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5.4 Hedgers (HGs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.5.5 Investors (IVs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.5.6 Overall Dynamics for All Different Parties . . . . . . . . . 82
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5 Quantum Price Levels—Basic Theory and Numerical
Computation Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
5.1 Quantum Price Levels (QPLs) . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.1.1 The Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.1.2 The Quantum Dynamics . . . . . . . . . . . . . . . . . . . . . . 91
5.1.3 QPLs Versus S & R Levels . . . . . . . . . . . . . . . . . . . 92
5.2 Schrödinger Equation Revisit . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.3 Quantum Finance Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.4 Physical Meaning of Wave Function W . . . . . . . . . . . . . . . . . . 95
5.5 Solving Quantum Finance Schrödinger Equation . . . . . . . . . . . 97
5.6 Numerical Computation of Quantum Anharmonic
Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .......... 97
5.7 Evaluate Quantum Price Levels Using Numerical
Computation Technique . . . . . . . . . . . . . . . . . . . .......... 99
xx Contents

5.8 Finite Difference Method (FDM) . . . . . . . . . . . . . . . . . . . . . . . 100

5.9 Finite Difference Method (FDM) to Evaluate QF Wave
Function . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
5.10 Numerical Evaluation of kjXAUUSD Using FDM . . . . . . . . . . . . . 102
5.11 Numerical Computation of Quantum Energy Levels En . . . . . . . 104
5.12 Numerical Computation of Quantum Finance Energy
Levels (MATLAB Version) . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
5.13 Depressed Cubic Equation Solver—Gerolamo Cardano . . . . . . . 107
5.14 Cardano’s Method for Calculating QFEL . . . . . . . . . . . . . . . . . 108
5.15 Numerical Computation of Quantum Energy Levels
(MQL Version) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
5.16 Numerical Algorithm to Calculate QPL Using MQL . . . . . . . . . 110
5.17 QPLs for XAUUSD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
5.18 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6 Quantum Trading and Hedging Strategy . . . . . . . . . . . . . . . . . . . . 119
6.1 Financial Trading and Hedging Strategies in a Nutshell . . . . . . 120
6.2 Traditional Trading and Hedging Strategies . . . . . . . . . . . . . . . 121
6.3 Latest R&D on Trading and Hedging Strategies . . . . . . . . . . . . 122
6.4 Technical Analysis and Program Trading . . . . . . . . . . . . . . . . . 123
6.5 Technical Analytical Tools and Indicators . . . . . . . . . . . . . . . . 124
6.6 Seven Major Financial Trading and Hedging Algorithms . . . . . 125
6.6.1 M1.1 Trend Trading Strategy—Key Patterns . . . . . . . 126
6.6.2 M1.2 Trend Trading Strategy—MA Crossing . . . . . . . 127
6.6.3 M1.3 Trend Trading Strategy—Signal Lines
Crossing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128
6.6.4 M2.1 Breakout Trading Strategy—Key Patterns . . . . . 129
6.6.5 M2.2 Breakout Trading Strategy—S & R Lines . . . . . 130
6.6.6 M3.1 Reversal Trading Strategy—Key Reversal
Patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
6.6.7 M3.2 Reversal Trading Strategy—S & R Lines . . . . . 132
6.6.8 M3.3 Reversal Trading Strategy—Technical
Indicator RSI . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133
6.6.9 M4.1 Channel Trading Strategy—Trend Channel
Trading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 134
6.6.10 M4.2 Channel Trading Strategy—S & R Channel
Trading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
6.6.11 M4.3 Channel Trading Strategy—Bollinger Bands . . . 136
6.6.12 M5 Averaging Trading Strategy . . . . . . . . . . . . . . . . 137
6.6.13 M6 Stop-Loss Trading Strategy . . . . . . . . . . . . . . . . . 139
6.6.14 M7.1 Hedge Trading Strategy—Single-Market
Single-Product (SM-SP) Hedge . . . . . . . . . . . . . . . . . 140
Contents xxi

6.6.15 M7.2 Hedge Trading Strategy—Single-Platform

Multi-product (SP-MP) Hedge . . . . . . . . . . . . . . . . . . 140
6.6.16 M7.3 Hedge Trading Strategy—Multi-platform
Multi-product (MP-MP) Hedge . . . . . . . . . . . . . . . . . 142
6.7 Quantum Trading and Hedging Strategy . . . . . . . . . . . . . . . . . . 143
6.7.1 Basic Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
6.7.2 Quantum Price Levels Versus S & R Analysis . . . . . . 144
6.7.3 Quantum Finance Forecast Center—Daily Forecast
and QPLs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
6.7.4 Basic Quantum Trading and Hedging Strategies . . . . . 147
6.7.5 QT#1.1 Inexperienced Investors—Quantum
Forecast Oscillation Trading Strategy (QFOTS) . . . . . 147
6.7.6 QT#1.2 Inexperienced Investors—Quantum
Forecast Oscillation Trading with Averaging
Strategy (QFOTAS) . . . . . . . . . . . . . . . . . . . . . . . . . 148
6.7.7 QT#2.1 Experienced Traders—Quantum Forecast
Breakout Trading Strategy (QFBTS) . . . . . . . . . . . . . 150
6.7.8 QT#2.2 Experienced Traders—Quantum Forecast
Breakout Trading with Averaging Strategy
(QFBTAS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
6.7.9 QT#3.1 Experienced Traders—Quantum Forecast
Single-Product Hedging Strategy (QFSPHS) . . . . . . . 152
6.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 157
7 AI Powerful Tools in Quantum Finance . . . . . . . . . . . . . . . . . . . . . 159
7.1 An Overview of Artificial Intelligence . . . . . . . . . . . . . . . . . . . 160
7.1.1 From Greek Mythology to Turing Test . . . . . . . . . . . 160
7.1.2 Definition of AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
7.1.3 Strong Versus Weak AI . . . . . . . . . . . . . . . . . . . . . . 162
7.1.4 Strong (Hard) AI . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
7.1.5 Weak (Soft) AI . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
7.1.6 Classification of Artificial Intelligence . . . . . . . . . . . . 165
7.2 Neural Networks—The Brain of AI . . . . . . . . . . . . . . . . . . . . . 166
7.2.1 Biological Neural Network—Our Brain . . . . . . . . . . . 166
7.2.2 Integrate-and-Fire Operations in Biological Neural
Network . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
7.2.3 A Neuron Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
7.2.4 Single Hidden Layer Artificial Neural Network
Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
7.2.5 Classification of ANNs by Machine Learning
Technique . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
xxii Contents

7.2.6 Classification of ANNs by Areas of Application . . . . . 172

7.2.7 Auto-associative Network—Network Architecture . . . 173
7.2.8 Auto-associative Network—Network Training
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
7.2.9 Hopfield Network—Network Training Algorithm . . . . 175
7.2.10 Hopfield Network—Network Training Algorithm . . . . 175
7.2.11 Feedforward Backpropagation Network
(FFBPN)—Network Architecture . . . . . . . . . . . . . . . . 177
7.2.12 Feedforward Backpropagation Network
(FFBPN)—Training Algorithm . . . . . . . . . . . . . . . . . 178
7.2.13 Neural Network—Where to Go? . . . . . . . . . . . . . . . . 179
7.3 Genetic Algorithms—The Optimization Engine . . . . . . . . . . . . 180
7.3.1 Nature of Evolution—Genetic Algorithms . . . . . . . . . 180
7.3.2 Nature of Evolution—Genetic Algorithms . . . . . . . . . 181
7.3.3 GA—Population Initialization . . . . . . . . . . . . . . . . . . 182
7.3.4 GA—Fitness Evaluation Function . . . . . . . . . . . . . . . 182
7.3.5 GA—Parents Selection Scheme . . . . . . . . . . . . . . . . . 183
7.3.6 GA—Crossover Operations . . . . . . . . . . . . . . . . . . . . 183
7.3.7 GA—Mutation Operations . . . . . . . . . . . . . . . . . . . . 184
7.3.8 GA—Implementation . . . . . . . . . . . . . . . . . . . . . . . . 185
7.3.9 GA—Applications . . . . . . . . . . . . . . . . . . . . . . . . . . 186
7.4 Fuzzy Logic—The Fuzzification Engine . . . . . . . . . . . . . . . . . . 187
7.4.1 The World of Fuzziness, Chaos, and Uncertainty . . . . 187
7.4.2 The Birth of Fuzzy Logic . . . . . . . . . . . . . . . . . . . . . 188
7.4.3 Fuzzy Theory and the Uncertainty Principle . . . . . . . . 189
7.4.4 ICMR Model—Fuzzy Identification Module . . . . . . . 190
7.4.5 ICMR Model—Fuzzy Categorization Module . . . . . . 191
7.4.6 ICMR Model—Fuzzy Modeling Module . . . . . . . . . . 192
7.4.7 ICMR Model—Fuzzification Versus
Defuzzification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193
7.4.8 ICMR Model—Fuzzy Reasoning Module . . . . . . . . . 193
7.4.9 Case Study—Fuzzy Air-Conditioning Control
Systems, FACS . . . . . . . . . . . . . . . . . . . . . . . . . . . . 196
7.4.10 Applications of Fuzzy Logic in Daily Life . . . . . . . . . 199
7.4.11 Fuzzy Expert Systems on Financial Trading . . . . . . . . 201
7.4.12 Fuzzy-neuro System in Quantum Finance . . . . . . . . . 201
7.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
Contents xxiii

8 Chaos and Fractals in Quantum Finance . . . . . . . . . . . . . . . . . . . . 209

8.1 Basic Concepts and Formulation on Chaos Theory . . . . . . . . . . 210
8.1.1 The World of Nonlinear Dynamics . . . . . . . . . . . . . . 210
8.1.2 Deterministic World Versus Probabilistic World . . . . . 211
8.1.3 Chaotic Flow Versus Chaotic Maps . . . . . . . . . . . . . . 212
8.1.4 Origin of Chaos Theory—Lorenz Attractor . . . . . . . . 213
8.2 Characteristics of Chaotic Systems . . . . . . . . . . . . . . . . . . . . . . 215
8.2.1 Sensitive Dependence on Initial Conditions . . . . . . . . 216
8.2.2 Highly Nonlinear but Deterministic in Nature . . . . . . 217
8.2.3 Bifurcation Phenomenon . . . . . . . . . . . . . . . . . . . . . . 218
8.2.4 Applications of Chaos Theory . . . . . . . . . . . . . . . . . . 218
8.2.5 Chaos Theory Versus Uncertainty Principle . . . . . . . . 219
8.3 Fractals in Quantum Finance . . . . . . . . . . . . . . . . . . . . . . . . . . 221
8.3.1 Fractal Versus Chaos—Two Sides of a Coin . . . . . . . 221
8.3.2 A Brief History of Fractals . . . . . . . . . . . . . . . . . . . . 222
8.3.3 Self-similarity in Mandelbrot Set Fractals . . . . . . . . . . 223
8.3.4 Fractal Dimension . . . . . . . . . . . . . . . . . . . . . . . . . . 223
8.3.5 Natural Phenomena with Fractal Features . . . . . . . . . . 225
8.4 Applications of Chaos Theory and Fractals in Quantum
Finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
8.4.1 Chaos Theory in Quantum Finance . . . . . . . . . . . . . . 227
8.4.2 Fractal Finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227
8.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
9 Chaotic Neural Networks in Quantum Finance . . . . . . . . . . . . . . . . 235
9.1 Quantum World of Chaotic Oscillation . . . . . . . . . . . . . . . . . . 236
9.1.1 Multiverse and Quantum Theory . . . . . . . . . . . . . . . . 236
9.1.2 Chaos in Brain—Brainwave . . . . . . . . . . . . . . . . . . . 237
9.2 Chaotic Neural Oscillators . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
9.2.1 An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
9.2.2 Wang-Oscillator—The Structure . . . . . . . . . . . . . . . . 240
9.2.3 Wang-Oscillator—Bifurcation Diagram . . . . . . . . . . . 241
9.2.4 Wang-Oscillator—Major Contributions
and Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
9.3 Lee-Oscillator . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
9.3.1 The Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243
9.3.2 Lee-Oscillator—Bifurcation Diagram . . . . . . . . . . . . . 244
9.3.3 Lee-Oscillator—Potential Applications . . . . . . . . . . . . 245
9.4 Quantum Finance Oscillators (QFO) . . . . . . . . . . . . . . . . . . . . 246
9.4.1 The Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
9.4.2 The Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 246
xxiv Contents

9.5 Applications of QFO in Quantum Finance . . . . . . . . . . . . . . . . 248

9.5.1 The Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
9.5.2 QFO for Real-Time Financial Prediction Using
Chaotic Neural Networks . . . . . . . . . . . . . . . . . . . . . 250
9.5.3 Chaotic Fuzzy-neuro Oscillators—Quantum
Finance Signals (QFS) . . . . . . . . . . . . . . . . . . . . . . . 251
9.5.4 Chaotic Deep Neural Networks in Quantum
Finance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252
9.5.5 Chaotic Multiagent Trading System . . . . . . . . . . . . . . 252
9.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

Part II Quantum Finance Applications

10 Quantum Price Levels for Worldwide Financial Products . . . . . . . 263
10.1 Financial Markets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
10.1.1 An Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264
10.1.2 Primary Versus Secondary Market . . . . . . . . . . . . . . . 265
10.1.3 Types of Financial Markets . . . . . . . . . . . . . . . . . . . . 266
10.1.4 OTC Market and Market Makers . . . . . . . . . . . . . . . . 266
10.1.5 Forex, OTC, and Market Maker . . . . . . . . . . . . . . . . 267
10.1.6 Forex Trading Characteristics . . . . . . . . . . . . . . . . . . 268
10.1.7 Retail Forex Trading Platforms . . . . . . . . . . . . . . . . . 271
10.2 Supports and Resistances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
10.2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
10.2.2 Rising and Declining Supports and Resistances . . . . . 273
10.2.3 Role Reversal on Support and Resistance . . . . . . . . . 273
10.2.4 Trend Reversal on Support and Resistance . . . . . . . . . 274
10.2.5 Support and Resistance Versus QPLs . . . . . . . . . . . . 274
10.3 The MetaTrader (MT) Platform and MQL . . . . . . . . . . . . . . . . 276
10.3.1 An Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
10.3.2 MQL—MetaQuotes Language . . . . . . . . . . . . . . . . . . 277
10.3.3 MQL—A Brief Overview . . . . . . . . . . . . . . . . . . . . . 277
10.3.4 MQL—Time Series and Indicators Access
Functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 278
10.3.5 MQL—Trade Functions . . . . . . . . . . . . . . . . . . . . . . 279
10.3.6 MQL—Technical Indicators . . . . . . . . . . . . . . . . . . . 280
10.3.7 MQL Programming Tips . . . . . . . . . . . . . . . . . . . . . . 281
10.4 System Implementation—QPL Evaluation for Worldwide
Financial Products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 282
Contents xxv

10.5 MQL Program for QPLs Evaluation of Worldwide Financial

Products (QPL2019.mq4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283
10.5.1 System Flowchart and Main Program Logic . . . . . . . . 283
10.5.2 Program Declaration . . . . . . . . . . . . . . . . . . . . . . . . . 286
10.5.3 Program Initialization: init() . . . . . . . . . . . . . . . . . . . 286
10.5.4 Process 1: Calculate All K Values . . . . . . . . . . . . . . . 286
10.5.5 Process 2: READ ALL Daily Time Series . . . . . . . . . 289
10.5.6 Process 3: Calculate Mean (mu) and Standard
Deviation (sigma) of Return Array . . . . . . . . . . . . . . 290
10.5.7 Process 4: Generate the Returns Wavefunction
Distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
10.5.8 Process 5: Evaluate Lambda Value for the Returns
Wavefunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 292
10.5.9 Process 6: Using QP Schrodinger Equation
to FIND the First 21 Energy Levels . . . . . . . . . . . . . 295
10.5.10 Process 7: Program End . . . . . . . . . . . . . . . . . . . . . . 295
10.5.11 Implementation Results . . . . . . . . . . . . . . . . . . . . . . . 295
10.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 298
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 299
11 Time Series Chaotic Neural Oscillatory Networks for Financial
Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
11.2 TSCNON—System Architecture . . . . . . . . . . . . . . . . . . . . . . . 302
11.3 TSCNON—System Implementation Framework . . . . . . . . . . . . 304
11.4 Workshop #2—Quantum Finance Forecast System
(TSCNON.Mq4) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
11.4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
11.4.2 Main Program Logic . . . . . . . . . . . . . . . . . . . . . . . . . 307
11.4.3 Process #1—Program Initialization . . . . . . . . . . . . . . 309
11.4.4 Process #2—Check for Forecast Day . . . . . . . . . . . . . 315
11.4.5 Process #3—Loop over ALL Trading Products . . . . . 315
11.4.6 Process #4—Read FC_DT/HL/PF Data Files . . . . . . . 315
11.4.7 Process #5—Create TSCNON and Initialize
ALL LEE-Oscillators . . . . . . . . . . . . . . . . . . . . . . . . 321
11.4.8 Process #6—Register Time Series Data . . . . . . . . . . . 321
11.4.9 Process #7—TSCNON Training . . . . . . . . . . . . . . . . 321
11.4.10 Process #8—TSCNON Forecast . . . . . . . . . . . . . . . . 322
11.4.11 Process #9—Check Forecast Within Range
or NOT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
11.4.12 Process #10.1—Forecast Completed, Store
Forecast Results and Total Time Taken . . . . . . . . . . . 325
xxvi Contents

11.4.13 Process #10.2—Write FC_HL Data File

and Close All Files . . . . . . . . . . . . . . . . . . . . . . . . . . 325
11.5 TSCNON—Implementation Results . . . . . . . . . . . . . . . . . . . . . 326
11.6 TSCNON—System Performance . . . . . . . . . . . . . . . . . . . . . . . 329
11.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336
12 Chaotic Type-2 Transient-Fuzzy Deep Neuro-Oscillatory
Network (CT2TFDNN) for Worldwide Financial Prediction . . . . . . 339
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
12.2 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 342
12.2.1 An Overview on Type-1 and Interval Type-2
Fuzzy Logic Systems on Financial Prediction . . . . . . . 342
12.2.2 Lee-Oscillators with Retrograde Signaling
(LORS) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 345
12.3 CT2TFDNN—The System . . . . . . . . . . . . . . . . . . . . . . . . . . . 346
12.3.1 CT2TFDNN—System Framework . . . . . . . . . . . . . . . 346
12.3.2 Financial Signal Generator Module (FSGM) . . . . . . . 350
12.3.3 Chaotic Type-2 Transient-Fuzzification Module
(CT2TFM) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 350
12.3.4 GA-Based TOP-10 Financial Signals Selection
Module (GAT10FSSM) . . . . . . . . . . . . . . . . . . . . . . 353
12.3.5 Chaotic T2 Transient-Fuzzy Deep Neural
Network Module (CT2TFDNNM) . . . . . . . . . . . . . . . 356
12.4 CT2TFDNN—System Implementation . . . . . . . . . . . . . . . . . . . 359
12.4.1 Quantum Finance Forecast Center . . . . . . . . . . . . . . . 359
12.4.2 129 Worldwide Financial Products . . . . . . . . . . . . . . 359
12.4.3 CT2TFDNN Network Parameter Settings . . . . . . . . . . 360
12.4.4 System Implementation and Pilot Run . . . . . . . . . . . . 361
12.5 CT2TFDNN—Performance Analysis . . . . . . . . . . . . . . . . . . . . 361
12.5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
12.5.2 System Training Performance Analysis . . . . . . . . . . . 362
12.5.3 System Forecast Simulation Performance
Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
12.5.4 Actual Daily Forecast Performance Analysis . . . . . . . 366
12.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373
13 Quantum Trader—A Multiagent-Based Quantum Financial
Forecast and Trading System . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
13.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
13.2 Quantum Forecaster—Chaotic FFBP-Based Time Series
Supervised-Learning Neural Networks for Financial
Prediction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
Contents xxvii

13.2.1 Time Series Prediction Using Supervised-Learning-

Based Artificial Neural Networks . . . . . . . . . . . . . . . 377
13.2.2 Quantum Finance CSL Network—Chaotic
FFBP-Based SL Network for Time Series Financial
Prediction with Quantum Field Signals (QFF) . . . . . . 377
13.3 Quantum Trader—Chaotic RBF-Based Actor-Critic
Reinforcement-Learning Networks—Optimization of Trading
Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
13.3.1 An Overview of Reinforcement Learning (RL) . . . . . . 378
13.3.2 Discrete-Time Actor-Critic RL Model . . . . . . . . . . . . 380
13.3.3 Radial Basis Function Neural Network (RBFN)
for RL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 381
13.3.4 Quantum Finance CRL Network—Chaotic
RBF-Based Actor-Critic RL Network
for the Optimization of Trading Strategy . . . . . . . . . . 382
13.3.5 Quantum Finance CRL Network System
Architecture for the Optimization of Trading
Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
13.3.6 Quantum Finance CRL Network Learning
Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 384
13.4 System Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
13.4.1 Quantum Finance—TEQNA 5-Tier Architecture . . . . 385
13.4.2 Quantum Trader System Implementation Using
129 Worldwide Financial Products . . . . . . . . . . . . . . 386
13.4.3 Quantum Trader—System Implementation . . . . . . . . . 387
13.4.4 Quantum Finance Daily Forecast at QFFC . . . . . . . . . 389
13.4.5 Quantum Forecaster Performance Test . . . . . . . . . . . . 390
13.4.6 Quantum Trader Trading Strategy Performance
Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
13.4.7 Quantum Trader Worldwide Trading Strategy
Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394
13.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
14 Future Trends in Quantum Finance . . . . . . . . . . . . . . . . . . . . . . . . 399
14.1 Introduction—The Dawn of Quantum Computing . . . . . . . . . . . 400
14.2 Two Sides of the Same Coin—Hard Quantum Computing
Versus Soft Quantum Computing . . . . . . . . . . . . . . . . . . . . . . . 400
14.3 From Quantum Finance to AI Fintech . . . . . . . . . . . . . . . . . . . 401
14.4 Quantum Finance System Development Platform . . . . . . . . . . . 402
xxviii Contents

14.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 403

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405

Appendix A: List of 139 Financial Products . . . . . . . . . . . . . . . . . . . . . . . 407

Appendix B: List of 39 Financial Indicators . . . . . . . . . . . . . . . . . . . . . . . 411
About the Author

Dr. Raymond S. T. Lee founder of quantum finance forecast center with over 20
years of IT consultancy, R&D experiences in AI, chaotic neural networks, intelli-
gent fintech system, quantum finance, and intelligent e-commerce systems, had
successfully commercialized his AI-fintech invention at business sectors in China
and Hong Kong.
He attained his B.Sc. (Physics) from Hong Kong University in 1989, and M.Sc.
(Information Technology) and Ph.D. (Computer Science) from Hong Kong
Polytechnic University in 1997 and 2000, respectively.
After graduation from Hong Kong University, he joined the Hong Kong
Observatory of the Government of the Hong Kong Special Administrative Region
as a meteorological scientist on weather forecasting and developing numerical
weather forecast system from 1989 to 1993.
From academic perspective, he had worked at the Department of Computing of
Hong Kong Polytechnic University (HKPolyU) as Lecturer, and promoted as
Assistant Professor in 2000 and Associate Professor in 2005, respectively. During
this time, he had published over 90+ publications and author of six textbooks and
monographs covering the fields at AI, chaotic neural networks, AI-based fintech
systems, intelligent agent technology, chaotic cryptosystems, ontological agents,
neural oscillators, biometrics, weather simulation, and forecasting systems.
From commercial perspective, he was invited to join Leanda Investment Group
in China (2012–2017) as Group CTO and Chief Analyst to implement his
self-invented and patented AI-fintech invention—quantum finance forecast system
on major commodities in China for 1,000+ investors.
In March 2017, he set up the quantum finance forecast center (aka QFFC) (http://
qffc.org), a nonprofit making, AI-fintech R&D, and worldwide financial forecast
center, that aims at the R&D and provision of a wholly FOC and open platform for
worldwide traders and individual investors to acquire free knowledge of worldwide
129 financial product forecasts based on the state-of-the-art AI, chaotic neural
networks and quantum finance technologies. At present, QFFC has over 10,000+
registered worldwide members, which consists of worldwide professional traders,

xxix
xxx About the Author

quants, and independent investors from major fund houses and financial institutions
using the free quantum finance daily forecasting services.
Upon the completion of the fully automated quantum finance forecast system, he
joined Beijing Normal University-Hong Kong Baptist University United
International College (UIC) in China as Associate Professor in 2018 to further
his R&D works on quantum finance and to contribute his knowledge on quantum
finance, AI-fintech, chaotic neural networks, and related intelligent forecast systems
to the fellow students and the community in China.
Abbreviations

ANN Artificial Neural Networks

CBTF Chaotic Bifurcation Transfer Function
CDNONRS Chaotic Deep Neuro-oscillatory Network with Retrograde
Signaling
CNON Chaotic Neural Oscillatory Network
CT1-FNON Chaotic T1 Fuzzy Neuro-oscillatory Network
CT2TFDNN Chaotic T2 Transient-Fuzzy Deep Neuro-oscillatory Network
with Retrograde Signaling
CT2TFL Chaotic T2 Transient-Fuzzy Logic
CT2TFLS Chaotic T2 Transient-Fuzzy Logic System
CT2TFMF Chaotic T2 Transient-Fuzzy Membership Function
CTU Chaotic Transfer Unit
DNN Deep Neural Network
EC System Evolutionary Computing System
FFBPN Feedforward Backpropagation Networks
FFS Fuzzy Financial Signals
FFSCM Fuzzy Financial Signals Crossover Module
FFSFEM Fuzzy Financial Signals Fitness Evaluation Module
FFSMM Fuzzy Financial Signals Mutation Module
FFSPGM Fuzzy Financial Signals Population Generation Module
FFSV Fuzzy Financial Signal Vector
FLS Fuzzy Logic System
FOU Footprint Of Uncertainty
FSGM Financial Signal Generator Module
GA Genetic Algorithms
HFAPT High-Frequency Algorithmic Program Trading
IT2FLS Interval Type-2 Fuzzy Logic System
IT2-FNN Interval Type-2 Fuzzy-neuro Network
LORS Lee-oscillator with Retrograde Signaling
MQL MetaQuotes Language

xxxi
xxxii Abbreviations

MT4 MetaTrader4
NGPSM New Generation Population Selection Module
PCA Principal Component Analysis
QAOM Quantum Anharmonic Oscillatory Model
QFFS Quantum Finance Forecast Centre
QPL Quantum Price Level
RMSE Root Mean Square Error
STR State Transition Region
SVM Support Vector Machine
T2FL Type-2 Fuzzy Logic
T2FLS Type-2 Fuzzy Logic System
T2FMF Type-2 Fuzzy Membership Function
TOP10-FFSSM TOP-10 Fitness FFS Selection Module
 Part I
Quantum Finance Theory
Chapter 1
Introduction to Quantum Finance

As far as the laws of mathematics refer to reality, they are not

certain; and as far as they are certain, they do not refer to reality.
Albert Einstein (1879–1955)

What is the world of reality?

Is there any physical laws governing our world?
Does everything happen simply by chance or predetermined?
Professor Albert Einstein’s quotation gave us an exceedingly admirable description
of the true nature as regards reality. The fact is that this fundamental question remains
unsolved throughout the evolution of human civilization. Does it mean that we can’t
do anything about it? The answer is definitely no.
Human beings by nature are exceptionally adequate with one trait in countless
years—live in our predefined models. Even though we never know the real aspect
of the world of reality, our minds can unconsciously create the model of reality
and make us believe that this is the real world (reality) we live in. In science, we
follow the same logic. To study any phenomenon that occurs in the real world (for
instance, weather), we build a so-called model to represent it. By observing and
studying object and matter dynamics within this model, we try to formulate some
laws of dynamics and/or equations to generalize their activities to find out some
patterns or even to predict what will happen in the future. In finance, we follow the
same footsteps. Quantum finance, as a new age of finance technology that tackles
the ultimate problem of finance—the quantum world of finance—is no exception.
In this chapter, we will begin with the origin of quantum finance—the way we
percept and model our physical world, from Newtonian’s classical physical world to
Heisenberg’s quantum world of subatomic particles. Next, we will study a unique
feature of the quantum world, the wave–particle duality phenomena of quantum

© Springer Nature Singapore Pte Ltd. 2020 3

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_1
4 1 Introduction to Quantum Finance

particles, and how such phenomena are related to quantum finance. After that, we
will study quantum field theory and the birth of quantum finance followed by the
Heisenberg’s uncertainty principle in quantum mechanics, and how it relates to two
important theories in AI—chaos theory and fuzzy logic. Lastly, we will explore major
components of quantum finance—the four-tier concentric sphere quantum finance
model.

1.1 A Tale of Two Worlds—From the World of General

Relativity to the World of Quantum Mechanics

From ancient Greeks to the dawn of science, philosophers, mathematicians, physi-

cists, psychologists, and natural scientists over centuries strive to better understand
the world of reality we live in. We observe all sorts of phenomena, ranging from
weather changes in our atmosphere to comets and supernova of the universe. Based
on these observations, we discover some common patterns and dynamics. Above
these common patterns and dynamics, we build certain physical and mathematical
models, equations, and laws of dynamics to make sense of them and to generalize
them to what we called knowledge. If we are fortunate to exert our sagacity, we can
derive from these laws of dynamics and mathematical models not only to extract the
patterns that had already occurred but also have a chance to glimpse the future.
Finance, especially the worldwide financial markets, is believed to be the phe-
nomena that resulted in high-level intellectual activities and the collective behav-
iors of worldwide investors and market regulators. Different from classical finance
which focuses on the statistical models and analysis of financial markets and market
patterns, quantum finance applies the latest research of quantum field theory and
quantum mechanics to study the actual dynamics of quantum particles (so-called
quantum financial particles, QFPs) in every financial market. By exploring these
quantum financial dynamic activities, we would have a better chance not only to
observe the financial patterns but also to predict their future movements.
To study the actual dynamics of financial particles, why do we use quantum model,
but not others?
To answer this question, we need to explore a more fundamental question: What
are the laws of dynamics governing the world of reality we inhibit?
Thanks to the prominent minds and scientific discovery in the past centuries, con-
temporary physics and cosmology tell us that the world of reality can be defined
by three physical models: world of classical mechanics discovered by Isaac New-
ton (1643–1727), cosmic world of general relativity discovered by Albert Einstein
(1879–1955), and the subatomic world of quantum mechanics discovered by Werner
Heisenberg (1901–1976), together with the three distinctive laws and equations to
govern these worlds as shown in Fig. 1.1.
1.1 A Tale of Two Worlds—From the World … 5

Einstein’s Newton’s Heisenberg’s

Cosmic world Physical world Subatomic world
General Relativity Classical Mechanics Quantum Mechanics
E = mc2 F = ma

Fig. 1.1 Laws of physical worlds (Tuchong 2019a, b and c)

1.1.1 Newton’s Physical World of Classical Mechanics

Sir Isaac Newton (1643–1727) was an English mathematician, physicist,

astronomer, and theologian. His book Mathematical Principles of Natural Philos-
ophy (Brackenridge 1999) published in 1687 laid out the foundations of classical
mechanics, together with his remarkable works on optics and calculus. These three
great contributions bestowed on him as one of the most important prominent minds
and influential scientists in the human history (Pourciau 2006; Stewart 2004).
In Newton’s model, the world of reality is called the physical world—the world
we live and experience physically. All matter and object motions in this physical
world are strictly governed by his three laws of dynamics known as Newton’s laws
of motion. Causes and effects (so-called causation) of these matter and object motions
must follow these laws agreeably in this world almost without any exceptions. Yet,
modern cosmology and quantum mechanics tell us that it is not always the case. The
fact is that the laws of physics only hold under two conditions:
(1) Objects and matters of normal size, what we called macroscopic objects; and
(2) Object motions at normal speed, as compared with the speed of light (c = 3 ×
108 ms−1 ).

1.1.2 Einstein’s Cosmic World of General Relativity

General relativity (GR, also known as the general theory of relativity) is the geo-
metric theory of gravitation coined by Professor Albert Einstein (1879–1955) at the
presentation of his work The Field Equations of Gravitation at the Prussian Academy
of Science in November 1915 (Einstein 1915), and The Foundation of the General
Theory of Relativity published by Annalen der Physik in 1916 (Einstein 1916), which
set the cornerstone of general relativity along with quantum mechanics as the two
pillars of modern physics (Woodhouse 2007). He is best known to the general public
6 1 Introduction to Quantum Finance

for his groundbreaking mass–energy equation E = mc2 , which has been considered
as “the world’s most famous equation” after Newton’s second law of motion F =
ma.
Different from centuries of beliefs that time as a special one-way, linear, and non-
reversible quantity, the groundbreaking idea of general relativity is the consideration
of time as a dimension identical to the three dimensions of space we are known
of. General relativity generalizes Einstein’s special relativity and Newton’s law of
universal gravitation, providing a unified description of gravity as a geometric prop-
erty of space and time, known as spacetime—a critical mass of modern cosmology
(Woodhouse 2007; Böhmer 2016; Carroll 2016).
In Einstein’s world of general relativity, space, time, and matters are inter-
connected to each other. Even his early work on special relativity already introduced
the notion of time dilation—in which time is not an absolute constant quantity but
can vary according to the observer’s so-called frame of reference (Chaturvedi et al.
2006). In layman’s term, to describe matters and objects with normal mass and veloc-
ity (i.e., much slower than the speed of light), Newton’s laws of motion are good
enough to do the job. But, when matters and objects with normal mass move close
to the speed of light, it is the time to apply special relativity. In case that matters and
objects are extremely massive, such as planets and black holes in which even light
and time can be distorted by their massive gravitational fields as visual interpretation
is shown in Fig. 1.1a, general relativity should be applied. In fact, general relativity
is the basis of current cosmological models of a consistently expanding universe and
Big Bang Theory.

1.1.3 Heisenberg’s Subatomic World of Quantum Mechanics

Professor Werner Karl Heisenberg (1901–1976) was a German theoretical physi-

cist and one of the key figures of quantum mechanics. He coined the term quantum
mechanics in his influential paper Quantum theoretical re-interpretation of kinematic
and mechanical relations published at Zeitschrift für Physik in September 1925
(Heisenberg 1925). Quantum mechanics (QM) and quantum field theory describe
nature at the smallest scales of energy levels of atoms and subatomic particles. Dif-
fering from Newtonian classical physics, quantum mechanics (Norsen 2017) has five
characteristics:
1. Objects and matters in this quantum world (or what we called subatomic world)
are called quantum particles, and their dynamics are governed by the famous
Schrödinger equation.
2. Quantization—a unique phenomenon in quantum mechanics—describes energy,
momentum, angular momentum, and related quantities in this subatomic world
as discrete states and energy levels.
3. Wave–particle duality states that objects and matters in this world have
characteristics of both particles and waves simultaneously.
1.1 A Tale of Two Worlds—From the World … 7

4. Heisenberg’s uncertainty principle states that there are natural limits of preci-
sion for paired quantity of every quantum particle, for example, position and
momentum of any quantum particle.
5. Quantum entanglement—a unique phenomenon in quantum mechanics—states
that multiple quantum particles are linked (interacted) together in a way such
that the measurement of one particle’s quantum state determines the possible
quantum states of the other particles, even they are separated by very far distance
(Sen 2017).
As a groundbreaking theory to describe all fundamental particles and their events,
quantum mechanics believes to be a perfect tool to model any fundamental phenom-
ena in the physical world such as finance, for example. Figure 1.1c shows a visual
interpretation of Heisenberg’s subatomic world of quantum mechanics.

1.2 Two Sides of the Same Coin—Wave–Particle Duality

in Quantum Finance

1.2.1 Wave–Particle Duality and Double-Slit Experiment

Wave–particle duality is the fundamental concept in quantum mechanics which states

that every quantum particle (or so-called subatomic particle) coexists in two different
states, waves and particles. In classical physics, we all learnt that waves and particles
are two totally different entities and notions governed by two distinct physical laws:
Newton’s laws of motion govern all motions of physical objects and matters, whereas
Maxwell’s wave equations govern all the physical dynamics of waves. As we will
explain the details in the coming chapters, Schrödinger equation is the foundation of
quantum mechanics that provides a unique way to visualize how wave and particle
dynamics are related and coexist from the mathematical perspective in this quantum
world.
Can we visualize wave–particle duality in the physical world?
The answer is yes: The famous double-slit experiment for photon particles.
Figure 1.2 shows the experiment setup of double-slit experiment.
The setup of double-slit experiment in quantum mechanics borrows from Profes-
sor Thomas Young’s famous double-slit experiment of interference phenomena in
classical wave theory. But this time, laser beam was used as the light source to emit
photons in the experiment as shown. Same as the classical double-slit experiments,
a wall with two slits of openings was set between the light source and the screen.
The whole experiment consists of two parts: part 1 experiment used photoelectric
detector to detect the motion of every photon emission reflecting its full path to the
destination (screen) and part 2 experiment was the same setup except that the detector
was switched off.
Although the two setups were basically the same, yet interesting effects occurred.
Part 1’s experiment that used photoelectric detector showed that particle motions of
8 1 Introduction to Quantum Finance

Experiment Part I Experiment Part 2

With Photoelectric Detector On With Photoelectric Detector Off

Random parƟcle paƩerns Interference wave paƩerns

Photoelectric
Detectors

Light Source (Laser Beam) Light Source (Laser Beam)

Fig. 1.2 Experimental setup of double-slit experiment of photon particles

light photon projection were randomly distributed as every photon path is a particle
motion-independent event. Part 2’s experiment that did not use photoelectric detector
showed that alterative bright-and-dark bands are similar in any typical wave interfer-
ence phenomena. What do they mean? What we watch (observe) can really affect the
experimental results? How can light particles (photon) be sometimes particle-like
and sometimes wave-like?
Quantum mechanics provides a simple but extraordinary explanation: first, all
quantum particles coexist in two totally different states: wave and particle states,
we called it wave–particle duality. Second, how they behave all depend on how we
observe (detect) their motions (phenomena). In other words, if we use photoelectric
detector to detect light, it will behave as photon particles, that is, photoelectric detec-
tor can only detect photons. On the other hand, if we don’t use any photoelectric
detector, it will behave as its usual state, that is, light waves with typical interference
phenomena, which means that the observer oneself (or the detector itself) will affect
the experimental results; it is rather against the traditional beliefs in science in which
the observer should not be part of the experiment. Is it true?
Despite the notion that was extraordinary among science community in the early
twentieth century, it was not a novel concept in philosophy. Professor Immanuel
Kant’s (1724–1804) remarkable work Critique of Pure Reason published in 1781
(Kant 2008) stated that: the world of reality we percept is just the model reality
shaped by the senses and mind. In layman’s term, it can be interpreted as wearing
1.2 Two Sides of the Same Coin—Wave–Particle Duality … 9

a pair of colored spectacles to see the world. Different colors of the world naturally
depend on the color of the spectacles we wear. As mentioned at the beginning of this
chapter, humans are good at creating models to reflect everything we see and percept
in our minds, and the observation results we obtain naturally depend on the model
we create.
In fact, the major significance of the wave–particle duality is that all behaviors of
light and matter can be explained through a single equation which relates quantum
particle’s wave dynamics with its particle dynamics—the Schrödinger equation. This
ability to describe reality in the form of waves is the core of quantum mechanics and
is also the quintessence of this book.

1.2.2 Wave–Particle Duality in Finance

In financial markets, can we observe any wave–particle duality?

The answer is definitely “yes” and “always”.
Experienced financial analyst or trader will tell us that financial markets appear
with both particle and wave properties constantly.
The two pillars of modern financial analytical tools, technical analysis and chart
analysis, are very good examples of such wave–particle duality.
Technical analysis (Murphy 1999) describes the financial markets as physical
motion of price (or index); we analyze the price movement in respect of moving
averages, momentums, and acceleration, just like any physical objects we encounter
in the physical world, whereas chart analysis shapes the financial markets as financial
waves and patterns. From that, it creates and defines various major reversal and trend
patterns such as golden ratio patterns, Fibonacci patterns, and the famous Elliott wave
patterns shown in Fig. 1.3 that gave credence to affect the market price trends, trading
patterns, and individual reasonings for investors to decide the pre-eminent time to
trigger the buy or sell decisions (Borden 2018; Brown 2012; Bulkowski 2005).

100% of Wave 1 50.0%, 61.8%

5 or 161.8% of Wave 4 B of Wave A

161.8% or 261.8%
3 of Wave 1

161.8%, 100%,
A
1 61.8%, or 50.0%
4 38.2%, 50.0% , of Wave 5
61.8% of Wave 4
C 100%, 161.8%
of Wave A, or
2 50.0% or 61.8%
161.8% of Wave B
of Wave 1

Fig. 1.3 Elliot waves chart principle in financial analysis

10 1 Introduction to Quantum Finance

In other words, in the financial world at any particular instance, when we can
observe (or measure) the price (or index value) of any financial product such as the
Dow Jones index (DJI), we treat this as a particle, and use technical analytical tools
to study and analyze it. At the same time, we can also observe not only on price
but also the entire financial time series as waves and patterns; use chart analysis to
discover the embedded waves and patterns.
In quantum finance, can we model financial market using quantum mechanics and
quantum field theory?

1.3 Quantum Field Theory and the Birth of Quantum

Finance

Quantum field theory is a unified theoretical framework to describe motion and

dynamics of fundamental particles. It consists of
• Maxwell field theory,
• Special relativity, and
• Quantum mechanics.
Quantum field theory regards all fundamental particles as excited states (quanta)
of their underlying fields (quantum fields).
In quantum field theory, atoms absorb and emit electromagnetic radiation, as tiny
oscillators with their energies can only take on a series of discrete values—quantum
anharmonic oscillators (QAHOs).
This process of restricting energies to discrete values is called quantization.
Their interactions and dynamics can be visually represented by the famous
Feynman diagrams as shown in Fig. 1.4.
Quantum finance (QF) is a newly developed interdisciplinary subject applying
quantum mechanics (QM) and quantum field theory (QFF)—econophysics. The first

Fig. 1.4 Feynman diagram

1.3 Quantum Field Theory and the Birth of Quantum Finance 11

published work An Introduction to Econophysics—Correlations and Complexity in

Finance was written by Mantegna and Stanley in (1999).
Modern econophysics view this as an application of Brownian motion—funda-
mental phenomenon of statistical physics.
During the past decades, various methods and theories had been proposed for
stock price/returns analysis, option pricing, and portfolio analysis.
In this book, we will study how to combine quantum mechanics and quantum field
theories including Feynman’s path integrals, Hamiltonians, quantum wave functions,
quantum oscillators with contemporary AI tools such as fuzzy logics, genetic algo-
rithms, and chaos and fractals to model and interpret worldwide financial markets
for financial forecast and quantum trading.

1.4 Heisenberg’s Uncertainty Principle in Finance

and Modern AI Technology

In 1927, German physicist Professor Werner Heisenberg stated that the more pre-
cisely the position of some particle is determined, the less precisely its momentum—
the famous Heisenberg’s uncertainty principle (Gribbin 1984).


σx σ p ≥ (1.1)
2
where  is the reduced Planck constant.
Unlike classical physics, uncertainty principle states that all events occur in quan-
tum states and will (only will) collapse into a particular reality when the observer
takes the measurement (or observe the event).
This phenomenon is associated with the famous Schrödinger’s cat paradox
(Gribbin 1984; Monroe et al. 1996).

1.4.1 Schrödinger’s Cat Paradox

• In this paradox as shown in Fig. 1.5, a cat, a flask of poison, and a radioactive
source are placed in a sealed box.
• If an internal monitor (e.g., GM counter) detects radioactivity (i.e., a single atom
decaying), the flask is shattered and releases poison which kills the cat.
• The interpretation of quantum mechanics implies that after a while, the cat is
simultaneously alive and dead. Yet, when one looks in the box, the cat is either
alive or dead (Bhaumik 2017).
• This poses the question of when exactly quantum superposition ends and reality
collapses into one possibility or the other (Stamper-Kurn et al. 2016).
12 1 Introduction to Quantum Finance

Fig. 1.5 Schrödinger’s cat paradox in quantum mechanics

1.4.2 Uncertainty Principle in Finance

One may wonder, does uncertainty principle (phenomena) occur in financial markets?
The answer is: always.
Any experienced traders can tell us that there is no 100% sure in financial market.
Anything can happen in every second. The only thing we are 100% sure and confirm
is the time when we look at the market, which is similar to what Einstein said in his
famous quote:
I like to think that the moon is there even if I am not looking at it
Albert Einstein (1879–1955)

In other words, without actually observing the market, anything can occur.
If that is the case, how can we model it by using the technology we have? Or, is
it merely a fancy thought?
In fact, there are two technologies that provide an excellent analog to uncertainty
principle in modern AI (Russell and Norvig 2015).
They are fuzzy logics and chaos theory.
Fuzzy logics (FLs): In the world of fuzzy logic, everything is uncertain. Fuzzy
logic (Zadeh and Aliev 2019) provides an easy-to-implement solution to model mul-
tiple attributes that can be occurred at the same time the so-called fuzzy variables
(FVs).
Chaos theory (CT): In the world of chaos theory, everything is a complex sys-
tem, so-called deterministic chaos (Schuste 2005) (Fig. 1.6). Chaos theory provides
a framework and mathematical model to simulate highly chaotic or random-like
phenomena, such as weather patterns and financial markets (Lee 2005).
1.4 Heisenberg’s Uncertainty Principle in Finance and Modern AI Technology 13

Fig. 1.6 Deterministic

chaos in chaos theory and
butterfly effect

We will study these two fascinating and useful topics in Chap. 7—AI powerful
tools in quantum finance.

1.5 Basic Components of Quantum Finance

In this book, we will explore how quantum mechanics and quantum field theory can
be used to model financial markets. More importantly, we will study how modern
AI technologies such as artificial neural networks, fuzzy logics, genetic algorithms,
chaos theory, and fractals can be integrated with quantum finance model to design
and implement intelligent real-time financial forecast and quantum trading systems.
Figure 1.7 shows the concentric sphere model of quantum finance.
1. First Tier—Energy Field Layer
• The core,
• Provides quantum field—quantum price field (QPF).
2. Second Tier—Chaotic Neural Network Layer
• Provides the neural dynamics in quantum finance,
• Supports neural oscillators and chaotic neural networks.
3. Third Tier—AI-fintech Layer
Provides AI-finance technology tools,
14 1 Introduction to Quantum Finance

Fuzzy
Logics

Chao c Neural
Networks

QUANTUM
Chaos
PRICE Fractals
Theory FIELD

Neural
Oscillators

Gene c
Algorithms

Fig. 1.7 Concentric sphere model of quantum finance

Supports fuzzy logics (FLs), genetic algorithms (GAs), chaos theory (CT),
fractals, support vector machine (SVM), etc.
4. Fourth Tier—Application Layer
• Quantum price levels (QPLs),
• Short-term price prediction,
• Long-term trend prediction, and
• Intelligent multiagent-based trading systems.

1.6 Conclusion

In this chapter, we introduced the foundation of quantum finance with exploration

of basic dynamics laws of the world we live in from three different perspectives,
ranging from macroscopic world of general relativity to microscopic world of quan-
tum mechanics. We also reviewed key properties and features of quantum mechanics
including quantization, wave–particle duality, and Heisenberg’s uncertainty prin-
ciple. More importantly, we studied how these unique and extraordinary quantum
phenomena are related and can be observed in real-world financial markets, which
constituted the motivation in developing quantum finance model in the following
chapters.
References 15

In the last section, we introduced quantum finance four-tier concentric sphere

model, the central framework to integrate quantum finance system core model with
modern AI technology as the neural network, AI-fintech, and application tiers which
constituted Part I of this book—quantum finance theory.
In Part II, we will study AI-fintech application including the implementa-
tion of quantum price level (QPL) for worldwide financial products in Chap. 10,
design and implementation of worldwide 120+ quantum finance forecast system in
Chap. 11, quantum finance prediction using chaotic transient-fuzzy deep network in
Chap. 12, and the implementation of intelligent agent-based quantum trading system
in Chap. 13.

Problems

1.1 What is quantum finance? State and explain briefly the major difference
between quantum finance and traditional finance theory.
1.2 What is the four-tier concentric sphere model of quantum finance? How can
it relate to contemporary AI technology?
1.3 What is wave–particle duality phenomenon in quantum theory? Give an exam-
ple of wave–particle duality phenomenon in financial market and explain how
it works.
1.4 State and explain the three major components of quantum field theory and
discuss how they are related to quantum finance.
1.5 State and explain the difference between quantum mechanics and quantum
field theory.
1.6 What are the basic limitations of Newtonian classical mechanics?
1.7 Why the understanding of nature of time is critical in modern finance?
1.8 State and explain the five characteristics of quantum mechanics, and how they
are related to quantum finance.
1.9 State and explain the double-slit experiment in quantum mechanics, and its
analog to the financial market.
1.10 State and explain how Kant’s theory provides a probable explanation to
wave–particle duality in quantum theory, and hence quantum finance.
1.11 What are technical analysis and chart analysis in finance engineering? Discuss
and explain how these analytical techniques are related to wave–particle duality
in quantum finance.
1.12 State and explain the Schrödinger’s cat paradox in quantum mechanics. So, at
any moment, is the cat alive or die? why? and how it relates to the financial
markets?
1.13 What are Elliot waves in finance? Discuss and explain how Elliot waves can
be interpreted by quantum finance.
1.14 If everything is uncertain according to uncertainty principle, how can we
predict the future such as financial prediction?
1.15 What are the roles of fuzzy logics and chaos theory in quantum finance?
16 1 Introduction to Quantum Finance

References

Bhaumik (2017) M. L. Is Schrödinger’s Cat Alive? Quanta 6(1): 70–80.

Böhmer (2016) C. G. Introduction To General Relativity And Cosmology, WSPC.
Borden, C. (2018) Fibonacci Trading: How to Master the Time and Price Advantage. UK: McGraw-
Hill Education.
Brackenridge, J. B. (1999) ISAAC NEWTON. the Principia: Mathematical Principles of Natural
Philosophy, 3rd Edition (1726). Newly Translated by I. Bernard Cohen and Anne Whitman.
Berkeley: University of California Press.
Brown, C. (2012) Elliot Wave Principle: Elementary Concepts, Wave Patterns, and Practice
Exercises. New York: Bloomberg Press.
Bulkowski, T. N. (2005) Encyclopedia of Chart Patterns. UK: Wiley.
Carroll (2016) Spacetime And Geometry: An Introduction To General Relativity. Pearson Education
India.
Chaturvedi, S. et al. (2006) Space, time and relativity. Resonance 11(7): 14–29.
Einstein, A. (1915) Die Feldgleichungen der Gravitation. Sitzungsberichte der Preussischen
Akademie der Wissenschaften zu Berlin: 844–847.
Einstein, A. (1916) The Foundation of the General Theory of Relativity. Annalen der Physik. 354
(7): 769.
Gribbin, J. (1984) In Search of Schrödinger’s Cat: Quantum Physics and Reality. Bantam Books;
Reprint edition.
Heisenberg, W. (1925) Über quantentheoretische Umdeutung kinematischer und mechanischer
Beziehungen. Zeitschrift für Physik. 33 (1): 879–893.
Kant, I. (2008) Critique of Pure Reason (Translated by M. Weigelt & M. Muller). Penguin Classics.
Lee, R. S. T. () Advanced Paradigms in Artificial Intelligence From Neural Oscillators, Chaos
Theory to Chaotic Neural Networks. Advanced Knowledge International, Australia, 2005.
Mantegna, R. N. and Stanley, H. E. (1999) Introduction to Econophysics: Correlations and
Complexity in Finance, Cambridge University Press.
Monroe et al, C. (1996) A “Schrödinger Cat” Superposition State of an Atom. Science 272(5265):
1131–1136.
Murphy, J. J. (1999) Technical Analysis of the Financial Markets: A Comprehensive Guide to
Trading Methods and Application. New York: New York Institute of Finance.
Norsen, T. (2017) Foundations of Quantum Mechanics: An Exploration of the Physical Meaning of
Quantum Theory (Undergraduate Lecture Notes in Physics). Springer.
Pourciau, B. (2006) Newton’s Interpretation of Newton’s Second Law. Archive for History of Exact
Sciences 60(2): 157–207.
Russell, S. and Norvig, P. (2015) Artificial Intelligence: A Modern Approach, 3rd edition. Pearson
Education India.
Schuste, H. G. and Just, W. (2005) Deterministic Chaos: An Introduction. Wiley-VCH.
Sen, A. (2017) Quantum Entanglement and its Applications. Current Science 112(7): 1361.
Stamper-Kurn, D. M. et al. (2016) Verifying quantum superpositions at metre-scales. Nature 537,
(7618): E1-E2.
Stewart, I. G. (2004) The principia: mathematical principles of natural philosophy. Studies in History
and Philosophy of Science Part A 35(3): 665–667.
Tuchong (2019a) Einstein’s Cosmic World of General Relativity. http://stock.tuchong.com/image?
imageId=289742135997235449. Accessed 22 Aug 2019.
Tuchong (2019b) Newton’s Physical World of Classical Mechanics. http://stock.tuchong.com/
image?imageId=217614207574867979. Accessed 21 Aug 2019.
Tuchong (2019c) Heisenberg’s Subatomic World of Quantum Mechanics. http://stock.tuchong.com/
image?imageId=423495114220503185. Accessed 22 Aug 2019.
Woodhouse, N. M. J. (2007) General Relativity (Springer Undergraduate Mathematics Series).
Springer.
Zadeh, L. A. and Aliev, R. A. (2019) Fuzzy Logic Theory and Applications: Part I and Part II.
World Scientific Publishing Company.
Chapter 2
Quantum Field Theory for Quantum
Finance

Anyone who is not shocked by quantum theory has not

understood it.
Niels Bohr (1885–1962)

What is quantum field theory?

How does it relate to quantum mechanics?
How does it relate to quantum finance?
The quote by Professor Niels Bohr, renowned Danish physicist and Nobel prizewin-
ner for Physics in 1922, is exceptionally accurate. Anyone studies in quantum theory
will soon discover that models, basic concepts, and the related phenomena associated
with quantum theory are completely different from what we have learnt in classical
physics. It involves many riveting topics and phenomena that have never been con-
sidered or even thought about from the world of classical physics. In fact, it shows
us an entirely new perspective to interpret the world of reality we live in.
As mentioned in Chap. 1, the major component in quantum theory (QT)—quan-
tum field theory—is originated from quantum mechanics (QM), which is an inte-
grated theoretical physics theory with the combination of classical field theory, quan-
tum mechanics, and special relativity.
In this chapter, we will first begin with a brief history of quantum mechanics.
Second, we will explore four unique features and phenomena in quantum mechan-
ics which include wave–particle duality, Heisenberg’s uncertainty principle, quanti-
zation of energy, and quantum entanglement.

© Springer Nature Singapore Pte Ltd. 2020 17

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_2
18 2 Quantum Field Theory for Quantum Finance

Lastly, we will study four main topics in quantum field theory: classical field
versus quantum field, Feynman’s path integral, and quantum oscillators which are
related to our study of quantum finance theory.
In Chaps. 4 and 5, basic understanding of quantum mechanics and quantum field
theory are vital not only to realize the motivation in acquiring quantum theory to
model the dynamics of financial markets but also to set out a concrete foundation for
further study and research on quantum finance theory with related theoretical and
mathematical models.

2.1 Quantum Mechanics—The Basics

Quantum mechanics (QM), known as quantum physics or quantum theory, is a the-

oretical physics coined by Professor Werner Heisenberg in 1925 (Heisenberg 1925)
which describes the nature and dynamics of matters at the smallest scales including
atoms and subatomic particles.
Prior QM, the whole world of physics (so-called physical world), is governed by
two distinctive laws of physics:
1. Newton’s law of motion governs nature, dynamics, and motions of all objects
(macroscopic objects);
2. Maxwell’s electromagnetic wave equation governs nature, properties, and the
propagation of electromagnetic waves within the material world.
Quantum mechanics differs from classical physics in four aspects:
1. Wave–particle duality—objects have characteristics of both particles and waves;
2. Quantization—all energy, momentum, angular momentum, and other quantities
of a bound system are restricted to discrete values;
3. Heisenberg’s uncertainty principle—limits to the precision with which quantities
can be measured;
4. Quantum entanglement—instantaneous interactions between subatomic particles
even a great distance.
See Fig. 2.1.
2.1 Quantum Mechanics—The Basics 19

Fig. 2.1 Wave functions of the electron in a hydrogen atom at different energy levels (PoorLeno
2019). Note Quantum mechanics cannot predict the exact location of a particle in space, only the
probability of finding it at different locations. The brighter areas represent a higher probability of
finding the electron

2.1.1 A Brief History Quantum Mechanics

Table 2.1 shows a snapshot of distinguished figures in the course of quantum mechan-
ics since 1800s.

2.1.2 Quantum Mechanics—Wave–Particle Duality

Phenomena

Prior quantum mechanics (QM), waves (governed by Maxwell’s electromagnetic

wave equation) and particles motion (governed by Newton’s laws of motion) are two
totally different realms in the physical world:
20 2 Quantum Field Theory for Quantum Finance

Table 2.1 A brief history of quantum mechanics

Professor Michael Faraday (1791–1867)
A British scientist who contributed to the study of electromagnetism and electrochemistry. He
discovered cathode rays in 1838, which triggered the study of blackbody radiation—the dawn
of quantum mechanics
Professor James Clerk Maxwell (1831–1879)
A Scottish scientist who formulated the classical theory of electromagnetic wave that integrated
electricity, magnetism, and light as different forms of the same phenomenon. His works are also
known as the second great unification in physics after Sir Isaac Newton
Professor Ludwig Boltzmann (1844–1906)
An Austrian physicist and philosopher whose greatest achievement was in the development of
statistical mechanics and second law of thermodynamics. In 1877, he proposed energy states of
a physical system that can be discrete
Professor Max Planck (1858–1947)
A German theoretical physicist whose discovery of energy quanta awarded the Nobel Prize for
Physics in 1918. Planck’s hypothesis that energy is radiated and absorbed in discrete quanta
precisely matched the observed patterns of blackbody radiation
Professor Albert Einstein (1879–1955)
A German-born theoretical physicist who developed the theory of relativity. In 1905, Albert
Einstein interpreted Planck’s quantum hypothesis realistically and used it to explain the
photoelectric effect awarded the Nobel Prize for Physics in 1921
Professor Erwin Schrödinger (1887–1961)
An Austrian physicist who developed a number of fundamental results in the field of quantum
theory: the Schrödinger equation provides the calculation on wave function of a system how it
changes dynamically in time awarded the Nobel Prize for Physics in 1933
Professor Werner Heisenberg (1901–1976)
A German theoretical physicist and one of the key pioneers of quantum mechanics. He
published his work in 1925 in a breakthrough paper for the creation of quantum mechanics
awarded the Nobel Prize for Physics in 1932
Professor Richard Feynman (1918–1988)
American theoretical physicist, best known for his work in the path integral formulation of
quantum mechanics and quantum electrodynamics. For his contributions to the development of
quantum electrodynamics awarded the Nobel Prize for Physics in 1965

(1) Maxwell electromagnetic wave equation (Becherrawy 2012), given by

∂B
∇ ×E =− (2.1)
∂t
(2) Maxwell electromagnetic wave equation, given by

F = m · ẍ (2.2)

Wave–particle duality is the central philosophy of QM which describes every

matter (quantum particle) not only in terms of particle dynamics but can also be
considered (more importantly observed) as waves, which is governed by Schrödinger
equation.
2.1 Quantum Mechanics—The Basics 21

Time-dependent Schrödinger equation, given by (Schrödinger 1926),

∂
i ψ(x, t) = Ĥ ψ(x, t) (2.3)
∂t
Such important property is clearly demonstrated by three important findings and
experiments in the history of quantum mechanics, and they are as follows:
(1) Planck’s blackbody radiation (Planck 1901) and the discovery of Planck’s law:

E = hv (2.4)

where E is Energy, h = 6.55x10−34 J-s (Planck’s constant), v-frequency of blackbody

radiation.
Figure 2.2 shows the intensity of light emitted from a blackbody at any given
wavelength. Each curve represents behavior at a different body temperature. Planck
was the first to explain the shape of these curves. His works on blackbody radiation
and the discovery of Planck constant provided the foundation of quantum mechanics.

Fig. 2.2 Blackbody radiation

22 2 Quantum Field Theory for Quantum Finance

Observing
Double-slit screen

Electron

Electron Gun

Interference
paƩern

Fig. 2.3 Double-slit experiment

(2) Double-Slit Experiment

Double-slit experiment of photon particle is an important experiment in quantum
mechanics that reveals the wave–particle duality of quantum particles (Fig. 2.3).
Moreover, it displays the fundamentally probabilistic nature of quantum-mechanical
phenomena. The experiment was first performed with light by Professor Thomas
Young (1802). In 1927, The Davisson–Germer experiment demonstrated that elec-
trons were shown the same behavior, which was later extended to atoms and
molecules (Davisson and Germer 1928).
(3) Einstein’s Photoelectric Experiment
Photoelectric effect is the emission of electrons when light (or laser beam) falls on
a metal plate in a vacuum tube, which can form a complete with the generation
of electricity. The electrons emitted in this manner can be called photoelectrons.
In 1905, Professor Albert Einstein published a paper (Einstein 1905) proposed that
light energy is carried in discrete quantized packets to explain experimental data
from the photoelectric effect (Fig. 2.4). This model contributed to the development
of quantum mechanics and awarded him the Noble Prize in 1921 for “The discovery
of the photoelectric effect.”
Quantum finance mentioned in Chap. 1 is a common belief in financial commu-
nity. In fact, the two major branches of modern technical analysis techniques—chart
pattern analysis versus technical indicators/oscillators analysis—are typical exam-
ples of how we percept and interpret worldwide financial markets by their wave
(market patterns)–particle (market indicators) duality. The difference is: in quantum
finance, we have concrete mathematical and computer algorithms to model and study
the wave–particle duality of financial markets.
2.1 Quantum Mechanics—The Basics 23

Fig. 2.4 Photoelectric experiment

2.1.3 Heisenberg’s Uncertainty Principle

The uncertainty principle (also known as Heisenberg’s uncertainty principle) is a

unique and the most important property in quantum mechanics (Plotnitsky 2010).
In the principle of uncertainty, Professor Werner Heisenberg (1901–1976)
described the fact that there is a natural limit of precision for the measurement
(observation) of certain pairs of physical properties on any matters, which known as
complementary variables (or what we called quantum pairs) such as position (x) and
its momentum (p), given by


p · x = (2.5)
2
where  is the reduced Planck’s constant = /2π .
As explained in his famous quote:
The uncertainty principle refers to the degree of indeterminateness in the possible present
knowledge of the simultaneous values of various quantities with which the quantum theory
deals; it does not restrict, for example, the exactness of a position measurement alone or a
velocity measurement alone.
Werner Heisenberg (1901–1976)

In fact, uncertainty principle is an inherent property of matter due to its own wave
property. It must be emphasized that measurement does not mean only a process in
24 2 Quantum Field Theory for Quantum Finance

which a physicist observer takes part but rather any interaction between classical and
quantum objects regardless of any observer.
This principle has profound influence in modern physics (especially experimental
physics) such as mesoscale numerical weather prediction (NWP) in computational
meteorology, superconductivity, and quantum optics.
In quantum finance, principle of uncertainty provides a way to describe the natural
limit of precision for the observation of intrinsic pair in financial market, which will
be revealed in Chap. 4.

2.1.4 Quantization of Energy

In quantum physics, quantization is the process of transition from a classical under-

standing of physical phenomena to a brand new interpretation in using quantum
mechanics. Quantization converts classical fields into operators acting on quantum
states of the field theory.
Rutherford–Bohr model (or Bohr model in short), presented by Professors Niels
Bohr and Ernest Rutherford in 1913, proposed the quantization of energy levels in
atoms (Bohr 1913) with three unique properties:
1. Discrete orbits (energy levels)—like planets in the solar system, electron is able to
revolve in certain stable orbits around the nucleus which contrary to what classical
electromagnetism suggests. These stable orbits are called stationary orbits and
are attained at certain discrete distances from the nucleus. The electron cannot
have any other orbit in between the discrete ones.
2. These discrete orbits are attained at distances for which the angular momentum
of the revolving electron is an integral multiple of the reduced Planck’s constant
given by.
3. Electrons can only gain and lose energy by jumping from one allowed orbit to
another, with the absorption or emission of discrete amount of electromagnetic
radiation given by.
Figure 2.5 shows the Bohr model that the planetary-like model of the hydrogen
atom (Z = 1), where the electron confined to an atomic shell encircles a small,
positively charged atomic nucleus; and where an electron jumps between orbits is
accompanied by an emitted or absorbed amount of electromagnetic energy (hf ). The
orbits in which the electron may travel are shown as gray circles, where n is the
principal quantum number. The 3 → 2 transition depicted here produces the first line
of the Balmer series, and for hydrogen (Z = 1) results in a photon of wavelength
656 nm (red light).
2.1 Quantum Mechanics—The Basics 25

Fig. 2.5 The Bohr model

2.1.5 Quantum Entanglement

Quantum entanglement (Schrödinger 1935) is a quantum phenomenon which occurs

when pairs or groups of quantum particles generate, interact, or share spatial prox-
imity in ways such that the quantum state of each particle cannot be described inde-
pendently of the state of the other(s), even when the particles are separated by a very
large distance—described by Einstein as spooky action at a distance (Einstein et al.
1935). Quantum entanglement has been demonstrated experimentally on photons,
neutrinos, electrons, and even on small diamonds.
Figure 2.6 shows a laser beam fired through a certain type of crystal that can cause
individual photons to be split into pairs of entangled photons.
The utilization of entanglement in communication and computation is a very active
research area. In fact, quantum entanglement has many applications in quantum
information theory (Sen 2017; Laloë 2012):
1. Quantum computing (Yanofsky and Mannucci 2008) is computed using quantum-
mechanical phenomena, such as superposition and entanglement. Such a com-
puter is completely different from binary digital electronic computers based on
transistors and capacitors. Whereas common digital computing requires that the
data be encoded into binary digits (bits), each of which is always in one of the
two definite states (0 or 1), quantum computation uses quantum bits or qubits,
which can be within the superpositions of states.
In January 2019, IBM launched IBM Q System One, its first integrated quantum
computing system for commercial use.
2. Quantum cryptography (Assche 2006) is an application of quantum entanglement
in real-time encryption and message communication between two remote parties
which share the same quantum state.
26 2 Quantum Field Theory for Quantum Finance

Fig. 2.6 Quantum

entanglement

3. Quantum teleportation is a process by which quantum information (e.g., the exact

state of an atom or photon) can be transmitted (exactly, in principle) from one
location to another, with the help of classical communication and previously
shared quantum entanglement between the transmitting and receiving locations.
4. Quantum finance is accompanied by the intrinsic property of quantum entangle-
ment. Operational researches were carried out in the previous 15 years on how
various financial markets, major forex markets, or major commodity markets
such as gold and crude oil markets are affected when quantum entanglement
occurred, and hence the prediction of major market reversal event.
Important applications of spin-off from quantum theory include
• Semiconductors and microprocessors;
• Laser technology;
• Quantum computing;
• Quantum cryptography;
• Quantum finance;
• Quantum optics;
• Superconductivity (high-speed transportation); and
2.1 Quantum Mechanics—The Basics 27

• Medical and images’ research such as magnetic resonance imaging (MRI) and
DNA editing.

2.2 From Quantum Mechanics to Quantum Field Theory

Quantum field theory is a theoretical framework that combines three important the-
oretical physics theories (Padmanabhan 2016).
They are
1. Quantum mechanics—provides basic mathematical, physical and philosophical
framework;
2. Classical field theory—provides mathematical, physical model, and framework
on quantum fields’ composition and dynamics;
3. Special relativity—provides mathematical, physical model and framework
regarding interchange between quantum particles and their intrinsic energies,
the creation and annihilation mechanisms of subatomic particles, and their math-
ematical interpretations.
4. Einstein’s famous mass–energy formulation:

E = mc2 (2.6)

Note: This formulation not only explains the intrinsic energy of a matter with mass
m or its mass–energy equivalent property, but more importantly it provides a way
to describe and interpret how fundamental (quantum) particles can be created or
annihilated to/from pure energy. The interpretation is not only the central concept of
special and general relativity but also is an important component in quantum field
theory to explain the existence and dynamics of various subatomic particles.
As there are a great deal of topics and phenomena in quantum field theory, this
book will focus on key topics which are essential for constructing the mathematical
and physical framework on quantum finance.
They are
• Classical field versus quantum field,
• Feynman path integral, and
• Quantum oscillators.

2.3 Classical Field Versus Quantum Field

Before we learn more in-depth understanding of what is quantum field theory, let’s
have a look on what is the major difference between a classical field versus quantum
field.
28 2 Quantum Field Theory for Quantum Finance

3.1 2.8 2.6

2.9 2.7 2.5

2.7 2.6 2.4

Fig. 2.7 Classical field

3.1 2.8 2.6

2.9 2.7 2.5

2.7 2.6 2.4

2.9 2.7 2.5

Fig. 2.8 Quantum field

2.3 Classical Field Versus Quantum Field 29

A classical field (field in short) can be thought of as the assignment of a physical

quantity at each point of space and time.
In the physical world, there are two major classical fields (nonrelativistic):
1. Newton’s gravitational field (G field)

GM
g(r) = − (2.7)
r2
where G = gravitational constant, M = mass of the massive body (e.g., Earth)
exerts on a comparably small object with mass m, and r = distance.
2. Maxwell’s electromagnetic field (EM field)

1 q
E=− (2.8)
4π 0 r 2

where q = charge of the test charged object, 0 = permittivity of free space, and
r = distance.
In quantum field theory, particles are described by quantum fields that satisfy the
Schrödinger equation, which is given by (nonrelativistic):
 
∂ 2 2
i (r, t) = − ∇ + V (r, t) (r, t) (2.9)
∂t 2μ

Figure 2.7 shows the 2D illustration of a classical field, in which there are discrete
numbers at every point of the space, whereas in a quantum field, the field value at
single point in the space is not an exact number but rather some sort of distribution—
probability density function (pdf) or wave function in terms of quantum mechanics
in Fig. 2.8.

2.4 Feynman’s Path Integral

2.4.1 A Thought Experiment

Developed by Professor Richard Feynman in 1948 (Feynman 1948; Feynman and

Hibbs 1965), the path integral formulation of quantum finance theory is concerned
with the direct computation of scattering amplitude of a certain interaction process,
rather than the establishment of operators and state spaces. To understand path inte-
gral, let’s begin with a thought experiment on standard double-slit thought experiment
(DSTE).
Figure 2.9 shows the configuration of a standard DSTE, photon particle emits from
source (S) (at T = 0) either go through opening O1 or O2 and strike the projection
screen at final point F (at T = t).
30 2 Quantum Field Theory for Quantum Finance

O1

F
S
O2

Fig. 2.9 Double-slit thought experiment (single layer)

By using the superposition principle in QM, the amplitude of detection at final

point F is the sum of the amplitude for the paths S-O1 -F and S-O2 -F.
How about if we drill another opening O3 ?
By common sense, it should be simply done by the superposition of one more
path amplitude S-O3 -F to the evaluation function, as shown in Fig. 2.10.
The resulted amplitude (ψF ) measured at F will be
n
ψF = ψF (S − Oi − F) (2.10)
i

Let’s move on to the next level.

How about if we add one more layer and start with three openings?
Figure 2.11 shows the multi-slit thought experiment with double layer.
Generally, the path integration (path integral) equation multi-slit thought experi-
ment with double layers will look like

O1

O3 F
S
O2

Fig. 2.10 Triple-slit thought experiment (single layer)

2.4 Feynman’s Path Integral 31

Fig. 2.11 Multi-slit thought experiment (double layers)

n n  
ψF = ψF S − O1i − O2j − F (2.11)
j i

In fact, it is what Feynman did in 1948 during his famous Feynman’s lectures on
path integral and quantum mechanics.
Now then, what would happen if we add infinite number of layers; at the same
time drill infinite number of openings onto it?
Figure 2.12 shows all layers between S and F will simply disappear into empty
space. Why do they behave like this?
The answer is: photons emit from point source S and arrive at detector F through
empty space. The wave function detected at F is the sum of infinite paths begin-
ning from source S to detector F—known as path integral formulation proposed by
Feynman in 1948.

ψF = ψAll Path (S − F) (2.12)
i

F
S

Fig. 2.12 Infinite-slit thought experiment (infinite layers)—path integral

32 2 Quantum Field Theory for Quantum Finance

2.4.2 Feynman’s Path Integral Formulation

In quantum field theory (Padmanabhan 2016), the wave function value detected the
path from point source qS to a final point qF in time T which is governed by the
unitary operator e−iHT where H is the Hamiltonian.
Using Dirac’s bra and ket notation, we have
 
qF |e−iHT |qS (2.13)

Using Feynman’s path integral formulation (Feynman and Hibbs 1965), we

divided the whole T into N segments of width δt = T /N . We have
   
qF e−iHT qS = qF e−iH δt e−iH δt e−iH δt . . . qS (2.14)

Then, we can break it down into N intervals:

 
N −1    
qF e−iHT qS = ∫ dqk qF e−iHT qN −1 . . . q1 e−iHT qS (2.15)
k=1

After relative calculation, we obtain the famous path integral formulation of QFT:
  T
qF |e−iHT |qS = ∫ Dq(t)ei 0 dt [ 2 mq̇ −V (q)]
1 2
(2.16)

  T
qF |e−iHT |qS = ∫ Dq(t)ei 0 dt L(q̇,q) (2.17)

where L(q̇, q) = 21 mq̇2 − V (q) and

     
im N /2 N −1
Dq(t) = lim − dqk (2.18)
n→∞ 2π δt k=1

2.4.3 Path Integral Formulation in Quantum Finance

In quantum finance, path integral formulation (Baaquie 2018) plays an important role
in the modeling of the most fundamental financial instrument—worldwide interest
rate (r) modeling.
Image the following scenario:
If we consider worldwide interest rate, say, US interest rate r as the propagation
of quantum particle Qr.
The change of interest rate Qr during a period of time T from state QrS to QrF
can be formulated by Feynman path integral formulation:
2.4 Feynman’s Path Integral 33

Fig. 2.13 US federal interest rate 1995–2015

T  . 
i dtL Qr,Qr
QrF e−iHT QrS = ∫ DQr(t)e 0 (2.19)

In fact, it is one of the major applications of QFT in finance for the past 15 years.
Besides path integral formulation, are there any other similar financial instru-
ment(s) can be used?
Figure 2.13 shows the US federal interest rate from 1955 to 2015 in four different
timeframes. In Chap. 3, we will explore how to model forward interest rates in differ-
ent timeframes with quantum finance by using Feynman’s path integral technique.

2.5 Quantum Oscillators

2.5.1 Quantum Harmonic Oscillator—The Basics

Quantum harmonic oscillator (QHO) is the quantum mechanics analog of a classical

harmonic oscillator (Blaise and Henri-Rousseau 2011).
Simple harmonic oscillator:

F = mẍ = −kx (2.20)

where k is the force constant.

Damped harmonic oscillator:

F = Fext − kx − cẋ (2.21)

 √
where c is the damping coefficient and ζ = c 2 mk is the damping ratio
 
ω0 = k m is the “Undamped angular frequency of the oscillator.”
(Recall conditions for critically damped/overdamped/underdamped oscillations)
34 2 Quantum Field Theory for Quantum Finance

QHO is one of the most important model systems in QM and QFT.

In one dimension QHO, the Hamiltonian Ĥ is given by

p̂2 1 p̂2 1
Ĥ = + k x̂2 = + mω2 x̂2 (2.22)
2m 2 2m 2

where m = mass ofthe quantum particle; k is the force constant; x̂ is the position
operator; p̂ = −i∂ ∂x is the momentum operator; and ω is the angular frequency.
The solution of 1D-QHO is given by

mω 1/4  
1 mω
ψn (x) = √ · · e−iωt · Hn x (2.23)
2n n! π 

n = 0, 1, 2, …
where n = 0, 1, 2, …; ψn is the wave function and n is the quantum number.
The corresponding discrete energy levels are given by
 
1 
En = ω n + = (2n + 1) ω, n = 0, 1, 2, . . . (2.24)
2 2

Note that E0 corresponds to the ground state and all n > 0 are discrete excited
states.
Figure 2.14 shows the different dynamics in classical versus quantum har-
monic oscillators. Figure 2.14A and B illustrate the classical harmonic oscillators;
Fig. 2.14C–H illustrate the quantum harmonic oscillators—solutions according to
Schrödinger equation, where the horizontal axis is particle position, and the vertical
axis is the real part (blue) or imaginary part (red) of the wave function.

2.5.2 N-Dimensional Quantum Harmonic Oscillators

Using the same logic of 1D QHO, the N-dimensional QHO is exactly analogous to
N independent one-dimensional harmonic oscillators with identical mass and spring
constant.
In this case, the quantities x1 , …, xN would refer to the positions of each of the N
quantum particles.
The Hamiltonian is given by

N  p2 1

H= i
+ mω xi
2 2
(2.25)
i=1 2m 2

For a particular set of quantum numbers (n), the energy eigenfunctions for the N-
dimensional oscillator are expressed in terms of the one-dimensional eigenfunctions
as
2.5 Quantum Oscillators 35

Fig. 2.14 Illustrations of classical versus quantum harmonic oscillators

N
x|ψn  = xi |ψn  (2.26)
i=1

The energy levels of the systems are given by

 
N
En = ω (n1 + n2 + · · · + nN ) + (2.27)
2

where ni = 0, 1, 2 … are the energy level in dimension i.

Figure 2.15 shows a 2D trampoline model of quantum harmonic oscillators. As
shown, the quantum anharmonic oscillator model (in 2D perspective) can be illus-
trated as trampoline in which every quantum oscillator is connected with four neigh-
boring oscillators in the 2D plane, resulted in the anharmonic vibration (from its
stationary level) as vertical movement which analog to the amplitude of the wave
function shown in the above QHO equations.
36 2 Quantum Field Theory for Quantum Finance

Fig. 2.15 2D trampoline

model of quantum harmonic
oscillators

2.5.3 Quantum Anharmonic Oscillators

Even though quantum harmonic oscillator (QHO) can explain many quantum phe-
nomena and wave–particle dualities for simple quantum particles, however, for real-
world applications, many quantum particles and quantum phenomena involve the
anharmonic oscillations (Gilyarov and Slutsker Slutsker 2010).
Figure 2.16 shows an HCl molecule as an anharmonic oscillator (Padilla and Pérez
2012) vibrating at energy level E3 . D0 is the dissociation energy, r0 is the bond length,
and U is the potential energy. Energy is expressed in wavenumbers. The hydrogen
chloride molecule is attached to the coordinate system to show bond length changes
on the curve.
The Hamiltonian of a typical quantum anharmonic oscillator (QAHO) is given by

p2 1
H= + mω2 x2 + λxk (2.28)
2m 2
where λ is the damping coefficient and k is the order of QAHO.
Commonly found QAHOs include
Cubic QAHO with k = 3:

p2 1
H= + mω2 x2 + λx3 (2.29)
2m 2
Quartic QAHO with k = 4:

p2 1
H= + mω2 x2 + λx4 (2.30)
2m 2
Sextic QAHO with k = 6:
2.5 Quantum Oscillators 37

Fig. 2.16 Quantum anharmonic oscillations of HCI molecules’ illustrations

p2 1
H= + mω2 x2 + λx6 (2.31)
2m 2

λx2m -class QAHO—a special class of QAHO which provides a genius way of
solving these high-order AHO equations.

p2 1
H= + mq ω2 x2 + λx2m (2.32)
2mq 2

In quantum finance theory, we have deduced that the quantum dynamics of price
return (r) for any financial market is in fact a kind of quartic quantum anharmonic
oscillations, which is also a special case of λx2m -class QAHO with m = 2. Detail
mathematical derivation along with numerical analysis to solve important quantum
finance equation will be studied in Chaps. 4 and 5.
38 2 Quantum Field Theory for Quantum Finance

2.6 Conclusion

In this chapter, we introduced a general overview of four major concepts and models
in quantum theory: (1) quantum mechanics, (2) quantum field theory, (3) Feynman’s
path integral, and (4) quantum anharmonic oscillators.
These four concepts with models in quantum theory form critical mass to estab-
lish the basic theoretical and mathematical models in quantum finance theory in
next chapter, namely, (1) Feynman’s path integral model of quantum finance (also
known as first generation of quantum finance); and (2) quantum anharmonic model
of quantum finance (also known as second generation of quantum finance).
To avoid complex mathematical derivations, this chapter focused on the basic
concepts of quantum theory with its related models which are critical to understand
the mathematical and quantum model in quantum finance. For readers who are inter-
ested to explore in-depth mathematical models and quantum theory, please refer to
the reference section at the end of this chapter.

Problems

2.1 Before quantum mechanics, what are the two main streams of theories govern
the laws of physics and how quantum mechanics are difference with these
theories?
2.2 We always say 1920s are the key years for the birth of quantum mechanics. Who
are the two key figures in the course of the development of the basic concepts
and theories in quantum mechanics and discuss their major contributions?
2.3 What is the physical importance of Schrödinger equation in terms of the inter-
pretation of wave–particle duality phenomena?
2.4 What are the three most important experiments in the 1920s for the interpreta-
tion and illustration of wave–particle duality in quantum mechanics? Discuss
and explain their findings and importance. Can they be interpreted in financial
markets? How?
2.5 Discuss and explain briefly Heisenberg’s uncertainty principle in quantum
mechanics. How can such phenomena relate to modern finance?
2.6 What is the basic difference between a quantum field and classical field? How
can quantum field be interpreted in terms of financial markets?
2.7 Discuss and explain quantization phenomena in quantum theory. How can it
have been related to quantum finance? What are the ground-state and various
excited energy levels in quantum finance?
2.8 Discuss and explain quantum entanglement in quantum theory. How can it be
interpreted in quantum finance?
2.9 What is quantum computing? Discuss and explain briefly the basic concept and
how it is related to quantum finance.
References 39

References

Assche, G. (2006) Quantum Cryptography and Secret-Key Distillation. Cambridge University Press;
1 edition.
Baaquie, B. (2018) Quantum Field Theory for Economics and Finance. Cambridge University Press.
1st edition.
Becherrawy, T. (2012) Electromagnetism: Maxwell Equations, Wave Propagation and Emission.
Wiley-ISTE; 1 edition, 2012.
Blaise, P. and Henri-Rousseau O. (2011) Quantum Oscillators. Wiley.
Bohr, N. (1913) On the Constitution of Atoms and Molecules, Part I” (PDF). Philosophical Maga-
zine. 26 (151): 1–24.
Davisson, C. J. and Germer, L. H. (1928) Reflection of Electrons by a Crystal of Nickel. Proceedings
of the National Academy of Sciences of the United States of America. 14 (4): 317–322.
Einstein, A. (1905) Über einen die Erzeugung und Verwandlung des Lichtes betreffenden heuristis-
chen Gesichtspunkt [On a Heuristic Viewpoint Concerning the Production and Transformation
of Light]. Annalen der Physik (Berlin) (in German). Hoboken, NJ (published 10 March 2006).
322 (6): 132–148.
Einstein, A., Podolsky, B., Rosen, N. (1935) Can Quantum-Mechanical Description of Physical
Reality Be Considered Complete? Physical Review 47 (10): 777–780.
Feynman, R. P. (1948) Space-Time Approach to Non-Relativistic Quantum Mechanics. Reviews of
Modern Physics. 20 (2): 367–387.
Feynman, R. P.; Hibbs, A. R. (1965) Quantum Mechanics and Path Integrals. New York: McGraw-
Hill.
Gilyarov V. L. and Slutsker, A. I. (2010) Energy features of a loaded quantum anharmonic oscillator.
Physics of the Solid State 52(3): 585–590.
Heisenberg, W. (1925) Über quantentheoretische Umdeutung kinematischer und mechanischer
Beziehungen. Zeitschrift für Physik. 33 (1): 879–893.
Laloë, F. (2012) Applications of quantum entanglement. In Do We Really Understand Quantum
Mechanics? Cambridge University Press, 150–167.
PoorLeno (2019) Hydrogen Density Plots. https://commons.wikimedia.org/w/index.php?title=File:
Hydrogen_Density_Plots.png&oldid=343305978. Accessed 21 Aug 2019.
Padmanabhan, T. (2016) Quantum Field Theory: The Why, What and How (Graduate Texts in
Physics). Springer.
Padilla A. and Pérez, J. (2012) A Simulation Study of the Fundamental Vibrational Shifts of HCl
Diluted in Ar, Kr, and Xe: Anharmonic Corrections Effects. International Journal of Spec-
troscopy, 2012: 1–7.
Planck, M. (1901) Ueber das Gesetz der Energieverteilung im Normalspectrum [On the law of the
distribution of energy in the normal spectrum]. Annalen der Physik. 4 (3): 553–563.
Plotnitsky, A. (2010) Epistemology and Probability: Bohr, Heisenberg, Schrödinger, and the Nature
of Quantum-Theoretical Thinking (Fundamental Theories of Physics). Springer.
Sen, A. (1926) Quantum Entanglement and its Applications. Current Science 112(7): 1361, 2017.
Schrödinger, E. An Undulatory Theory of the Mechanics of Atoms and Molecules. Physical Review
28: 1049–1070.
Schrödinger, E. (1935) Discussion of probability relations between separated systems. Mathematical
Proceedings of the Cambridge Philosophical Society. 31 (4): 555–563.
Yanofsky, N. S. and Mannucci, M. A. (2008) Quantum Computing for Computer Scientists. Cam-
bridge University Press; 1 edition.
Young, T. (1802) The Bakerian Lecture: On the Theory of Light and Colours. Philosophical Trans-
actions of the Royal Society of London. 92: 12–48.
Chapter 3
An Overview of Quantum Finance
Models

It is a curious historical fact that modern quantum mechanics

began with two quite different mathematical formulations: the
differential equation of Schrödinger and the matrix algebra of
Heisenberg. The two apparently dissimilar approaches were
proved to be mathematically equivalent.
Richard P. Feynman (1918–1988)

What is quantum finance?

What are the major quantum finance models?
Quantum finance – which way to go?
In this famous quotation on quantum mechanics, Professor Feynman pointed out
the vital feature of quantum mechanics mathematical derivation, the two totally
different approaches to formulate and model the quantum world which end up being
mathematically equivalent.
Did it sound familiar?
As we learnt from quantum mechanics’ basic concept, whether we observe the
quantum world as particle dynamics or wave dynamics, it only affects the way we
model the quantum world, and the mathematical tool along with the model we are
using. The reality of the quantum world doesn’t change, and from the mathematical
perspective, they are mathematical equivalent.
As mentioned in Chap. 1, quantum finance is a newly developed interdisciplinary
subject with the integration of
1. Quantum theory—the provision of quantum mechanics and quantum field theory
as the quantum finance dynamics with field analysis; and
2. Computational finance—the provision of computational finance models as the
system framework and the application basis.
To meet the demand from the exponential growth of program trading and intel-
ligent financial forecast service, in 2012, the author had integrated contemporary
quantum finance theory with AI technology to implement quantum finance forecast
system for the real-time forecast of worldwide commodity markets.
In this chapter, we will first review a brief history of quantum finance.

© Springer Nature Singapore Pte Ltd. 2020 41

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_3
42 3 An Overview of Quantum Finance Models

Second, we will study the two main branches of quantum finance models—Feyn-
man’s path integral model and the quantum anharmonic oscillator model.
Lastly, we will explore the future of quantum finance, and how it relates to the
development of intelligent finance systems.

3.1 A Brief History of Quantum Finance

Quantum finance is a newly developed interdisciplinary subject introduced in 1990s

by applying quantum mechanics and quantum field theory to theoretical economics—
so-called econophysics.
Nevertheless, econophysics style of R&D was established much earlier.
In 1900, Professor Louis Jean-Baptiste Alphonse Bachelier (1870–1946), a French
mathematician in his Ph.D. thesis Théorie de la speculation (translated as The Theory
of Speculation) published by Annales Scientifiques de l’École Normale Supérieure,
set the foundation of a mathematical model with Brownian motion in valuing stock
options (Bachelier 1900).
It was historically the first paper to use advanced mathematics in the study of
finance. He is also considered as the forefather of mathematical finance and also a
pioneer in the study of stochastic processes.
His thesis contained three different versions of the first mathematical theory of
Brownian motion. In modern terminology, Brownian motion was characterized as
1. A process with independent homogeneous increments whose paths
are continuous;
2. The continuous-time process, which is the limit of symmetric random walks; and
3. The Markov process whose forward Kolmogorov equation is the heat equation.
Owing to the above reasons, most mainstream econophysicists consider finance
as an application of Brownian motion—the fundamental phenomenon of statistical
physics for the modeling of financial market.
The first published work on Econophysics—An Introduction to Econophysics -
Correlations and Complexity in Finance—was written by Professors Rosario Man-
tegna and Eugene Stanley in 1999 (Mantegna and Stanley 1999). This pioneering
text explored the use of statistical physics concepts such as stochastic dynamics,
short- and long-range correlations, self-similarity, and scaling concepts of financial
systems’ description. These were the dynamic new specialty of econophysics.
For the last two decades, various methods and theories were proposed for stock
price/returns analysis, interest rate modeling, option pricing, and portfolio analysis.
3.2 Major Works and R&D in Quantum Finance 43

3.2 Major Works and R&D in Quantum Finance

Although statistical physics is the mainstream theory of Econophysics, active R&D

with the adoption of quantum mechanics, quantum field theory with related concepts,
and frameworks such as Feynman’s path integral model and quantum oscillator model
to model financial markets can be found which include the following:
1. B. Baaquie in his book Quantum Finance (Baaquie 2004) reviewed the applica-
tion of Feynman’s path integral theory for option pricing and interest rate mod-
eling. Professor Baaqui is also the first scholar who consolidated a complete
concept and theory of quantum finance using quantum field theory (Baaquie
2007, 2013, 2018);
2. Other research works on path integral including the sensitivity analysis using
path-independent quantum finance model by Kim et al. (2011);
3. Quantum anharmonic oscillator modeling on finance analysis included Gao and
Chen (2017) work on quantum anharmonic oscillator model for the stock mar-
ket; Ye and Huang (2008) work on nonclassical oscillator model for persistent
fluctuations in stock markets; Meng et al. (2015) work on quantum spatial-
periodic harmonic model for daily price-limited stock markets;
4. Quantum wave function for stock market analysis by Ataullah et al. (2009);
5. Quantum statistical approach to simplified stock markets by Bagarello (2009);
6. A finite-dimensional quantum model for the stock market by Cotfas (2013);
7. Nakayama (2009) works on gravity dual for Reggeon field theory and nonlinear
quantum finance;
8. Piotrowski and Sładkowski (2005) studied the quantum diffusion model of
prices and profits;
9. Probability wave approach on security transaction volume–price behavior anal-
ysis by Shi (2006);
10. Bohmian quantum potential approach on stock market credibility analysis by
Nasiri et al. (2018);
11. Schaden (2002) applied quantum theory to model secondary financial markets;
12. Zhang and Huang (2010) defined wave functions and operators of the stock
market to establish the Schrödinger equation for stock price.

3.3 Two Pillars of Quantum Finance Models

In respect of quantum finance R&D of the past decades, there are basically two main
quantum models with significant impacts and research results in financial modeling
and applications. They are
1. Feynman’s path integral approach (Zee 2011; Swanson 2014) and
2. Quantum anharmonic oscillator approach (Bloch 1997).
44 3 An Overview of Quantum Finance Models

Fig. 3.1 Feynman’s path integral approach in quantum finance for government bond and interest
rate modeling

In the next two sections, we will review the basic concept and framework of
these two approaches in quantum finance. Figures 3.1 and 3.2 show the typical
applications of these two mathematical models on the modeling of forward interest
rate and quantum price levels in secondary financial markets, respectively.

3.4 Feynman’s Path Integral Approach in Quantum

Finance

3.4.1 Forward Interest Rate in a Nutshell

Before we study the Feynman’s path integral approach in quantum finance for the
modeling of forward interest rate, let’s have an overview of forward interest rate in
finance.
From computational finance perspective, forward interest rate f (t, x) is the interest
rate fixed at time t ≤ x for an instantaneous (overnight) loan taken at future time x.
So, at any instant t, f (t, x) is a function of x (not t) that can be visualized as a 1D
path in the future time domain (x) satisfy (x ≥ t), as shown in Fig. 3.3.
3.4 Feynman’s Path Integral Approach in Quantum Finance 45

Fig. 3.2 Quantum anharmonic model in quantum finance for quantum price level modeling (QFFC
2019)
Calendar Time

Future Time

Fig. 3.3 Forward interest rate function domain (green area)

46 3 An Overview of Quantum Finance Models

With the initial condition f (tS , x) is fixed by the financial market, each point (t, x)
manifests itself in the functional domain (green area) and corresponds to one forward
interest rate incidence f (t, x).

3.4.2 Zero Coupon Bond (Treasury Bond)

Treasury bond is a risk-free financial instrument which has a single cash flow con-
sisting of a fixed payoff of, say, $1 at some future time T; its price at time t < T is
denoted by P(t, T ), with P(T , T ) = 1.
So, for a bond maturing at time T, its value P(t, T ) before maturity is given by
discounting P(T , T ) = 1 to the time t by the spot interest rate, given by
 T   T 
P(t, T ) = E e− t dsr(s) P(T , T ) = E e− t dsr(s) (3.1)

where
1. E[] is the expectation value;
2. r(s) = f (s, s) is the spot interest rate.
Figure 3.4 shows the forward interest rate of Eurodollar futures from 1990 to 1996
with eight different maturities, ranging from 3 months up to 7 years.

3.4.3 Quantum Field Theory of Interest Rate Model

To model the interest rate in respect of quantum field theory using Feynman’s path
integral, first take a look at the forward interest rate of Eurodollar futures from 1990
to 1996 with eight different maturities, ranging from 3 months up to 7 years as shown
in Fig. 3.5.
If we look carefully, we can find certain interesting features from this chart:
1. All these forward interest rate time series appeared to be highly chaotic but have
a certain shared pattern. In Chap. 8—chaos and fractals in quantum finance—we
will study such kind of chaotic phenomena—so-called deterministic chaos, a
vital feature and phenomena in financial engineering.
2. All these curves are basically randomly evolved in time, but in a highly co-related
manner, basically without any line crossings.
In other words, the interest rate function f (t, x) can be assumed to be independent
variable w.r.t. to all x and t values.
Let A(t, x) be a two-dimensional quantum field, and the quantum interest rate
model is given by Baaquie (2004)
3.4 Feynman’s Path Integral Approach in Quantum Finance 47

Fig. 3.4 Forward interest rate of Eurodollar Futures from 1990 to 1996

Interest Rate ψr

tS tF
Time (t)

Fig. 3.5 Feynman’s path integral model for interest rate

∂f (t, x)
= α(t, x) + σ (t, x)A(t, x) (3.2)
∂t
48 3 An Overview of Quantum Finance Models

3.4.4 Feynman’s Path Integral Formulation of Forward

Interest Rate

The Feynman’s path integral for the forward interest rate is given by Baaqie (2013)

1
∫ DAeS[A] F[A]
E(F[A]) = (3.3)
Z
   +∞ 
where DA = −∞ dA(t, x) and Z = DAeS[A]
(t,x)∈B
The time-dependent Hamiltonian is given by
 ∞  ∞
−1   δ2 δ
H (t) = dxdx M x, x ; t

− dxα(t, x) (3.4)
2 t δf (x)δf (x ) t δf (x)
     
where M x, x ; t = σ (t, x)D x, x ; t σ t, x
Note:
1. S[A] is the stiff action function, given by Baaqie (2013)
 

2 
+∞ +∞
1 1 ∂A 2 1 ∂ 2A
S[A] = − dt dz A (t, z) + 2
2
(t, z) + 4 (t, z)
2 tS 0 μ ∂z λ ∂z 2
(3.5)

2. D is the domain of (t, x) for the path integration;

3. The initial value of the forward interest rate f (tS , x) is an important boundary
condition for Feynman’s path integral calculation.
4. As shown in Eq. 3.3, to calculate the expectation value E(F[A]), one has to
perform integration of F[A] over all possible functions A(t, x) of all possible
(t, x) in the domain space D.
5. In other words, to evaluate the forward interest rate using Feynman’s path inte-
gral
 method, one has to perform many integrations infinitely, one integration of
dA(t, x) for each state variable (t, x), which is natural as the core concept of
Feynman’s path integral is the integration of (superposition in quantum theory)
all possible paths that lead from S to F, as indicated in Chap. 2.
3.5 Quantum Anharmonic Oscillator Approach in Quantum Finance 49

3.5 Quantum Anharmonic Oscillator Approach

in Quantum Finance

3.5.1 Motivation

One of the major characteristics of quantum theory is the wave–particle duality which
can be described by the Schrödinger equation.
Time-dependent Schrödinger equation:

∂ 2 ∂ 2
i Ψ (x, t) = − + V (x, t) Ψ (x, t) (3.6)
∂t 2m ∂ 2 x

Time-independent Schrödinger equation:


2 ∂ 2
− + V (x) Ψ (x) = EΨ (x) (3.7)
2m ∂ 2 x

Instead of solving quantum dynamics’ problem by integrating all (in fact is infin-
ity) the possible paths from starting point S to end point F, why not focus on solving
the Schrödinger equation directly using numerical method (numerical approxima-
tion)?
Imagine: if we can solve this equation numerically (i.e., instead of analytically),
say, into an n-order polynomial with differential operators, together with the inte-
gration of finite difference method (FDM). Technically speaking, we can implement
such algorithm into a computer program for actual implementation (not simulation).
Actually, such notion is more important to tackle real-world problems with numer-
ous time series data such as weather and finance.
As we will see, anharmonic oscillation (AHO) provides an excellent analog of
Schrödinger equation for the modeling of quantum dynamics (3.6) and quantum
energy levels (3.7).
In fact, many atomic/subatomic particles’ dynamics in the real world can be
described in terms of quantum anharmonic oscillation (QAHO) as shown in Fig. 3.6.
It shows the quantum anharmonic oscillator model for HCI molecules and their
corresponding quantum energy well.

3.5.2 QAHO Modeling—Prices Versus Returns

Different from Feynman’s path integral method, quantum anharmonic oscillator

(QAHO) method directly models the wave–particle dynamics—Schrödinger equa-
tion.
Owing to this reason, QAHO method tailored for the modeling of secondary
financial markets such as the trading of stocks and futures among investors or OTC
50 3 An Overview of Quantum Finance Models

Fig. 3.6 Quantum anharmonic oscillations of HCI molecules

trading such as commodities, financial indices, and forex between market maker
(MM) and investors.
In other words, the objective of QAHO modeling in quantum finance targets at
quantum dynamics and quantum energy levels for market instruments involve in the
secondary markets, forex market, for example (Fig. 3.7).
So, the natural and most fundamental question is: What market instrument(s) we
should model? Prices, indices, price returns, price deviations…?
To answer this question, we should ask this vital question:
If financial market is really a quantum phenomenon, that is, the motions, fields,
and dynamics of quantum particles, what are the financial particles we are talking
about? Or what is the best analog of the displacement x (from the equilibrium state)?
Price or price returns or …?
Why?

3.5.3 Quantum Anharmonic Oscillations of Returns

The answer is … price returns (r or return in short).

3.5 Quantum Anharmonic Oscillator Approach in Quantum Finance 51

Fig. 3.7 Worldwide financial products

If we want to make the best analog of quantum particle displacement from the
equilibrium position, naturally we should choose price return (r) with two major
reasons:
1. Return (r) in finance is an important financial instrument to reflect the relative
change of financial behavior (say stock price) within a fixed period of time (say,
daily returns), which directly analog to physical meaning of quantum theory for
the definition of x which relates to displacement to the equilibrium position; and
2. Use of return (r) to model financial market in quantum finance is totally coherent
classical and statistical finance theories which all use price return for financial
market modeling and statistical analysis.
So how about Prices? Is there any significance of price in quantum finance?
The answer is: very important.
As analog to gravitational field, one can imagine if all financial instruments are
quantum financial particles, what will be the quantum field?
52 3 An Overview of Quantum Finance Models

The natural logic should be the quantum financial energy field, QFEF (or quantum
energy field in short) created by itself (just like gravitation field created by graviton)
and/or QFEF created by other financial particles (in other financial markets).
So, to analog the potential energy in gravity and electromagnetic field (EM field),
quantum price energy levels (or quantum price levels, QPLs in short) are the quantum
finance quantity to describe the quantized energy field acting onto the quantum
financial particles.

3.5.4 A Quantum Anharmonic Oscillator Model of Financial

Markets

Let’s begin with QAHO modeling in quantum finance.

First, define quantum financial particle r—price return (e.g., stock).
For numerical computation, we define the fractional price change as

p(t) − p(t − Δt)

r(t, Δt) = (3.8)
p(t − Δt)

where p(t) is the price (say stock price) at time t and Δt denote the time interval.
The analog motions and dynamics of (stock) price to a quantum particle (namely,
quantum financial particles, QFPs), price return (r) will be a direct analog to the
displacement (x) of a quantum particle from its equilibrium state, assuming they are
performing anharmonic oscillation (as we will see shortly).
So, quantum dynamics of the quantum financial particle are given by Schrödinger
equation:

∂
i ψ(r, t) = H ψ(r, t) (3.9)
∂t
H is the Hamiltonian operator given by

2 ∂ 2
H =− + V (r, t) (3.10)
2m ∂ 2 r

3.5.5 Physical Meaning of Hamiltonian in Quantum Finance

The Hamiltonian (H) in Schrödinger equation

2 ∂ 2
H =− + V (r, t) (3.10)
2m ∂ 2 r
3.5 Quantum Anharmonic Oscillator Approach in Quantum Finance 53

consists of two parts, K.E. (first part) and P.E. (second part). K.E. corresponds to
quantum motion of the quantum financial particle (QFP), and P.E. corresponds to
potential energy (well) generated by the quantum energy field of QFP itself (or other
QEF generated by other QFPs in multidimensional financial markets).
The reduced Planck’s constant  can be considered as the uncertainty of irrational
transaction and m represents the intrinsic properties of financial market such as
market capital in a stock market.
Corresponding to the stationary market environment in classical computational
finance theory (CFT), we can analog such quantum dynamics as time-independent
Schrödinger equation, given by

H (r)ϕ(r) = Eϕ(r) (3.11)

Or

2 ∂ 2
− + V (r) ϕ(r) = Eϕ(r) (3.12)
2m ∂ 2 r

where Ψ (r, t) = ϕ(r)e−iEt/ ; E is the eigenenergy of financial market; and ϕ(r) is

the corresponding eigenfunction.
Inherent from uncertainty principle in quantum theory, return (r) should be existed
as wave function until we make the observation, which in fact will be affected by
time frame of measurement that is again consistent with modern finance and fractal
theories in particular.

3.5.6 Probability Density Function in Quantum Finance

In an efficient market and stationary market environment, quantum dynamics of the

QFP r can be visualized as wave function with probability density function (pdf)
given by

Q(r, t) = |ψ(r, t)|2 = |ψ(r)|2 (3.13)

where Q(r, t) is the probability density function for a series of observations (mea-
surements) of a financial market.
Figure 3.8 shows the statistics (i.e., probability density function) of XAUUSD
(gold w.r.t. USD) daily returns in the past 2048 time series observations. We will
study in detail quantum price level (QPL) evaluation in Chap. 5.
In an efficient market with stationary market environment, the probability density
function should behave as normal distribution coherent classical statistical finance
theory. The distribution itself should be characterized by its statistical indicators such
as mean, variance, and kurtosis.
What is the physical meaning of distribution in terms of quantum theory?
54 3 An Overview of Quantum Finance Models

Quantum Price Return Wavefunc on Q(r) of XAUUSD

(For the past 2048 trading- day sta s cs)
0.05

0.045

0.04

0.035

0.03
Q(r)

0.025

0.02

0.015

0.01

0.005

0
0.97 0.98 0.99 1 1.01 1.02 1.03
Daily Return (r)

Fig. 3.8 Probability density function of XAUUSD

As recalled from the famous double-slit experiment of photon particles, the

observed (detected) pattern in the backdrop screen is, in fact, the measurements
(observations) of the quantum behavior of the wave-to-particle realization and
resulted as probability density function (pdf). In other words, in quantum theory
(quantum finance theory in particular), classical statistical (finance) theory is one of
the realizations of the quantum (finance) theory, which can help us to examine the
characteristics of the quantum (finance) behaviors.

3.5.7 Quantum Anharmonic Oscillator Formulation

In computational finance, there is an important financial instrument which is the

excess demand (z), defined as follows.
At any time t, z+ (t) and z− (t) denote the instantaneous demand and supply for
the financial asset.
The excess demand (z) at any instance is given by

Δz = z+ − z− (3.14)

Let r(t) is the instantaneous returns, which is given by

dp
= r(t) = F(Δz) (3.15)
dt
3.5 Quantum Anharmonic Oscillator Approach in Quantum Finance 55

For small Δz, F can be approximated by a scaling factor γ and become

Δz
r(t) = (3.16)
γ

where γ represents market depth, the excess demand z required to move quantum
price p by one single quantum.
So we have

dr d 2p 1 d (Δz)
= 2 = (3.17)
dt dt γ dt

After learning the financial dynamics of various parties in an efficient market, we

will deduce that
dr d Δz
=γ = −γ δr + γ υr 3 (3.18)
dt dt
where δ and υ are the damping terms and volatility factors of financial market.
The Brownian price return can be described by Langevin equation:

d 2r dr d V (r)
mr = −η − (3.19)
dt 2 dt dr
where
mr mass of the financial particle p;
η damping force factor;
V (r) time-independent quantum potential.
d 2r
For overdamping case where dt 2
= 0, we have

d V (r) dr
− = η = −γ ηδr + γ ηυr 3 (3.20)
dr dt

  γ ηδ 2 γ ηυ 4
V (r) = −γ ηδr + γ ηυr 3 dr = r − r (3.21)
2 4

3.5.8 Quantum Finance Schrödinger Equation—QFSE

Combining Eqs. (3.17)–(3.21), we obtained the Quantum Finance Schrödinger Equa-

tion (aka QFSE) of a quantum finance particle as
56 3 An Overview of Quantum Finance Models


− d 2 γ ηδ 2 γ ηυ 4
+ r − r ϕ(r) = Eϕ(r) (3.22)
2m dr 2 2 4

Certain main characteristics can be found in QFSE:

1. The time-independent Schrödinger equation of a quantum finance contains both
K.E. and P.E. terms.
2. Different from classical quantum harmonic oscillator, quantum finance oscillator
is an anharmonic quantum oscillator which consists of two high-order P.E. terms
to represent (1) damping (trading restoration and market absorption) potential
and (2) volatility (risk control) potential.
3. Although market is visualized
 (observed) as price, the quantum dynamics are con-
 
strained by price return r = dp dt , which is consistent with classical financial
theory.

3.5.9 Significance of QFSE in Quantum Finance

The significance of QFSE in quantum finance includes the following:

1. Once we formulate QFSE in terms of quantum anharmonic oscillation (QAHO),
solving the quantum anharmonic oscillator will become clear. It is one of the
popular mathematical and high-energy physics topics. The solution is also applied
to modern physics technology which includes superconductivity, quantum theory
on laser and optics, quantum computing, and naturally quantum finance.
2. This exquisite equation sets an important path to (actually) evaluate the “holy
grail” of quantum finance—quantum finance price energy level (or quantum price
level, QPL) for all financial instruments in the secondary financial markets.
3. In Chap. 4, we will study the complete quantum finance theory’s mathematical
modeling based on quantum anharmonic oscillation theory.
4. In Chap. 5, we will explore how to formulate, implement quantum finance theory
to calculate quantum price levels (QPLs) for worldwide financial products, and
more importantly, build a workable MT4 program for actual calculation in any
PC machines.

3.6 Quantum Finance—Which Way to Go

3.6.1 Feynman Feynman’s Path Integral Approach Versus

Quantum Anharmonic Oscillator Approach
 Feynman’s path integral Quantum anharmonic oscillator
PROS • Technically speaking, this method can be used to evaluate basically • Technically speaking, this method can be used to evaluate any
all financial instruments in primary financial market, such as forward financial instruments in secondary financial market including
interest rate and, option pricing financial derivative, forex prices, and financial indices such as DJI
• Strict implementation of Feynman’s path integral technique on and FTSE
finance modeling and technically sensible • With proper integration of numerical computational technique(s),
QAHO can also be used to calculate the quantum energy
levels—so-called quantum price levels (QPLs) for any financial
markets
• With the adoption of numerical computational techniques and finite
difference method (FDM), QAHO can be easily implemented by
computer/trading program, MQL/MT4, for example
3.6 Quantum Finance—Which Way to Go

CONS • To calculate expectation value of wave function, integrations of all • As QAHO is based on polynomial approximation of differential
the possible paths are required equation(s) of QM and QFT, calculation results are basically the best
• Technically possible, but mathematically too complex to calculate approximation of these quantum finance instrument values
and implement for commercial use • Owing to the polynomial approximation nature and sensitivity to
initial conditions (which is approximation in natural), error-prone for
market forecast, especially when the time horizon extends
57
58 3 An Overview of Quantum Finance Models

3.6.2 Quantum Finance—Which Way to Go

The comparison chart listed in Sect. 3.6.1 showed that both Feynman’s path integral
approach and quantum anharmonic oscillator approach have pros and cons. Further,
both approach methods head to different area applications in finance engineering.
Owing to its intrinsic property of modeling the detail paths with their dynamics of
every quantum financial particle, Feynman’s path integral approach is well suited for
modeling quantum dynamics on primary financial markets (PFMs) such as interest
rates, forward interest rates of contracts and bonds, plus issue new securities in
stock markets. It is the first generation of quantum finance modeling technique being
studied and explored in the finance engineering realm, also known as first-generation
quantum finance. However, the intrinsic property of its computational complexity
entails Feynman’s path integral method, especially when dealing with financial big
data which consist of infinite numbers of possibility (i.e., paths), such methods are
difficult to be applied to real-time worldwide financial engineering (Fig. 3.9).
Quantum anharmonic oscillator approach, on the other hand, models quantum
financial world as anharmonic oscillations of quantum financial particles. The major
breakthrough is the conversion of classical Schrödinger equation into simpler quan-
tum anharmonic equations. Thus, the equations can be solved by simple numerical
calculation using digital computational method to be revealed in Chaps. 4 and 5. Such
major breakthrough leads us to apply quantum finance technology into real-world
financial engineering, with the integration of contemporary AI tools to implement
real-time financial prediction and quantum trading systems. This approach is the
latest mathematical technique for quantum finance modeling, also known as second
generation of quantum finance. However, since this model is tailored for solving
QFSE and the evaluation of all related quantum financial energy level known as
quantum price levels (QPLs) to be revealed in Chaps. 4 and 5, it is well suited for the
application to all financial products in secondary financial markets (SFMs) such as
worldwide forex markets, commodity markets, financial indices, and cryptocurrency
market.
Like any other theory and technology, quantum finance—as a newly emerging
theory—should and ought to adopt any possible and workable theory, methodology,
and technology to make it more useful, practical, and comprehensible, not solely on
research but also for public and financial community.
Especially for the exponential growth of program trading, there are tremendous
needs for intelligent financial forecasting and trading tools, not solely for the insti-
tutional clients and fund houses but more importantly for independent traders and
investors—to provide an open and fair environment for worldwide investment.
As one would expect, the implementation of quantum price levels calculation is
a good start, but it is also just the beginning of the journey.
The major challenges we are facing are how to make use of it, together with other
AI tools and technology, and to design and build intelligent financial forecast and
smart trading programs, which will be studied in the following chapters.
3.7 Conclusion 59

Fig. 3.9 Quantum

finance—which way to go?
(Tuchong 2019)

3.7 Conclusion

Every discipline of technology has its own course and history. In this chapter, we
discussed the brief history of quantum finance, from Professor L. Bachelier’s Theory
of Speculation in 1900 to the latest works on quantum anharmonic oscillator theory of
quantum finance. We also studied the latest research and studies of quantum finance
in the past 30 years.
Through these, we focused on the discussion of two major mathematical
approaches and models of quantum finance—(1) Feynman’s path integral approach
and (2) quantum anharmonic oscillator approach. We reviewed their basic concepts,
principal features, and financial and mathematical implications with their signifi-
cances in the modeling of real-world financial markets.
The applicability of any new theory and technology must fulfill three challenges:
1. physically sound,
2. mathematically logical, and
3. computationally feasible.
60 3 An Overview of Quantum Finance Models

Physically sound means that such model and theory must be aligned with the
physical world of reality it represents, in our case, the quantum world of worldwide
financial markets.
Mathematically logical means that all the mathematical models and frameworks
should be aligned with contemporary theoretical, mathematical research and studies.
Computationally feasible means that the new model and theory must be practically
feasible to be adopted as computer models, algorithms, and computer systems for
real-time modeling of financial market.
In next chapter, we will study how quantum anharmonic oscillator of quantum
finance can be physically and mathematically sound for the modeling of worldwide
secondary financial markets. In Chap. 5, we will further explore how such technology
is computationally feasible and effective to model and evaluate quantum price levels
(QPLs), which provides a critical mass to implement real-time worldwide financial
predication and quantum trading systems.

Problems

3.1 Why Richard Feynman in his famous quotation said the two apparently dissim-
ilar approaches in quantum mechanics (1) differential equation formulation of
Schrödinger and (2) matrix algebra formulation of Heisenberg are in fact to
be mathematically equivalent?
3.2 What is computational finance? Discuss and explain what are the major dif-
ferences between computational finance and classical finance.
3.3 State and explain the two major branches of quantum finance models and how
they are related with quantum field theory and modern quantum mechanics.
3.4 What is econophysics? Discuss and explain how modern physics affect and
contribute to the development of modern economic and finance theories. Give
three modern economic/finance theories as example.
3.5 What is/are the major contribution(s) of L. Bachelier in modern finance?
3.6 Why mainstream econophysicists traditionally considered finance as an appli-
cation of Brownian motion? And how such notion is important to the devel-
opment of finance and economic theories in the past century?
3.7 State and discuss three applications and R&D works of quantum finance done
in the past two decades.
3.8 What is Feynman’s path integral approach in quantum finance? And how it
can be applied to forward interest rate modeling?
3.9 What are the differences between primary versus secondary financial markets?
Discuss and explain why Feynman’s path integral approach is best to be applied
in primary financial markets instead of secondary financial markets.
3.10 What is forward interest rate? Why is it important to financial markets?
3.11 Figure 3.5 shows the forward interest rates of Eurodollar futures from 1990 to
1996, with eight different maturities ranging from 3 months to 7 years. State
and explain three characteristics found from Fig. 3.5 in terms of
(i) the coherence of the eight chart patterns;
3.7 Conclusion 61

(ii) the ordering of these patterns in terms of the period of maturity; and
(iii) the crossings of chart patterns.
3.12 State and explain the formulation of forward interest rate modeling using
Feynman’s path integral in quantum finance.
3.13 Describe and explain the physical meaning and logic behind interest rate mod-
eling using Feynman’s path integral in quantum finance.
3.14 What are the major pros and cons for the modeling of forward interest rate
using Feynman’s path integral in quantum finance?
3.15 In addition to the modeling of forward interest rate, what are the other potential
applications of Feynman’s path integral in quantum finance?
3.16 What is quantum anharmonic oscillator approach in quantum finance? State
and explain how it works.
3.17 For financial markets modeling, we usually use price returns (r) instead of
price (p) directly, why? Give two examples of financial markets to support
your explanation.
3.18 What is the major difference between quantum anharmonic oscillator approach
versus Feynman’s path integral approach in quantum finance?
3.19 Discuss and explain why quantum anharmonic oscillator approach is tailored
for the modeling of secondary financial markets instead of primary financial
markets.
3.20 Discuss and explain physical and mathematical meanings of price (p), price
returns (r), and quantum price levels (QPLs) in quantum finance using anhar-
monic oscillator model, using Dow Jones Index (DJI) as example to support
your explanation.
3.21 What is the physical meaning of Hamiltonian (H) in Schrödinger equation of
quantum finance? What are K.E. and P.E. parts? And how they are related in
financial market’s dynamics in quantum finance?
3.22 What is the physical meaning and importance of eigenenergy values of the
time-independent Schrödinger equation in quantum finance?
3.23 What is the physical meaning and importance of probability density function
(and its values) in quantum finance? And how can we observe/ measure it in
real-world financial markets?
3.24 How can we interpret classical statistical finance theory using quantum finance
theory? And how it is related to wave–particle duality phenomena in quantum
theory?
3.25 What is market depth in modern finance? What is its importance? What is the
mathematical interpretation of market depth in quantum finance?
3.26 State and explain the formulation and physical meaning of each mathematical
term of Langevin equation in quantum finance.
3.27 State and explain the main characteristics of quantum finance Schrödinger
equation (QFSE).
3.28 What are the significances of quantum finance Schrödinger equation (QFSE)?
Give two examples of how to apply QFSE for the modeling of financial mar-
kets.
62 3 An Overview of Quantum Finance Models

3.29 Discuss and explain the pros and cons for the modeling of quantum finance
using (1) Feynman’s path integral approach and (2) quantum anharmonic oscil-
lator approach.
3.30 For the modeling of forex/cryptocurrency market, such as short-term financial
prediction, which approach is more referable? Why?

References

Ataullah, A., Davidson I. and Tippett, M. (2009) A wave function for stock market returns. Physica
A: Statistical Mechanics and its Applications. 388(4): 455–461.
Baaquie, B. E. (2004) Quantum Finance. Cambridge University Press.
Baaquie, B (2007) Quantum Finance: Feynman’s Path Integrals and Hamiltonians for Options and
Interest Rates. Cambridge University Press, 1st edition.
Baaquie, B. E. (2013) Financial modeling and quantum mathematics. Computers and Mathematics
with Applications. 65(10): 1665–1673.
Baaquie, B. (2018) Quantum Field Theory for Economics and Finance. Cambridge University Press.
1st edition.
Bachelier, L. (1900) Théorie de la speculation. Annales Scientifiques de l’École Normale Supérieure,
3(17): 21–86.
Bagarello, F. (2009) A quantum statistical approach to simplified stock markets. Physica A: Statis-
tical Mechanics and its Applications. 388(20): 4397–4406.
Bloch, S. C. (1997) Introduction to Classical and Quantum Harmonic Oscillators. Wiley, 1st edition.
Cotfas, L. (2013) A finite-dimensional quantum model for the stock market. Physica A: Statistical
Mechanics and its Applications. 392(2): 371–380.
Gao, T. and Chen, Y. (2017) A quantum anharmonic oscillator model for the stock market. Physica
A: Statistical Mechanics and its Applications. 468: 307–314.
Kim, M. J., Kim, S. Y. Hwang, D. I., Lee, S. Y. (2011) The sensitivity analysis of propagator for
path independent quantum finance model. Physica A: Statistical Mechanics and its Applications.
390(5): 847–863.
Mantegna, R. M. and Stanley, H. E. (1999) Introduction to Econophysics: Correlations and Com-
plexity in Finance. Cambridge University Press, 1st edition.
Meng, X., Zhang, J. and Guo, H. (2015) Quantum spatial-periodic harmonic model for daily price-
limited stock markets. Physica A: Statistical Mechanics and its Applications. 438: 154–160.
Nakayama, Y. (2009) Gravity dual for Reggeon field theory and nonlinear quantum finance. Inter-
national Journal of Modern Physics A. 24(32): 6197–6222.
Nasiri, S., Bektas, E. and Jafari, G. R. (2018) The impact of trading volume on the stock mar-
ket credibility: Bohmian quantum potential approach. Physica A: Statistical Mechanics and its
Applications. 512: 1104–1112.
Piotrowski, E. W. and Sładkowski, J. (2005) Quantum diffusion of prices and profits. Physica A:
Statistical Mechanics and its Applications, 345(1–2): 185–195.
QFFC (2019) Official site of Quantum Finance Forecast Center. http://qffc.org. Accessed 22 Aug
2019.
Schaden, M. (2002) Quantum finance. Physica A: Statistical Mechanics and its Applications.
316(1–4): 511–538.
Shi, L. (2006) Does security transaction volume-price behavior resemble a probability wave? Phys-
ica A: Statistical Mechanics and its Applications. 366: 419–436.
Swanson, M. S. (2014) Feynman’s Path Integrals and Quantum Processes (Dover Books on Physics).
Dover Publications.
References 63

Tuchong (2019) Quantum Finance – which way to go? http://stock.tuchong.com/image?imageId=

383418791542128669. Accessed 22 Aug 2019.
Ye, C. and Huang, J. P. (2008) Non-classical oscillator model for persistent fluctuations in stock
markets. Physica A: Statistical Mechanics and its Applications. 387(5): 1255–1263.
Zee, A. (2011) Quantum Field Theory in a Nutshell. Princeton University Press, 2nd edition.
Zhang, C. and Huang, L. (2010) A quantum model for the stock market. Physica A: Statistical
Mechanics and its Applications. 389(24): 5769–5775.
Chapter 4
Quantum Finance Theory

The mathematical framework of quantum theory has passed

countless successful tests and is now universally accepted as a
consistent and accurate description of all atomic phenomena.
Erwin Schrödinger (1887–1961)

If financial markets really exhibit quantum properties

How can we model it??
What do we want to achieve?
Since Professor Erwin Schrödinger proposed his influential Schrödinger equation in
1925 (Schrödinger 1926)—the core of quantum mechanics and quantum field the-
ory—the studies and research of this “spooky” quantum world were thrived. The
advance of computing technology nurtured the study of this unique subatomic world
with the possibility of solving these highly complex quantum equations using con-
temporary computer technology is no more a reverie.
If we say 1920s was quantum theory’s first golden age, the launch of IBM System
ONE—first commercial-used quantum computer road show in Las Vegas at CES2019
early this year coined quantum theory’s second golden age—the age of quantum
computing.
Why now?
As there is a saying, things wouldn’t happen simply by chance. The only thing is
whether we know their cause-and-effect—the causality.
If we say quantum computer is the machine that drives this second golden age
of quantum theory, the quantum dynamics model we will study in this chapter—the
quantum anharmonic oscillator (QAOH) model—can be considered as the soul of
this giant machine.
Based on concepts and theories of quantum mechanics and quantum field theory
studied in previous chapters, in this chapter, we will introduce the core concept of
this book—quantum finance theory. First, we will study basic conceptual model of
quantum finance theory—the quantum anharmonic oscillator model and introduce
quantum price level (QPL) concept for the modeling of quantum finance energy
levels. After that, we will revisit Schrödinger equation and combine with the author’s

© Springer Nature Singapore Pte Ltd. 2020 65

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_4
66 4 Quantum Finance Theory

latest work on quantum dynamics modeling of various parties in a secondary financial

market (SFM), along with a step-by-step mathematical derivation of quantum finance
Schrödinger equation (QFSE).

4.1 Quantum Finance Theory

4.1.1 The Concept

In quantum finance, we model the dynamics of financial instruments (such as curren-

cies, financial indices, cryptocurrencies) of worldwide financial markets as quantum
financial particles (QFPs) with wave–particle duality characteristics.
The motions and dynamics significance of these quantum financial particles are
subject to their intrinsic quantum energy fields, the so-called quantum price fields
(QPFs) and appeared to us as quantum price levels (QPLs) in financial markets.
They are similar to quantum particles that are affected by the superposition of their
own energy levels and the energy field generated by other neighboring quantum
particle(s). Figure 4.1 illustrates the quantum financial energy fields created by three
types of financial particles: US dollars, Euros, and British pounds.
From technical finance perspective, these quantum price levels correspond to the
support and resistance (S & R) levels as we know of.
In other words, one of the major objectives of quantum finance theory is to estab-
lish an effective and logical quantum finance model and help us to locate all these

Fig. 4.1 Illustration of quantum financial field for worldwide financial products
4.1 Quantum Finance Theory 67

QPLs of worldwide financial markets using quantum mechanics and quantum field
theories. Such quantum finance model must be logically sound and should be a
coherent body of classical finance concepts and models.
If dynamics and motions of all fundamental particles found in the physical world
are governed by quantum mechanics and quantum field theory as mentioned in
Chap. 3, it is reasonable and rational to say that the worldwide financial markets—
more precisely—are the price/index movements of all different financial instruments
in every single time step, which are believed as results of the collective conscious,
behavior, and decision-making of all participants in the markets that can also be
modeled by quantum theory without exception.
The only problem is: How can we model them in terms of quantum mechanics
and quantum field theory?
More importantly, how to implement these mathematical models in a logical,
sensible, and effective manner so that they can apply and implement them in any
digital computers?

4.1.2 The Notion of Quantum Price

There is one pivotal analog of quantum finance theory and classical quantum theory
is the notion of the fundamental particles in quantum finance—quantum price.
From the finance perspective, we are all familiar with that the current price (spot
price) of any financial product not only represents instantaneous price value but also
the price level (a kind of energy level in finance) that resides at every particular
moment as shown in Fig. 4.2.

Fig. 4.2 Stock market price movement (Tuchong 2019a)

68 4 Quantum Finance Theory

In other words, in classical finance, price itself is not only a scalar value, but in
some way; it also represents the intrinsic energy it possesses (analog to the potential
energy in classical mechanics). In terms of classical finance, the meaning of price
for any financial market has three different interpretations:
1. It represents the instantaneous price value of that particular financial product at
a particular time;
2. It represents the time of the particular financial product which reaches that par-
ticular price value. That’s why when we say the price of a particular financial
product (e.g., stock), we also speak about the time;
3. It represents that intrinsic energy of that financial product at a particular time,
which similar to the interpretation for the potential energy of objects and matters
in the physical world.
As analog to quantum theory, it is similar to the particle existed and influenced
by its own quantum potential well (Harrison and Valavanis 2016). Figure 4.3 shows
a typical example of the quantum potential energy well generated by a quantum
particle (e.g., an atom).
The truth is: the analog of financial particles in quantum finance theory using
quantum theory is even more sensible and understandable in the finance community
than the notion of quantum particles in its quantum field. The only concern is: How
can we make use of classical finance concepts and beliefs to establish quantum
dynamics of quantum finance particles?

Fig. 4.3 Quantum potential energy level well generated by a quantum particle (e.g., an atom)
4.1 Quantum Finance Theory 69

4.1.3 Revisit of Schrödinger Equation

It was recalled that in quantum theory, any quantum particle shown in Fig. 4.4 at
position x, the time-dependent Schrödinger equation (Zee 2011) is given by

∂
i ψ(x, t) = Ĥ ψ(x, t) (4.1)
∂t

where ψ(x, t) is the wave function, Ĥ is the Hamiltonian operator, and  is the
reduced Planck constant, whereas the corresponding time-independent counterpart
is given by

Ĥ ϕ(x) = Eϕ(x) (4.2)

where

ψ(x, t) = ϕ(x) e−iEt/

Note that the time-independent Schrödinger equation is a standard eigenfunction

where E is the eigenenergy with discrete energy levels, and the Hamiltonian operator
Ĥ is given by

− ∂ 2
Ĥ = + V (x) (4.3)
2m ∂x2

where Ĥ comprises kinetic energy, K.E. (first term) + potential energy, P.E. (second
term).

Fig. 4.4 Illustration of quantum particle in the subatomic world (Tuchong 2019b)
70 4 Quantum Finance Theory

4.2 Quantum Finance Model

4.2.1 The Mathematical Model

Let r be the price return of a particular quantum financial particle (QFP) at time t
(say USD/CAD or simply US Index).
We can rewrite Schrödinger equation in (4.1) as

∂
i ψ(r, t) = Ĥ ψ(r, t) (4.4)
∂t

and the corresponding Ĥ is given by

− ∂ 2
Ĥ = + V (r, t) (4.5)
2m ∂r 2
where
1. Ĥ comprises K.E. (kinetic energy, the first term) and P.E. (potential energy, the
second term);
2.  is the Planck constant representing the uncertainty of financial behavior;
3. m is the mass representing the intrinsic potential of financial market, such as the
market capitalization of a particular financial product in financial market.
As mentioned in Chap. 3, the usage of price returns (r) instead of price (p) itself
is threefold:
1. Use price returns (r) instead of price (p) on mathematical and statistical modeling
is a kind of de facto standard in classical financial and statistical finance theory;
2. It is more sensible and easier to implement, especially if we want to compare
the performance of two (or more) financial products. By using price return (r)
instead of price (p), even when the two financial products have totally different
orders of magnitude, e.g., Dow Jones Index versus CADUSD, one is at order
1 × 104 and the other at order of 1 × 101 ;
3. In terms of quantum theory, the physical meaning of x in time-dependent
Schrödinger equation (4.1) is the displacement of quantum particle relative to its
equilibrium position, and not the distance of travel in classical mechanics as in
Newton’s laws of motion. Therefore, it is obvious and more sensible to use price
return (r) instead of price (p) as physical interpretation of r (say daily return) can
be considered as the relative price change of that particular financial product at
present compared with its closing price of the previous trading day.
4.2 Quantum Finance Model 71

4.2.2 Physical Meaning of Wave Function ψ

Wave function (ψ) is the most important component in the entire quantum theory. It is
the realization of wave–particle duality of quantum particles, which can be considered
as the bridge to link up between particle and wave realms in the subatomic world as
shown in Fig. 4.5.
What is the wave function (ψ) in quantum finance?
Can we observe it, or measure it?
As recalled that in the “double-slit experiment” (Aharonov et al. 2017), what we
saw in the backdrop screen is the realization of wave function from quantum particle
projections.
However, if we measure it “closely”, all wave-like phenomena will disappear. Can
we also recall that?
So, what shall we do?

Fig. 4.5 Solution of Schrödinger’s equation for quantum harmonic oscillators (left) with their
amplitudes (right)
72 4 Quantum Finance Theory

In quantum theory, we measure the pdf (probability density function, ρ) of obser-

vations instead, which is given by

ρ(x, t) = |ψ(x, t)|2 = |ϕ(x)|2 (4.6)

The wave function in quantum finance can be observed by using similar method.
For example, for a time series of 1000 trading days, we measure the closing price
return of these 1000 time steps and plot the probability density function (pdf ) of
price return r versus pdf of occurrence  [0, 1] in order to analog ψ.
That is,

ρ(r, t) = |ψ(r, t)|2 = |ϕ(r)|2 (4.7)

4.3 Financial Dynamics

4.3.1 Key Players in Secondary Financial Markets

Once we have the financial model, the next step is to explore the dynamics which
means all motions and activities occur inside the model. That is: what are the dynam-
ics in a typical secondary financial market such as forex, commodity, or cryptocur-
rency?
In other words, what are the major participants in a financial market? What are
their behaviors?
For example, in forex market—the biggest OTC (over the counter) market in the
worldwide finance (Hull 2016), what are the key participants?
Figure 4.6 shows a framework in a typical secondary financial market (SFM) such
as worldwide forex markets (Strumeyer 2017).

4.3.2 Market Maker (MM)

Market maker (MM), also known as liquidity provider, is a company or an individual

that quotes both a buy and a sell price in a financial instrument or commodity held in
inventory, hoping to make a profit on the bid-offer spread, or turn (Čekauskas et al.
2012; Zhu et al. 2009).
The U.S. Securities and Exchange Commission defines market maker as firms
that stand ready to buy and sell stock on a regular and continuous basis at a publicly
quoted price.
Most foreign exchange trading firms are market maker and so are many banks.
4.3 Financial Dynamics 73

Fig. 4.6 Key Players in a typical secondary financial market (SFM)

The most common type of market maker is brokerage houses that provide purchase
and sale solutions for investors, in an effort to keep financial markets liquid.
Market maker can also be an individual intermediary, but due to the size of secu-
rities needed to facilitate the volume of purchases and sales, the vast majority of
market makers work on behalf of large institutions.
They sell to and also buy from clients, compensated by means of price differentials
for services in providing liquidity, reducing transaction costs, and facilitating trade.
They earn profit by the difference between the price at which they are willing to
buy a stock and the price that the firm is willing to sell it, also known as the market
spread or bid-ask spread.
Market maker also provides liquidity to their own firm’s clients, for which they
earn a commission.
In terms of investment dynamics, the main function is to maintain healthy market
liquidity or facilitate the efficient absorption of buy/sell orders (Fig. 4.7).

4.3.3 Arbitrageurs (ARs)

Arbitrageurs are traders that take advantage of a price difference between two or
more markets.
They look for an imbalance in the pricing of a security in two different markets
and buy it at the cheap price in one market to immediately sell it at the higher price
in the other market, making a profit on the difference (Hong et al. 2012; Matsushima
2013).
This kind of operation is called arbitrage.
74 4 Quantum Finance Theory

Fig. 4.7 The Market Maker (MM)

Arbitrageurs are typically very experienced investors since arbitrage opportunities

are difficult to find and require relatively fast trading. Arbitrageurs also play an
important role in the operation of capital markets, as their efforts in exploiting price
inefficiencies keep prices more accurate than they otherwise would be.
However, in an efficient market nowadays with open information and high-speed
trading, there are basically no rooms for arbitragers trading (Malkiel 2003).
Example of Arbitrage Trading
Suppose that the exchange rates of EUR/USD in Frankfurt quotes at $1.25, while in
New York it quotes at $1.3 at the same time.
Hence, the euro is undervalued in Frankfurt and overvalued in New York. An
arbitrageur would exploit this imbalance to exchange dollars for euros in Frankfurt
and immediately exchange the euros for dollars in New York.
The profit for each dollar in this double exchange is 0.8 − 0.769 = 0.031 or a 3%
on the amount invested (Fig. 4.8).

4.3.4 Speculators (SPs)

Speculators take risk on purpose by betting on future movements of the security’s

price (Brunetti et al. 2016; Yi 2017).
For example, a speculator buys a stock under the conviction that its price will rise
without any logical and rational reasons most of the time.
Many speculators pay little attention to the fundamental value of a security and
instead purely focus on price movements.
In finance, speculation is also the practice of engaging in risky financial transac-
tions in an attempt to profit from short-term fluctuations in market value of a tradable
financial instrument—rather than attempting to profit from the underlying financial
attributes embodied in the instrument such as capital gains, dividends, or interest.
4.3 Financial Dynamics 75

Fig. 4.8 The Arbitrageurs

(ARs)

Speculators are important to market because they bring liquidity besides assume
market risk and, conversely, they can also have a negative impact on markets when
their trading actions result in a speculative bubble that drives up an asset’s price
to unsustainable levels. If a speculator believes that a particular asset is going to
increase in value, they may choose to purchase as much of the asset as possible.
This activity, based on the perceived increase in demand, drives up the price of the
particular asset. If this activity is seen across the market as a positive sign, it may
cause other traders to purchase the asset as well, further elevating the price. This can
result in a speculative bubble, where the speculator activity has driven the price of
an asset above its true value.
In terms of investment behavior, speculators differ from common investors in
the sense that they don’t have any risk control mindset. In other words, there is no
damping factor against market volatility in their investment strategies (Fig. 4.9).

4.3.5 Hedgers (HGs)

Hedgers trade so to reduce or eliminate the risk of taking a position on security.

The main goal is to protect the portfolio from losing value at the expense of
lowering the possible benefits.
This attitude (or trading strategy) is called hedging.
A hedging strategy usually involves taking contrarian positions in two or more
securities. That’s why many people also call hedgers as contrarians, but actually they
are not always 100% acting against the market trends (Radalj 2006; Röthig 2011).
Speculators and hedgers are different terms that describe traders and investors.
Speculation involves trying to make a profit from a security’s price change, whereas
76 4 Quantum Finance Theory

Fig. 4.9 Speculators (SPs)

hedging attempts to reduce the amount of risk, or volatility, associated with a secu-
rity’s price change (Yung and Liu 2009; Lin et al. 2009).
Hedging involves taking an offset position in a derivative in order to balance any
gains and losses of the underlying asset. Hedging attempts to eliminate the volatility
associated with an asset price by taking offset positions contrary to what the investor
currently has. The main purpose of speculation, on the other hand, is to profit from
betting on the direction in which an asset will be moving.
For example, taking a long position on a stock and a short position on another
stock with inverse price behavior or the portfolio risk control strategy we have just
mentioned (Fig. 4.10).

4.3.6 Investors (IVs)

An investor is a person who allocates capital with the expectation of a future financial
return.
Types of investments include equity, debt securities, real estate, currency, com-
modity, token, derivatives such as put and call options, futures, forwards, etc.
That is, someone who provides a business with capital and someone who buys a
stock are both investors.
There are two types of investors, individual (independent) investors and institu-
tional investors.
Individual investor is a nonprofessional investor who buys and sells securities,
mutual funds, or exchange-traded funds (ETFs) through traditional or online broker-
age firms or savings accounts. Retail investors invest much smaller amounts than large
institutional investors, such as mutual funds, pensions and university endowments,
4.3 Financial Dynamics 77

Fig. 4.10 Hedgers (HGs)

(Tuchong 2019c)

and trade less frequently. But wealthier retail investors can now access alternative
investment classes like private equity and hedge funds.
Institutional investors (Davis and Steil 2004; Gabaix et al. 2006) are major players
in the financial market. They are the pension funds, mutual funds, and also some
private equity investors. Institutional investors account for about three-quarters of
the volume of trades on the New York Stock Exchange. They move large blocks of
shares and have tremendous influence on stock market’s movements.
In terms of investment dynamics, investors normally act as Trend Followers
together with certain degree of sense of risk control (i.e., certain degree of damping
factor against market volatility in their investment strategies) (Fig. 4.11).

4.4 Financial Dynamics and the Notion of Excess Demand

Once we have the key player in financial market, how can we model their financial
dynamics?
The notion of excess demand (z).
In classical finance and microeconomics, excess demand is a function express-
ing excess demand for a product—the excess of quantity demanded over quantity
supplied—in terms of the product’s price and possibly other determinants. In a math-
ematical perspective, it is the product’s demand function minus its supply function.
78 4 Quantum Finance Theory

Fig. 4.11 Investors (IVs)

In a pure exchange economy, the excess demand is the sum of all agents’ demands
minus the sum of all agents’ initial endowments (Tian 2016; Momi 2010).
The price of the product is said to be the equilibrium price if it is such that the
value of the excess demand function is zero: which means, when the market is in
equilibrium, the quantity supplied equals the quantity demanded. In this case, it is
said that the market clears. If the price is higher than the equilibrium price, excess
demand will normally be negative, meaning that there is a surplus (positive excess
supply) of the product, and not all of it being offered to the marketplace is being
sold. If the price is lower than the equilibrium price, excess demand will normally
be positive, meaning that there is a shortage of demand in the financial market.
Mathematical Derivations
At any time t, z+ (t) and z− (t) denote the instantaneous demand and supply for the
financial asset.
The excess demand (z) at any instance is given by

Δz = z+ − z− (4.8)

Let r(t) is the instantaneous returns, which is given by

dp
= r(t) = F(Δz) (4.9)
dt
For small Δz, F can be approximated by a scaling factor γ and become

Δz
r(t) = (4.10)
γ
4.4 Financial Dynamics and the Notion of Excess Demand 79

where γ can be used to represent the market depth, the excess demand z required
to move the quantum price p by one single quantum. Note that, when γ is high, it
means that the market has a higher absorbability to excess demand z against price
changes.
So, we have

dr d 2p 1 d (Δz)
= 2 = (4.11)
dt dt γ dt

4.5 Quantum Dynamics in Financial Markets

According to the investment behaviors of all these five key participants in financial
markets, their corresponding quantum dynamics can be interpreted as follows.

4.5.1 Market Makers (MMs)

The quantum dynamics for market makers is given by


dz+ 
= −α+ z+ and (4.12a)
dt MM

dz− 
= −α− z− (4.12b)
dt MM

Note:
1. Market makers (MMs) provide market facilitator services to absorb ALL out-
standing excess order z+ and z− .
2. α+ and α− are the market absorbability factors.
3. In terms of quantum dynamics, basically it is a quantum harmonic oscillator
(QHO) with ż± ∝ z± .
Combining with Eq. (4.8), we have
 
d Δz  d (z+ − z− ) 
=
dt MM dt 
 MM

dz+  dz− 
= −
dt MM dt MM
= −α+ z+ + α− z− (4.13)

For an efficient market, we can assume

80 4 Quantum Finance Theory

α+ = α− = αMM (4.14)

So, we have

d Δz 
= −αMM Δz (4.15)
dt MM

Using Eq. (4.10), we have


d Δz 
= −γ αMM r (4.16)
dt MM

4.5.2 Arbitrageurs (ARs)

As mentioned, in an efficient market nowadays with open information and high-speed

trading, there are basically no rooms for arbitrageurs’ trading.

Principle of No Arbitrage on Security Markets

There are no-arbitrage opportunities in Reality.

Extended No Arbitrage on security markets states:

1. Arbitrage is not possible.
2. There are no transaction costs, no taxes, and no restrictions on short selling.
3. It is possible to borrow and lend at equal risk-free interest rates.
4. All securities are perfectly divisible.

The principle of no-arbitrage conditions states that market participants cannot

claim risk-free return/profit unless there are conditions that facilitate to those claims.
The no-arbitrage rule maintains a key position in analyzing price and price move-
ments in financial markets as well as being a central component of such theories and
concepts as efficient market hypothesis.
So, assume Principle of No Arbitrage is in every quantum time step, there is no
quantum dynamics for arbitrageurs.
In reality, even though there might have a chance of arbitrageurs for an instance
of time, but within an efficient market and high-speed trading, such dynamics will
soon be died-down and absorbed by the market makers (MMs).
In respect of quantum finance dynamics of the quantum financial particles (QFPs)
and the high speed of electronic transactions in every moment of time, it is practically
impossible nowadays for arbitrageurs to make profit by looking for the price differ-
ence between different financial markets of the same product. The quantum dynamics
4.5 Quantum Dynamics in Financial Markets 81

of the same product between different markets synchronize to their equilibrium states
at every moment.

4.5.3 Speculators (SPs)

Contradictory, speculators always emerge in every financial market.

Their quantum dynamics are given by

d Δz 
= −δSP r (4.17)
dt SP

Since speculators have no idea of risk control, their quantum dynamics only
contain the harmonic oscillator term (the delta term, δSP r), without any higher order
volatility term.
Note that, although speculators might happen to be trend follower (+δSP ), most
of the time their risk-taking nature will drive them to irrational speculation of market
reversals and act against the market (−δSP ).
As shown in Eq. (4.17), the existence of the only first-order return r means that
the entire quantum dynamics of speculators (SPs) is a simple quantum harmonic
oscillator (QHO) without any anharmonic component.
From the physical meaning perspective, the existence of the only first order of
return (r) in Eq. (4.17) means that dynamics of the speculators only exist the force
constant which is the same as the force constant for spring in Newton’s mechanics.

4.5.4 Hedgers (HGs)

Hedgers (HGs) represent experienced and skillful traders (also known as sophisti-
cated traders) that apply sophisticated hedging strategies across different products
and markets. Further will be studied in Chap. 6.
Although they do not always act against the trend, their skills usually demonstrated
by reverse trading or prediction of market reversal and act before common investors.
So, their quantum dynamics are given by

d Δz   
 = − δHG − υHG r 2 r (4.18)
dt HG

Note that the quantum dynamics for a hedger has two terms: (1) quantum harmonic
oscillatory term (delta term)—proportion to return (r) and (2) quantum anharmonic
term (υ) stands for the market volatility, risk control factor proportional to r 2 .
82 4 Quantum Finance Theory

In terms of quantum dynamics, the existence of both first-order and third-order

terms of return r cause it to become a typical quantum anharmonic oscillator (QAOH)
(Bhargava et al. 1989; Gao and Chen 2017).

4.5.5 Investors (IVs)

As mentioned, investors represent common investors and rational investors with

certain degree of risk control.
Their unusual strategies (1) follow the trend to gain profit and (2) minimize risk.
So, their quantum dynamics are given by

d Δz   
 = δI V − υI V r 2 r (4.19)
dt I V

Note that, similar to hedgers, the quantum dynamics of an investor has two terms:
(1) quantum harmonic oscillatory term (delta term)—proportion to return (r) and
(2) quantum anharmonic term (υ) stands for the market volatility, risk control factor
proportional to r 2 .
But different from hedgers, common investors are usually trend followers (TFs),
so they are basically acting toward returns (r).

4.5.6 Overall Dynamics for All Different Parties

By combining quantum dynamics of all key participants in the financial market, the
overall quantum dynamics in a typical financial market is given by
   
d Δz d Δz  d Δz  d Δz  d Δz 
= + + + (4.20)
dt dt MM dt SP dt HG dt I V
d Δz    
= −γ αMM r − δSP r − δHG − υHG r 2 r + δI V − υI V r 2 r (4.21)
dt
That is,

d Δz
= −δr + υr 3 (4.22)
dt
Combining Eq. (4.11), we have

dr d Δz
=γ = −γ δr + γ υr 3 (4.23)
dt dt
4.5 Quantum Dynamics in Financial Markets 83

where
δ = γ αMM + δSP + δHG − δI V (damping term) and
υ = υHG − υI V (volatility term)
The Brownian price return can be described by Langevin equation:

d 2r dr d V (r)
mr 2
= −η − (4.24)
dt dt dr
where
mr mass of the financial particle p;
η damping force factor;
V (r) time-independent quantum potential.
For the consistency of Eqs. (4.23) and (4.24), i.e., overdamping case where
d 2r
dt 2
= 0, we have

d V (r) dr
− = η = −γ ηδr + γ ηυr 3 (4.25)
dr dt

  γ ηδ 2 γ ηυ 4
V (r) = −γ ηδr + γ ηυr 3 dr = r − r (4.26)
2 4

So, the time-independent Schrödinger Eqs. (4.4)–(4.5) of a quantum finance par-

ticle can be written as follows:

Quantum Finance Schrödinger Equation, QFSE

  
− d 2 γ ηδ 2 γ ηυ 4
+ r − r ϕ(r) = Eϕ(r) (4.27)
2m dr 2 2 4

Note:
1. The time-independent Schrödinger equation of a quantum finance contains both
KE and PE terms.
2. Different from classical quantum harmonic oscillator, the quantum finance oscil-
lator is a quantum anharmonic oscillator which consists of two high-order PE
terms that represent (1) damping (trading restoration and market absorption)
potential and (2) volatility (risk control) potential.
3. Although the market is visualized (observed) as price, the quantum dynamics

are controlled by price return (r = dp dt ), which is consistent with classical

financial theory.
84 4 Quantum Finance Theory

4.6 Conclusion

In this chapter, we introduced the main theme of this book: quantum finance theory
using quantum anharmonic oscillator (QAOH) model. We began with quantum finan-
cial particles (QFPs) concept and their intrinsic quantum energy fields, the so-called
quantum price field (QPF). We also studied the notion of quantum price in quantum
finance and its relationship with quantum energy level. After that, we examined the
Schrödinger equation and explored the physical meaning of wave function (ψ).
We then studied the core of this chapter—the five key players in secondary
financial markets (SFMs)—market makers, arbitrageurs, speculators, hedges, and
investors. First, we analyzed their roles, characteristics, and behaviors in the finan-
cial market. After that, we introduced classical financial dynamics and the notion of
excess demand.
Based on the definition of excess demand, we derived quantum dynamics of
these five key parties and deduced the overall quantum dynamic equation of their
combined behaviors in secondary financial market. By combining with Eq. (4.11) and
the assumption of overdamping case, we derived the quantum finance Schrödinger
equation (QFSE) in form of an order-4 quantum anharmonic oscillator system.
An important issue of our step-by-step derivation of the QFSE is that we began
the entire mathematical modeling quantum dynamics of secondary financial market
based on classical belief and understanding in terms of classical finance and microe-
conomic theory. By doing so, we can ensure that the QFSE derived can be consistent
with basic financial concept and theory, despite we are now using completely new
perspective and tool to model the dynamics for secondary financial markets.
In the next chapter, we will explore how to use QFSE, in conjunction with numeri-
cal computational technique and finite difference method (FDM) to derive and evalu-
ate quantum finance energy level (QFEL), together with quantum price levels (QPLs)
for any financial products in secondary financial markets.

Problems

4.1 In the past, quantum theory including quantum mechanics and quantum field
theory are queried by many mainstream scientists and believed to be some
sort of “spooky” science. However, nowadays many scientists and general
community accept the general idea of quantum theory and the existence of
subatomic world. Why?
4.2 What is quantum computer? What is the major difference between quantum
computer and the contemporary computers we are using?
4.3 What is the analog between quantum particles in quantum theory and the
dynamics of the financial particles in financial markets? Give a live example
such as forex market for illustration.
4.6 Conclusion 85

4.4 As a basic concept in finance, “prices” in a financial market, such as Dow

Jones index (DJI), have three physical and mathematical meanings and inter-
pretations, what are they? How can these three characteristics of prices relate
to quantum finance theory?
4.5 What are the major similarities and differences between prices in financial
markets versus quantum particles in quantum theory?
4.6 Quantization phenomena are easily found in financial markets and commonly
accepted in technical analysis. Why? Give three examples of technical analyt-
ical tools and explain how they work in the realm of quantum finance.
4.7 State the time-dependent Schrödinger equation and its Hamiltonian operator.
(i) Identify kinetic energy (KE) and potential energy (PE) terms in the for-
mulation;
(ii) State and explain how these two terms are interpreted in quantum finance.
4.8 In computational finance and quantum finance, we usually use price returns
(r) instead of prices (p) for the mathematical formulation and financial anal-
ysis. Why? What is/are the advantage(s) of using returns instead of prices
for mathematical formulation? Use two financial products as examples for
illustration.
4.9 What are the physical meanings of wave function (ψ) in (1) quantum theory
of quantum particles; (2) quantum finance of quantum financial particles?
4.10 How can we observe and measure wave function (ψ) values (and distribution)
in terms of (1) quantum theory of quantum particles such as photons; (2)
quantum finance theory of quantum financial particles such as Dow Jones
index?
4.11 The below figure shows the solutions of Schrödinger’s equation for quantum
harmonic oscillators (left) and their amplitudes (right).
86 4 Quantum Finance Theory

(i) Write a MATLAB program to solve the Schrödinger’s equation for quan-
tum harmonic oscillators and plot these figures;
(ii) Describe their physical meanings in terms of quantum finance theory of
financial particles such as Dow Jones index (DJI);
(iii) Can we simply model quantum finance particles such as DJI using quan-
tum harmonic oscillators in reality? Why?

4.12 What is a secondary financial market (SFM)? What are the major differences
between primary financial market (PFM) and SFM? In each case, please give
two examples for illustration.
4.13 What is a market maker (MM)? Discuss and explain the role(s) and importance
of market maker in financial markets.
4.14 What are arbitrageurs? What are the major differences between arbitrageurs
and common investors in terms of (1) trading/investment strategy; (2) risk
control? Give two examples of financial markets for illustration.
4.15 What are speculators? What are the major differences between speculators and
common investors in terms of (1) trading/investment strategy; (2) risk control?
Why speculators are important in financial markets? Give two examples of
financial markets for illustration.
4.16 What are hedgers? What are the major differences between hedgers and com-
mon investors in terms of (1) trading/investment strategy; (2) risk control? Why
hedgers are important in financial markets? Give two examples of financial
markets for illustration.
4.17 Can an investor be both a hedger and speculator at the same time? Why or
why not? Give two examples of financial markets for illustration.
4.18 What are common investors (or investors in short)? What are the characteris-
tics of investors in terms of (1) trading/investment strategy; (2) risk control?
Why common investors are vital in financial markets? Give two examples of
financial markets for illustration.
4.19 Discuss and explain excess demand (z) in terms of (1) physical meaning in
financial markets; (2) mathematical formulation. How can it relate to quantum
finance?
4.20 State and explain the quantum finance formulations of the five key participants
in a typical financial market.
4.21 State and explain the principle of no arbitrage. Explain why it is important in
the formulation of the quantum dynamics of arbitrageurs in quantum finance.
4.22 What is an efficient market? Explain why it is important in the formulation of
the quantum dynamics of market maker in quantum finance.
4.23 What is quantum anharmonic oscillator (QAHO)? What is the major difference
between quantum harmonic oscillator (QHO) and its anharmonic counterpart
QAHO in terms of (1) quantum dynamics and (2) energy levels?
4.24 For the formulation of real-world financial market, should we choose quantum
harmonic oscillator (QHO) or quantum anharmonic oscillator (QAHO) for
system modeling and mathematical formulation? Why?
4.6 Conclusion 87

4.25 Combine the quantum finance mathematical formulation of all five key partici-
pants in a typical financial market and derive the quantum finance Schrödinger
equation.
4.26 State the time-independent quantum finance Schrödinger equation (QFSE).
(i) Describe and explain the physical meaning of each term in QFSE;
(ii) What is the physical meaning of (1) Planck constant, (2) mass, (3) damp-
ing factor η, gamma γ , and (4) delta δ terms in QFSE?
4.27 Discuss and explain the importance of quantum finance Schrödinger equation
(QFSE) in terms of (1) modeling of secondary financial markets and (2) real-
time financial prediction. Give two examples of secondary financial markets
as explanation.

References

Aharonov, Y. et al. (2017) Finally making sense of the double-slit experiment. Proceedings of the
National Academy of Sciences of the United States of America 114(25): 6480–6485.
Bhargava, V.T.A. et al. (1989) Anharmonic oscillators in higher dimension: Accurate energy eigen-
values and matrix elements. Pramana 32(2): 107–115.
Brunetti, C. et al. (2016) Speculators, Prices, and Market Volatility. Journal of Financial and
Quantitative Analysis 51(5): 1545–1574.
Čekauskas, K. et al. (2012) The Effects of Market Makers and Stock Analysts in Emerging Markets:
Market Makers and Stock Analysts. International Review of Finance 12(3): 305–327.
Davis, E. P. and Steil, B. (2004) Institutional Investors (The MIT Press). MIT Press.
Gabaix, X. et al. (2006) Institutional Investors and Stock Market Volatility. The Quarterly Journal
of Economics 121(2): 461–504.
Gao, T. and Chen, Y. (2017) A quantum anharmonic oscillator model for the stock market. Physica
A: Statistical Mechanics and its Applications. 468: 307–314.
Harrison, P. & Valavanis, A. (2016) Quantum Wells, Wires and Dots: Theoretical and Computational
Physics of Semiconductor Nanostructures, John Wiley & Sons.
Hong, H. et al. (2012) Do arbitrageurs amplify economic shocks? Journal of Financial Economics
103(3): 454–470.
Hull, J. C. (2016) Fundamentals of Futures and Options Markets. Pearson, 9th edition.
Lin, B. et al. (2009) Who bets against hedgers and how much they trade? A theory and empirical
tests. Applied Economics 41(27): 3491–3497.
Malkiel, B.G. (2003) The Efficient Market Hypothesis and Its Critics. The Journal of Economic
Perspectives 17(1): 59–82.
Matsushima, H. (2013) Behavioral aspects of arbitrageurs in timing games of bubbles and crashes.
Journal of Economic Theory 148(2): 858–870.
Momi, T. (2010) Excess demand function around critical prices in incomplete markets. Journal of
Mathematical Economics 46(3): 293–302.
Radalj, K.F. (2006) Hedgers, speculators and forward markets: Evidence from currency markets.
Environmental Modelling and Software 21(9): 1381–1386.
Röthig, A. (2011) On speculators and hedgers in currency futures markets: who leads whom?
International Journal of Finance & Economics 16(1): 63–69.
Schrödinger, E. (1926) An Undulatory Theory of the Mechanics of Atoms and Molecules. Physical
Review 28: 1049–1070.
88 4 Quantum Finance Theory

Strumeyer, G., Swammy, S. (2017) The Capital Markets: Evolution of the Financial Ecosystem
(Wiley Finance). Wiley.
Tian, G. (2016) On the existence of price equilibrium in economies with excess demand functions.
Economic Theory Bulletin 4(1): 5–16.
Tuchong (2019a) Stock Market Price Movement. http://stock.tuchong.com/image?imageId=
455757697366360685. Accessed 21 Aug 2019.
Tuchong (2019b) Illustration of quantum particle in the subatomic world. http://stock.tuchong.com/
image?imageId=488506530928001038. Accessed 21 Aug 2019.
Tuchong (2019c) Hedgers. http://stock.tuchong.com/image?imageId=75952820146313693.
Accessed 21 Aug 2019.
Yi, M. (2017) Speculator-triggered crisis and interventions. Journal of Macroeconomics 52:
135–146.
Yung, K. & Liu, Y. (2009) Implications of futures trading volume: Hedgers versus speculators.
Journal of Asset Management 10(5):318–337.
Zee, A. (2011) Quantum Field Theory in a Nutshell. Princeton University Press, 2nd edition.
Zhu, M. et al. (2009) Does the market maker stabilize the market? Physica A: Statistical Mechanics
and its Applications 388(15): 3164–3180.
Chapter 5
Quantum Price Levels—Basic Theory
and Numerical Computation Technique

It is true that in quantum theory we cannot rely on strict

causality. But by repeating the experiments many times, we can
finally derive from the observation’s statistical distributions, and
by repeating such series of experiments, we can arrive at
objective statements concerning these distributions.
Werner Heisenberg (1901–1976)

What are quantum price levels?

Do they really exist?
How can we model them?
How can we make use of them?
Professor Werner Heisenberg in this famous quotation stated an important fact: To
prove or disprove any new concept or theory, simple concepts and ideas are just
not enough. What we need are some hard evidences from repeating the experiments
many times to explore the general observation’s statistical distributions. Based on
these statistical distributions, we discover some critical patterns. From that, we can
conclude with some objective statements and findings to prove or disprove our pro-
posed concept or theory.
It is consistent with German philosopher Immanuel Kant’s remarkable quotation
regarding the relationship between theory and experience:
Experience without theory is blind, but theory without experience is mere intellectual play.
Immanuel Kant (1724–1804)

In his famous quotation, Immanuel Kant pointed out that the important relation-
ship between concepts and ideas versus experiences and experimental observations
is that although concepts, ideas, and theories are the foundation and critical mass for
new knowledge, without solid objective experimental observations and experiences,
they are merely intellectual fantasies.
The merit of science against pseudoscience is that scientific theories and concepts
can be proved, deduced, and observed in the form of mathematical proven, deduction,
and tested objectively through experiments. Thus, scientific attitude is even more
critical in quantum theory because basic concept and theory of subatomic world and

© Springer Nature Singapore Pte Ltd. 2020 89

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_5
90 5 Quantum Price Levels—Basic Theory and Numerical Computation …

its quantum phenomena are very different from our general belief of this physical
world.
The truth is: the mathematical framework of quantum theory has passed countless
successful tests with experiments and is now accepted universally as a uniform and
accurate description of all quantum phenomena in the subatomic world.
In this book, we follow the same footsteps.
In Chaps. 1–3, we introduced the world of quantum finance and studied its basic
concept and theory. Based on the quantum anharmonic oscillator (QAOH) model
learnt in Chap. 4, we examined its quantum dynamics and mathematical formulation
in a typical secondary financial market.
In this chapter, we discuss how such mathematical formulation can be applied
in terms of objective experimental results with distribution to evaluate the quantum
finance energy levels (QFELs) and quantum price levels (QPLs).
First, we will begin with the investigation of QPL as an analog to quantum energy
levels (QELs) in classical quantum world.
Next, we will explore Schrödinger equation and the physical meaning of wave
function in this equation. From that, we will study the experiments (time series trading
results) of financial products for over 2000 trading days and obtain the observation’s
statistical distributions. By repeating such series of experiments onto various financial
products, we can arrive at objective statements concerning these distributions. After
that, we will combine these observation results with finite difference method (FDM)
to derive the value of λ in the quantum finance Schrödinger equation (QFSE).
Finally, we will employ the latest research on λ2m QAOH model, together with
Cardano’s depressed cubic equation solver to solve quantum finance Schrödinger
equation (QFSE) and calculate all quantum finance energy levels (QFELs); and hence
the corresponding QPL using numerical computation method.

5.1 Quantum Price Levels (QPLs)

5.1.1 The Concept

What are quantum price levels?

As a direct analog to energy levels in an atom shown in Fig. 5.1, quantum finance
energy levels (QFELs) can be considered as invisible energy levels that exist in
every financial market; and quantum price levels (QPLs) can be interpreted as the
realization of these financial energy levels shown in every secondary financial market.
They are similar to quantum energy levels in an atom; these quantum price levels
exist intrinsic in nature as shown in Fig. 5.2.
In other words, they coexist with the continuance of financial particles in every
financial market.
From the finance perspective such as forex market, which means when market
opens every time, the quantum financial particles of various currency pairs (e.g.,
5.1 Quantum Price Levels (QPLs) 91

Fig. 5.1 Quantum energy levels in an atom

CADUSD, AUDEUR, JPYUSD) will automatically exist and generate their quantum
energy field instantaneously, visualized as quantum price levels (QPLs) shown in
every financial market and start their quantum finance anharmonic oscillations and
motions.
More importantly, these QPLs exist in discrete states with level 0 as the ground
states (E 0 ) (during market open) and all excited states E n in discrete energy levels.

5.1.2 The Quantum Dynamics

In terms of quantum dynamics, these energy states are eigenenergy levels described
in Schrödinger equation of quantum financial particles—the Quantum Finance
92 5 Quantum Price Levels—Basic Theory and Numerical Computation …

Fig. 5.2 Quantum price levels in quantum financial energy field

Schrödinger Equation (QFSE) derived in Chap. 4, which is given by the follow-

ing order-4 quantum anharmonic oscillator (5.1):
  
− d 2 γ ηδ 2 γ ηυ 4
+ r − r ϕ(r ) = Eϕ(r ) (5.1)
2m dr 2 2 4

Like an electron in an atom, without external force in action, the electron will
remain in its stable ground-state orbit.
Similarly, without external financial stimulus such as financial events, major finan-
cial news, significant worldwide events, the release financial index figures such as
PMI, CPI, PPI, etc., these financial particles will remain in their stable and equilib-
rium states (so-called ground states) and vibrates in the form of random-like quantum
anharmonic oscillations as described by QFSE.
However, when there is a significant financial stimulus (either positive or negative
events and stimulus), these financial particles will jump to another higher or lower
energy levels (i.e., QPLs), not continuous, but in the form of discrete jumps.
After that, market makers (MMs) will absorb the stimulus and financial particles
will be back to equilibrium states but reside at a new QPL.

5.1.3 QPLs Versus S & R Levels

Do such quantum price levels (QPLs) really exist?

5.1 Quantum Price Levels (QPLs) 93

Fig. 5.3 Support and resistance levels in technical analysis

The truth is … we don’t know.

But if we ask any experienced financial analysts and traders, they will all tell us that
financial markets are neither random motions nor continuous movements. Discrete
and quantized price levels do exist in all financial markets, ranging from stock, forex,
worldwide commodity, and even the latest cryptocurrency markets. That’s why we
have the concept of support and resistance levels (S & R levels) for over centuries
in the financial realm.
In fact, the technical analysis developed since the eighteenth century is based on
this fundamental concept and idea (Murphy 1999).
The difference is that, with the exponential growth of program trading, especially
the popularity of high-frequency program trading (HFPT), the financial markets
become so complex and chaotic that we don’t believe one can observe (locate) all
these hidden energy levels manually (Fig. 5.3).

5.2 Schrödinger Equation Revisit

As recalled in quantum dynamics (Zee 2011; Schmitz 2019), for a particle at position
x, the time-dependent Schrödinger equation is given by

∂ 

i ψ(x, t) = H ψ(x, t) (5.2)

∂t


where ψ(x, t) is the wave function, H is the Hamiltonian operator, and  is the reduced
Planck constant.
Note that the position x in Schrödinger equation is the displacement of quantum
particle from its equilibrium states, rather than the displacement stated in the classical
94 5 Quantum Price Levels—Basic Theory and Numerical Computation …

Newtonian dynamics, whereas the corresponding time-independent counterpart is

given by


H ϕ(x) = Eϕ(x) (5.3)

where ψ(x, t) = ϕ(x)e−i Et/ .

Note that the time-independent Schrödinger equation is a standard eigenfunction
where E is the eigenenergy with discrete energy levels and the Hamiltonian operator


H is given by


− ∂ 2
H= + V (x) (5.4)
2m ∂ x 2


where H composes of kinetic energy, K.E. (kinetic energy, the first term) + potential
energy, P.E. (potential energy, the second term).

5.3 Quantum Finance Model

Let r be the price return of a particular quantum financial particle (QFP) at time t
(say USD/CAD or simply US Index).
In typical secondary financial markets such as forex or stock market with over
2000-trading day historical records. Normally, we can use the daily returns to repre-
sent r in QFSE. In that case, the daily return r will be given by

T oday Closing Price

r= (5.5)
Y ester day Closing Price

Of course, in case of real-time high-frequency program trading (HFPT), such

return price r can use the hourly or 5-min returns as the basis of timeframe.
So, we can rewrite Schrödinger equation in (5.2) as

∂ 

i ψ(r, t) = H ψ(r, t) (5.6)

∂t


and the corresponding Hamiltonian operator H is given by


− ∂ 2
H= + V (r, t) (5.7)
2m ∂r 2
Note:


• The Hamiltonian operator H composes of K.E. and P.E.;

5.3 Quantum Finance Model 95

•  is the Planck constant representing the uncertainty of investors’ behavior in the

financial market;
• m is the mass representing the intrinsic potential of financial market (such as the
overall market capitalization of secondary financial market).

5.4 Physical Meaning of Wave Function Ψ

In quantum mechanics and quantum field theory, wave function (ψ) is the most
important component in the complete mathematical model, as it is the realization of
wave–particle duality of quantum particles in this unique subatomic world of reality.
What is the corresponding wave function in quantum finance?
Can we observe it, or measure it?
Remember in the double-slit experiment of quantum particle, what we have seen
in the backdrop screen is the realization of the wave function of quantum particle
projections.
However, if we measured it closely at every single path of motion of these quantum
particles, all the wave-like phenomena will disappear. Remember that?
In other words, if finance particles really possess such quantum duality property,
technically speaking, we cannot observe or measure these discrete quantum energy
levels (quantum price levels, QPLs, in our case) intentionally and directly. What we
can observe is the random-like price fluctuations in market movement chart.
So, what should we do?
The answer is to
1. follow Professor Heisenberg’s suggestion in his famous quote cited at the begin-
ning of this chapter; and
2. do the experiment many times and observe wave functions’ distribution.
When we studied classical quantum mechanics in Chap. 2, remember we talked
about evaluating the quantum wave function of a quantum particle; all we can do is
by measuring pdf (probability distributed function, ρ) of the observations instead,
which is given by

ρ(x, t) = |ψ(x, t)|2 = |ϕ(x)|2 (5.8)

In quantum finance, wave function can be observed and evaluated by using similar
method.
For example, for a time series of 2048 trading days of Gold versus US Dollar
(XAUUSD), we measure the daily closing price returns (r) of these 2048 time step
and plot the pdf of r versus pdf of occurrence
[0, 1] to analog the quantum finance
wave function ψ of XAUUSD.
That is,
96 5 Quantum Price Levels—Basic Theory and Numerical Computation …

Quantum Price Return Wavefunc on Q(r) of XAUUSD

(For the past 2048 trading-day sta s cs)
0.05

0.045

0.04

0.035

0.03
Q(r)

0.025

0.02

0.015

0.01

0.005

0
0.97 0.98 0.99 1 1.01 1.02 1.03
Daily Return (r)

Fig. 5.4 Quantum price return wave function Q(r) of XAUUSD for the past 2048-trading day (As
of January 6, 2019 market information)

ρ(r, t) = |ψ(r, t)|2 = |ϕ(r )|2 (5.9)

Figure 5.4 shows the quantum price return wave function Q(r) of XAUUSD for
the past 2048 trading days (as of January 6, 2019 market information).
It is calculated by evaluating the distribution function of daily price returns (r)
and plot against the total number of occurrences, given by

no. o f occur ences o f event r

r= (5.10)
total no. o f events E

where E is the total number of events, 2046 events in our case (with the exclusion of
the boundary cases).
Several features can be found:
1. The wave function falls into a standard normal distribution.
2. Maximum (also the mean) occurrence of r is close to 1, which is also the ground
state (without any stimulus). Such finding is quite surprising in the financial per-
spective. As one may think of the returns with the most frequency events should
be either greater than or less than 1, it won’t be 1 which means no change from
yesterday’s price. But the truth is, after calculating all the 129 worldwide finan-
cial products, almost all of them have similar patterns, in which the maximum
returns occur when r is close to (or equal to) 1.
3. Basically, the wave function is symmetric with the central axis which corresponds
to the mean (which is also the maximum) wave function value that can be reflected
5.4 Physical Meaning of Wave Function Ψ 97

by the fact that quantum finance Schrődinger equation (QFSE) is symmetric with
respect to the ground state. In other words, once we evaluate the excited states
of quantum finance energy levels (E1 , E2 , E3 , …) which corresponds to positive
QPL (the QPLs above the ground-state QPL when market open), all negative
QPL (the QPL below the ground-state QPL) can be deducted automatically, as
we will see in the upcoming numerical calculations.

5.5 Solving Quantum Finance Schrödinger Equation

Quantum finance Schrödinger equation (QFSE) is given by

  
− d 2 γ ηδ 2 γ ηυ 4
+ r − r ϕ(r ) = Eϕ(r ) (5.11)
2m dr 2 2 4

Once we have the method to evaluate ϕ(r ) (i.e., Q(r) in the last section), the center
question is to find all corresponding quantum price levels (QPLs).
That is, all the eigenenergy values in QFSE.
Numerous physicists and mathematicians in the past 50 years had devised many
methods and techniques to solve this important equation, such as the Hill determi-
nant, Bargmann representation, coupled-cluster method, and variation–perturbation
expansion (Grosse and Martin 2005; Muller-Kirsten 2012; Popelier 2011; Bouard
and Hausenblas 2019; Rampho 2017; Kisil 2012; Witwit 1996; Rohwedder 2013;
Carbonnière 2010).
However, most of them are either technically or mathematically complex in numer-
ical computations.
In 2007, Dasgupta et al. in their paper Simple systematics in the energy eigen-
values of quantum anharmonic oscillators published in the Journal of Physics A:
Mathematical and Theoretical provided a genius numerical method to solve a class
of Schrődinger equations known as “λx 2m quantum anharmonic oscillators.”
In the next section, let’s take a look at how we can adopt this numerical calculation
method to solve QFSE within seconds by simple PC computers.
We will be surprised by how genius and simple numerical methods can be by
changing the reference horizon.

5.6 Numerical Computation of Quantum Anharmonic

Oscillators

A typical λx 2m quantum anharmonic oscillators (aka AHO) is given by (Dasgupta

et al. 2007):
98 5 Quantum Price Levels—Basic Theory and Numerical Computation …

d 2ψ  2 
H (m) (λ)ψ = − + x + λx 2m ψ = Eψ (5.12)
dx2
in which the excited energy levels can be closely approximated by the following
polynomials:
 (m+1)  (m−1)
(m+1)
E (m,n) E (m,n)
− = K 0(m,n) λ (5.13)
2n + 1 2n + 1

where E (m,n) is the nth excited state energy of λx 2m AHO and K 0(m,n) are constants.
If we look QFSE closely, it is in fact a typical quartic anharmonic oscillator in
Chap. 2 (Sect. 2.5.3) with a quartic term in P.E. dynamics.
So, we can convert QFSE (5.11) into a λx 2m AHO:

d 2 ϕr  2 
2
+ r + λr 2m ϕr = Eϕr (5.14)
dr
Put m = 2, we have

d 2 ϕr  2 
2
+ r + λr 4 ϕr = Eϕr (5.15)
dr

Note: We normalize quadratic term r 2 with quartic term r 4 and combine it to coeffi-
cient λ. Note that this λ is different from the previous one in the original QFSE. Put
in order to honor with Dasgupta’s work, the author intends to keep it in here (and
also in the book). Besides, K.E. is also normalized with coefficient set to 1, which is
a usual practice in numerical derivation of Schrödinger equation as one may find out
that we can discard K.E. during the course of evaluation of different energy levels.
Once we have QFSE in the form of (5.15), we can make use of numerical solution
of quantum energy levels in (5.13) by setting m = 2 and further simplify the equation
into
 3  
E(n) E(n)
− = (K 0 (n))3 λ or
2n + 1 2n + 1
 3  
E(n) E(n)
− − (K 0 (n))3 λ = 0 (5.16)
2n + 1 2n + 1

where
 1/3
1.1924 + 33.2383n + 56.2169n 2
K 0 (n) = (5.17)
1 + 43.6196n

Note:
5.6 Numerical Computation of Quantum Anharmonic Oscillators 99

Fig. 5.5 Education corner of QFFC.org (QFFC 2019)

The numerical Eqs. (5.16) and (5.17) can be easily calculated by using MATLAB
(Chapra 2017; Marghitu and Dupac 2012) and any other computational packages,
and of course, using MQL (Young 2015) and integrated with MT (metatrader) plat-
forms—the biggest international online trading and program development platform
for real-world financial applications. Details can be found from the education corner
of quantum finance forecast center official site, QFFC.org.
In summary, once we know coefficient λ, all the quantum energy levels (or quan-
tum price levels) can be found (Fig. 5.5).
The question is: How can we find λ?

5.7 Evaluate Quantum Price Levels Using Numerical

Computation Technique

In order to find λ, we have to revisit QFSE (5.15), given by

d 2 ϕr  2 
+ r + λr 4
ϕr = Eϕr (5.15)
dr 2
As mentioned, this quantum finance Schrödinger equation (QFSE) has four char-
acteristics:
100 5 Quantum Price Levels—Basic Theory and Numerical Computation …

1. QFSE is a complex quantum anharmonic oscillator  with the combination of

harmonic term (Eϕr ) + anharmonic term (i.e., r 2 + λr 4 ϕr )—also known as
perturbation term.
2. It consists of two parts: K.E. part (the first term, kinetic energy) and P.E. part (the
second term, potential energy).
3. The quantum finance wave function ϕr . If we observe (measure) it with sufficient
samples (say over 2000 samples, e.g., 2000 trading days), it will become a typical
normal distribution with the maximum value appears at the mean, which is also
the stable (ground) state with K.E. = 0.
4. The quantum finance wave function is symmetric with the ground state as the
central axis when the samples are sufficiently large.
Using computer to solve this complex equation, we can apply numerical technique
with finite difference method (FDM).

5.8 Finite Difference Method (FDM)

With the advance of computer technology, finite difference method (aka FDM) pro-
vides an intuitive practical method to solve complex differential and high-order equa-
tions by numerical approximation and implemented by computer algorithms (Dimov
et al. 2004; Duffy 2006; Tavella and Randdall 2000).
Although the technique is using numerical approximation, when the number of
samplings/observations is sufficiently large, the finite step (x) is sufficiently small
(as compared with sampling/observation horizon x). Such method can give a very
close approximation to the “actual” analytical solution.
In fact, FDM is commonly used in many large-scale real-time problems such as
worldwide weather prediction (also known as numerical weather prediction), earth-
quake modeling, traffic control systems, and naturally worldwide financial analysis.
The finite difference method is a threefold process:
1. For any function y = Q(x), subdivide the x-axis into N subdivisions, i.e., x1 , x2 ,
x3 , … x N .
2. Set the value of y at xi as yi .
3. The finite difference formulations for all differential operators will be

d x = x (5.18a)

dy yi+1 − yi
(xi ) = (5.18b)
dx x
d2 y yi+1 + yi−1 − 2yi
(xi ) = (5.18c)
dx2 x 2
5.8 Finite Difference Method (FDM) 101

X0 = 0 X1 X2 X3 X4 X5 = 1
Fig. 5.6 Basic concept of FDM

Let’s take a look at how we can apply FDM to solve our QFSE. Figure 5.6
illustrates the basic idea of FDM.

5.9 Finite Difference Method (FDM) to Evaluate QF Wave

Function

Figure 5.7 illustrates the quantum price return wave function Q(r) (ϕr in our QFSE)
mentioned in Sect. 5.4, that is, the wave function distribution statistic of XAUUSD
by plotting the pdf of daily returns in the past 2048 trading days.
But the difference is that this time we display this pdf function by dividing the
x-axis (r) into 100 equal divisions, with each width x given by

3σ
x = (5.19)
50
where σ is the standard deviation of r for the past 2048 trading days (totally, we have
2046 r sample observations by excluding the boundary records).
This figure also shows the regression curve of the wave function for illustration
purpose.
As mentioned, when the number of observations/sampling is sufficiently large,
all four characteristics of QFSE mentioned in Sect. 5.7 hold.
Besides, certain important findings can be concluded from Fig. 5.7:
1. ϕ Max at r ∼
= 1 (ground state, denotes as r0 ).
2. ϕr = ϕ(r0 ) is symmetric with r ∼
= 1 as symmetry axis, especially when r-segment
close to the symmetry axis.
3. So, we can take the first left and right r-segment for calculation, denoted as
r−1 and r+1 , respectively.
102 5 Quantum Price Levels—Basic Theory and Numerical Computation …

Quantum Price Return Wavefunc on Q(r) of XAUUSD

(For the past 2048 trading-day sta s cs)
0.05

0.045

0.04

0.035

0.03
Q(r)

0.025

0.02

0.015

0.01

0.005

0
0.97559986
0.97657587
0.97755187
0.97852788
0.97950388
0.98047989
0.9814559
0.9824319
0.98340791
0.98438391
0.98535992
0.98633592
0.98731193
0.98828793
0.98926394
0.99023994
0.99121595
0.99219196
0.99316796
0.99414397
0.99511997
0.99609598
0.99707198
0.99804799
0.99902399
1
1.00097601
1.00195201
1.00292802
1.00390402
1.00488003
1.00585603
1.00683204
1.00780804
1.00878405
1.00976006
1.01073606
1.01171207
1.01268807
1.01366408
1.01464008
1.01561609
1.01659209
1.0175681
1.0185441
1.01952011
1.02049612
1.02147212
1.02244813
1.02342413
Daily Return (r)

Fig. 5.7 Quantum price return wave function Q(r) of XAUUSD

Figure 5.8 shows three major r-segments in the QF wave function of r; they are
r0 , r−1 , and r+1 .
It also corresponds to ground state ϕ(r0 ) and first +ve and –ve r states, ϕ(r−1 )
and ϕ(r+1 ), respectively.
According to QFSE (5.15), the wave function is symmetric with r0 as the symmetry
axis and ϕ(r0 ) is the max wave function values.
Since we have all 2046 r observations and also their distribution, i.e., ϕ(r’s),
technically we have sufficient information to evaluate λ for every financial product.
To evaluate λ using finite difference method, we have two options:
1. Using FDM to evaluate the boundary continuously of r0 and r+1 segments (i.e.,
The rightmost approximation of ϕ(r0 ) ∼
= leftmost approximate of ϕ(r+1 )).
2. Using FDM and symmetric property of ϕr , we make use of the symmetric prop-
erty of r−1 -segment and r+1 -segment to “cancel out” the K.E. term, and then
using FDM to evaluate λ using the symmetric property of their P.E. component.
Which option we could choose? Why?

5.10 Numerical Evaluation of λ| X AU U S D Using FDM

For illustration purpose, we will continue to use XAUUSD as an example to demon-

strate how to evaluate λ using FDM.
5.10 Numerical Evaluation of λ| X AU U S D Using FDM 103

Fig. 5.8 Illustration of FDM calculation in quantum finance

For the 2048 trading days of XAUDUSD (as of January 16, 2019), we have the
following statistics information in Table 5.1.
As recalled from the QFSE, we have

d 2 ϕr  2 
2
+ r + λr 4 ϕr = Eϕr (5.15)
dr
Since QFSE (5.15) is symmetric with respect to the central axis r0 , when we con-
sider quantum dynamics for r+1 and r−1 segments, their K.E. terms can be canceled
out, so we have
 2   2 
r+1 + λr+1
4
ϕr+1 = r−1 + λr−1
4
ϕr−1 or

Table 5.1 Statistic results of the quantum price return wave function Q(r) of XAUUSD (as of
January 16, 2019)
Product: XAUUSD
No. of r = 2046 r0 = 0.999604 ϕ(r0 ) = 0.047785
r = 0.000793 r+1 = 1.000396 ϕ(r+1 ) = 0.038825
Max(ϕ) = 0.047785 r−1 = 0.998811 ϕ(r−1 ) = 0.039821
Max(ϕ) N o = 50 μ = 0.999821 σ = 0.013213
Data source Forex.com MT4 system
104 5 Quantum Price Levels—Basic Theory and Numerical Computation …
 2 
 r−1 ϕr−1 − r+1
2
ϕr+1 

λ= 4 (5.20)
r+1 ϕr+1 − r−1
4
ϕr−1 

For XAUUSD, after calculation, we have λ = 1.16813758.

Table 5.2 shows λ values for all 120 forex products using MQL program (One
can visit QFFC.org education corner for detailed implementation procedure).

5.11 Numerical Computation of Quantum Energy

Levels En

Once we have λ, we can use (5.16)–(5.17) to evaluate all energy levels En .

   
E(n) 3 E(n)
− − (K 0 (n))3 λ = 0 (5.21)
2n + 1 2n + 1
 1/3
1.1924 + 33.2383n + 56.2169n 2
K 0 (n) = (5.22)
1 + 43.6196n

Note that (5.16) is a typical cubic polynomial which can be easily solved by
MATLAB using “root” command.
For XAUUSD, by using λ = 1.16813758, we can write a simple MATLAB
program (in the next section) to calculate all first 21 energy levels. Table 5.3 shows
the experimental results for the calculation of the first 21 quantum finance energy
levels (QFELs) of XAUUSD.

5.12 Numerical Computation of Quantum Finance Energy

Levels (MATLAB Version)

In this MATLAB program shown in Fig. 5.9, we use λ value 1.16813758 to calculate
the first 21 K and quantum finance energy levels (QFELs) of XAUUSD.
To solve the cubic equation, MATLAB command roots() is used to evaluate the
solution set. Since (5.21) is a typical cubic equation, normally it will have solution
set of three values, with one real number and two complex solutions.
So, we set the real number solution as QFEL value.
5.12 Numerical Computation of Quantum Finance Energy Levels (MATLAB Version) 105

Table 5.2 λ values for ALL 120 forex products using MQL program
CODE λ values CODE λ values CODE λ values
XAGUSD 1.16813758 US2000 1.01691648 GBPDKK 0.50015095
CORN 0.98147439 AUDCAD 0.99800233 GBPHKD 0.49946476
US30 1.00927814 AUDCHF 1.00650666 GBPJPY 1.02721719
AUDUSD 1.01090471 AUDCNH 0.99788161 GBPMXN 1.00743969
EURCHF 0.9922947 AUDJPY 1.01297607 GBPNOK 0.97866528
GBPCAD 0.98033867 AUDNOK 1.01297576 GBPNZD 1.01766392
NZDJPY 0.99409385 AUDNZD 0.99883417 GBPPLN 0.98982647
USDCNH 1.00129406 AUDPLN 0.99972703 GBPSEK 1.00541074
XAUAUD 0.97310053 AUDSGD 0.99145652 GBPSGD 0.99543469
XAUCHF 1.28307613 CADCHF 1.05615292 GBPUSD 0.99737283
XAUEUR 1.03339416 CADJPY 0.97655725 GBPZAR 0.99672306
XAUGBP 1.10858157 CADNOK 0.9981341 HKDJPY 1.01256568
XAUJPY 1.20798503 CADPLN 1.02762915 NOKDKK 1.00723481
XAUUSD 0.87114449 CHFHUF 0.99232627 NOKJPY 1.00878002
COPPER 0.98546677 CHFJPY 0.94512371 NOKSEK 0.99266368
PALLAD 0.97495035 CHFNOK 1.00053241 NZDCAD 1.0105619
PLAT 0.93898709 CHFPLN 1.00582673 NZDCHF 0.97178881
UK_OIL 1.01219635 CNHJPY 1.00000253 NZDUSD 1.00708451
US_OIL 1.07644811 EURAUD 1.00721165 SGDHKD 1.0028679
US_NATG 1.76511177 EURCAD 0.96576866 SGDJPY 0.95994849
HTG_OIL 0.90630263 EURCNH 1.01233192 TRYJPY 0.5018959
COTTON 1.02930805 EURCZK 0.99233097 USDCAD 1.00299693
SOYBEAN 0.50226883 EURDKK 0.99994162 USDCHF 0.96609929
SUGAR 0.99525331 EURGBP 0.99797319 USDCZK 0.99456678
WHEAT 0.99615377 EURHKD 1.00358691 USDDKK 0.99719426
IT40 1.01850019 EURHUF 0.98265533 USDHKD 1.00178794
AUS200 0.99426146 EURJPY 0.91939902 USDHUF 1.01153898
CHINAA50 0.9806911 EURMXN 1.02025986 USDILS 1.0047121
ESP35 0.93834053 EURNOK 0.99508525 USDJPY 0.50079764
ESTX50 1.00351004 EURNZD 0.50156959 USDMXN 0.99266275
FRA40 1.00704187 EURPLN 1.06863464 USDNOK 0.9984592
GER30 1.03777101 EURRON 0.99952845 USDPLN 1.01260473
HK50 0.99188819 EURRUB 0.99533066 USDRON 1.00335247
JPN225 0.9884408 EURSEK 1.03002348 USDRUB 0.98921247
N25 0.98915404 EURSGD 1.00701412 USDSEK 1.0196364
NAS100 0.99279678 EURTRY 1.01094015 USDSGD 1.00527642
(continued)
106 5 Quantum Price Levels—Basic Theory and Numerical Computation …

Table 5.2 (continued)

CODE λ values CODE λ values CODE λ values
SIGI 1.0158226 EURUSD 1.01141223 USDTHB 0.9990309
SPX500 1.00436699 EURZAR 1.04497648 USDTRY 1.02358136
SWISS20 1.00564252 GBPAUD 1.0092642 USDZAR 0.9828679
UK100 0.98794556 GBPCHF 0.99417175 ZARJPY 1.08799327
Source Experimental results using MQL program, 2019

Table 5.3 K values and QFEL values of the first 21 quantum finance energy levels
Product: XAUUSD (λ = 1.16813758)
Energy level K QFEL
0 1.060410426 1.409932766
1 1.266594551 4.744287679
2 1.491211949 8.908118719
3 1.663522514 13.59094957
4 1.806129863 18.6925368
5 1.929228428 24.15086474
6 2.038364753 29.92294434
7 2.136927359 35.97686567
8 2.227155031 42.28781818
9 2.310613024 48.8358504
10 2.388443595 55.60450183
11 2.4615088 62.57991521
12 2.530477086 69.75023292
13 2.595878459 77.10517084
14 2.658141083 84.635708
15 2.717616385 92.33385484
16 2.77459678 100.1924762
17 2.829328496 108.2051536
18 2.882021043 116.3660765
19 2.932854345 124.6699548
20 2.981984198 133.1119479
Source Experimental results using MATLAB development tool, 2019
5.13 Depressed Cubic Equation Solver—Gerolamo Cardano 107

% *****************************************************************************************************************
%
% Date: 5.1.2019
% Subject: Calculation of Quantum Price Level using QFSE
% Author: Dr. Raymond LEE
% Version; 1.0
%******************************************************************************************************************
% Define parameters
Maxj = 10000; % Max no. of iteration (approx of infinity)
MaxN = 21; % Max Quantum Price Energy Levels
l = 1.16813758; % Lambda value for the QFSE of XAUUSD
oldE = 0.0; % Aux E variable
%
% Define and initialize K and Energy Level arrays
%
E = zeros(3,MaxN); % Energy levels array
K = zeros(1,MaxN); % K parameter array

% Calculate ALL K2 parameters

for n=1:1:MaxN
K(n) = ((1.1924+33.232383*(n-1)+56.22169*(n-1)*(n-1))/(1+43.6106*(n-1)))^(1/3);
end

% Calculate ALL Energy Levels

for n=1:1:MaxN
a = (1/(2*(n-1)+1))^3;
b = 0;
c = -1/(2*(n-1)+1);
d = -1*l*(K(n)^3);
E(:,n) = roots([a,b,c,d]);
end

Fig. 5.9 MATLAB program for the numerical computation of QFEL for XAUUSD

5.13 Depressed Cubic Equation Solver—Gerolamo

Cardano

Actually, the MATLAB version is just for illustration.

In real-time situation, we will implement quantum energy level calculation using
MQL.
Why?
The answer is simple, so that we can fully automate the time series financial data
extraction process with statistic calculation of quantum finance wave function and
calculate all quantum finance energy levels in a single MT program.
The only problem is: Without the fantastic cubic solver like roots() in MATLAB,
how can we solve the cubic (5.21) in MQL?
The solution is … Cardano’s method.
Gerolamo Cardano (1501–1576) was a close friend of Leonardo da Vinci, an Ital-
ian mathematician, physician, biologist, physicist, chemist, astrologer, astronomer,
philosopher, writer, and gambler.
108 5 Quantum Price Levels—Basic Theory and Numerical Computation …

He was one of the most influential mathematicians of the Renaissance and was one
of the key figures in the foundation of probability as well as the earliest introducer
of binomial coefficients and binomial theorem in the western world.
Cardano was the first mathematician to make systematic use of negative numbers.
One of his most influential discoveries is the solution to one particular case of
the cubic equation, so-called “depressed cubic equation” (cubic equation without
quadratic term):

ax 3 + bx + c = 0 (5.23)

5.14 Cardano’s Method for Calculating QFEL

Published in his book Ars Magna in 1545, Cardano devised a genius way to solve
the depressed cubic equation as follows:
Given a depressed cubic equation:

t 3 + pt + q = 0 (5.24)

Cardano introduced variables u and v such that

u+v =t (5.25)

With additional condition,

3uv + p = 0 (5.26)

By simple substitution of (5.19) into (5.18a, 5.18b, 5.18c), we have

u 3 + v 3 + (3uv + p)(u + v) + q = 0 (5.27)

Using (5.26), we have

u 3 + v 3 = −q and (5.28a)

p3
u 3v3 = − (5.28b)
27
The tricky point is that the combination of (5.28a and 5.28b) will lead to the
following quadratic equation where u 3 and v 3 are the roots of the equation:

p3
z 2 + qz − =0 (5.29)
27
5.14 Cardano’s Method for Calculating QFEL 109

By solving (5.29), we can find u 3 and v 3 :

 
q q2 p3 q q2 p3
u =− +
3
+ , v3 = − − + (5.30)
2 4 27 2 4 27
Finally, we have
 
3 q q2 p3 3 q q2 p3
t= − + + + − − + (5.31)
2 4 27 2 4 27

At the time of Cardano, there is no concept of complex number. As we now know,

the solution of this cubic equation will normally has three roots, one real and two
symmetric complex roots. In our case, we take the first real root as solution, which
can be done easily by any programming platform, MQL naturally.

5.15 Numerical Computation of Quantum Energy Levels

(MQL Version)

By substituting (5.21) with Cardano’s formula,

 3  
E(n) E(n)
− − (K 0 (n))3 λ = 0 (5.21)
2n + 1 2n + 1

Finally, we have
 
3 q q2 p3 3 q q2 p3
E(n) = − + + + − − + (5.31)
2 4 27 2 4 27

where

p = −(2n + 1)2 (5.32a)

q = −λ(2n + 1)3 [K 0 (n)]3 and (5.32b)

 1/3
1.1924 + 33.2383n + 56.2169n 2
K 0 (n) = (5.32c)
1 + 43.6196n

Figure 5.10 shows the program section in MQL for the calculation of first 21
energy levels of a financial product (with given λ determined in the previous section).
Once all QF energy levels are calculated, the determination of the quantum price
levels will be a straightforward scaling problem.
110 5 Quantum Price Levels—Basic Theory and Numerical Computation …

//******************************************************************************************************************
// 6. Using Quantum Finance Schrődinger EQT to find the first 21 Quantum Energy Levels
//
// By solving the Quartic Anharmonic Oscillator as cubic polynomial equation of the form
// a*x^3 + b*x^2 + c*x + d = 0
//
// Using (Dasqupta et. al. 2007) QAHO solving equation:
// (E(n)/(2n+1))^3 - (E(n)/(2n+1)) - K(n)^3*L = 0
//
// and Solving the above Depressed Cubic Eqt using Cardano's Method
//******************************************************************************************************************

for (eL=0;eL<21;eL++)
{
p = -1 * pow((2*eL+1),2);
q = -1 * L * pow((2*eL+1),3) * pow(K[eL],3);

// Apply Cardano's Method to find the real root of the depressed cubic equation
u = MathPow((-0.5*q + MathSqrt(((q*q/4.0) + (p*p*p/27.0)))),p3);
v = MathPow((-0.5*q - MathSqrt(((q*q/4.0) + (p*p*p/27.0)))),p3);

// Store the QPE

QPE[eL] = u + v;

// Printout the QF Energy Levels

Print("TP",nTP+1," ",TP_Code[nTP]," Energy Level",eL," QPE = ",QPE[eL]);
FileWrite(QPLog_FileHandle," Energy Level",eL," QPE = ",QPE[eL]);
}

Fig. 5.10 Program section in MQL for the calculation of first 21 energy levels of a financial product

Figure 5.11 shows the overall algorithm for the determination of first 21 quantum
finance energy levels (QFELs) and hence the quantum price levels for 120 financial
products using MQL.
For details, please visit the programming workshop at the Education Corner of
QFFC.org.

5.16 Numerical Algorithm to Calculate QPL Using MQL

For each financial product, do the following:

• Read the daily time series and extract (Date, O, H, L, C, V)
• Calculate dally price return r(t)
• Calculate quantum price return wave function Q(r) (size 100)
• Evaluate λ value for the wave function Q(r) using FDM and Eq. (5.21) and evaluate
other related parameters:
– sigma (std dev of Q)
5.16 Numerical Algorithm to Calculate QPL Using MQL 111

Fig. 5.11 Flowchart for the

determination of the first 21
QFELs and QPLs
112 5 Quantum Price Levels—Basic Theory and Numerical Computation …

– maxQPR (max quantum price return—for normalization)

• Once λ is found, using Quantum Finance Schrodinger Equation (numerical
solution) by solving the depressed cubic equation using Cardano’s method
(5.32a–5.32c) to calculate first 21 quantum finance energy levels, QFEL(n), n
= [1 … 20]
• Calculate quantum price return, QPR(n)

p = −(2n + 1)2 (5.33)

Q F E L(n)
QPR(n) = (5.34)
Q F E L(0)

where n = [1 … 20]
• Calculate normalized QPR(n)

NQPR(n) = 1 + 0.21 ∗ sigma ∗ QPR(n) (5.35)

where n = [1 … 20]
• Save two level of data files:
– For each financial product, save the QPL table contains QPE, QPR, and NQPR
for the first 21 energy levels
– For all financial product, create a QPL summary table containing NQPR for all
FP, which will be used for financial prediction using recurrent neural networks.

5.17 QPLs for XAUUSD

Table 5.4 shows QPE, QPR, and NQPR for the first 21 energy levels of XAUUSD
by using the 2048 daily time series data from Forex.com.
According to quantum finance theory and the symmetric property of QFSE, at the
beginning of each trading day, the first 21 QPL+ is calculated by

Q P L 0 = PO pen ∗ N Q P R(0) (5.30a)

Q P L +n = PO pen ∗ N Q P R(n), n = [1 . . . 20] (5.30b)

Q P L −n = PO pen /N Q P R(n), n = [1 . . . 20] (5.30c)

In real application, every day at 08:00 HKT/UTC0, QFFC will calculate the fore-
cast H/L for worldwide 129 financial products, together with daily eight closest QPLs
for each FP, upload onto QFFC official site for public access.
5.18 Conclusion 113

Table 5.4 QPE, QPR, and

Product: XAUUSD (λ = 1.16813758)
NQPR for the first 21 energy
levels of XAUUSD by using Energy level QPE QPR NQPR
the 2048 daily time series 0 1.40993277 1 1.00277473
data from Forex.com
1 4.7443013 3.36491314 1.00933673
2 8.90806181 6.3180756 1.01753097
3 13.590797 9.63932275 1.02674654
4 18.69227098 13.25756193 1.03678619
5 24.15047183 17.12881096 1.04752787
6 29.9224128 21.22258132 1.05888698
7 35.97618549 25.51624187 1.07080074
8 42.28698048 29.99219642 1.08322032
9 48.83484717 34.63629495 1.09610645
10 55.60332572 39.43686325 1.10942674
11 62.57855942 44.38407341 1.12315392
12 69.74869114 49.4695157 1.13726466
13 77.10343711 54.68589633 1.15173872
14 84.63377674 60.0268174 1.16655835
15 92.33172074 65.48661249 1.18170782
16 100.1901342 71.06022114 1.19717309
17 108.2025989 76.74309121 1.21294154
18 116.3633045 82.53110164 1.22900172
19 124.666961 88.42050054 1.24534322
20 133.108728 94.40785487 1.26195653
Source Experimental results using MQL for implementation, 2019

5.18 Conclusion

In this chapter, we studied the basic theory of quantum price level (QPL).
More crucially, this chapter shows how to evaluate the quantum finance energy
level (and hence the QPL) by solving quantum finance Schrödinger equation (QFSE)
using simple numerical computation methods.
Prior to this book, quantum finance and path integral analysis seem to be a mathe-
matically complex and far-reaching techniques for data scientists, quants, and finan-
cial analysts to comprehend; let’s alone with the actual calculation/ program design
of QPL.
But now, quantum finance and hence QPL can be easily and practically applied
to real-world financial situation by using common PC systems.
Any quants and analysts can now use their own PC/MAC systems to program and
calculate any QPL, and QPL they want to integrate into their own trading systems.
114 5 Quantum Price Levels—Basic Theory and Numerical Computation …

In fact, quantum price level (QPL) evaluation is only the first step of quantum
finance.
More important is how to make use of it, together with other AI tools to design
and implement real-time financial prediction and intelligent trading systems.

Problems

5.1 Why experiments and experimental results are key factors in quantum theory?
And how can we do that in quantum finance? Give two examples in financial
markets in your explanation.
5.2 What is pseudoscience? What is the major difference between science and
pseudoscience? Give two examples of pseudoscience examples in finance and
explain why they are not scientific methods.
5.3 What are the three major similarities and differences between quantum energy
levels (QELs) in quantum theory versus quantum price level (QPL) in quantum
finance?
5.4 What are support and resistance levels (S & R) in technical analysis? Why
they are important in modern finance and financial trading?
5.5 Discuss and explain the similarities and differences between quantum price
levels (QPLs) and support and resistance levels (S & R). Give two examples
in financial markets in your explanation.
5.6 What is the physical meaning of wave function (ψ) in quantum finance? And
how wave function can be interpreted by double-slit experiment?
5.7 What is a λx 2m quantum anharmonic oscillator? What is the major difference
between a typical quantum anharmonic oscillator and a λx 2m quantum anhar-
monic oscillator? What is the mathematical significance of λx 2m quantum
anharmonic oscillator?
5.8 State and explain the physical meaning of each term in the mathematical
formulation of excited energy levels in a λx 2m quantum anharmonic oscillator.
5.9 State and explain the mathematical derivation of the quantum finance
Schrödinger equation (QFSE) with the integration of λx 2m quantum anhar-
monic oscillator formulation. Explain how it can be used to evaluate quantum
price levels (QPLs) in quantum finance.
5.10 How can we evaluate the wave function (ψ) of any financial product in quan-
tum finance? The following figure shows the wave function distribution of
Gold (XAUUSD) using the past 2048-trading day time series.
5.18 Conclusion 115

(i) Using MQL (tutorials can be found in QFFC.org official site) or R to plot
wave function distribution of (1) CORN; (2) Dow Jones Index (DJI) and
crude oil (US Oil);
(ii) Compare these three wave function distribution charts with XAUUSD,
what are the major similarities and difference between these four distri-
bution charts? Why?

5.11 Given the following experiment results for the wave function distribution of
Silver (XAGUSD) in 2046-trading days:

No. of r = 2046 r0 = 0.999604 ϕ(r0 ) = 0.047785

r = 0.000793 r+1 = 1.000396 ϕ(r+1 ) = 0.038825
Max(ϕ) = 0.047785 r−1 = 0.998811 ϕ(r−1 ) = 0.039821
Max(ϕ) N o = 50 μ = 0.999821 σ = 0.013213

Calculate λ term for XAGUSD.

5.12 Write the flowchart to show the steps and processes to calculate the 20 closest
quantum price levels (QPLs) of XAGUSD using MT4 system with 2000 past
trading day time series.
5.13 Based on the λ term calculated for XAGUSD in Question 5.11, and the numer-
ical formulation of QAHO given by
 3  
E(n) E(n)
− − (K 0 (n))3 λ = 0
2n + 1 2n + 1
 1/3
1.1924+33.2383n+56.2169n 2
where K 0 (n) = 1+43.6196n
116 5 Quantum Price Levels—Basic Theory and Numerical Computation …

And today opening price for XAGUSD is 14.31, calculate the 21 closest quan-
tum price levels (QPLs) for XAGUSD today.
5.14 Below given the quantum finance Schrödinger equation in QAHO format.

d 2 ϕr  2 
2
+ r + λr 4 ϕr = Eϕr
dr
(i) State and explain the physical meanings of each term in terms of quantum
finance theory;
(ii) What is importance of the anharmonic term in the QFSE?
(iii) Discuss and explain the major characteristics of QFSE in QAHO format;
(iv) What is the importance of QFSE in QAHO format in terms of (1) model-
ing of quantum financial markets; (2) computational and implementation
perspective for the evaluation of QPL?

5.15 What is finite difference method (FDM)? How it works?

5.16 State and explain the basic assumptions and formulations of finite difference
method.
5.17 Discuss and explain how to apply finite difference method (FDM) for the
evaluation of quantum finance wave function.
5.18 Based on the wave function distributions evaluated in 5.10 for CORN, DJI,
and US Oil:
(i) Choose the best dr to evaluate their λ values.
(ii) How to choose the best dr interval? Why? and What is the importance?
5.19 What is a depressed cubic equation? Discuss and explain Cardano’s technique
for solving depressed cubic equation.
5.20 By solving the depressed cubic equation to evaluate QFEL, normally three
roots will be evaluated with two of them are imaginary roots.
(i) What is the physical meaning of these three roots of QFEL in terms of
quantum theory and quantum finance?
(ii) Is it possible to have three roots are all real numbers? Why? and what is
the physical meaning in terms of quantum finance?
(iii) Which root we should take as the solution of QFEL? Why?
5.21 MQL programming assignment.
Study the MQL tutorial from QFFC.org official site. Write an MQL program
to
(i) Read the previous 2000-trading day time series of CORN, DJI, and US
Oil.
(ii) Plot their wave function distribution charts.
(iii) Evaluate λ values of their wave function distribution.
(iv) Implement Cardano’s technique to evaluate QFELs and QPRs of these
products.
5.18 Conclusion 117

(v) Using the opening price of these financial products, modify your MQL
to calculate the first 41 quantum price levels (i.e., QPL−20 , QPL−19 , …,
QPL0 , …, QPL+19 , QPL+20 ).
(vi) State and explain how you can test the accuracy and performance of
using these QPLs to replace support and resistance levels (S & R) for
financial analysis.
(vii) Modify your MQL to implement (vi) for the performance evaluation
of these QPLs for CORN, DJI, and US Oil and compare their overall
performance.
(viii) Perform the same procedure as (vii) to write an MQL program to cal-
culate and evaluate the performance of QPLs for ALL forex products
in your MT4 platform. Check for any patterns and characteristics you
can find out and explain why.

References

Bouard, A. & Hausenblas, E. (2019) The nonlinear Schrödinger equation driven by jump processes.
Journal of Mathematical Analysis and Applications 475 (1): 215–252.
Carbonnière, P. et al. (2010) The VCI-P code: an iterative variation–perturbation scheme for effi-
cient computations of anharmonic vibrational levels and IR intensities of polyatomic molecules.
Theoretical Chemistry Accounts 125(3): 543–554.
Chapra, S. C. (2017) Applied Numerical Methods with MATLAB for Engineers and Scientists.
McGraw-Hill Education, 4th edition.
Dasgupta, A. et al. (2007) Simple systematics in the energy eigenvalues of quantum anharmonic
oscillators. Journal of Physics A: Mathematical and Theoretical 40(4): 773–784.
Dimov, I. et al. (2014) Finite difference methods, theory and applications. Lecture Notes in Computer
Science book series (LNCS, volume 9045), Springer.
Duffy, D. J. (2006) Finite Difference Methods in Financial Engineering: A Partial Differential
Equation Approach. Wiley.
Grosse, H. and Martin, A. (2005) Particle Physics and the Schrödinger Equation (Cambridge Mono-
graphs on Particle Physics, Nuclear Physics and Cosmology). Cambridge University Press.
Kisil, V. V. (2012) Hypercomplex Representations of the Heisenberg Group and Mechanic. Inter-
national Journal of Theoretical Physics 51(3): 964–984.
Marghitu, D. B. and Dupac, M. (2012) Advanced Dynamics: Analytical and Numerical Calculations
with MATLA. Springer.
Muller-Kirsten, H. J. W. (2012) Introduction to Quantum Mechanics: Schrödinger Equation and
Path Integral. World Scientific.
Murphy, J. J. (1999) Technical Analysis of the Financial Markets: A Comprehensive Guide to
Trading Methods and Applications. New York Institute of Finance.
Popelier, P. (2011) Solving the schrodinger equation: has everything been tried? World Scientific
Pub Co Pte.
QFFC (2019) Official site of Quantum Finance Forecast Center. http://qffc.org. Accessed 21 Aug
2019.
Rampho, G. J. (2017) The Schrödinger equation on a Lagrange mesh. Journal of Physics: Confer-
ence Series 905: 12037.
Rohwedder, T. (2013) The continuous Coupled Cluster formulation for the electronic Schrödinger
equation. ESAIM: Mathematical Modelling and Numerical Analysis 47(2): 421–447.
118 5 Quantum Price Levels—Basic Theory and Numerical Computation …

Schmitz, W. (2019) Particles, Fields and Forces: A Conceptual Guide to Quantum Field Theory
and the Standard Model (The Frontiers Collection). Springer.
Tavella, D. & Randall, C. (2000) Pricing financial instruments: the finite difference method. John
Wiley & Sons, New York.
Witwit, M. R. M. (1996) Energy Levels for Nonsymmetric Double-Well Potentials in Several
Dimensions: Hill Determinant Approach. Journal of Computational Physics, 123(2): 369–378.
Young, A. R. (2015) Expert Advisor Programming for MetaTrader 4: Creating automated trading
systems in the MQL4 language. Edgehill Publishing.
Zee, A. (2011) Quantum Field Theory in a Nutshell. Princeton University Press, 2nd edition.
Chapter 6
Quantum Trading and Hedging Strategy

Self-discipline is what separates the winners and the losers.

Thomas Peterffy (born at 1944)

What are financial trading and hedging strategies?

What is quantum trading?
What is trading discipline?
Anyone working in program trading may be familiar with Mr. Thomas Peterffy. Mr.
Peterffy was born in 1944, a Hungarian-born American billionaire businessman. He
was the founder, chairman, CEO, and the largest shareholder of Interactive Brokers.
He emigrated to United States and worked as an architectural draftsman and later
became a computer programmer. In 1977, he purchased a seat at the American Stock
Exchange and played a role in developing the first electronic trading platform for
securities in US. He stated in his inspiring quotation a very important fact in financial
investment—self-discipline. The truth is: if a trader has neither self-discipline nor
self-control, one may incur vast losses in the long run.
As the author always mentioned in investment talks and seminars, humans are
species of habit and arrogance. Subsequent profits were yielded from financial mar-
ket, we inclined to be arrogant and persist in investing without any self-discipline
and self-reflection, until one day it was lost in all respects. It is similar to an old
saying in Wall Street: “Failure is the common factor of successful investment. The
only one who can truly win (profit) from the market as a long run is the one that has
self-discipline and know when to cut loss and stop.”
How can we maintain self-discipline in financial trading?
Although there are many ways to achieve this, the most effective method is scien-
tific trading techniques and strategies. The logic behind is: If we perform trading or
hedging by our own subjective judgements, we won’t incessantly stop loss when the
market goes to the opposite direction. As regards to behavior psychology, we always
have the belief and see what we believe. For instance, if we believe the stock market
is bearish and do a short sell by our own judgement (or feeling) without any scientific

© Springer Nature Singapore Pte Ltd. 2020 119

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_6
120 6 Quantum Trading and Hedging Strategy

backup and evidence, even though the market goes to the opposite direction, we tend
to always find evidence either from market charts or financial news to support our
judgement that “the market will go down very soon, just wait for a couple of hours”,
until we have already lost a fair amount.
The merits of scientific trading strategy and hence program trading is that we
won’t let our subjective opinions affect our trading activities. If the market goes to
the opposite direction, we simply follow the objective trading strategy to stop loss
or trigger the hedging algorithm.
In this chapter, quantum finance forecast center (QFFC.org) provides a more
preferable solution—daily/weekly financial forecast for worldwide 120+ financial
products. Every day at 00:00UTC (08:00 HKT), QFFC will compute the daily
high/low of worldwide 120+ financial product including worldwide forex, major
commodity, international financial indices, major cryptocurrencies, and broadcast via
QFFC official site QFFC.org. Worldwide investors and traders can base on these fore-
cast results, which together with quantum trading and hedging strategies is revealed
in this chapter to perform their quantum trading.
This chapter is organized as follows: First, we begin with basic concept of finan-
cial trading and hedging strategies, with an overview of latest R&D on AI-based
trading and hedging strategies. Second, we introduce seven major classical trading
the hedging strategies, which is the author’s collection of over 20 years of experi-
ences on financial analysis and trading in various financial markets. In fact, they are
also the building blocks of quantum trading methodology. Third, we explore basic
concepts and techniques on quantum trading—the main theme of this chapter. In the
conclusion section, we study investment attitude and the importance of objectivity
in trading and investment.

6.1 Financial Trading and Hedging Strategies in a Nutshell

In classical finance, trading strategy refers to the design and implementation of

buying/selling algorithm (or composition of these actions) to gain profit in financial
market.
Conventional trading strategies are based on fundamental or technical analysis
(Murphy 1999) or both. They are usually verified by backward testing, where the
process should follow scientific method, and by forward testing performed in a sim-
ulated trading environment.
Technical trading strategies are relied on technical indicators to generate trading
signals. For example, a simple trading strategy may be a moving average (MA)
crossover whereby a short-term moving average crosses above or below a long-term
moving average (Kaufman 2013).
Hedging strategy (or hedging in short), on the other hand, refers to the design of
composite trading algorithm to provide risk control and/or possible profit in adverse
market movements.
6.1 Financial Trading and Hedging Strategies in a Nutshell 121

Fig. 6.1 Intelligent trading and market analysis (Tuchong 2019)

Owing to the requirement of complex financial analysis and computations

in 70s–80s, hedging was solely performed by major fund houses and financial
institutions.
But thanks to the advance of computer technology, there is an improvement on
personal computer and workstations along with the popularity of online trading plat-
forms; nowadays, hedging becomes a possible and useful tool even for independent
investors and traders.
In fact, modern finance includes hedging being one of the major strategies for
the design and implementation of trading strategy, and it becomes a standard trading
tool for the design of high-level trading algorithms and strategies (Leshik and Cralle
2011; Nelken 1999; Rhoads 2011) (Fig. 6.1).

6.2 Traditional Trading and Hedging Strategies

Traditional finance involved trading and hedging strategies in 70s–80s (i.e., before the
age of online and program trading) on investment funds, bonds, and stocks which was
mainly performed by major fund houses and financial institution in major financial
markets.
Conventional trading and hedging strategies include the following:
1. Single-product trading/hedging strategy—long buy/short sell of stocks, or com-
position of these actions with their options counterparts.
2. Mutual product trading/hedging strategy—also known as pair-trading—is the
long buy/short sell of one particular stock (e.g., Coca-Cola) with the counter-
direction buy/sell or a mutual stock (highly correlated stock with different
expected returns/variance returns) as safeguards and risk controls (e.g., Pepsi
in this example).
122 6 Quantum Trading and Hedging Strategy

Fig. 6.2 Program trading

3. Trading of a particular stock together with risk control by counteraction on co-

related financial index, e.g., the trading of GM Inc. with hedging on DJIA.
4. Multi-market trading/hedging strategy (portfolio trading) become favored by
major fund houses from 80s due to the widespread of portfolio analysis, portfolio
trading which involves buy/sell composition of a batch of financial products
(called portfolios) to maximize returns on one hand, and minimize risks—return
variances—on the other hand.
Also, the surface of major investment funds, mutual funds products, and services
began in 80s (or commonly called wealth investment) in financial community-enabled
worldwide investment. The importance of financial values in portfolio investment
resulted in a mass of R&D conducted in major universities and financial institutions
in the past decades. Typical examples include R&D on chaos and fractals on finance,
as well as quantum finance (Fig. 6.2).

6.3 Latest R&D on Trading and Hedging Strategies

Over the years, financial engineering ranging from financial signals, chart patterns
study for the modeling of financial prediction and trading systems is one of the most
stimulating topics for both academia and financial community. Not only because of
its utmost importance in terms of financial and commercial values, but it vitally poses
a real challenge to worldwide researchers and quantitative analysts (a.k.a. quants)
throughout the world.
The flourishing of AI technology in the past 30 years such as various hybrid
intelligent financial trading and hedging systems with the integration of various AI
and related technology was proposed. Latest R&D on intelligent trading and hedging
system include the following:
1. Artificial neural network (ANN) was proposed by Li et al. (2017) for the devel-
opment of quantitative trading strategy.
6.3 Latest R&D on Trading and Hedging Strategies 123

2. AI clustering technique was proposed by Hsu et al. (2009) for the design of
profit refiner on futures trading strategy.
3. Genetic algorithm (GA) technology was proposed by Allen and Karjalainen
(1999) for the determination of technical trading rules and Shiau (2009) using
multi-objective GA for the optimization of reservoir hedging rules.
4. Principle component analysis (PCA) technology was proposed by Byun et al.
(2015) for multi-currencies trading in the forex market.
5. Fuzzy logic system (FLS) was proposed by Naranjo et al. (2015) for the devel-
opment of fuzzy capital management and trading system; Vella and Ng (2015)
used dynamic fuzzy money management approach for controlling the intraday
risk-adjusted performance of AI trading algorithms.
6. Data mining system was proposed by Pauna (2019) for time price series for
algorithmic trading systems.
7. Decision support system (DSS) was proposed by Schmidt et al. (2010) for
experimental analysis of online trading algorithms.
8. Reinforcement-learning system (RLS) was proposed by Yang et al. (2018) for
the design of investor sentiment reward-based trading systems.
9. Stochastic technique was proposed by Helseth (2016) for the design of linearized
progressive hedging algorithm on secondary financial markets.
10. Hybrid AI system was proposed by Chang and Lee (2017) with the integration
of Markov decision process (MDP) and GA to formulate trading strategies for
stock markets; Evan et al. (2013) with the integration of ANN and GA developed
algo-trading model for intraday foreign exchange speculation; Liu et al. (2017)
with the integration of fuzzy logic rules and GA developed a method to quantify
moving average strategy of crude oil future markets; Zhang and Maringer (2016)
integrated recurrent reinforcement learning (RRL) with GA for equity trading.

6.4 Technical Analysis and Program Trading

Although both fundamental and technical analyses are the most popular and com-
monly used in financial analysis, the design of trading algorithm, technical analytical
tools—indicators such as RSI, MA, and MACD are commonly used for program
trading because they are
1. Easy to implement and program for program trading;
2. Fully integrated to time series financial data which are now openly provided
by many online trading platforms (e.g., MT platform—the largest forex online
trading platform in the world with over hundreds of financial service plat-
forms/operators);
3. Easy to understand and learn, only basic financial and statistical knowledge is
required to understand technical analysis;
124 6 Quantum Trading and Hedging Strategy

Fig. 6.3 Program trading using technical indicators on MT4 trading platform

4. Easy to program and visualize on the trading platform. Unlike traditional stock
market in 60s–70s, only financial analysts and brokers can access real-time fig-
ures (patterns) of financial markets. Nowadays, every investor and trader with
basic computing (programming) knowledge can not only visualize the financial
markets, but can also design and program their own trading and hedging algo-
rithms into their investment portfolio. The additional knowledge they require are
methods and techniques such as financial forecasts, AI-based trading algorithms,
etc. which is also the aim of this book (Fig. 6.3).

6.5 Technical Analytical Tools and Indicators

Technical analysis is a trading discipline employed to evaluate investments and iden-

tify trading opportunities by analyzing statistical trends gathered from trading activ-
ity, such as price movement and volume.
Technical analysts (TA) believe that all knowledge regardless of type (fundamen-
tal, economic, political, psychological, or other) is already reflected and discounted
in market prices.
Unlike fundamental analysts, they believe that it is meaningless to study company
financial statements, earnings and dividend reports, industry developments, and other
data in an attempt to determine the intrinsic value of a stock or other market instru-
ment since it is common to have a wide divergence between intrinsic value and actual
market price.
For example, it is not unusual for a company’s stock to be trading at a price well
above or below its book value per share.
6.5 Technical Analytical Tools and Indicators 125

Price movements are simply the reflection of changes in supply and demand
without concern of their underlying forces. Technical analysts are only interested in
what happens to the price according to supply and demand. Say, if demand is greater
than supply, price will increase. On the other hand, if supply is greater than demand,
price will decline. The study of market prices is all that is necessary.
The components of technical analysis are as follows:
• Major market patterns (e.g., major reversal patterns),
• Trend lines and channels,
• Supports and resistance (S & R) analysis, and
• Technical indicators and oscillators (e.g., MA, RSI, MACD, stochastics, Bollinger
Bands).
Nowadays, the design of trading and hedging algorithms (and hence programs)
mainly makes use of technical indicators, oscillators, supports, and resistance lines
for easy implementation, since most of them have already built-in functions bundled
with the program trading development platform, such as MQL trading indicator
function library in MT platform.

6.6 Seven Major Financial Trading and Hedging

Algorithms

In previous chapters, we have studied basic concepts and theory of quantum finance,
especially on how to apply quantum finance theory to model and evaluate quantum
price (energy) levels (QPLs) which is an important tool in technical analysis.
Before we move on to learn how to design and implement AI-based quantum
financial forecast and trading system, let’s study a practical financial investment
topic—trading and hedging strategy.
First, we will review seven major trading and hedging techniques—which is the
author’s collection of over 20 years of trading experience in stock, commodity, and
forex trading, along with the numerous technical trading courses conducted in China
and Hong Kong.
Traders study various kinds of trading techniques and strategies throughout the
years which are summarized into following seven types:
1. Trend trading strategy,
2. Breakout trading strategy,
3. Reversal trading strategy,
4. Channel trading strategy,
5. Averaging trading strategy,
6. Stop-loss trading strategy, and
7. Hedge trading strategy.
126 6 Quantum Trading and Hedging Strategy

Fig. 6.4 Snapshot of program trading using RSI oscillator

Note that these seven trading strategies despite are independent in nature, and most
of the time they can be combined to use in order to produce the best performance.
Figure 6.4 shows the snapshot of program trading using RSI oscillator.

6.6.1 M1.1 Trend Trading Strategy—Key Patterns

Trend trading is also known as forward trading. The main concept is to trade following
the market trend, if exists. Trend trading is a trading strategy that attempts to capture
gains through the analysis of a security’s momentum in a particular direction. Trend
traders enter into a long position when a security is trending upward and/or enter a
short position when a security is trending downward.
Trend trading strategy assumes that a financial instrument will continue to move
along its current trend and often contain a take-profit or stop-loss provision if there
are any signs of a reversal. It can be used by short-, intermediate-, or long-term
trading. Regardless of their chosen timeframe, traders will remain in their position
until they believe the trend has reversed, although reversals may occur at different
times for each timeframe.
In technical analysis, we have studied all major patterns and trends identification
techniques.
Trend trading strategy #1 using major key patterns include the following:
• Nine major reversal patterns such as head-and-shoulder patterns, round-top/bottom
patterns, ascending/descending triangles, rectangles, etc.;
6.6 Seven Major Financial Trading and Hedging Algorithms 127

Fig. 6.5 M1.1 Trend trading strategy—key patterns

• Trend lines and channels;

• Consolidation patterns;
• Gaps; and
• Supports and resistances.
Once a trend is clearly identified, we have to execute the sell/buy command accord-
ing to the trend direction.
Let’s have a look at the up/down trend lines as shown in Fig. 6.5.
Note that the buy and sell should execute
• When the trend is well developed, i.e., at least hits two low/high.
• Execute at least 10% above (uptrend)/below (downtrend) price.

6.6.2 M1.2 Trend Trading Strategy—MA Crossing

In technical analysis, most of the time financial patterns are moving in trends, either
bullish or bearish (or oscillation in some cases).
In other words, once we have identified the current pattern is bullish/bearish, we
can trigger the buy/sell command according to the trend.
Typical example is moving average (MA) trend identification.
The decision of using MA crossing is simple:
1. If day-line down-crossing MA line is bearish, trigger sell signal;
2. If day-line up-crossing the MA line is bullish, trigger buy signal.
Again, one should place the selling signal at the price at least 10% below the
down-cross point as a kind of buffer.
Same cases apply to other technical indicators and oscillators.
Figure 6.6 shows a typical example of the price (S&P 500) bar chart down-cross
the 200-day MA line. It shows the fast signal line (i.e., the day-line) down-crossing
128 6 Quantum Trading and Hedging Strategy

Fig. 6.6 M1.2 Trend trading strategy—MA crossing

the slow 200-day simple moving average (SMA)—means bearish, trigger the sell
signal.
As an exercise, based on what we have learnt here, try to figure out other
bullish/bearish trends from any market chart and determine the buy/sell thresholds.

6.6.3 M1.3 Trend Trading Strategy—Signal Lines Crossing

Instead of single technical indicator up-/down-crossing, many TAs are using multiple
technical signal crossing method, in which one short-term indicator acts as fast-
moving signal and the other long-term indicator acts as slow-moving signal.
The decision of using signal crossing is simple:
1. If a fast signal line down-crossing a slow signal line—bearish, trigger sell signal
and
2. If a fast signal line up-crossing a slow signal line—bullish, trigger buy signal.
Figure 6.7 shows two MA signal lines: 5-day MA (fast-moving) versus 20-day
MA (slow-moving) lines. When 5-day fast-moving MA line down-crosses a 20-day
slow-moving MA line from above, a bearish trend appears.
6.6 Seven Major Financial Trading and Hedging Algorithms 129

Fig. 6.7 M1.3 Trend trading strategy—signal lines crossing

Again, can place a sell command 10% below the down-cross point.
One may wonder: For MA signal lines crossing, what are the two MA lines being
used?
Actually, there is no definitive answer. But many TAs prefer to use the Fibonacci
number sequence (i.e., 1, 1, 2, 3, 5, 8, 13, 21, 34, 55, 89, 144…) to select the MA lines.
For example, the signal crossing of MA-5 versus MA-21 (or MA-20 for simplicity)
is used to simulate the comparison between weekly (short-term) trend with monthly
(mid-term) trend.

6.6.4 M2.1 Breakout Trading Strategy—Key Patterns

Breakout trading strategy, similar to trend trading strategy is to trade by following

the trend.
But different from trend trading strategy, it will only execute when the price has
reached the breakout point.
In technical analysis, we have studied certain patterns such as flag consolidation
pattern with well-defined breakout price as shown in Fig. 6.8.
130 6 Quantum Trading and Hedging Strategy

Fig. 6.8 M2.1 Breakout trading strategy—key patterns (e.g., flag consolidation)

It shows a typical flag consolidation pattern in technical analysis with well-defined

channel lines. Once the pattern is confirmed, one can trigger a buy order while the
market breaks from the breakout price.
Similar to trend trading strategy, the breakout price must be at least 10% away
from the breakout price.
This strategy can also be used once we have clearly identified the sup-
port/resistance lines. In other words, when the price of an asset moves above a
resistance area or moves below a support area. Breakouts indicate the potential for
the price to start trending in breakout direction.
For example, a breakout to the upside from a chart pattern could indicate the price
will start trending higher. Breakouts that occur on high volume (relative to normal
volume) show greater conviction which means the price is more likely to trend in
that direction.

6.6.5 M2.2 Breakout Trading Strategy—S & R Lines

Figure 6.9 shows two typical cases of breakouts from support and resistance lines.
Note that, once we identify the support/resistance lines, we can execute the buy/sell
command when the price breakout from breakout thresholds of, say, 10% buffer.
In fact, breakout trading using supports and resistance lines is commonly exercised
on stock and commodity trading using the day chart to identify support and resistance
lines by applying standard technical analysis techniques. However, it may be difficult
6.6 Seven Major Financial Trading and Hedging Algorithms 131

Fig. 6.9 M2.2 Breakout

trading strategy—S & R
lines

to identify at forex and cryptocurrency trading as they are more chaotic, especially
during US trading time zone, i.e., US time 09:00–16:00.

6.6.6 M3.1 Reversal Trading Strategy—Key Reversal Patterns

Reversal trading as its name is executed at least at one of these three scenarios:
1. Clearly identify the key reversal pattern(s);
2. Clearly identify effective support/resistance; and
3. Reversal occurs in over-buy/over-sell regions of technical indicator such as RSI.
Naturally, reversal trading strategy is a perfect tool once we have identified the
eight key reversal patterns by technical analysis, which include key reversals, head-
and-shoulders tops and bottoms, rounding tops and bottoms (saucers), ascending and
descending triangles, rectangles, double and triple tops and bottoms, diamond, rising
and falling wedges, V-shape formations (spikes).
132 6 Quantum Trading and Hedging Strategy

Fig. 6.10 M3.1 Reversal trading strategy—key reversal patterns

Figure 6.10 shows a typical head-and-shoulder top pattern.

Once the pattern is clearly identified, we can place a sell command 10% below
the return move of the head-and-shoulder pattern.
As an exercise, try to figure out the position of sell/buy signal point for all the key
reversal patterns using technical analysis.

6.6.7 M3.2 Reversal Trading Strategy—S & R Lines

Reversal trading strategy using support and resistance lines (S & R lines) is exactly
the counterpart of breakout trading strategy #2—S & R lines in the sense that the S
& R lines continue as an effective price level to support or resist from price breakout.
In many worldwide financial products trading such as forex and commodity trad-
ing, this situation usually occurs during Asia market period, i.e., 09:00 HKT–16:00
HKT in which the financial markets are oscillating within the daily S & R lines.
Figure 6.11 shows a typical reversal trading based on S & R Lines.
Once the S & R lines are identified, buy/sell command can be triggered after third
reversal from the S & R. (Note: at least two times of reversals are required for S/R
confirmation).
Again, a 10% threshold is required to dispatch signal.
6.6 Seven Major Financial Trading and Hedging Algorithms 133

Fig. 6.11 M3.2 Reversal

trading strategy #2—S & R
lines

6.6.8 M3.3 Reversal Trading Strategy—Technical Indicator

RSI

Relative strength index (RSI) is a price momentum indicator developed by Mr. J.

Welles Wilder in 1978. Since its introduction of Wilder’s New Concepts in Technical
Trading Systems book published in 1978, it has been widely used by worldwide
technical analysts, particularly those who are commodities and future oriented.
From the technical analysis perspective, relative strength index (RSI) is a very
reliable technical indicator to indicate the strength of supply versus demand in sec-
ondary financial market. RSI compares bullish and bearish price momentum plotted
against the graph of an asset’s price.
Traditional interpretation and usage of RSI are that reading values of 70 or above
indicate that a security is becoming overbought or overvalued and may be primed
134 6 Quantum Trading and Hedging Strategy

Fig. 6.12 M3.3 Reversal trading strategy—technical indicator RSI

for a trend reversal or corrective pullback in price. RSI reading values of 30 or below
indicate an oversold or undervalued condition.
The fact is: most TAs believe that 70–30 is sufficient to reflect the over-buy/over-
sell in the market.
Figure 6.12 shows the daily bar chart of Dow Jones Index, DJI (top) versus 14-day
RSI (bottom).
It is apparent in using 70–30 as selling/buying threshold is an acceptable choice.
For experienced traders, they usually trigger selling signal only when RSI rise to
above 70, reach the max RSI, and return back to 70 RSI line. It is called second-
time-hit-top.
Similar to RSI 30 case, it is called second-time-touch-base.

6.6.9 M4.1 Channel Trading Strategy—Trend Channel

Trading

A trend channel (or channel in short) is a channel drawn on a security price series
chart by graphing two trend lines drawn at resistance and support levels. Trading
channels can be drawn using a variety of methodologies. Generally, traders believe
that security prices will remain within a trading channel. Therefore, traders use
trading channels to develop buy-and-sell trading signals.
In many instances, prices repeatedly move about the same distance away from
a trend line before returning to the trend line. In these cases, a straight line can be
drawn connecting the peaks of rallies in an uptrend or the bottoms of declines in
a downtrend. That line is often parallel to the trend line and is called a return or
channel line. Between the channel line and trend line creating a trend channel is a
range within which prices move.
6.6 Seven Major Financial Trading and Hedging Algorithms 135

Fig. 6.13 M4.1 Channel trading strategy—channel patterns

Well-defined trend channels appear most frequently in charts of actively traded

securities. Thinly traded securities offer little opportunity for trend channels to
develop. Inexperienced TAs often use trend channels to determine good profit-taking
levels. For example, in an uptrend, they will sell a stock when it reaches the upper
level of its trend channel. Experienced TAs watch price movements within the two
boundary lines of the trend channel looking for a warning signal that the trend direc-
tion is changing.
Channel trading strategy can be applied when one of the following occurs:
• Clearly identify the channel pattern and S/R lines.
• Clearly identify the price oscillator between S/R lines pair.
• Clearly identify the Bollinger Band.
Figure 6.13 shows typical cases for channel trading using uptrend and downtrend
channels.
Again use at least 10% as threshold and trigger the signal when the price oscillates
at least two times to confirm the effectiveness of the channel.

6.6.10 M4.2 Channel Trading Strategy—S & R Channel

Trading

As mentioned in previous section, prices oscillating between supports and resis-

tance lines (S & R lines) are commonly found during Asia trading period, i.e., HKT
09:00–16:00 in which most international forex prices “oscillating” between daily
S-R lines, especially when there is no significant market news and financial figures
release.
Figure 6.14 shows the day-trade chart for AUDCAD in MT platform
136 6 Quantum Trading and Hedging Strategy

Fig. 6.14 M4.2 Channel trading strategy—S & R channel trading

of Forex.com.
We can clearly identify an effective S-R line pair to trigger two buy/sell signals.
However, at the third time of oscillation, a breakout occurs in which the previous
support line now becomes the resistance (i.e., support → resistance). In this case, a
sell signal can be triggered when the price arrived at the breakthrough 10% threshold
of the breakout line.
The fact is: in traditional technical analysis, support and resistance lines are com-
pletely two different kinds of trading patterns. In other words, a support cannot turn
into resistance lines (or vice versa). However, due to the popularity of program and
algorithm trading, support and resistance lines mentioned in previous chapter are
now considered as some sorts of discrete energy levels, similar to quantized energy
levels in an atom. In other words, a support line for a financial product at a particular
moment can become a resistance line at another moment due to the change of market
stimulus.

6.6.11 M4.3 Channel Trading Strategy—Bollinger Bands

Bollinger Bands technique (BB in short) was developed by Mr. John Bollinger in
1980s for the purpose of factoring in price volatility. Bollinger Bands are effective
for virtually any security or market and for any investment time horizon. Rather
than placing bands at a certain percentage distance from a moving average line as is
done with trading bands, Bollinger Bands are placed two moving standard deviations
above and below a simple moving average line.
6.6 Seven Major Financial Trading and Hedging Algorithms 137

Fig. 6.15 M4.3 Channel trading strategy—Bollinger bands

More specifically, the bands are placed above and below a simple moving average
at a distance of 1.5–2.0 times the root mean square of the deviations (σ) from the
average. The amount of data used in the calculation is equal to the number of periods
used for the simple moving average.
For example, if we use a 20-day simple moving average, all calculations should
be based on 20 days of data. Similar to moving average, Bollinger Bands are so
important and commonly used that almost all financial trading software/systems are
bundled with the BB tools.
In fact, Bollinger Bands are very reliable with effective technical signal and trading
tool, especially in day trade of forex products which are usually oscillating between
upper–lower Bollinger Bands (so-called Bollinger channel).
Figure 6.15 is the same H1 chart of AUDCAD with period-20 Bollinger Band.
Note that it provides least three chances (maybe more) to trigger buy/sell signals.
But remind that such channel always breakthrough during Europe–US trading period.
It is wise to set stop-loss as risk control.

6.6.12 M5 Averaging Trading Strategy

Averaging method (or averaging in short) is a very important and commonly used
trading strategy for experienced traders and hedge fund traders.
The logic is very simple: instead of putting 100% of investment at a single price
(either buy or sell), we divide the investment into at least two (or more, usually three)
batches, each of equal amount (or different proportion, according to the trader’s
preference).
138 6 Quantum Trading and Hedging Strategy

Fig. 6.16 M5 averaging trading strategy

So even when the price moves in reverse direction, the overall execution price
will be the average of two (or three) triggering prices.
The merits of averaging trading are threefolds:
1. Provide risk sharing especially during adverse market situation;
2. Provide more chances for investors and traders to enter market at one hand, and
locate a good averaging position even the market swings sideway; and
3. Provide a good trading strategy with the integration of S & R lines or/and quantum
price levels (we will discuss that in detail in the upcoming section—quantum
trading strategy).
Figure 6.16 shows a realistic case of forex trading chart of AUDCAD using aver-
aging trading strategy. As shown in Fig. 6.15, if we use S1 and S2 as support lines,
we can execute 50% buy at T1-S2 price. When price reaches S1 at T2, at this “better
position”, we trigger the other 50% buy. Once it reaches S3, we close the order. In
that case, the average gain will be S3 − (S2 + S1)/2, which is much better than solely
investing 100% at S2-T1. Same method can be applied and through a combination
to any kind of trading strategy.
In fact, professional traders and investors usually combine averaging trading strat-
egy with other trading and hedging methods to provide a more flexible, dynamic
trading and risk control strategy.
6.6 Seven Major Financial Trading and Hedging Algorithms 139

6.6.13 M6 Stop-Loss Trading Strategy

Stop-loss strategy seems to be very simple, but it is the most important trading
strategy especially for derivatives trading and margin trading products such as forex
and cryptocurrency trading.
Usually, stop-loss strategy is combined with target profit in which the ratio is
normally 1:2. For example, if the target profit is 240 pts, the stop-loss should be
around 120–150 pts.
How to set stop-loss?
Traditionally, there are two methods:
1. Fixed stop-loss according to the investor’s risk control preference. It varies for
different products. For forex trading, normally we set 150–200 points. It can also
be the proportion of the average daily range (i.e., range between day-high and
day-low) and
2. Dynamic stop-loss according to some technical indices such as S & R lines.
In the next section, we will see quantum trading to provide a third solution for
stop-loss strategy using quantum price levels (QPLs).
Using previous example on AUDCAD trading in Fig. 6.15, if we trigger a buy
order at time T1 0.96251, we set the stop loss of 125 pts (i.e., 0.96126). So even
though price finally drops 464 pts to S1 0.95787 at time T2, the loss is controlled to
125 pts only. Figure 6.17 shows another example based on the trading of Dow Jones
Index (DJI) in MT platform. As shown, DJI breakouts at T1-S2 level (23180) and
drop 1728 points to S1 (21452) before it finally rebound. So, if we set the stop loss
with 200 points, we can control the potential loss to an acceptable level.

Fig. 6.17 M6 Stop-loss trading strategy

140 6 Quantum Trading and Hedging Strategy

As an old saying in Wall Street: The best and most successful trader is not the one
that can always win, but the one that has self-discipline to follow stop-loss strategy
in every trading battle.

6.6.14 M7.1 Hedge Trading Strategy—Single-Market

Single-Product (SM-SP) Hedge

Hedge trading (or hedging in short) is the second kind of risk control-based trading
strategy. A hedge is an investment position intended to offset potential losses or
gains that may be incurred by a companion investment. A hedge can be constructed
from many types of financial instruments including stocks, forex, insurance, forward
contracts, swaps, and options.
Technically speaking, hedging is the practice of taking a position in one market
to offset and balance against the risk adopted by assuming a position in a contrary
or opposing market or investment.
It is a common tool for experienced traders as it involves more technical skill to
so-called release the hedge.
Hedge trading basically is of three types as given below:
1. single-product hedging,
2. multi-product hedging, and
3. cross-trading-platform hedging.
First, let’s have a look at a basic case: single-market single-product (SM-SP)
hedge.
In a single-product trading, when we execute a buy/sell order at a certain price P,
we execute the same amount of reverse order at a price “P” at the stop-loss level to
control loss when the market goes in opposite direction.
Same example of AUDCAD trading is used for illustration in Fig. 6.18. Instead
of using stop loss, we set a sell hedge at the stop-loss level below S2. When price
drops to S1 and rebound, we harvest the hedge order and release the hedge.

6.6.15 M7.2 Hedge Trading Strategy—Single-Platform

Multi-product (SP-MP) Hedge

With the advance in computer technology and online trading, real-time calculation
of co-relation and co-variance between different financial products under the same
trading platform, e.g., forex, is no more a reverie.
Single-platform multi-product (financial products) hedging strategy is a kind of
composite hedging strategy with market momentum evaluation of two (or more)
similar products in a financial platform. It is commonly used in forex platform which
contains over 100+ forex products to study.
6.6 Seven Major Financial Trading and Hedging Algorithms 141

Fig. 6.18 M7.1 Hedge trading strategy—single-market single-product (SM-SP) hedge

SP-MP hedge ranging from simple two products hedge (i.e., traditional paired-
hedge) to multi-product hedge with a portfolio of products (so-called portfolio-hedge)
with the help of AI tools such as fuzzy logics, neural networks, and genetic algorithms
for portfolio weighting factor “ω” evaluation (we will study all major AI tools in the
next chapter).
Figure 6.19 shows a typical five-product cyclic hedge (5PC-hedge) for the cyclic
hedge of top five foreign currencies: AUD, CAD, EUR, GBP, and USD. Try to figure
out how it works and the logic behind.
Note that most often such kind of high-level and composite hedging strategy
can only be valid and effective within a period of time and restricted to number of

Fig. 6.19 Five-product

cyclic hedge (5PCH) in forex
AUD

CAD USD

EUR GBP
142 6 Quantum Trading and Hedging Strategy

financial products. That’s why all major fund houses and laboratories recruited top
mathematicians and AI scientists to develop complex trading and hedging algorithms.
In the same way that the author always mentioned in investment talks; nowadays,
the design and implementation of trading and hedging algorithms resemble more to
mathematical and AI problems than classical financial problems.

6.6.16 M7.3 Hedge Trading Strategy—Multi-platform

Multi-product (MP-MP) Hedge

Considering the complexity of world financial market, one of the most important and
complex hedging systems is the multi-platform multi-product (MP-MP) hedge which
involves multiple-financial products across different financial trading platforms.
The logic behind came from the fundamental concept of financial theory that all
financial markets are the currency flow, e.g., USD behind different major worldwide
financial markets. So, in the perfect case of hedging for risk control, the best solution
should be the hedging across different products and platforms to protect adverse
market change of a particular product/financial market.
Traditional wisdom tells us that USD currently remains as major clearance cur-
rency for three types of worldwide major financial and commodity markets which
include (a) US Index, (b) crude oil, and (c) precious and colored metals gold (or
silver) along with agricultural products such as cotton and sugar that they are of
great consideration as shown in Fig. 6.20.

Fig. 6.20 MP-MP hedge of

world three major markets
(gold, crude oil, and US
index)
6.6 Seven Major Financial Trading and Hedging Algorithms 143

Many fund houses and financial institutions have active research for the mining of
such golden combination using various AI techniques such as fuzzy-neuro networks,
hybrid EC (evolutionary computing) systems, deep neural networks, etc. Active R&D
on this challenging topic is being carried out at worldwide fund houses, financial
institutions, and universities (Iba and Aranha 2012).

6.7 Quantum Trading and Hedging Strategy

6.7.1 Basic Concept

In previous chapters, we have learnt that basic concepts of quantum finance stem
from two different aspects:
• Path integral approach and
• Quantum anharmonic oscillator approach.
Path integral approach is based on Feynman’s path integral by the modeling and
integration of all possible paths of the quantum financial particles. Major applications
are exercised on the modeling and primary financial markets evaluation such as
forward interest rate and option pricing.
Quantum anharmonic oscillator (QAHO) approach is based on the modeling and
numerical approximation of the quantum finance Schrödinger equation (QFSE) into
quantum anharmonic oscillation—a quartic polynomial equation that can be solved
numerically by simple numerical computational model. Major applications are exer-
cised on the evaluation of quantum finance energy levels (so-called quantum price
(energy) levels, QPL) for any financial products in the secondary financial markets
(Fig. 6.21).

Fig. 6.21 Illustration of

quantum finance energy
fields in financial markets
144 6 Quantum Trading and Hedging Strategy

Which one should we employ to design quantum trading and hedging strategy?
and why?

6.7.2 Quantum Price Levels Versus S & R Analysis

The answer is clearly to employ quantum anharmonic oscillator (QAHO) approach

since all abovementioned trading and hedging strategies fall within the domain of
secondary financial markets ranging from stocks and options to online forex and
cryptocurrency trading activities.
In addition, one of the major breakthroughs in quantum finance using QAHO
model is the quantum price (energy) level (QPL) evaluation, which forms a direct
analog to supports and resistances (S & R) in technical analysis.
If we consider the financial world is a quantum finance energy field formed by
financial particles, it is a natural logic to presume that these inherited QPL are exactly
the same concept of support and resistance we defined in classical technical analysis.
But there is only one difference. If quantum dynamics of the financial particles is
strictly followed from quantum theory, there should be no difference between support
and resistance levels. They should all be the same as discrete energy levels that the
financial particles can jump from one to another (in both directions). In fact, this
logic is entirely equivalent to nowadays assumption on support and resistance lines
in program trading, which we generally assume that support and resistance lines are
simply quantized price levels in price motions.
In other words, once we adopt quantum finance method to evaluate all QPLs
for any financial product automatically, we literally evaluate all the possible S & R
levels, which can be applied in turn to the seven trading and hedging strategies learnt
previously (Fig. 6.22).

6.7.3 Quantum Finance Forecast Center—Daily Forecast

and QPLs

Since December 1, 2017 with patented quantum finance technology, quantum finance
forecast center (QFFC.org) provides daily (and weekly) free financial forecasts for
worldwide 120+ financial products which include the following:
• 9 major cryptocurrencies,
• 84 forex,
• 19 major commodities, and
• 17 worldwide financial indices.
Up to December 31, 2018, QFFC has over 10000+ registered users which include
worldwide professional traders and investors.
Figures 6.23 and 6.24 show snapshots of daily forecast results of BTCUSD (Bit-
6.7 Quantum Trading and Hedging Strategy 145

Fig. 6.22 Support and resistance levels in financial analysis

Fig. 6.23 Quantum finance forecast of BTCUSD on April 15, 2019

146 6 Quantum Trading and Hedging Strategy

Fig. 6.24 Quantum finance forecast of USDCAD on April 15, 2019

coin versus USD) and forex pair USDCAD on April 15, 2019. At 08:00 HKT (UTC
00:00), QFFC calculates the following for 120+ worldwide financial products:
• Forecast high/low and
• Eight closest QPLs.
Professional traders and investors can base on these forecasts with QPLs and
integrate with their program trading strategies for intelligent trading.
QFFC members can also make use of QFFC’s API to integrate the daily fore-
casts/QPLs into their MT4 EA programs for intelligent trading.
In the next section, we will study some basic quantum trading and hedging tech-
niques/strategies.
6.7 Quantum Trading and Hedging Strategy 147

6.7.4 Basic Quantum Trading and Hedging Strategies

In comparison with technical and fundamental analysis used in the past 50 years, the
major breakthroughs of quantum finance technology are the scientific computations
of
• Daily (weekly) high/low for day-trade and mid-term trading, with
• QPLs to replace subjective evaluation of support and resistance lines that can be
used for breakout analysis and mid-term/long-term trading/market trend analysis.
The following quantum trading and hedging techniques/strategies are the theoret-
ical characteristics collection of these quantum financial instruments (i.e., forecast
high/low and QPLs) and feedbacks from professional traders and investors using
QFFC forecast services in the past 2 years.
As a friendly advice, for any new trading and hedging, please study carefully and
thoroughly, together with sufficient simulations and test run (at least 100 trading days
as industrial norm) before putting into actual practice.
Different from the major seven trading and hedging strategies, at quantum finance
trading, since we already have scientific daily forecasts and QPLs, the quantum
trading (includes hedging) techniques are categorized in terms of users’ trading
experiences, which is also the way we have conducted investment seminars in the
financial community nowadays.

6.7.5 QT#1.1 Inexperienced Investors—Quantum Forecast

Oscillation Trading Strategy (QFOTS)

Just like driving, financial trading is a kind of practical activity that not only require
professional knowledge but also practical experience. Any experienced investor with
over decades of trading experience must realize that trading (and especially “fail-
ure”) experiences are very important elements for every investor and trade. Espe-
cially nowadays program trading is accompanied by monitoring and handling trading
strategies of multiple products in multiple-financial markets.
As an industrial norm, we set 3–5 years (around 1000 transactions) to define
whether an investor is experienced or inexperienced.
For inexperienced investors who don’t have sufficient skills on high-level trading
techniques such as breakout trading, hedging, and release-hedge skills, they are
suggested to perform basic quantum forecast oscillation trading strategy (QT#1-
QFOTS).
QT#1-QFOT are based on daily forecasts, together with daily calculated QPLs as
SL (stop loss) and TP (target profit) to perform day trade.
The logic behind is: for any financial market, over 80% of time are undergoing so-
called oscillation between the day-high and day-low. Since we already have evaluated
the forecast high/low, we can simply use them as approximation of the day-high/low
148 6 Quantum Trading and Hedging Strategy

Fig. 6.25 Quantum finance forecast of XAUUSD on April 15, 2019

for reversal trading. That is, set the buy limit at the forecast low and sell limit at the
forecast high.
Also, for SL and TP, we can simply use daily QPLs as reference levels as example
shown in Fig. 6.25 for XAUUSD on April 15, 2019.

6.7.6 QT#1.2 Inexperienced Investors—Quantum Forecast

Oscillation Trading with Averaging Strategy (QFOTAS)

As we have learnt from the beginning of this chapter, averaging is a simple and good
technique not only as risk control but also for investment diversification.
Naturally, we can adopt averaging technique into our quantum forecast oscillation
trading strategy.
Just like working S/R lines as averaging levels, we can now employ QPLs for
averaging levels, which, in fact, is entirely inherent to quantum theory and quantum
field concepts also learnt in previous chapters.
6.7 Quantum Trading and Hedging Strategy 149

Fig. 6.26 Quantum finance forecast of UK oil on April 15, 2019

Use UK oil forecast on April 15, 2019 as example shown in Fig. 6.26. If we set first
buy limit at the forecast low 7149, we can set the second buy limit at the lower QPL,
i.e., 7128, and the third one at 7069, each with one-third of the day-trade investment.
Of course, remember to set SL, say, at the fourth QPL 6997.
By doing so, we can, on one hand, self-guard investment with a reasonable level
of oscillations for the financial product, and on the other hand, provide a reasonable
investment diversification with reasonable level of stop-loss mechanism.
In fact, not only inexperienced investors, many professional traders usually per-
form this strategy when they identify a particular financial market has a very high
chance of oscillation that always occurs during Asia trading period.
150 6 Quantum Trading and Hedging Strategy

6.7.7 QT#2.1 Experienced Traders—Quantum Forecast

Breakout Trading Strategy (QFBTS)

Among the seven major trading and hedging techniques, breakout trading is an
important trading skill for experienced traders and investors.
In actual fact, the success in identifying financial market whether it is an oscillating
market or high chance of market breakout defines whether one is an experienced or
inexperienced trader.
According to quantum finance theory, if a financial market undergoes severe
excitations, either by some financial news outbreaks or some important financial
figures/indices (e.g., GDP, PMI, etc.) announcement, the financial quantum particle
should jump from one QPL to another one, in a discrete manner. In fact, the author
customarily observed this phenomenon while working as chief analyst in a fund
house and conducting R&D on quantum finance.
The truth is: The more excited the stimulus is, the more accurate and precise is the
quantum jump. That’s why breakout trading is so important in financial investment.
Use UK100 index forecast on April 15, 2019 as example shown in Fig. 6.27. For a
day-low breakout, we can use the lower three QPLs (or more) for breakout strategy,
i.e., 7395.2, 7364.6, and 7330.5. Similar case for day-high breakout.
One important reminder, according to quantum finance theory, the range between
day-high and day-low is the ranging of normal daily oscillation of the financial
market. We should determine the market as breakout if the market jump/fall beyond
the oscillating region.

6.7.8 QT#2.2 Experienced Traders—Quantum Forecast

Breakout Trading with Averaging Strategy (QFBTAS)

The integration of averaging technique and breakout strategy is a direct enhancement

of quantum finance breakout strategy. That is, naturally, the integration of averaging
technique at every QPL breakouts, such as 30019, 29871 for Forecast Low breakouts
and 30482, 30691 for Forecast High breakouts, is shown in Fig. 6.28 for HSI.
It is the usual (in fact standard) technique being adopted by almost every experi-
enced trader. Even if we can identify it is a breakout market, it is impossible for us
to predict how much it can breakout.
Thus, averaging is the most viable solution and is also a smart move.
One important reminder is that: as we all know, once the market breakout, it can
be very serious.
So, (1) it is important to set (and calculate) as many QPLs as possible and (2)
Don’t try to do reverse trading during a severe market breakout. It is a very dangerous
act to do so.
6.7 Quantum Trading and Hedging Strategy 151

Fig. 6.27 Quantum finance forecast of UK100 index on April 15, 2019

The truth is: once market breakout, anything can occur. So, for those use breakout
strategy, all we need to do (and should) is to choose a good position to harvest the
gain. That’s all. Nothing else (traditional wisdom).
In fact, this also differentiates between an experienced versus inexperienced
trader. For an experienced trader, one will know that statistically speaking it is very
difficult to harvest a good breakout gain and followed by a reversal gain. So, one will
harvest the big gain and take a rest. But for inexperienced investor, one will make a
usual human mistake, hubris—thinking one is invincible and continue with reverse
trading until a huge loss had incurred.
152 6 Quantum Trading and Hedging Strategy

Fig. 6.28 Quantum finance forecast of HSI (Heng Seng Index) on April 15, 2019

6.7.9 QT#3.1 Experienced Traders—Quantum Forecast

Single-Product Hedging Strategy (QFSPHS)

As mentioned, hedging is one of the most important trading techniques in modern

finance. It does not only function as risk control but is also a vital technique in
high-level trading activity. It is an important component in quantum trading.
According to quantum finance theory and actual practice, 80% of time the market
undergoes oscillation mode between the forecast high–low region. However, what
we are interested in is actually the other 20% breakout cases that are usually trig-
gered by excited events such as financial news outbreaks or unexpected financial
figures announcement, which will push/pull the market from one QPL to another
QPL (normally with more than two to three QPL jumps/drops).
So, experienced traders can perform quantum forecast single-product hedging
strategy (QFSPHS) like this:
6.7 Quantum Trading and Hedging Strategy 153

Fig. 6.29 Quantum finance forecast of EURCAD on April 15, 2019

• To capture the 80% oscillation market, set buy limit and sell limit at the forecast
low and forecast high, respectively, such as 1.57314 and 1.59218, respectively, as
shown in Fig. 6.29.
• To cater for the 20% breakout scenario, do the SP hedging, i.e., set sell at each
QPL beyond the forecast low, combine with averaging and hedging techniques for
breakout market. Similar for buy-side breakout.
Depends on the trading skills and/or actual market movement, professional traders
can have some variations of QFSPHS. For example, one can release the hedge using
QPL(s), i.e., do a hedge at each breakout QPL, and partial cut-loss (release the hedge)
simultaneously. That is to conform with quantum finance theory and is a wise move
to partial release the hedge.
By using the same technique, one can naturally design and implement more com-
plex, cleverness multi-product and multi-platform quantum hedging strategies.
As an old saying, anything is possible. It all depends on our imagination.
154 6 Quantum Trading and Hedging Strategy

6.8 Conclusion

In this chapter, we have studied seven major trading strategies which can be applied
to financial products and trading platform.
Also, we have studied how to apply quantum finance to design quantum trading
and hedging strategy.

Note that these methods can be applied effectively only if one has sufficient
knowledge and actual experience of technical analysis and quantum finance.
In addition, trading activities can be (and should be) carried out by a combination
of these techniques and strategies to yield the best performance.
A final reminder: risk control strategy is the most important strategy that we
must always bear in mind. The fact is: The biggest failure is always the one at most
confident and lack of risk awareness.

As an old saying in Wall Street:

There is no genius in trading.
The truth is: the common factor for all “successful traders” are “Failures”.
Only though numerous Trials-and-Errors one can learnt and perfect one’s
trading skills.
So, be patient. A successful trading strategy is the one that can be tested
against TIME.
6.8 Conclusion 155

Problems

6.1 Why self-discipline is important to financial trading as mentioned in the open-

ing quotation of this chapter?
6.2 How to maintain self-discipline in financial trading? State three techniques or
methods and explain how they work.
6.3 Discuss and explain why program trading (or algorithmic trading) becomes so
popular nowadays.
6.4 What are the major similarities and differences between cryptocurrency trading
versus traditional forex trading?
6.5 Why conventional trading strategies are mainly based on technical analysis
tools and techniques? Give any three technical analysis tools and techniques
to support your explanation.
6.6 What is the definition of hedging strategy? Describe briefly how it works in a
financial market such as forex market.
6.7 What are the major difficulties and challenges to design a good trading/hedging
strategy? And how can we overcome them by using quantum finance technol-
ogy?
6.8 State and explain four commonly used technical analysis tools and indicators
for the design of trading and hedging strategies.
6.9 State and explain seven major types of trading and hedging techniques. For
each type of technique, use a financial product and its corresponding financial
charts as example for explanation.
6.10 Trend trading is the most commonly used trading method, why? What is/are
the major limitation(s)?
6.11 Trend trading using multiple MA signal line crossing technique:
(i) What is Fibonacci number sequence?
(ii) How it works for the design of trading strategy using multiple MA signal
lines crossing techniques?
(iii) If we want to use three MA signal lines crossing technique, what is the
best combination of MA number from the Fibonacci to represent: daily,
weekly, and monthly pattern crossing? Why?
(iv) What are major advantages and shortcomings using Fibonacci number
sequence for the selection of MA signal lines?
6.12 Discuss and explain breakout trading strategy. Why it is commonly used by
experienced traders and hedgers?
6.13 What are the major challenges in breakout trading strategy and how quantum
finance can be helped to solve these problems?
6.14 How can we relate and interpret market breakouts in terms of quantum finance
theory? And how we can use QPL to design a scientific breakout strategy?
6.15 Discuss and explain what is reversal trading in finance? And how quantum
finance technology can be used for reversal trading?
156 6 Quantum Trading and Hedging Strategy

6.16 Discuss and explain why portfolio trading becomes a popular strategy for the
design of trading and hedging strategy. How it works?
6.17 Discuss and explain any three latest R&D of trading and hedging strategies
using AI and related technology.
6.18 Why AI technologies such as fuzzy logic and genetic algorithms are commonly
used in the design of intelligent trading and hedging systems?
6.19 What are the major limitations of using technical analytical techniques and
tools for the design of trading strategies? and how quantum finance provides
an alternative solution?
6.20 In the old days, financial trading and hedging are totally two different concepts
and methodologies, but now the distinction between them becomes obscure.
Why? Is it related to the popularity of program trading?
6.21 What are Bollinger Bands? Discuss and explain how to use Bollinger Bands
to design trading strategy. What are the major limitations and how quantum
finance can be applied to resolve these problems?
6.22 What is stop-loss strategy? Why it is important especially for the design of
program trading strategies?
6.23 State and discuss three major hedging strategies. Give an example of financial
product for explanation.
6.24 Discuss and explain the major concepts and approach of quantum trading.
Which one is more preferable for the application to program trading? Why?
6.25 Discuss and explain the major similarities and differences between QPL and
S & R in technical analysis. How we can use QPL to replace S & R for the
design of program trading and hedging algorithms?
6.26 State and explain three levels of quantum finance trading and hedging strate-
gies. Why they are divided into three levels? Each of them uses a financial
product as example to explain how it works.
6.27 For inexperienced investors, it is preferable to use daily forecasts together with
simple stop-loss strategy instead of hedging technique. Why?
6.28 Quantum trading program.

(i) Study MQL workshop from QFFC.org to learn the basic skill of design
and writing MQL-based trading programs.
(ii) Finish workshop #1 to learn how to design and implement an MQL
program to calculate all the QPR and QPL of any financial product
based on the past (says) 2000-trading day time series.
(iii) Finish workshop #2 to learn how to design and implement quantum
finance forecast system to calculate the daily forecast high/low of any
financial product.
(iv) Based on the level 1 quantum trading strategies to implement: (1)
QT#1.1 quantum forecast oscillation trading strategy (QFOTS) and (2)
QT#1.2 quantum forecast oscillation trading with averaging strategy
(QFOTAS) using QPL for at least five forex products (e.g., AUDCAD,
EURJPY, USDCAD, EURGBP, etc.).
6.8 Conclusion 157

(v) First, use MT4 simulator to test for the trading performances of these
five products for at least 1 year.
(vi) Second, use demo account to perform the mock trading for at least 20
trading days (i.e., 1 month) and compare their trading performance and
overall monthly returns.
(vii) Once you are experienced in quantum trading, implement the level 2
and level 3 quantum trading strategies and compare their trading per-
formances.

References

Allen, F. and Karjalainen, R. (1999) Using genetic algorithms to find technical trading rules. Journal
of Financial Economics. 51(2): 245–271.
Byun, H. W. et al. (2015) Using a principal component analysis for multi-currencies-trading in the
foreign exchange market. Intelligent Data Analysis. 19(3): 683–697.
Chang, Y. and Lee, M. (2017) Incorporating Markov decision process on genetic algorithms to
formulate trading strategies for stock markets. Applied Soft Computing. 52: 1143–1153.
Evans, C. et al. (2013) Utilizing artificial neural networks and genetic algorithms to build an algo-
trading model for intra-day foreign exchange speculation. Mathematical and Computer Mod-
elling. 58(5–6): 1249–1266.
Helseth, A. (2016) Stochastic network constrained hydro-thermal scheduling using a linearized
progressive hedging algorithm. Energy Systems. 7(4): 585–600.
Hsu, Y. et al. (2009) Profit refiner of futures trading using clustering algorithm. Expert Systems With
Applications. 36(3): 6192–6198.
Iba, H. and Aranha, C. C. (2012) Practical Applications of Evolutionary Computation to Financial
Engineering: Robust Techniques for Forecasting, Trading and Hedging (Adaptation, Learning,
and Optimization). Springer.
Kaufman, P. J. (2013) Trading Systems and Methods. Springer.
Leshik, E. and Cralle, J. (2011) An introduction to algorithms for stock trading on the NASDAQ
and New York Stock Exchange, Wiley-Blackwell.
Li, Z. et al. (2017) Research on Quantitative Trading Strategy Based on Neural Network Algorithm
and Fisher Linear Discriminant. International Journal of Economics and Finance. 9(2): 133.
Liu, X. et al. (2017) Quantified moving average strategy of crude oil futures market based on fuzzy
logic rules and genetic algorithms. Physica A: Statistical Mechanics and its Applications. 482:
444–457.
Murphy, J. J. (1999) Technical Analysis of the Financial Markets: A Comprehensive Guide to
Trading Methods and Applications. New York Institute of Finance.
Naranjo, R. et al. (2015) An Intelligent Trading System with Fuzzy Rules and Fuzzy Capital
Management. International Journal of Intelligent Systems. 30(8): 963–983.
Nelken, I. (1999) Pricing, Hedging, and Trading Exotic Options: Understand the Intricacies of
Exotic Options and How to Use Them to Maximum Advantage. McGraw-Hill.
Pauna, C. (2019) Data Mining Methods on Time Price Series for Algorithmic Trading Systems.
Informatica Economica. 23(1): 26–36.
Rhoads, R. (2011) Trading VIX Derivatives: Trading and Hedging Strategies Using VIX Futures,
Options, and Exchange-Traded Notes. Wiley.
Schmidt, G. et al. (2010) Experimental Analysis of an Online Trading Algorithm. Electronic Notes
in Discrete Mathematics. 36: 519–526.
158 6 Quantum Trading and Hedging Strategy

Shiau, J. (2009) Optimization of Reservoir Hedging Rules Using Multiobjective Genetic Algorithm.
Journal of Water Resources Planning and Management. 135(5): 355–363.
Tuchong (2019) Intelligent trading and market analysis. http://stock.tuchong.com/image?imageId=
533733773505462352. Accessed 21 Aug 2019.
Vella, V. and Ng, W. L. (2015) A Dynamic Fuzzy Money Management Approach for Control-
ling the Intraday Risk-Adjusted Performance of AI Trading Algorithms. Intelligent Systems in
Accounting, Finance and Management. 22(2): 153.
Yang, S. Y. et al. (2018) An investor sentiment reward-based trading system using Gaussian inverse
reinforcement learning algorithm. Expert Systems with Applications. 114: 388–401.
Zhang, J. and Maringer, D. (2016) Using a Genetic Algorithm to Improve Recurrent Reinforcement
Learning for Equity Trading. Computational Economics. 47(4): 551–567.
Chapter 7
AI Powerful Tools in Quantum Finance

A computer would deserve to be called intelligent if it could

deceive a human into believing that it was human.
Alan Turing (1912–1954)

What is intelligence?
What is artificial intelligence?
What are the core AI technologies?
How can we apply AI to quantum finance?
All who have heard about artificial intelligence (AI) must be familiar with Sir Alan
Turing—the father of AI with his famous Turing test. He was an English mathe-
matician, computer scientist, cryptanalyst, and theoretical biologist. He was highly
influential in the development of theoretical computer science, provided a formal-
ization on algorithm concepts and computation with his famous invention—Turing
machine was deemed as original computer prototype. His famous quotation brought
out the foundation concept and AI notion—machine or software to mimic and exem-
plification of human behaviors, which is also the core concept of Turing test.
In previous chapters, we have learnt the basic concepts of quantum finance.
More importantly, we have learnt how to make use of time series returns, together
with quantum anharmonic oscillator model of quantum finance Schrödinger equation
to evaluate quantum price levels (QPLs) on any financial market, which provide us
a new way and concrete solution to model financial markets.
We have also learnt how to apply QPLs, together with the seven major financial
trading (hedging) techniques to design and implement quantum trading and hedging
systems. The only component left behind is: how to make use of time series financial
information, together with QPLs to forecast financial markets using state-of-the-art
AI techniques?

© Portions of this chapter are reprinted from Lee (2006), with permission of Springer Nature.

© Springer Nature Singapore Pte Ltd. 2020 159

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_7
160 7 AI Powerful Tools in Quantum Finance

In this chapter, we will study the three pillars of AI:

• Artificial neural networks,
• Fuzzy logics, and
• Genetic algorithms.
In Chap. 8, we will study other two important computational intelligence (CI)
techniques in finance: chaos theory and fractals, how they can be used in quantum
finance. Prior to that, let’s review a brief history of artificial intelligence.

7.1 An Overview of Artificial Intelligence

7.1.1 From Greek Mythology to Turing Test

Many textbooks on artificial intelligence named Dartmouth conference in 1956 as

the Birth of AI.
Was it?
The answer is yes and no.
“Yes” in the sense that five AI founders: Professors Allen Newell (CMU), Herbert
Simon (CMU), John McCarthy (MIT), Marvin Minsky (MIT), and Arthur Samuel
(IBM) coined the word of artificial intelligence (aka AI) at this conference. Since
then, R&D on AI had commenced officially as a newly emerged discipline. They also
became leaders of AI research in different AI domains for the following decades.
“No” in the sense that: the notion of man-made intelligence, or what we called
artificial intelligence (AI) is not a new concept.
According to Greek mythology (Greek epic poet Hesiod’s Theogony 700B.C.),
Prometheus (Hansen 2005), a titan and culture hero who created first human from
clay (“humanoid-robot” in today’s term) and defied Zeus to steal fire (intelligence)
from the chariot of the sun for further progress and comfort of human civilization.
If we review the formal conceptualization of AI theory, one may think of Sir Alan
Turing and his celebrated invention—Turing test.
In 1950, he published Computing Machinery and Intelligence (Turing 1950). In
his paper, he designed a method to evaluate whether a machine can be characterized
as human and pondered the way to measure technology progression. To study the
basic problem: Can machines think? He initiated a series of tests that were meant
to help human answer the question—Turing test (at that time he called this test as
imitation game).
The test comprised of an interrogation room conducted by a judge. Figure 7.1
illustrates that there are two unseen subjects in individual terminals being judged:
a human and a computer program. Based on the quality of conversation, the judge
makes dialogue to ensure both subjects within a designated timeframe and attempts
to identify persona between human and computer program. He concluded that if the
7.1 An Overview of Artificial Intelligence 161

Fig. 7.1 Illustration of Turing test

judge is likely to pick out either subject, it showed that the computer has succeeded
in revealing to think with human intelligence.
Modern version of the Turing test has employed more than one human judge
to converse and interrogate with both subjects. The test is marked success if more
than 30% of the judges contemplated the computer program as human subsequent to
five minutes’ conversation. The Loebner Prize is an annual Turing test competition
that was started by American inventor and activist Dr. Hugh Loebner (1942–2016)
in 1991. He created additional rules to the original Turing test, calling for human
and computer program to have 25-min conversation with each of the four judges
presiding over the competition. The winner of the competition (not necessarily one
that passes the Turing test) is the one whose computer bot received the most votes
and highest ranking from the judges, regardless of whether there were 30% judges’
votes. Despite Sir Alan Turing predicted that by year 2000 machines would eventually
pass the Turing test, this did not occur until 2014. In 2014, Professor Kevin Warwick
organized a Turing test competition to mark the 60th anniversary of Sir Alan Turing’s
death. A computer chatbot called Eugene Goostman with the persona of a 13-year-
old boy had passed the Turing test in this competition. He secured 33% judges’ votes
who were convinced that he was human.

7.1.2 Definition of AI

Although there is no official definition of AI, one refers to definition of AI from Web-
ster’s Dictionary and the American Association for Artificial Intelligence (AAAI)
as
162 7 AI Powerful Tools in Quantum Finance

Fig. 7.2 Illustration of AI (Tuchong 2019a)

The capacity of computers or programs to operate in ways to mimic human thought processes,
such as reasoning and learning.
—Webster’s New College Dictionary
The scientific understanding of the mechanisms underlying thought and intelligent behavior
and their embodiment in machines.
—AAAI

Professors Stuart Russell and Peter Norvig in their remarkable book: Artificial
Intelligence—A Modern Approach (2015) provided a substantial survey on AI defi-
nition in four major categories:
• Systems that think like human;
• Systems that think rationally;
• Systems that act like human; and
• Systems that act rationally.
In author’s book—Fuzzy-Neuro Approach to Agent Applications published by
Springer in 2006 (Lee 2006) defined AI as (Fig. 7.2)
The exemplification of human intellectual thoughts, acts and behaviors for the design and
implementation of intelligent systems, software objects (agents) and robotic systems.

7.1.3 Strong Versus Weak AI

Since Dartmouth conference, many AI scientists at that time predicted that a machine
with human intelligent would exist in no more than a generation and they were given
7.1 An Overview of Artificial Intelligence 163

millions of dollars to make this vision come true. However, it became apparent that
the difficulties of constructing human-like robot were underestimated.
In 1973, in response to criticism of prolonged progress, shortages of practical and
promising AI applications, both U.S. and British Governments terminated funding
undirected research on artificial intelligence subsequent to winter of AI. Seven years
later, a visionary initiative by the Japanese Government inspired governments and
industry to provide billions of dollars on AI research, but by the late 80s investors
became disillusioned by insufficient computer capacity (hardware) and withdrew
funding again.
It was not until the first decades of twenty-first century, investment and interest in
AI revived, while machine learning was successfully applied to many problems in
academia and industry due to the presence of powerful computer hardware. It was
called the second golden age of AI.
One special feature of AI is the conformity with AI. There were debates after
Dartmouth conference even ongoing at present, on AI general definition and classi-
fication.
Professor Robert Wilensky (1951–2013), an influential AI scholar mentioned
in his work—Planning and Understanding: A Computational Approach to Human
Reasoning stated that:
Artificial Intelligence is a field renowned for its lack of consensus on fundamental issues.

One typical example was the argument between strong (hard) AI versus weak
(soft) AI. Some AI scientists believed that AI should focus on the design and sys-
tems/programs implementation that mimic and simulate how human think and act—
the so-called weak AI. Opponents believed that AI systems should not only think
and act like human, but rather should think and behave consciously like human—
so-called strong AI.
In fact, Turing test is an exemplification example of strong AI.

7.1.4 Strong (Hard) AI

Strong AI refers to the study of machine intelligence that could successfully perform
any intellectual task as per human. It is also called artificial general intelligence
(AGI).
Key characteristics of strong AI include the ability to reason, solve puzzles, make
judgments, plan, learn, and communicate. It should have consciousness, objective
thoughts, and self-awareness.
Some AI scientists argue that a machine with strong AI should be able to go
through the same human development process, starting with a childlike mind through
learning in developing an adult mind. It should be able to interact with the world,
learn from it, acquire its own common sense, and language. Another argument is that
we will not know when we have developed strong AI (if it can indeed be developed)
because there is no consensus on what constitutes intelligence.
164 7 AI Powerful Tools in Quantum Finance

Core areas of strong AI research include the following:

• Construct intelligent machines, robotic in particular;
• R&D on generalized rule-based and case-based systems, e.g., IBM Deep Blue
supercomputer on chess playing; and
• R&D on consciousness, objective thoughts with self-awareness.
Owing to its complexity, the progress of strong AI was prolonged between
1950–1980s.
The celebrate AI versus human contests—Deep Blue versus Mr. Garry
Kasparov—six-game chess matches between world chess champion and an IBM
supercomputer Deep Blue. The first match was played in Philadelphia and won by
Kasparov in 1996. The second was played in New York City and won by Deep Blue
in 1997. The 1997 match was the first defeat of a reigning world chess champion by
a computer under tournament condition.

7.1.5 Weak (Soft) AI

The emergence of computer technology such as software engineering is the software

design (programs) that can be executed by computer (systems) to work out problems.
In this respect, we have a new area on AI R&D which is not focused on general intel-
ligence implementation, but rather mimic human behaviors, especially on thinking
and problem-solving—so-called weak (soft) artificial intelligence.
Weak AI simulates human cognition and benefits mankind such as data analysis
and automates time-consuming functions in ways that human is unable to do on
occasions. In order to surpass this limitation, mimic human intelligence grew expo-
nentially in the last 20 years. In fact, weak AI helps big data transform into usable
information by detecting patterns and making predictions—data mining. Examples
include Facebook’s news feed, Amazon’s suggested purchases, Apple’s Siri, and
iPhone’s technology that answer users’ verbal enquiries.
Major R&D on soft AI include the following:
• Artificial neural networks (ANNs)—using computer models to mimic human brain
(neural) structure on thinking and problem-solving.
• Genetic algorithms (GAs)—using computer models to mimic human genetic evo-
lution, also called evolutionary computing (EC) or evolutionary programming
(EP).
• Fuzzy logic (or fuzzy logic system, FLS)—using computer models to mimic
human imprecise (so-called fuzzy) determination of matters and events, typical
applications on real-time control systems.
As we can see, the boundary between hard and soft AI becomes obscure. Modern
AI nowadays entails integrated disciplines including cognitive science, neuroscience,
biological and artificial neural networks, evolutionary computing technology, robotic
(engineering), data mining and deep learning, active vision, natural language pro-
cessing (NLP), etc. (Fig. 7.3).
7.1 An Overview of Artificial Intelligence 165

Fig. 7.3 Weak AI (Tuchong 2019b)

7.1.6 Classification of Artificial Intelligence

Owing to the advance of computing technology and the diversification of AI appli-

cation in different areas, the author had defined AI classification into two main
categories: macroscopic versus microscopic AI (Fig. 7.4).
Macroscopic AI
• Also called behavioral approach (BA) AI refers to the acts and behaviors.
• Focus on the design and implementation of intelligent machines (systems) that
mimic high-level (appears to be seen) human intellectual behavior.
• BA-AI is further sub-divided into two sub-categories: mind-BA—high-level men-
tal activities and body-BA—high-level body and control activities.
Microscopic AI
• Also called cognitive approach (CA) AI refers to the thoughts and mental pro-
cesses.
• Focus to mimic human low-level mental activities, referred by cognitive scientists
as human cognitive mental activities.
• According to various types of mental process, CA-AI approach is sub-divided
into four major types: neural network, fuzzy logics, evolutionary computing, and
chaotic systems.
166 7 AI Powerful Tools in Quantum Finance

Fig. 7.4 Classification of artificial intelligence

7.2 Neural Networks—The Brain of AI

7.2.1 Biological Neural Network—Our Brain

Artificial neural networks (ANNs) (or neural networks in short) is one of the major
components of microscopic AI, which focus on the study and modeling of intelligent
system to mimic one of the most important human organs—the brain.
The first scientist to work in the area of brain science was Italian physician Profes-
sor Camillo Golgi (1843–1926) who invented the stain method to investigate neural
activities inside the brain (Bentivoglio 2014). By using this method, he proposed
that the brain is made up of syncytium—a sponge-like tissue that is activated by the
staining operation. Based on his discovery, Spanish neuroanatomist Professor San-
tiago Ramón y Cajal (1852–1934) proposed an innovative idea that “these staining
tissues were not sponge-like elements, but rather the collections of the brain cells
called neurons”, which were interlinked together to form a complex group—neural
networks as shown in Fig. 7.5. Both esteemed medical professionals were awarded
the Nobel Prize for Physiology or Medicine in 1906.
Figure 7.5 shows each neural cell (neuron) consists of
• nucleus—central body of the neuron;
• axon—prolonged filament which connects to other neurons;
• dendrites—tree-like structures which branch from the neuron; and
• synapse—the axon tips (junctions) that make contact with other neurons by attach-
ing to the dendrites of these neighboring neurons.
7.2 Neural Networks—The Brain of AI 167

Fig. 7.5 Biological neural network

Latest neuroscience revealed that human brain consists of around 1011 neurons.
The enigma is: the volume of neurons in human brain is so massive that we can only
make use of less than 10% of neurons throughout our lives, which is rather against
the basic evolution theory.
In fact, how such biological neural network works was still a mystery at that time.

7.2.2 Integrate-and-Fire Operations in Biological Neural

Network

Prior 1943, almost all neuroscientists believed that the sole purpose of neurons was
to process energy. Yet the way they work such as how to process information and
store memory was unsolved.
In 1943, neurophysiologist Professor Warren McCulloch (1898–1969) and mathe-
matician Professor Walter Pitts (1923–1969) published an influential paper A logical
calculus of the ideas immanent in nervous activity (McCulloch and Pitts 1943) which
set off the birth of artificial neural networks, ANN in short (Abraham 2016).
In the paper, they proposed that the main function of neural activities was to
process information—not energy storage. They maintained that the functions of the
neurons were just like logical switches. Signal transmission from one neuron to
another at synapses is the result of a complex chemical process in which specific
transmitter substances are released from the sending points of the junctions. If the
potential reaches a certain threshold, a pulse will be generated down the axon, known
as firing, as shown in Fig. 7.6. More importantly, in the paper, they demonstrated
168 7 AI Powerful Tools in Quantum Finance

Fig. 7.6 3D illustration of integrate and fire in neural network (Tuchong 2019c)

how their proposed network (now called artificial neural network) could be used to
perform basic logical operations such as and, or, and not operations.
This breakthrough has not only merely solved the century mystery of how biologi-
cal neural network works, but also provided a solid foundation for digital computing
technology development.
Although we now understand that the neural activities in our brains are quite
different from logical switches such as transistors in digital computer, they behave
somewhat nonlinear (or even chaotic) integrate-and-fire operators for information
process and transmission. This discovery in 1943 coined the first golden age of
artificial neural networks.

7.2.3 A Neuron Model

As a direct analog of the biological neuron, the schematic diagram of neuron structure
can be interpreted as a computational model in which synapses are represented by
weights that modulate the effect of associated input signals, the formulation is given
by
n 
y=f wi xi (7.1)
i=1

where x is the input signals, w’s are the weights, and y is the output.
7.2 Neural Networks—The Brain of AI 169

Fig. 7.7 Biological neuron versus neuron model

The nonlinear characteristics exhibited by the neuron is represented by a transfer

function f (x) such as a binary or bipolar sigmoid function, given by
Binary sigmoid function:

1
f (x) = (7.2)
1 + e−σ x

Bipolar sigmoid function:

1 − e−σ x
f (x) = (7.3)
1 + e−σ x

in which σ is the steepness parameter to control the curvature of the transfer function.
The learning capability of an artificial neuron is achieved by adjusting the weights
in accordance with a predefined learning algorithm, usually in the form of

wj = ασ xj (7.4)

where α is the learning rate and σ is the learning momentum.

Figure 7.7 illustrates the analog between the biological neuron and its neuron
model.

7.2.4 Single Hidden Layer Artificial Neural Network Model

Artificial neural networks (ANNs, or neural network in short) are the pieces of a
computing system designed to simulate the way human brain analyzes and pro-
cesses information. They are the foundations of artificial intelligence (AI) and solve
170 7 AI Powerful Tools in Quantum Finance

problems that would prove impossible or difficult by human or statistical standards.

ANN has self-learning capabilities that enable them to produce better results as more
data become accessible.
A typical artificial neural network consists of intermediate layer(s) known as
hidden layers to facilitate nonlinear computational capabilities of the network system.
Classical ANNs, such as feedforward neural network (FFNN) shown in Fig. 7.8,
allow signals (information) to flow from input units to output units in a forward
direction.
Other basic ANNs include classical Kohonen self-organizing map (SOM) and
learning vector quantization (LVQ) that are based on competition, along with the
adaptive resonance theory (ART), and also our main theme—feedforward backprop-
agation neural network (FFBPN).
From mathematical perspective, ANNs can be regarded as multivariate nonlinear
analytical tools. They are known to be superior at recognizing patterns from noisy,
complex data, and estimating their nonlinear relationships.
Many studies revealed that ANNs possess distinguished learning proficiency in
the underlying mechanics of time series problems ranging from stocks prediction
and foreign exchange rates in various financial markets to weather forecast.

Fig. 7.8 A typical single hidden layer artificial neural network model
7.2 Neural Networks—The Brain of AI 171

7.2.5 Classification of ANNs by Machine Learning

Technique

Following half a century of development, numerous neural networks have been pro-
posed with over 20 different types of commonly used ANNs.
Generally, ANNs are commonly classified by (1) machine learning techniques
(Fig. 7.9) and (2) areas of application (Fig. 7.10).
In terms of machine learning techniques, ANNs can be categorized into three
main types:
1. Supervised-learning (SL) neural networks—network learning (training) based on
input–output (target output) pairs. Typical examples include feedforward back-
propagation neural network (FFBPN), Hopfield network, support vector machine
(SVM), radial basis function (RBF) network, etc.
2. Unsupervised-learning (UL) neural networks—neural networks that do not
need any supervised learning and training strategies, include all kinds of self-
organizing, self-clustering, and learning networks such as SOM, ART (adaptive
resonant theory), etc.
3. Reinforcement-learning (RL) neural networks—different from supervised-
learning counterpart with well-defined input–output pairs, reinforcement learn-
ing trains the neural network to adopt with feedback signals, namely, reinforce-
ment signal (RS). For the right behavior, the network will respond with a posi-
tive RS to award the RL network, whereas the wrong behavior, the network will
respond with a negative RS to punish the RL network. This method is particu-
larity useful to tackle optimization problem without exact target solutions such
as trading strategy optimization. We will study the advanced topics in detail on
multiagent-based intelligent trading strategies in Chap. 13.

ANN
(By network learning classifica on)

Supervised Learning Unsupervised Learning Reinforcement Learning

ANN ANN ANN

- FFBPN - Actor-CriƟc MulƟagents

- Self Organizing Map SOM
- Hopefield Network - Q-Learning Systems
- AdapƟve Resonant Theory ART
- Support Vector Machine - Temporal-diīerence (TD) sys.
- Autoencoders
- Radial-Basis FuncƟon - Deep Reinforcement Learning
- GeneraƟve adversarial network
(DRL) System

Fig. 7.9 Classification of ANNs by machine learning technique

172 7 AI Powerful Tools in Quantum Finance

ANN
(By Applica on Areas)

Pa ern Associa ve
Classifica on Predic on Op miza on
Recogni on Memory

Adaline Adaline LVQ AAM Adaline

LVQ FFPBN ART Autoencoder Hopfield Net
FFBPN RBFN FFPBN HAM SOM
RBFN LSTM Net RBFN Hopfield Net Boltzmann
SVM SVM SOM BAM Machine

Key:
AAM Auto-associaƟve Memory LTSM Net. Long Short -Memory Term Network
ART AdapƟve Resonant Theory RBFN Radial Basis FuncƟon Network
BAM BidirecƟonal AssociaƟve Memory SOM Self-Organizing Map
FFBPN Feedforward BackpropagaƟon Network SVM Support Vector Machine
HAM Hetero-associaƟve Memory

Fig. 7.10 Classification of ANNs by areas of application

7.2.6 Classification of ANNs by Areas of Application

In terms of areas of application, ANNs can be categorized into five major types:
• classification,
• prediction,
• pattern recognition,
• associative memory, and
• optimization.
Figure 7.10 shows the classification of ANNs by areas of application. As shown
in Fig. 7.10, there are certain ANNs such as feedforward backpropagation network
(FFBPN) and radial basis function network (RBFN) that can be used on multiple area
applications such as classification, prediction, and pattern recognition, whereas other
ANNs such as bidirectional associative memory (BAM) and Boltzmann machine are
specialized to a particular application area.
In this chapter, we will study three basic and commonly used ANNs, they are as
follows:
• Associative network,
• Hopfield network, and
• Feedforward backpropagation network (FFBPN).
7.2 Neural Networks—The Brain of AI 173

7.2.7 Auto-associative Network—Network Architecture

Associative learning is one of the major fundamental characteristics of human intel-

lectual behavior.
It is also widely used by human and machines for pattern recognition such as
visual pattern identification and recognition include recalling human faces, voices,
and music. Significantly, it relates to recalling knowledge and memory association.
Figure 7.11 shows the system architecture of an auto-associative network. It is
a single-layered neural network to store a set of patterns for pattern association (or
what we call recalling).
The training of the associative network is conducted by stored patterns iterative
presentation updated by weights in accordance with the training algorithm.
Once the training is completed, the network can associate not only with the stored
pattern but can also associate with the correct incomplete or noisy query stored
pattern.
Basically, there are two major kinds of associative networks:
1. auto-associative networks—in which input (and query) patterns are the same
type (and nature) as the associated pattern; and
2. hetero-associative networks—in which input (and query) patterns are completely
different types (and nature) from the associated patterns.

Fig. 7.11 System Input nodes Output nodes

architecture of
auto-associative network w11
xi y1

x2 y2

xi yj

wnm
xn ym
174 7 AI Powerful Tools in Quantum Finance

7.2.8 Auto-associative Network—Network Training

Algorithm

The Hebb rule (Hebb 1949) is commonly used for network training. Figure 7.12
shows the network training algorithm of auto-associative network (Fausett 1994;
Patterson 1996).
Binary and bipolar vectors can be used in associative networks.
  Both training
vectors will be a set of training input and target output pairs x , y .
Hebb rule is simple to exercise on weight adjustments, and other methods such
as the delta rule can also be adopted. In that case, the weight adjustment formula (in
Step 2.3) will be replaced by
  
wij (new) = wij (old ) + α yj − yj xi where i ∈ [1 . . . n], j ∈ [1 . . . m] (7.5)

in which α is the learning rate of the network.

Since associative network architecture is plain to discern complex patterns such
as human face association and character recognition, it does provide an innovative
means to apply neural networks in association with memory storing and pattern
recalling.
Inspired by these discoveries in neuroscience and neurophysiology, the author
(Lee, 2004) proposed a chaotic neural associative network—the Lee-associator—to
model chaotic and progressive human memory recalling mechanisms.

Fig. 7.12 Network training algorithm of auto-associative network

7.2 Neural Networks—The Brain of AI 175

Fig. 7.13 System

architecture of discrete Zi+1 Zi+2
Hopfield network ZN

Zk+1 Zk+2 Zl

Z1 Z2 Zk

w1l

w1N

7.2.9 Hopfield Network—Network Training Algorithm

Professor John Hopfield published his influential paper in 1984, Neurons with graded
response have collective computational properties like those of two-state neurons
(Hopfield 1984). In this paper, he described how a simple recurrent auto-associative
network can be used for content-addressable memory systems. Hopfield network also
provided a model for the understanding of human memory, which can also be used
in pattern recognition to tackle complex optimization problems such as the typical
traveling salesman problem (TSP).
The architecture of Hopfield network (Fig. 7.13) is similar to a classical auto-
associative network but with three basic differences:
1. Hopfield network is a recurrent network in the sense that output nodes in one
time step are fed as input in the next time step.
2. In classical associative network, all neurons update their activations simultane-
ously, whereas in Hopfield network, one neuron selects solely at a time to update
its activation and then broadcast its new state to other members of the network.
3. Each neuron keeps on receiving the stimulus from external signal during the
entire auto-association process.

7.2.10 Hopfield Network—Network Training Algorithm

Hopfield network’s key aspect is that it demonstrated how a powerful memory stor-
age and retrieval device can be composed of simple auto-associative network mod-
ification. Figure 7.14 shows the training algorithm of a classical discrete Hopfield
network.
176 7 AI Powerful Tools in Quantum Finance

Fig. 7.14 Network training algorithm of Hopfield Network

In Hopfield network’s associative memory, there are two types of association:

auto-association and hetero-association. The former is a vector in association with
itself, whereas the latter are two different vectors in association with storage. It is
important to note that Hopfield’s network learning model used the identical to Hebb
rule (1949), which basically strived to show that learning occurs as a result of the
weights strengthening when the auto-association activity occurs.
In reality, Hopfield network’s vast application areas also set off the rebirth of
ANNs and explored on how neural networks can be applied to complex problems in
daily operations.
Hopfield network also provided a viable model for the understanding and model-
ing of human memory—a significant component in artificial intelligence.
7.2 Neural Networks—The Brain of AI 177

7.2.11 Feedforward Backpropagation Network

(FFBPN)—Network Architecture

Different from the previous two neural networks, feedforward backpropagation net-
works (FFBPNs) provide a multilayer network architecture (Lee 2006; Silva et al.
2017).
A typical FFBPN consists of an input layer, a hidden layer, and an output layer
as shown in Fig. 7.15. Although FFBPN can consist of several hidden layers, one or
maximum two hidden layers are usually sufficient in most of the cases.
The network training of FFBPNs consists of three main processes:
1. The feedforward process of network training;
2. The error evaluation process to calculate the errors between calculated output
values and target output values; and
3. The backpropagation process of the errors for weight adjustments.
Similar to most other networks, training terminates when errors are bound within
the tolerance level.
Note:
• In the network architecture, w’s denote network weights between the input and
hidden layers, and u’s denote network weights between the hidden and output
layers.
• The total numbers of neurons located in input (n), hidden (t), and output layers
(m), respectively.
• For activation functions, normally a sigmoid function is adopted.
Figure 7.18 shows the training algorithm of a typical FFBPN.

Fig. 7.15 System

architecture of FFBPN
178 7 AI Powerful Tools in Quantum Finance

7.2.12 Feedforward Backpropagation Network

(FFBPN)—Training Algorithm

An FFBPN can model various kinds of problems such as weather prediction and
stock forecasting, pattern recognition such as character recognition, classification,
and even optimization. In quantum finance stock (or forex) prediction, after time
series financial data (e.g., Daily O, H, L, C, V) are normalized, they are used together
with QPLs as input signals for network training in order to calculate next day’s H/L
forecasts. Figure 7.16 shows the training algorithm of FFBPN.

Fig. 7.16 Network training algorithm of FFBPN

7.2 Neural Networks—The Brain of AI 179

However, FFBPN has certain intrinsic limitation such as trapping in local minima
along with difficulty in choosing optimal parameter settings (and input vectors).
Lengthy convergence rate is also another major concern. A possible solution is the
integration with chaotic neural oscillator technique will be studied in Chap. 9.

7.2.13 Neural Network—Where to Go?

In this section, we have studied an overview of neural networks, their basic structure,
and mechanisms.
Current trends of neural networks research and development are focused on three
major areas:
1. The integration of other AI techniques to remedy certain classical neural net-
works intrinsic limitations includes fuzzy logic—fuzzy-neuro systems, genetic
algorithms (GAs) to overcome parameter selection problem and to fine-tune net-
works;
2. The investigation and study in neural dynamics, especially in neural oscillators,
neural oscillatory models; and
3. The investigation and study in chaotic neural dynamics of chaotic neural networks
to model complex AI problems (Figs. 7.17 and 7.18).

Fig. 7.17 Neural network and AI (Tuchong 2019d)

180 7 AI Powerful Tools in Quantum Finance

Fig. 7.18 Illustration of DNA and chromosome in a eukaryotic cell

7.3 Genetic Algorithms—The Optimization Engine

7.3.1 Nature of Evolution—Genetic Algorithms

Evolution refers to encode the operations of biological entities (chromosomes) rather

than the engineering of the whole living beings themselves.
According to Darwin’s evolution theory (Darwin and Huxley 2003), natural selec-
tion is based on survival of the fittest—in effect, chromosomes with high fitness values
will reproduce more than those with low fitness values.
Professor John Holland (1929–2015) introduced genetic algorithms (GAs) in 1960
based on Darwin’s concept of evolution theory. In 1975, he wrote a groundbreaking
book on genetic algorithms: Adaptation in Natural and Artificial Systems (Holland
1992). Afterward, his student Professor David E. Goldberg extended his work on GA
and published his distinguished book: Genetic Algorithms in Search, Optimization
and Machine Learning, which becomes one of the most cited books in computer
science and AI (Goldberg 1989).
Basic entity of GAs is the chromosome, which is a sequence of values/states.
Genetic algorithms are commonly used to generate high-quality solutions to opti-
mization and search problems by relying on bio-inspired operators such as mutation,
crossover, and selection (Kramer 2017).
7.3 Genetic Algorithms—The Optimization Engine 181

7.3.2 Nature of Evolution—Genetic Algorithms

Figure 7.19 shows a typical GA with the following operations:

1. Initialization of the population.
2. Parent selection process.
3. Reproduction process involving crossover and mutation operations.
4. Fitness value evaluation.
5. Iterative execution on the new population until satisfactory performance is
attained.
According to this evolutionary theory, an offspring is normally fitter if its ancestors
are better. Also, chromosomes will grow better as the generations grow.

Fig. 7.19 Flowchart of

genetic algorithms (GAs) Ini alize Popula on

Decode Chromosome String

Compute Fitness Value

Is performance OK/
STOP
?

Selected Parents for Reproduc on

Generate New Popula on

Using Crossover and Muta on
182 7 AI Powerful Tools in Quantum Finance

Fig. 7.20 Populations of chromosomes (Tuchong 2019e)

7.3.3 GA—Population Initialization

A population is a collection of chromosomes with the representation of a parameter

set {x1 , x2 , x3 , …, xm }. This parameter set is to be encoded as a finite length string
over an alphabet of finite length.
It is usually encoded as a binary value of zeros and ones, or list of normalized
real number between 0 and 1 in parameters optimization problem.
The population size depends on the problem nature, but typically contains several
hundreds or thousands of possible solutions. Often, the initial population is generated
randomly, allowing the entire range of possible solutions (the search space). Occa-
sionally, the solutions may be seeded in areas where optimal solutions are likely to
be found.
For a chromosome of length m, the possible number of different chromosome
strings is 2m (Fig. 7.20).

7.3.4 GA—Fitness Evaluation Function

An evaluation function f (x) is applied to the population to compute the fitness value
of a chromosome.
This evaluation function varies among different problems.
7.3 Genetic Algorithms—The Optimization Engine 183

For example, in a parameter optimization problem of a trading system, f (x) might

be the accumulated returns in a certain period of time, say, 2-month period.
It is vital that a complete GA remains in a form as it is the only selection criterion
for chromosome performance of the entire population and increases the possibility
for reproduction.
To justify stopping criterion of the GA, it usually depends on whether the best
chromosome in the population has attained a sufficient level or evolution (i.e., itera-
tion of reproduction) which exceeded the generation limit (say a maximum of 1000
generations).
The fitness function must not only correlate closely with the designer’s goal,
but it must also be computed rapidly. Execution speed is very important. A typical
genetic algorithm must be iterated many times in order to produce a usable result for
a nontrivial problem.
Fitness approximation may be appropriate, especially in the following cases:
• Fitness computation time of a single solution is extremely high.
• Precise model for fitness computation is missing.
• The fitness function is uncertain or noisy.

7.3.5 GA—Parents Selection Scheme

For parents selection, a fitness proportionate selection (also known as roulette-wheel

selection scheme) is commonly adopted. Fitness proportionate selection is a genetic
operator used in genetic algorithms to select prospective beneficial recombination.
The fitness function assigns a fitness to all chromosomes. This fitness level is used
to associate a probability of selection with each individual chromosome.
If f i is the fitness value (FV) of a member i of the population, its probability of
being chosen (i.e., reproduction) will be

fi
pi = m (7.6)
j fj

Those chromosomes selected for the possibility of reproduction are directly pro-
portional to their fitness value, and they conform to the basic feature of natural
selection that: A fitter organism has a higher chance of survival, hence reproduction.

7.3.6 GA—Crossover Operations

In GA, there are two main operators for reproduction, namely, crossover and muta-
tion.
184 7 AI Powerful Tools in Quantum Finance

One Point Crossover

1 0 1 1 0 0 0 1 1 0 1 1 1 0 1 0

0 0 1 1 1 0 1 0 0 0 1 1 0 0 0 1

Mul -point Crossover

1 1 1 0 1 1 0 0 0 1 0 1 0 1 0 0 1 0

1 0 1 1 1 0 1 1 0 1 1 1 1 1 1 1 0 0

Fig. 7.21 One-point versus multi-point crossover operation

In typical crossover operation, a pair of parent chromosomes is selected from the

population.
In one-point crossover, a random location is selected from the chromosome
strings, and chromosome elements beyond this crossover point are exchanged to
form a pair of new offspring according to the crossover rate.
Similarly, for two-point and uniform crossover, multiple points are selected for
crossover operations.
Figure 7.21 illustrates one-point versus multi-point crossover operations.
Different algorithms in evolutionary computation may use different data structures
to store genetic information, and each genetic representation can be recombined with
different crossover operators.
Typical data structures that can be recombined with crossover are bit arrays,
vectors of real numbers, or trees.

7.3.7 GA—Mutation Operations

For mutation operations, a single chromosome is selected from the population which
will be screened throughout the entire list.
Basically, there are five basic types of mutation (Fig. 7.22):
• Deletion,
• Duplication,
• Inversion,
• Insertion, and
• Translocation.
7.3 Genetic Algorithms—The Optimization Engine 185

Fig. 7.22 Five basic types of mutation operations

A particular element will be changed according to the mutation rate, which is

normally much lower than the crossover rate, say 1–5% of probability.
The classic example of a mutation operator involves a probability that an arbitrary
bit in a genetic sequence will be changed from its original state. A common method of
implementing the mutation operator involves generating a random variable for each
bit in a sequence. This random variable tells whether or not a particular bit will be
modified. This mutation procedure, based on the biological point mutation, is called
single point mutation. Other types are inversion and floating-point mutation. When
the gene encoding is restrictive as in permutation problems, mutations are swaps,
inversions, and scrambles.
The main purpose of crossover is to exchange information between randomly
selected parent chromosomes with the aim of not losing any improvement of infor-
mation. The purpose of mutation in GAs is preserving and introducing diversity.
Mutation should allow the algorithm to avoid local minima by preventing the pop-
ulation of chromosomes from becoming too similar to each other, thus slowing or
even stopping evolution.
This reasoning also explains the fact that most GA systems avoid only taking the
fittest of the population in generating the next but rather a random (or semi-random)
selection with a weighting toward those that are fitter.
Although the main objective of mutation is to introduce some genetic diversity
into the population, it must remain at a slow rate (normally less than 5%) in order
not to disrupt the genetic characteristic of the “good genes”.

7.3.8 GA—Implementation

Based on different parent selection criteria, reproduction scheme, crossover, and

mutation methods, there are numerous versions of schema for GAs implementation.
The fundamental one is reproduction to replace all parent population, using one-
point crossover and bit mutation.
For parent selection, the roulette-wheel parent selection scheme based on parent
fitness value is applied.
186 7 AI Powerful Tools in Quantum Finance

In the elitism scheme, parents with the highest fitness value will be retained in
next generation in order to guarantee the performance of the population at a certain
standard.
For crossover operation, a two-point crossover in the other extreme to uniform
crossover can be applied for other GA schemes.
For setting GA parameters, besides using a fixed crossover and mutation rates
throughout the whole evolutionary process, a dynamic crossover and mutation rate
assignment scheme can also be used.
The ratio of mutation rate will normally be set to a higher value when the number
of generations increases to a higher level, such as 500 iterations.
The main reason for this is to induce a higher diversity of the chromosomes when
the entire population evolves to a more mature stage, whereas a higher mutation rate
can bring more “freshness” to the population.

7.3.9 GA—Applications

As one can see, genetic algorithm (GA) is based on the foundation of evolution theory
that focuses on target system improvement after generations of evolution.
Naturally speaking, GA is particularly useful to tackle two main categories of
problem:
• Optimization problem and
• Scheduling problem.
Besides, GAs have been widely used with neural networks in two specific areas:
• Topology optimization
• Genetic training algorithms
In topology optimization, GAs are used to select the optimal topology (or param-
eter setting) for neural network which in turn is trained using some fixed training
scheme such as FFBP training.
In genetic training algorithms, the learning of a neural network is formulated as
a weight optimization problem, usually using the inverse mean square error as the
fitness evaluation scheme.
In Part II of the book—applications of quantum finance, we will study how GA
can be used for the selection of top-10 fuzzy financial signals in a quantum finance
transient-fuzzy DNN (deep neural network) for financial prediction as in Fig. 7.23.
7.4 Fuzzy Logic—The Fuzzification Engine 187

TOP-10 Fitness FFS Selec on Module

(T10FFSSM)

New Genera on Popula on Selec on Module

(NGPSM)

Fuzzy Financial Signals Fitness Evalua on Module

(FFSFEM)

Fuzzy Financial Signals Muta on Module

(FFSMM)

Fuzzy Financial Signals Crossover Module

(FFSCM)

Fuzzy Financial Signals Popula on Genera on Module

(FFSPGM)

Fig. 7.23 GA-based top-10 fuzzy financial signals selection scheme

7.4 Fuzzy Logic—The Fuzzification Engine

7.4.1 The World of Fuzziness, Chaos, and Uncertainty

In our daily lives, we have to handle numerous things all the time—solving problems,
asking and answering questions, etc.
At the end of the day, if we sit down and recall how many things were handled
that day, we will find that this number is surprisingly large.
188 7 AI Powerful Tools in Quantum Finance

If you take a closer look at all these things, we will find that most (or almost all)
of them have the following three basic properties:
• They are fuzzy and highly uncertain in nature.
• We need our knowledge and/or experience to handle them.
• The methods and/or solutions to handle them are usually highly dynamic (or
sometimes chaotic) in nature.
In fact, we might be aware that things that happen always seem to be highly
chaotic and unpredictable.
Does this mean that we cannot do anything to predict them, or cannot model all
these phenomena to be solved and handled automatically?

7.4.2 The Birth of Fuzzy Logic

The notion of fuzzy is not a new idea. The first scientific and mathematical inter-
pretation for the phenomenon of fuzzy (or one can say the fuzzy property of matter)
attracted little attention from scientists until the feverish study of Heisenberg Uncer-
tainty Principle of matter in the late 1920s and 1930s.
In 1937, Professor Max Black (1909–1988) published his paper Vagueness: an
exercise in logical analysis in the distinguished journal Philosophy of Science pro-
posed to use continuous logic component-wise techniques in sets and lists of elements
(Black 1937).
However, instead of using the name fuzzy sets to describe these vague sets, he
referred to these structures as vagueness. Nevertheless, he was considered to be the
forerunner to propose and define the first fuzzy membership function.
The term fuzzy set was coined by Professor Lotfi A. Zadeh (1921–2017) in 1965 in
his influential paper fuzzy sets published in the journal Information & Control (Zadeh
1965). In this paper, he proposed a modified new kind of set theory (so-called fuzzy
set theory), which was used to describe the belonging degree of an individual member
that can exist as any real value between 0 and 1, instead of just either 0 or 1 (Lee
2006; Ross 2016). Later, in his distinguished 3-Part journal papers The Concept of
a Linguistic Variable and Its Application to Approximate Reasoning published by
Information Sciences in 1975, he established a concrete theoretical and mathematical
foundation on type-1 and type-n fuzzy logics which began the age of fuzzy logic in
the following half-century (Zadeh 1975a, b, c).
Moreover, he also demonstrated how these fuzzy sets could be operated, including
fuzzy set union and intersection, and also developed a consistent framework for
dealing with all of these structures and operations.
He believed that fuzzy sets, instead of crisp sets, should intuitively exist in our
world of experience (Zadeh and Aliev 2019).
More importantly, these kinds of existence were even more naturally accepted by
us and consistent with what we understood as our world of existence.
7.4 Fuzzy Logic—The Fuzzification Engine 189

7.4.3 Fuzzy Theory and the Uncertainty Principle

In fact, Professor Zadeh’s observations were very sensible.

Most people might think that a fuzzy set tries to turn everything in our perceived
world blurry, unclear, or uncertain, while a crisp set is the reality.
Actually, it is just the opposite situation.
Lack of crispness should instead be the truth face of a real world.
According to the Heisenberg uncertainty principle learnt in previous chapter, we
cannot exactly (100%) determine (measure) the position and momentum of an object
at the same moment.
From fuzzy theory point of view, what Heisenberg’s uncertainty principle tells us
is that all measured qualities appearing in our world of perception are fuzzy in the
sense that it is not the quality itself that is fuzzy, but rather that we cannot always
measure these physical qualities with 100% certainty.
In fact, there are three types of uncertainties that are now recognized:
1. fuzziness (or vagueness), which results from the imprecise boundaries of fuzzy
sets;
2. non-specificity (or imprecision), which is connected with sizes (cardinalities) of
relevant sets of alternatives; and
3. strife (or discord), which expresses conflicts among the various sets of alterna-
tives.
Figure 7.24 shows the taxonomy of uncertainty and decisions. Vagueness is a
form of uncertainty where the analyst is unable to clearly differentiate between two
different classes, such as person of average height and tall person. This form of
vagueness can be modeled by some variation on fuzzy logic. We will return to this
interesting topic again when we study chaos theory in Chap. 8.

Epistemological Knowledge guided

Uncertainty decision
Objec ve
Uncertainty
Ontological Quasi-ra onal
Uncertainty decision
Uncertainty

Moral Rule guided

Subjec ve Uncertainty decision
Uncertainty

Rule Intui on guided

Uncertainty decision

Fig. 7.24 A taxonomy of uncertainty

190 7 AI Powerful Tools in Quantum Finance

7.4.4 ICMR Model—Fuzzy Identification Module

Structurally speaking, fuzzy logic can be decomposed into four basic components—
the so-called ICMR structural framework of fuzzy logic as shown in Fig. 7.25.
Fuzzy Identification Module
This module focuses on determination and identification of fuzzy variable(s) (FV )
being used in the system.
It is the most important and foremost stage in the entire system design, because an
incorrect (or inappropriate) identification of the FV(s) will defeat the whole purpose
of system modeling.
One important fact is that the same quality (variable) can play totally different
roles in different scenarios for the design of fuzzy systems.
For example, if one wants to use fuzzy logic to represent the following statement:
John is tall—the fuzzy variable being used is the quality height. Although this quality
does not appear in the statement, conceptually speaking it is exactly what we mean
by common sense, or what we called priori knowledge.
However, consider another statement: Most of the students in the class are tall.
In this case, if we want to represent this statement using fuzzy logic, the FV being
used should not be the quality height, but rather it should be the quality most—it
depicts the quality of students’ proportion who are tall in this class! In other words,
we should be very careful in selecting the FV(s) being used in a fuzzy system.

Fig. 7.25 Fuzzy

identification module of
ICMR model
7.4 Fuzzy Logic—The Fuzzification Engine 191

7.4.5 ICMR Model—Fuzzy Categorization Module

Based on FV(s) being identified in the last module, this module focuses on how to
categorize the selected FV(s) into different fuzzy sets.
By using the example in previous case, for FV height, the possible fuzzy sets will
be short, average, and tall—in which each of them is a fuzzy set with a range of
values:
• short—for student height below 1.55 m.
• average—for student height between 1.5 and 1.65 m.
• tall—for student height over 1.6 m.
As shown in the above example, one major characteristic of fuzzy sets is that there
exists fuzziness in their boundaries.
In other words, there is not always a clear boundary between one fuzzy set and
another, which is totally unacceptable in a traditional crisp set.
Of course, the categorization of fuzzy sets is up to system designer in the sense
that there are no strict criteria for one to decide the number, the type, and nature of
fuzzy sets being used. But it should be logical and make sense.
For instance, instead of using three fuzzy sets, one can further categorize height
into five categories: very short, short, average, tall, and very tall (Fig. 7.26).

Fig. 7.26 Fuzzy

categorization module of
ICMR model
192 7 AI Powerful Tools in Quantum Finance

7.4.6 ICMR Model—Fuzzy Modeling Module

For each fuzzy set, the designer has to define its membership function (MF). The
membership function is one of the most important components in fuzzy logic. It
defines fuzzy mapping—the fuzzy belongings of each element within its fuzzy set.
From the mathematical point of view, let X be a collection of objects denoted by
{ x|x ∈ X } which is the so-called universe of discourse.
A fuzzy set A in this universe of discourse X is characterized by a membership
function P A(x) which takes values in the interval [0, 1].
For instance, if we define the fuzzy set A as

A = {(x, μA (x)|x ∈ X )} (7.7)

If X is continuous, the fuzzy set A will be written as

A= μA (x)/x (7.8)
X

Otherwise, if X is discrete, the fuzzy set A will be written as


A= μA (xi )/xi (7.9)
N

where N is the total number of possible discrete values in A.

By using the above example, one can model the fuzzy membership functions of
each fuzzy element for the FV height as shown in Fig. 7.27.

(x)

1.0

0.63
Short Average Tall

0.23

1.4 1.5 1.6 1.63 1.7

Height (x)

Fig. 7.27 Fuzzy membership function. Note According to the above fuzzy membership functions.
If we say: the height of a student (Jack) is 1.63, the membership functions for fuzzy sets average and
tall will be 0.23 and 0.63, respectively. The implications for these fuzzy values are: Jack is tall,
but he is not among the tallest ones in the class. But overall speaking, he is taller than the average
students in the class
7.4 Fuzzy Logic—The Fuzzification Engine 193

Fig. 7.28 Fuzzification versus defuzzification

7.4.7 ICMR Model—Fuzzification Versus Defuzzification

The above example also demonstrates a very important concept in fuzzy logic—
the fuzzification versus defuzzification scheme. As said, the main concept of fuzzy
logic is to reflect the reality experience of our world (our perceived world), which is
uncertain and fuzzy.
In other words, the main job of fuzzy logic is to act as a bridge (or a crystal ball
as shown) for us to convert precise values (and measurements) from machine world
(the so-called digital world) to imprecise values in real world (the so-called perceived
world) which is full of fuzziness and uncertainty.
From the operational point of view, we denote the forward operation, which is
the conversion of digital measurements from machine world to fuzzy values in the
real world as the fuzzification scheme, while the reverse one is the defuzzification
scheme. From the mathematical point of view, fuzzification is the forward mapping of
membership function, whereas defuzzification is reverse mapping (inverse function)
of the membership function as shown in Fig. 7.28.

7.4.8 ICMR Model—Fuzzy Reasoning Module

This module is the brain of the entire fuzzy system that aims to explore system’s
heuristics by applying the so-called fuzzy reasoning technique.
This module consists of three main processes:
• fuzzy rules construction process (FRCP),
• fuzzy knowledge-based construction process (FKCP), and
• fuzzy inference process (FIP).
Fuzzy Rules Construction Process—Fuzzy Relations (FRs)
During the construction process, it is essential to define a set of rules. Fuzzy rules
are basically set of fuzzy relations (FRs) that normally appear in the form of if-then
propositions. They are also called fuzzy implication functions (FIFs) (Fig. 7.29).
194 7 AI Powerful Tools in Quantum Finance

Fig. 7.29 Fuzzy reasoning

module of ICMR model

Mathematically speaking, a FIF is denoted in the following form:

If x is A then y is B, where x, A ∈ X , y, B ∈ Y (7.10)

or written as

R = A → B where R is the fuzzy rule (7.11)

Fuzzy Knowledge-Based Construction Process—Fuzzy Knowledge Base (FKB)

The fuzzy knowledge base (FKB) is the knowledge center of a fuzzy system.
In a typical fuzzy system, the FKB normally appears as a collection of fuzzy rules
to represent a particular domain (or domains) of knowledge.
Following the FIF’s notation used in Table 7.1, the FKB of a fuzzy system is
denoted by a collection of fuzzy rules:

Rk = Ak → Bk , k ∈ [1 . . . N ] (7.12)

where N is the total number of fuzzy rules used in the FKB.

From the implementation point of view, there are various approaches to implement
the fuzzy relation operations.
It is up to the system developers to select the most suitable one for their applica-
tions.
The following figure shows the four commonly used fuzzy rule implementation
schemes.
7.4 Fuzzy Logic—The Fuzzification Engine 195

Table 7.1 Fuzzy rules and methods being used

Methods used Fuzzy rule
“If x is A then y is B” where x, A ∈ X, y, B ∈ Y

Mini rule Rmini = A × B = X ×Y μA (x) ∧ μB (y)/(x, y)

Product rule Rproduct = A × B = X ×Y μA (x) · μB (y)/(x, y)
Max-min rule Rmax−min = (A × B) ∪ (not A × Y )

= (μA (x) ∧ μB (y))∨(1 − μA (x))/(x, y)

X ×Y

Arithmetic rule Rarith = (not A × Y ) ⊕ (X × B)

= 1 ∧ (1 − μA (x) + μB (y))/(x, y)
X ×Y

Fuzzy Inference Process (FIP)

Fuzzy inference refers to the fuzzy reasoning (heuristic) operations based on facets
(or fuzzy facets) and knowledge (FKB) in the fuzzy systems.
There are two kinds of fuzzy inference methods:
1. The generalized modus ponens (GMP) and
2. The generalized modus tollens (GMT).
Owing to their characteristics of inference operations, GMP and GMT are also
called direct reasoning and indirect reasoning, respectively.
For example, given two groups of fuzzy set A, A ∈ X while B, B ∈ Y:
GMP—For any two fuzzy variables x and y, suppose we are given:
(a) Implication (the fuzzy rule): if x is A then y is B.
(b) Premise: x is A .
(c) Conclusion y is B .
which can be denoted by

B = A o R (where R is the relation). (7.13)

GMT —For any two fuzzy variables x and y, suppose we are given:
(a) Implication (the fuzzy rule): if x is A then y is B.
(b) Premise: y is B .
(c) Conclusion x is A .
which can be denoted by

A = R ◦ B (where R is the relation). (7.14)

Fuzzy Compositional Rule

One of the most important and powerful functions of fuzzy reasoning is the fuzzy
compositional rule (FCR) at fuzzy inference.
196 7 AI Powerful Tools in Quantum Finance

This situation arises when there is a complex knowledge base containing a list of
fuzzy rules (Rk ) that correspond to the measured input signals A k for each fuzzy set
Ak .
That is,
    
x = (A1 , A2 , . . . Ak , . . . An ), y = Bj , j = [1 . . . n] (7.15)

Based on the GMP defined in (7.13), the fuzzy variable B can be inferred as the
composition operation between the fuzzy set Ak and the fuzzy relation R:
   
B = (A1 , A2 , . . . Ak , . . . An ) ◦ R (7.16)

From the implementation point of view, two composition schemes are commonly
adopted: (1) the sup-min operation and (2) the sup-product operation.
In the sup-min operation, the membership function value μB after the composition
operations, is given by
n   n   n 
μB = suprx μAi μAki →μBk (7.17)
k=1 i=1 i=1

Basically, a composite fuzzy reasoning process consists of the following steps:

1. Identify each fuzzy rule and determine the connectives for each condition in the
rule base, in which an OR clause represents a union operation and an AND clause
represents an intersection operation.
2. For each fuzzy rule (Rk ), locate the fire strength αk .
3. Use either the sup-min or sup-product method and infer the overall defuzzification
control action.

7.4.9 Case Study—Fuzzy Air-Conditioning Control Systems,

FACS

In this section, we will use a simple fuzzy air-conditioning control system (FACS)
to illustrate how fuzzy composite reasoning works.
First of all, suppose the FACS can vary its power control (Power) by two environ-
mental factors, namely, air temperature (Temp) and relative humidity (RH), which
are given by two sets of fuzzy variables:
Temp = {Very Cold, Cold, Cool, Mild, Warm, Hot, Very Hot}
RH = {Very Dry, Dry, Humid, Very Humid}
7.4 Fuzzy Logic—The Fuzzification Engine 197

and ranges of temperatures and relative humidity for each fuzzy variable are given
by Table 7.2. The membership function of Temp (x) and Relative Humidity (y) are
shown in Figs. 7.30 and 7.31.
Suppose that the power setting of FACS has three different fuzzy states, given by
Power = {Low, Medium, High}
with the membership function shown in Fig. 7.32.

Table 7.2 Ranges of temperatures and relative humidity for each fuzzy variable
Temperature Fuzzy temp range Relative humidity Fuzzy RH range (%)
description (Fuzzy (°C) (Fuzzy variable)
variables)
Mild 17–25 Dry ≤60
Warm 23–31 Moderate 55–85
Hot 25–33 Humid 65–95
Very hot ≥28 Very humid >75

Fig. 7.30 Fuzzy

membership function of
temperature

Fig. 7.31 Fuzzy

membership function of
relative humidity
198 7 AI Powerful Tools in Quantum Finance

Fig. 7.32 Fuzzy

membership function of
power

Assume this power switch is controlled by the following three fuzzy rules:
Rule 1: If Temp is very hot and RH is very humid then switch to high.
Rule 2: If Temp is hot and RH is humid then switch to medium.
Rule 3: If Temp is warm or RH is moderate then switch to low.
Figure 7.33 illustrates a conceptual diagram of a fuzzy logic air-conditioning
system.
Now, given that the temperature is 30 °C and the relative humidity is 80%, the
defuzzification schemes of the power by using the sup-min method and sup-product
method are illustrated in Figs. 7.34 and 7.35.

Fuzzy Logic Air Condi oner

Target Fuzzy Logic

Temp, RH Control System

Room
Temp, RH

Fig. 7.33 Conceptual diagram of fuzzy logic air-conditioning system

7.4 Fuzzy Logic—The Fuzzification Engine 199

Fig. 7.34 Fuzzy reasoning of FACS by using the sup-min method

7.4.10 Applications of Fuzzy Logic in Daily Life

Nowadays, fuzzy logic (FL) is widely adopted in many real-world applications rang-
ing from fuzzy electronic appliances such as rice cookers, air conditioners, and
washing machines as in Fig. 7.36 to sophisticated control systems such as robotics
and fuzzy-based ABS (automatic braking systems).
In reality, many Japanese car manufacturers incorporate fuzzy systems for antilock
braking, active suspension systems, automatic transmissions, and engine emission
controls into their automobiles.
Fuzzy systems are easy to set up, typically require less processing power than
alternative approaches and provide robust performance.
200 7 AI Powerful Tools in Quantum Finance

Fig. 7.35 Fuzzy reasoning of FACS by using the sup-product method

Fig. 7.36 Smart home with fuzzy logic electronic home appliances (Tuchong 2019f)
7.4 Fuzzy Logic—The Fuzzification Engine 201

7.4.11 Fuzzy Expert Systems on Financial Trading

From the implementation point of view, current fuzzy logic is mainly integrated
with other AI technologies in order to solve complex problems and produce more
sophisticated and comprehensive systems (Lee 2006; Siler and Buckley 2004). This
type of hybrid system integration technique is so-called “hybridization”.
Basically, there are two major kinds:
• Fuzzy expert systems (or fuzzy-ES for short) and
• Fuzzy-neuro systems.
Fuzzy Expert Systems on Financial Trading
Owing to the capability of fuzzy systems that can handle imprecise concepts and
values, fuzzy logic has been widely used in many practical commercial and industrial
applications. Fuzzy-ES focus on the integration of fuzzy with expert knowledge in
classical expert systems.
When fuzzy logic applies in financial trading, a typical schematic diagram of
a fuzzy expert system is required to incorporate a fuzzy knowledge base (KBS)
together with the fuzzification of technical indicators, in order to drive expert advice
and perform day trade.
Figure 7.37 illustrates the conceptual diagram of fuzzy expert systems on financial
trading.

7.4.12 Fuzzy-neuro System in Quantum Finance

The major shortcoming of the fuzzy system is the lack of learning capability.
The design of fuzzy sets and the assignment of all fuzzy relations are done by
system designers (or experts) without any way of acquiring membership functions
and inference rules automatically.
To solve this problem, hybridization with other AI models that provide automatic
machine learning such as neural networks—the so-called fuzzy-neuro systems were
developed.
Nowadays, fuzzy-neuro systems are widely adopted for solving complex prob-
lems such as stock prediction, severe weather forecasting (e.g., rainstorm prediction),
adaptive tracking systems (e.g., tropical cyclone (TC) tracking systems, missile track-
ing systems), robot vision, etc.
Figure 7.38 shows the integration of fuzzy logic with chaotic neural network
for stock prediction, and we will study quantum finance advance applications in
Chap. 12.
202 7 AI Powerful Tools in Quantum Finance

Fig. 7.37 Conceptual diagram of fuzzy expert systems on financial trading

Input Layer
8-Bifurca on - Level Chao c Neuro- oscillatory Deep Learning Hidden Layers
TOP-10 FFS Vectors
CT2TFM

Output Layer
NEXT DAY Open, High, Low, Close

Forecast
Databank
FTS Vectors
FTSM

Fig. 7.38 System architecture of chaotic type-2 fuzzy deep neural network for financial prediction
(CT2TFDNN)
7.5 Conclusion 203

7.5 Conclusion

In this chapter, we have learnt the basic concept of artificial intelligence, what it is
and how it works.
We had also studied three major components of modern AI technology:
• Artificial neural networks,
• Genetic algorithms, and
• Fuzzy logic.
As mentioned, owing to the advancement of computer technology, the popularity
of Internet technology along with social medium and mobile computing, the needs
of AI and intelligent systems are more than anytime in this new era.
In fact, there are many new AI technologies, algorithms, techniques, and appli-
cations emerging on daily basis.
This chapter is only an introductory course for the path of AI.
The truth is: Different from other disciplines, AI is an integrated and cross-
discipline doctrine ranging from neuroscience and visual psychology to quantum
computing without any boundary. The only limitation is imagination and thoughts.
In next chapter, we will discuss two innovative AI technologies which are related
to financial engineering, and also important components of quantum finance—chaos
theory and fractals.

Problems

7.1 What is artificial intelligence (AI)? Give any three applications of AI in finance
industry and explain how they work.
7.2 What is Turing test? Discuss and explain how it works?
7.3 What discipline of AI does Turing test focus on? Why such test is believed
to be the best method to evaluate a machine (robot) is truly intelligence or
not? Give a live example in daily human interactions and communication to
support your explanation.
7.4 John Searle in 1980 proposed a thought experiment so-called “Chinese room
experiment” to challenge the validity of Turing test for the evaluation of
machine intelligence.
204 7 AI Powerful Tools in Quantum Finance

Chinese Room Thought Experiment

Similar to the setting of Turing test, but this time all Q&A between the
invigilator and the challenger are in Chinese in which the challenger (hu-
man) nothing about Chinese. So, the challenger in the Chinese room
follows English instructions in his computer program for manipulating
Chinese symbols, where a computer “follows” a program written in a
computing language to reply with Chinese symbols. The challenger pro-
duces the appearance of understanding Chinese by following the symbol
manipulating instructions but does not thereby come to understand Chi-
nese. Since a computer just does what the human does—manipulate sym-
bols on the basis of their syntax alone, merely by following a program,
comes to genuinely understand Chinese to pass the test.

(i) What is the main argument of John Searle’s Chinese room experiment
to challenge the validity of Turing test for the evaluation of machine
intelligence in AI?
(ii) Does the argument and challenge by John Searle’s Chinese room exper-
iment against Turing test is valid or not? Why?

7.5 What is an intelligent agent in AI? State and explain the three main difference
between an intelligent agent versus traditional computer program (system)?
Give two examples of how intelligent agents can be applied to finance.
7.6 Discuss and explain the major differences between strong AI versus weak AI.
Give two examples of strong AI and weak AI applications and explain how
they work.
7.7 In 1997, IBM Deep Blue defeated world chess champion Mr. Garry Kasparov.
It is also the first time AI computer program defeat world chess champion.
(i) Why this is an important event and milestone in AI history? In other
words, why chess playing with chess master is an important challenge
in AI community?
(ii) In fact, IBM Deep Blue system at that time is a classical AI-based expert
system tailored for chess playing. Discuss and explain what is an expert
system. How it works?
(iii) However, after the world chess contest in 1997, Kasparov in the TV
interview complained that it was not a fair game and not really a kind
of human versus machine intelligence. What was his argument behind?
Do you agree? How this event influences the future development of AI?
7.8 Why was the past 20 years can be claimed as the second golden age of AI?
and what kind of AI is focused on—strong AI or weak AI? What are the logics
behind?
7.9 State and explain the three main disciplines of weak (soft) AI. For each disciple,
give one real-world application of soft AI and explain how it works.
7.5 Conclusion 205

7.10 What is soft AI? Discuss and explain why soft AI is also known as cognitive
approach (CA) AI technology. Name the major four soft AI technologies and
briefly explain how they work.
7.11 What is artificial neural network (ANN)? What is the main difference between
ANN and contemporary computer system in terms of (1) memory storage and
(2) machine learning and computational methods.
7.12 What are integrate-and-fire operations in biological neural networks? Why it
is important in machine learning? And how artificial neural networks model
these operations?
7.13 Describe and state the roles and importance of transfer function in typical
artificial neural networks (ANNs). How can we model transfer function in
traditional ANN?
7.14 Describe and explain the major pros versus cons of using sigmoid function as
transfer function in typical ANN. Give a real-time example in finance predic-
tion to support your explanation.
7.15 What is a multilayer feedforward backpropagation network (FFBPN)?
Describe and explain how it works? State and explain two examples for the
application of FFBPN in finance engineering.
7.16 In a typical multilayer FFBPN, how can we determine the: (1) number of input
and output nodes? (2) number of hidden layers and the number of neurons in
each hidden layer?
7.17 Describe and explain the three major categories of artificial neural networks
(ANNs) in terms of machine learning technique. For each category of ANN,
give one example of application in financial engineering and describe briefly
how it works.
7.18 In machine learning, pattern recognition and time series prediction are sides
of the same coin. Discuss and explain why. Give an example in finance engi-
neering to support your explanation.
7.19 Programming exercise I
(i) Write the flowchart and system training algorithm of feedforward back-
propagation network (FFBPN).
(ii) Write a MATLAB program to implement the FFBPN training algorithm
by using the time series of any financial product (e.g., DJI) exported from
MT4/R platform for short-term financial prediction.
(iii) First, using 5 days to predict next day open/high/low/close.
(iv) Second, using 10 days to predict next day open/high/low/close.
(v) Third, using 20 days to predict next day open/high/low/close.
(vi) Compare their performance in terms of (1) training performance and
(2) forecast performance.
(vii) Which one is better? Why?
7.20 Describe and explain what is genetic algorithm (GA)? What kind(s) of AI
applications and problems GA tackles with? Why? Draw the flowchart of GA
and explain how it works.
206 7 AI Powerful Tools in Quantum Finance

7.21 What is the core idea and innovation of genetic algorithms (GA) in machine
learning? Give two examples of how we can apply GA on finance engineering
and hence quantum finance.
7.22 Discuss and explain the roles and functions of crossover and mutations in GA.
As compared with crossover, explain why mutation must keep at a much lower
rate. What is the logic behind?
7.23 State and explain two typical applications for the integration of ANN with GA
in financial engineering.
7.24 Discuss and explain the relationship between fuzzy logic and uncertainty prin-
ciple in quantum theory. Use two financial markets as example to support your
explanation.
7.25 Discuss and explain how fuzzification process works in fuzzy logic? How can
we apply fuzzification in financial engineering? Give two examples to support
your explanation.
7.26 State and discuss two main types of fuzzy expert systems. For each type of
fuzzy expert system, give one potential application on (1) finance engineering
and (2) quantum finance and explain how they work.
7.27 Discuss and explain why the integration of fuzzy logic with artificial neural
networks (or recurrent neural networks) is commonly used in AI applica-
tions. Give two examples of how fuzzy-neuro systems can be used in quantum
finance.
7.28 Programming exercise II
(i) Repeat 7.19, but this time using MQL to implement the FFBPN in MT4
platform.
(ii) For the input nodes, in addition to the time series information, add
10 commonly used technical indicator as input nodes, which include
iAlligator (alligator oscillator), iMA (moving average), iBands
(Bollinger bands), iMomentum (momentum indicator), iMACD (MACD
oscillator), iRSI (RSI indicator), iRVI (RVI indicator), iStdDev (stan-
dard deviation indicator), iStochastic (stochastic oscillator), and iWPR
(Williams’ percent range indicator).
(iii) Apply fuzzy logic to fuzzification of these 10 technical indices, using
the fuzzification scheme for RSI discussed in this chapter as example.
(iv) Apply genetic algorithm (GA) to choose the TOP 5 technical indicators
for live application.
(v) Implement this GA-based fuzzy-neuro network and compare its perfor-
mance with the traditional FFBPN implemented in (i).
(Note: For detail training workshop on MQL, please visit QFFC.org official site.)
References 207

References

Abraham, T. Rebel (2016) Genius: Warren S. McCulloch’s Transdisciplinary Life in Science. MIT
Press, Cambridge.
Black, M. (1937) Vagueness: An Exercise in Logical Analysis. Philosophy of Science 4(4): 427–455.
Bentivoglio, M. (2014) “Golgi, Camillo”. In Daroff, Robert B.; Aminoff, Michael J. Encyclopedia
of the Neurological Sciences (Second edition. ed.). Burlington: Elsevier Science. 464–466.
Darwin, D. and Huxley, J. (2003) The Origin of Species: 150th Anniversary Edition. Signet; Reprint,
Anniversary edition.
Fausett, L. (1994) Fundamentals of Neural Networks Architectures Algorithms and Applications.
Prentice Hall.
Goldberg, D. E. (1989) Genetic Algorithms in Search, Optimization, and Machine Learning.
Addison-Wesley Professional.
Hansen, W. Classical Mythology: A Guide to the Mythical World of the Greeks and Romans. Oxford
University Press, 2005.
Hebb, D.O. (1949) The Organization of Behavior. New York: Wiley & Sons.
Holland, J. H. (1992) Adaptation in Natural and Artificial Systems: An Introductory Analysis with
Applications to Biology, Control, and Artificial Intelligence. A Bradford Book; Reprint edition.
Hopfield, J.J. (1984) Neurons with Graded Response Have Collective Computational Properties like
Those of Two-State Neurons. Proceedings of the National Academy of Sciences of the United
States of America. 81(10): 3088–3092.
Kramer, O. (2017) Genetic Algorithm Essentials (Studies in Computational Intelligence). Springer.
Lee, R. S. T. Fuzzy-Neuro Approach to Agent Applications (From the AI Perspective to Modern
Ontology). Springer-Verlag, Heidelberg Germany, 2006.
Lee, R. S. T. A Transient-chaotic Auto-associative Network (TCAN) based on LEE-oscillators.
IEEE Transactions on Neural Networks. 15(5): 1228–1243, 2004.
McCulloch, W. and Walter, P. (1943). A Logical Calculus of Ideas Immanent in Nervous Activity.
Bulletin of Mathematical Biophysics. 5 (4): 115–133, 1943.
Patterson, D. W. (1996) Artificial Neural Networks. Prentice Hall.
Ross, T. J. (2016) Fuzzy Logic with Engineering Applications. Wiley.
Russell, S. and Norvig, P. (2015) Artificial Intelligence: A Modern Approach, 3rd edition. Pearson
Education India.
Siler, W. and Buckley, J. J. (2004) Fuzzy Expert Systems and Fuzzy Reasoning. Wiley-Interscience.
Silva, et al. (2017) Artificial Neural Networks: A Practical Course. Springer.
Tuchong (2019a) Illustration of AI. http://stock.tuchong.com/image?imageId=
286594758192922638. Accessed 21 Aug 2019.
Tuchong (2019b) Weak AI. http://stock.tuchong.com/image?imageId=467459610627670027.
Accessed 21 Aug 2019.
Tuchong (2019c) 3D illustration of integrate-and-fire in neural network. http://stock.tuchong.com/
image?imageId=483388184041619483. Accessed 21 Aug 2019.
Tuchong (2019d) Neural network and AI. https://stock.tuchong.com/image?imageId=
490703320800559171. Accessed 21 Aug 2019.
Tuchong (2019e) Populations of chromosomes. http://stock.tuchong.com/image?imageId=
453246275369566360. Accessed 21 Aug 2019.
Tuchong (2019f) Smart Home with Fuzzy Logic Electronic Home Appliances. http://stock.tuchong.
com/image?imageId=267053275471413372. Accessed 21 Aug 2019.
Turing, A. (1950) Computing Machinery and Intelligence. Mind, LIX (236): 433–460.
Zadeh, L. A. and Aliev, R. A. (2019) Fuzzy Logic Theory and Applications: Part I and Part II.
World Scientific Publishing Company.
Zadeh, L. A. (1975a) The Concept of a Linguistic Variable and Its Application to Approximate
Reasoning-I. Information Sciences. 8(1): 199–249.
208 7 AI Powerful Tools in Quantum Finance

Zadeh, L. A. (1975b) The Concept of a Linguistic Variable and its Application to Approximate
Reasoning-II. Information Sciences. 8(4): 301–357.
Zadeh, L. A. (1975c) The Concept of a Linguistic Variable and its Application to Approximate
Reasoning-III. Information Sciences. 9(1): 43–80.
Zadeh, L. A. (1965) Fuzzy sets. Information and Control. 8 (3): 338–353.
Chapter 8
Chaos and Fractals in Quantum Finance

The theory of chaos and theory of fractals are separate but have
very strong intersections. That is one part of chaos theory is
geometrically expressed by fractal shapes.
Benoit Mandelbrot (1924–2010)

What is chaos theory?

What are fractals?
What is the relationship between chaos and fractals?
How to apply chaos theory and fractals in quantum finance?
In previous chapters, we have learnt the basic concepts of quantum theory.
Wave–particle duality—as the foundation and basics of quantum theory—relates
to the nature of all fundamental particles (quantum particles) that exist in two kinds of
dynamics—particles (motions and dynamics) and waves (energy field and patterns)
can be narrated agreeably by Schrödinger equation.
In modern mathematics and computational physics, we also have two mathemati-
cal models that are agreeably analog to this unique phenomenon—chaos and fractals,
like a coin with both sides, such as Professor Benoit Mandelbrot—the father of fractal
said in his remarkable quotation.
More importantly, these two mathematical models have important implications
in computational finance and modern AI, which are classified as a branch of com-
putational intelligence (CI) and a major component in quantum finance model.
As mentioned in Chap. 1, financial markets refer to a range from stock markets
to cryptocurrency markets. If we look closely at their market dynamics, it is not
difficult to discover that they exist in both particle (market price movements) and
wave (market pattern movements) nature.
By applying numerical computation method in quantum finance Schrödinger
equation (QFSE), we have successfully evaluated all the quantum energy levels
(quantum price levels—QPLs).
So, can we model the particle and wave dynamics as well, without the need to
solve the complex path integral formulation?
The answer is yes—chaos and fractals.

© Springer Nature Singapore Pte Ltd. 2020 209

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_8
210 8 Chaos and Fractals in Quantum Finance

8.1 Basic Concepts and Formulation on Chaos Theory

8.1.1 The World of Nonlinear Dynamics

What is the nature of our world?

This problem has puzzled philosophers, scientists, poets, and scholars of various
disciplines for centuries.
What is the importance of knowing the dynamics of all matter in the world (and
universe)?
To understand this question, let us have a look at this topic of dynamics which
include the following:
• The dynamics of free-falling objects—governed by Newton’s laws of motion.
• The formation and dynamics of black holes—governed by Einstein’s theory of
general relativity.
• The dynamics of subatomic particles such as neutrinos and positrons—governed
by Heisenberg’s quantum mechanisms.
• The formation and dynamics of tropical cyclones, hurricanes, tornadoes, and earth-
quakes.
• The fluctuation of stocks.
• The chance of catching the bus this morning.
• The theory of causation.
• Our fate.
• Our future.
The fact is that this world of dynamics covers not only the physical motion of
matter in the universe, but also the nonphysical occurrence of events, notions, and
ideas.
Chaos theory is the study of such dynamics, which is believed to be highly non-
linear—chaotic. In other words, if one wants to explore the nature of the dynamics
governing matter (and notions) in the universe, one must enter the world of chaos
theory.
Figure 8.1 shows the famous analog in chaos theory—the butterfly effect—the
phrase was first coined by Professor Edward Lorenz (1917–2008) the father of chaos
theory gave a talk at a scientific meeting in 1972 (Krishnamurthy 2015; Lorenz 1995)
on his work regarding the chaotic phenomena found from his weather prediction
model—the Lorenz equations. The phrase suggests that the flap of a butterfly’s wings
in Africa could create a small change in the atmosphere that might eventually lead to a
hurricane in Florida. The butterfly effect concept has become important in the finance
world due to the continuous interdependence between capital markets subsequent to
increased globalization of the world economy. The volatility of a minor sector at
international markets can dilate rapidly and bleed into other markets. Broadened
Internet access and technology improvements also contribute to episodes of extreme
market volatility.
8.1 Basic Concepts and Formulation on Chaos Theory 211

Fig. 8.1 The butterfly effect in chaos theory

8.1.2 Deterministic World Versus Probabilistic World

As everything that happens in the world is highly dynamical (or chaotic, as we said)
and uncertain, does this mean that chaos theory is a study of probability?
The answer is absolutely not.
The truth is: chaos theory has a basis that is entirely different from probability.
Although chaos theory is the study of the world of dynamics, it is not a world of
uncertainty or a matter of chance.
Strictly speaking, chaos theory deals with the world of determinism rather than
probabilism.
Chaos theory maintains that all the dynamics in the world, no matter the complex-
ities involved, can be (and must be) somehow modeled deterministically by certain
chaotic motions. But whether we can find these chaotic motions is another question.
In fact, the formal name for chaos is deterministic chaos in order to reflect this
characteristic (Schuste and Just 2005; Lee 2005).
From the mathematical point of view, no matter the complexity of a physical
phenomenon, the challenge of chaos theory is to model its dynamics by a set of
equations (of time) as follows:
In discrete format:

 
 
xt+1 = f xt (8.1)
212 8 Chaos and Fractals in Quantum Finance

Fig. 8.2 Lorenz attractor (upper right) and the butterfly flapping wings (lower left)

 
where xt is n-dim vectors (with n-variables) and f is its dynamics.
In continuous format:
˙
  
x = f (x, t) (8.2)

 
where x is n-dim vectors of variable x and f is its dynamics.
Figure 8.2 shows the Lorenz’s attractor and butterfly effect—the second imple-
mentation of chaos theory, in which the chaos dynamics of classical Lorenz’s attractor
resembles the flipping wings of a butterfly.

8.1.3 Chaotic Flow Versus Chaotic Maps

In chaos theory, the discrete chaotic motions are also called chaotic maps, and the
continuous motions are called chaotic flows.
8.1 Basic Concepts and Formulation on Chaos Theory 213

Fig. 8.3 Chaotic flow versus chaotic map

From the physical point of view, the discrete maps can be regarded as taking
snapshots (or cross sections) of the continuous flow on a regular (or timely) basis as
shown in Fig. 8.3.
What is the significance of the findings of maps and flows?
The answer is: It is extremely important!
For example, if the variables under consideration are weather elements such as
air temperature or rainfall, and the time step (say for a discrete map) is 5 days, then if
we can find such chaotic function(s), what we have really done is solving a weather
forecasting problem!
Imagine if functions are used to model the next-day stock market or the 15-day
financial pattern!
We can expect that what chaos theory is telling us is not something uncertain (or
chaotic), but rather how to identify the uncertain event in a chaotic way!

8.1.4 Origin of Chaos Theory—Lorenz Attractor

Chaotic phenomena—the first time they were noticed by scientists was in the late
nineteenth century by Professor Henri Poincaré (1854–1912) during his exploration
of planetary motions in 1886.
However, the major breakthrough in chaos theory was made by mathematician and
research meteorologist Professor Edward Lorenz (1917–2008) during his experiment
214 8 Chaos and Fractals in Quantum Finance

on a miniature weather modeling system (Lorenz 1963), using three simple differ-
ential equations (the so-called Lorenz equations) instead of using complex statistical
models which were commonly used by meteorologists at that time.
The Lorenz equations are

ẋ = −σ x + σ y (8.3a)

ẏ = −x z + r x − y (8.3b)

ż = x y − bz (8.3c)

These equations were based on the fundamental Navier–Stokes equations of fluid

dynamics, which describe weather formation as the fluid circulation of air upon
distant heating by the sun.
In Lorenz equations, x is the fluid stream function, whose value is proportional to
speed of motion of the fluid due to convection. y is proportional to the temperature
gradient between rising and falling parts of the fluid at a given height—the so-called
horizontal flow. z is proportional to the deviation from temperature linearity as a
function of vertical position—the so-called vertical flow.
In such a highly simplified weather model, he unexpectedly discovered two impor-
tant things about this nonlinear model:
1. The nature of dynamics for the system itself can be entirely altered with an adjust-
ment in parameters. In his original work, he discovered that when parameters σ
and b are set to 10 and 8/3, respectively, the entire system will turn out to be highly
chaotic without any steady states (and limit cycles), whenever the Rayleigh num-
ber (r) exceeds critical value of 24.74. In other words, all the solutions would
evolve to be neither periodic nor convergent to any solution or equilibrium state.
2. The weather solution of this model, under such chaotic conditions, will be
extremely sensitive to its initial conditions. In other words, within the chaotic
region, only a very small deviation of initial conditions, say minor variations
(or a mistake) in inputting the initial weather conditions (even of the order of
10−5 ), will result in an entirely different weather situation in the following days,
as shown in Fig. 8.4.
Such discovery is essential to the scientific world because for any system with
dynamic equations y = f (x, a), where x is the variable and a is the parameter.
Common sense tells us that only variable x will change the dynamics of the system
(that’s why we called it variable), the changes of parameter a will only cause some
minor changes of the overall system, the overall physical behavior of system will
not be changed. Besides, traditional experimental science tells us that the overall
experimental errors of a system (e.g., prediction system) should be proportional to
the errors of variables. However, his discovery told us another story of reality. In
which for some systems, just like Lorenz’s equations, if we change the parameters
to certain range or values, the overall physical behavior and dynamics of the system
8.1 Basic Concepts and Formulation on Chaos Theory 215

Fig. 8.4 Lorenz attractor (3D illustration)

will be changed entirely into what we called deterministic chaos; or chaos systems.
In that situation, merely a very minor change on initial condition of the system will
cause a huge difference of experimental results in future.

8.2 Characteristics of Chaotic Systems

In summary, there are several features and characteristics that can be found in a
typical chaotic system.
They are
1. sensitivity to initial condition;
2. highly nonlinear but deterministic in nature;
216 8 Chaos and Fractals in Quantum Finance

3. consisted of bifurcation points that lead to chaos; and

4. consisted of self-similarity and fractals.

8.2.1 Sensitive Dependence on Initial Conditions

It is one of the most remarkable features of chaotic system, but it alone does not
mean that a system is chaotic, but every chaotic system must possess this property.
Dependence of initial conditions means that for a very slight change in initial
conditions or measurements (observations) of a chaotic system, the result produced
differ tremendously after exhibiting for a period of time.
People like to quote butterfly effect to express such characteristics.
In layman’s term, it states that a butterfly wings’ flapping in Africa could result
in hurricane in Florida.
Figure 8.5 illustrates the sensitivity to initial condition of Lorenz’s equation of
two systems, one with x(0) = 1.00001 and the other with x(0) = 1.00000, given
σ = 10; r = 28; b = 8/3.
One should note that the sensitivity to initial conditions, although are very similar
to the concept of % errors of measurements in experimental physics, they are two
totally different concepts.

Fig. 8.5 Lorenz’s equation—sensitivity to initial conditions

8.2 Characteristics of Chaotic Systems 217

In classical % error of measurements problem, the resulted % errors of a dynam-

ical system are normally proportional to the degree-of-discrepancy for the measure-
ment of observations. However, in chaos theory, the sensitivity dependence nature is
unrelated to such discrepancy, it is the intrinsic nature of chaotic system itself.

8.2.2 Highly Nonlinear but Deterministic in Nature

Even though a slightest difference in initial conditions can reach extremely different
results, all chaotic systems are deterministic in nature.
Deterministic means that equal causes having equal effects.
Or, from another viewpoint, equal effects have equal causes.
Equal causes having equal effects means predictability. That is, given e1 is the
effect of cause c1, if there is another cause c2 which is equal to c1, we can predict
that c2 will have an effect e2 equals to e1.
Equal effects have equal causes means reversibility. That is, given e2 and e1 are
two equal effects, if e1 is caused by c1, then we can say the cause c2 equals to c1 of
e2. Figure 8.6 illustrates two time series: one is deterministic chaos and the other is a
random walk. As shown, although they look alike, however, they are totally different
in terms of dynamics behavior.

Fig. 8.6 Deterministic chaos versus random walk

218 8 Chaos and Fractals in Quantum Finance

Fig. 8.7 Bifurcation diagram of a logistic map

8.2.3 Bifurcation Phenomenon

Does it mean that only complex nonlinear equations such as the Lorenz equations
have chaotic effect?
The answer is no.
As mentioned, chaotic phenomena are so common and usual that even simple
nonlinear dynamics as in the logistic Eq. (8.4) can have this chaotic phenomenon.

xt+1 = r xt (1 − xt ) (8.4)

Figure 8.7 depicts a typical bifurcation diagram for the logistic equation with
control parameter r ranging from 0 to 4 (Kapitaniak 2000; Lee 2005).
Bifurcation is a common phenomenon found in chaotic systems which indicates
sudden, qualitative changes in the dynamics of the system under investigation either:
1. From one kind of periodic case (with limit cycle(s) and fixed point(s)) to another
kind of periodic situation, such as the change in bifurcation of the logistic map
from r = 3 (from one fixed point to two fixed points) or
2. From a periodic stage to a chaotic stage, such as dramatic changes in the dynamics
of the logistic map for r increasing from 3.5 to 4.

8.2.4 Applications of Chaos Theory

Although chaos theory is rather new and its technology yet emerging, there are
already numerous applications in various industries which adopted this innovative
technology.
8.2 Characteristics of Chaotic Systems 219

Table 8.1 Applications of chaos theory

Application areas Potential applications
Consumer electronics (Chau and Wang 2011) • Chaotic kerosene fan heater (Katayama
1993)
• Chaotic dishwasher
• Chaotic air conditioners
• Chaotic mixer
Computer systems • Lee-associator for progressive memory
recalling (Lee 2006)
• Temporal pattern search (Davis 1990)
• Chaotic mobile robot navigation (Sambas
et al. 2016)
• Chaotic network design and implementation
(Soriano-Sanchez et al. 2018)
• Chaotic cryptosystems (Lee 2005)
Health care and life science • EEG analysis of heart rhythms (Freeman
2001)
• Chaos-aware defibrillator (Munakata 1998)
• Chaotic brain wave analysis (Freeman 2001)
Engineering (Kapitaniak 2000) • Circuit stability analysis
• Vibration control and study
• Heat combustion
Business and finance • Fractal market pattern analysis (Mandelbrot
1997)
• Financial prediction (Lee 2005; Kazem et al.
2013)
• Market prediction (Peters 1994)

• consumer electronics,
• computer systems,
• health care,
• life sciences,
• engineering, and
• business and finance.
Table 8.1 summarizes the potential applications derived from this fascinating
technology.

8.2.5 Chaos Theory Versus Uncertainty Principle

In Chap. 7, we have learnt the implication of Heisenberg’s uncertainty principle on

fuzzy logic. In fact, the principle is also closely related to chaos theory.
Professor Heisenberg mentioned from his work Physics and Philosophy
(Heisenberg 2007) that:
220 8 Chaos and Fractals in Quantum Finance

The Revolution in Modern Science, quantum mechanisms and the Uncertainty

Principle should not be unusual or unique phenomena, but rather should be universal
phenomena that exist in all matter (at the subatomic level).
If this is the case, the subatomic phenomena somehow might be coherent with
chaos theory.
Taking weather forecasts as an example, we know that most of the international
weather centers use numerical weather prediction (NWP) techniques to perform fore-
casts. For every single forecast, we need to know the so-called boundary conditions,
that is, grid-point weather observations include wet- and dry-bulb temperatures, mean
sea-level pressure (MSLP), upper level wind speed and direction, and so on. Accord-
ing to chaos theory, slight differences (errors) in the measured values may result in
completely different weather situations in the future.
Moreover, according to the Heisenberg’s uncertainty principle, it is quite impos-
sible for us to obtain 100% accuracy in measuring all different weather conditions
at the same time.
The same situation applies to finance, uncertainty principle in quantum finance
tells us that we cannot be 100% sure about the quantum finance pair: return (r ) and
return-rate (ṙ ) given by

Δr · ṙ ≥ /2 (8.5)

Does this really mean that we can never predict our future accuracy?
The answer is yes and no.
Yes, in the sense that if Heisenberg’s uncertainty principle is really true, we should
have no way to “measure” the quantum variables for any chaotic system 100%
correctly that will lead to tremendous deviation according to first rule of chaotic
system.
No, in the sense that although we cannot be 100% sure about the initial condition,
we have one mathematical model that can help us to solve this forecast problem, one
model that can predict the future without knowing the exact initial condition, which
is
Neural network, time series neural network to be exact.
In the coming chapters on quantum finance application, we will study a special
kind of neural network, so-called chaotic neural oscillator, that can solve this problem
agreeably with the integration of chaos theory and neural network in quantum finance
prediction.
8.3 Fractals in Quantum Finance 221

8.3 Fractals in Quantum Finance

8.3.1 Fractal Versus Chaos—Two Sides of a Coin

The bifurcation diagram of logistic map learnt in last section also revealed other
interesting phenomena of a typical chaotic system—namely, the existence of self-
similarity in chaotic dynamics.
If we take a closer look at the bifurcation pattern in the first chaotic region (i.e.,
3.5 < r < 4), we will discover that similar bifurcation patterns can be found within
the original pattern, but in a smaller or scaled-down size. Figure 8.8 shows self-
similarity between bifurcation at µ1 , µ2 , and µ3 with same pattern but decrease in
scale to infinity.
More surprisingly, if we continue to zoom into these bifurcation patterns, we will
discover that they, in turn, contain even more scaled-down features with completely
same shapes and patterns—what we called the self-similarity nature of chaotic sys-
tems.
Based on these interesting findings, a group of scientists established a new doctrine
of applied mathematical geometry—fractal geometry (or simply fractals).
In fact, fractals are the graphical (visual) presentations of chaos while chaos is
the physical dynamics of fractals.
In this section, we will explore what fractals are, their formulations, and how they
can be applied to quantum finance.

Fig. 8.8 Self-similarity in logistic map

222 8 Chaos and Fractals in Quantum Finance

8.3.2 A Brief History of Fractals

The study of self-similarity in graphs and patterns is not a new thing. German
polymath and philosopher Gottfried Leibniz (1646–1716) studied recursive self-
similarity and used the term fractional exponent to describe such phenomena.
In 1872, German mathematician Professor Karl Weierstrass (1815–1897) pre-
sented the first definition of a function with a graph that would be considered as
fractal today.
In 1883, German mathematician Professor Georg Cantor (1845–1918) who
attended lectures by Weierstrass, published examples of subsets of the real line known
as Cantor sets—one of the fundamental fractals.
At the same period, German mathematician Professor Felix Klein (1849–1925)
and French mathematician Professor Henri Poincaré (1854–1912) introduced a cat-
egory of fractal that has come to be called self-inverse fractals.
In 1904, Swedish mathematician Professor Helge von Koch (1870–1924)
extended Poincaré’s ideas and gave a more geometric definition of fractal geometric
with famous fractal discovery known as the Koch snowflake.
By 1918, two French mathematicians, Professors Pierre Fatou (1878–1929) and
Gaston Julia (1893–1978) discovered fractal behavior associated with mapping com-
plex numbers and iterative functions leading to further ideas about attractors and
repellers.
In 1967, Professor Benoit Mandelbrot (1924–2010) published a paper about self-
similarity: How Long Is the Coast of Britain? (Mandelbrot 1967) aroused public
interests in fractal geometry in nature. Later in 1985, he published the remarkable
work in New Scientist: Fractals—Geometry Between Dimensions and coined the
term fractals and illustrated his mathematical definition with striking computer-
constructed visualizations—the Mandelbrot set (Mandelbrot 1982).

(a) Koch snowflake fractals (b) Julia set fractals

Fig. 8.9 Koch snowflake and Julia set fractals

8.3 Fractals in Quantum Finance 223

Figure 8.9a and b illustrates the Koch snowflake fractals and Julia set fractals,
respectively.

8.3.3 Self-similarity in Mandelbrot Set Fractals

Four main properties of fractals are as follows:

• self-similarity;
• simple but non-differentiable;
• fractal dimension; and
• one-to-many, many-to-one.
Figure 8.10a–c shows the self-similarity of Mandelbrot set fractals under different
scales.
The Mandelbrot set is the set of c values in the complex plane for which the orbit
of z 0 = 0 under iteration of the quadratic map remain bounded:

z n+1 = z n + c, c ∈ M ⇔ lim supn→∞ |z n+1 | ≤ 2 (8.6)

That is, a complex number c is part of the Mandelbrot set if and only if when
starting with z0 = 0, and applying the iteration repeatedly, the absolute value of zn
remains bounded by a number, say, 2.
If c = 1, it will give the sequence 0, 1, 2, 5, 26, …, which tends to infinity. As this
sequence is unbounded, 1 is not an element of the Mandelbrot set.
If c = −1 gives the sequence 0, −1, 0, −1, 0, …, which is bounded, and so −1
belongs to the Mandelbrot set.
By using this simple method, we can construct 2D Mandelbrot fractals by plot-
ting all possible c with x–y-axis as the real–imaginary components (1 belongs to
Mandelbrot, whereas 0 does not belong).

8.3.4 Fractal Dimension

A fractal dimension is a ratio providing a statistical index of complexity comparing

how detail in a pattern (fractal pattern) changes with the scale at which it is measured.
Counter-intuitive is that a fractal dimension does not have to be an integer.
The essential idea of fractured dimensions has a long history in mathematics, but
the term itself was brought to the fore by Professor Benoit Mandelbrot based on his
1967 paper on self-similarity in which he discussed fractional dimensions.
The concept of a fractal dimension rests in unconventional views of scaling and
dimension.
For example, in traditional (discrete) dimension, given N is the number of sticks,
ε is the scaling factor, and D is the dimension, using box-counting technique, we
have (Fig. 8.11)
224 8 Chaos and Fractals in Quantum Finance

(b) Mandelbrot set x1 (c) Mandelbrot set x6

(a) Mandelbrot set (full scale)

Fig. 8.10 Mandelbrot set fractals (self-similarity under different scales)

1
N= or D = − log(N )/log(ε) (8.7)
εD
As shown in Fig. 8.12, when
ε = 1/3, N = 27 => D = 3.
Using this method to calculate the fractal dimension of Koch fractal, we have
For i = 2:
Dkoch = −log(ε)
log(N )
= −log(4)
log( 13 )
= 1.261859507
−log(16)
Similarly, Dkoch = log 1 = 1.261859507
(9)
8.3 Fractals in Quantum Finance 225

(a) Discrete Dimension (b) Fractal Dimension

Fig. 8.11 Discrete versus fractal dimension

8.3.5 Natural Phenomena with Fractal Features

How many natural phenomena can be found in fractals?

The answer is … many.
It is similar to the golden ratio on computational finance mentioned in Chap. 6,
fractals can be found anywhere in Mother Nature.

• Algae • Blood vessels • Brain waves

• Brownian motions • Cloud formation • Corals
• Coastlines • DNA and RNA • Crystals
• Financial patterns • Heartbeats • Lightning bolts
• River networks • Mountain ranges • Sand and soil
• Seashells • Snowflakes • Saturn rings
• Trees and branches • Turbulent flows • Tornadoes and hurricanes

Why?
The answer is obvious, if we are clever enough ….
Let’s come back to our old friend … quantum theory and subatomic world
Such one-to-many and many-to-one phenomena not only trigger modern graphics
and CGI such as the fantasy in sci-fi motion pictures, but also …
Worlds of simulation ….!
226 8 Chaos and Fractals in Quantum Finance

Snowflakes Blood vessels

Landscape Saturn ring

Trees Sea shell

Corals Cloud formation

Financial patterns Lightning bolts

Fig. 8.12 Fractals in nature (Tuchong 2019a, b, c, d, e, f, g, h, i, j)

8.4 Applications of Chaos Theory and Fractals in Quantum Finance 227

Fig. 8.13 System architecture of CDNOMS chaotic deep supervised-learning (CDSL) network for
time series financial prediction

8.4 Applications of Chaos Theory and Fractals in Quantum

Finance

8.4.1 Chaos Theory in Quantum Finance

Owing to the chaotic property of financial market, chaos theory becomes one of the
hottest topics in financial engineering, especially on financial prediction.
In the second part of this book, we will study quantum finance application on
worldwide financial prediction, explore how to integrate QPL; chaos theory, and
neural oscillators for real-time financial prediction.
Figure 8.13 shows a snapshot of the chaotic deep neural network for time series
financial prediction.

8.4.2 Fractal Finance

Self-similarity and recursive property of fractals led to active study application in

finance engineering since late 1980s.
As said, although fractals can generate very complex pattern, the interesting thing
is, it can be (should be) originated from very simple pattern (equation).
Let’s review a simple example of how fractals can be applied to technical finance.
In the most fundamental case, basic financial fractals are composed of five or
more bars as shown in Fig. 8.14.
The rules for identifying fractals are as follows:
228 8 Chaos and Fractals in Quantum Finance

Fig. 8.14 Bearish versus

bullish fractals

1. A bearish turning point occurs when there is a pattern with the highest high in
the middle and two lower highs on each side.
2. A bullish turning point occurs when there is a pattern with the lowest low in the
middle and two higher lows on each side.
Actually, in standard MT platform, MQL already provides such indicator iFractals
for real-time detection of bearish/bullish fractals.
Fractals are best used in conjunction with other indicators or forms of analysis
such as alligators or MAs.
Figure 8.15 shows a long-term uptrend with price staying predominantly above
the alligator’s teeth (middle moving average). Since the trend is up, bullish signals
can be used to generate buy signals.
In addition to identify buying and selling signals, can we use fractals to forecast
financial market, say, stock markets?
The answer is yes!

Fig. 8.15 Fractals in financial pattern

8.4 Applications of Chaos Theory and Fractals in Quantum Finance 229

Current research in financial forecast using fractals in two directions:

1. Integration of fractals (as technical signals) with chaotic neural network for finan-
cial prediction and
2. Simulation of multidimensional market pattern by fractal equations, if one may
do this. We can probably predict the future. Why?

8.5 Conclusion

In this chapter, we have learnt chaos and fractals—the two contemporary mathemat-
ical models with vast application on nonlinear dynamics and financial engineering.
As mentioned, chaos and fractals are sides of the same coin. They show us two
ways of looking (modeling) the world’s dynamics and all physical phenomena.
Just like the particle–wave duality in quantum theory, we are talking about the
same scenario in different (mathematical) languages.
But if we look at them from different perspective, chaos and fractals, in fact,
can be considered as mathematical (and hence computational) way to describe the
particle–wave duality in terms of nonlinear equations, which is totally coherent with
our computational model of quantum finance and price level (QPL) evaluation to
form an important component of quantum finance.
In fact, the entire puzzle of quantum finance is almost complete, but with one
piece of component—how to integrate chaos theory into neural network for chaotic
modeling of financial time series?
In next chapter, we will study the last component of quantum finance—chaotic
neural oscillator.

Problems

8.1 Why scientists always say chaos and fractals are two sides of the same coin?
What is the logic behind it? Use two financial markets as examples to support
your explanation.
8.2 What is the meaning of deterministic chaos? Is it equivalent to random event?
Why or Why not?
8.3 What kind of systems or phenomena in nature are categories as chaotic sys-
tems? State five examples and explain why they are considered as chaotic
systems.
8.4 What is the relationship between chaotic flow versus chaotic map?
8.5 What is butterfly effect in chaos theory? Can we find any examples of such
phenomena in finance industry? Please give three typical examples of finance
and explain how they work.
8.6 Programing Exercise I—Lorenz attractor
230 8 Chaos and Fractals in Quantum Finance

(i) Write a MATLAB program to implement the Lorenz attractor by using

the three Lorenz Eqs. (8.3a–8.3c);
(ii) Plot the 3D plot of the butterfly effect, as below.
(iii) What are the key parameters and their values in Lorenz equation in order
to generate the 3D chart of butterfly effect in chaos?
(iv) Explain why such findings are important in the modeling of complex
systems such as the modeling of weather or financial markets.

8.7 State and explain the four major characteristics of chaotic systems. Use finan-
cial market such as forex market to explain these four characteristics in finan-
cial engineering.
8.8 Programming exercise II—sensitivity to initial conditions of Lorenz system
(i) First, implement the MATLAB system in 8.6.
(ii) Make a very minor change of initial condition x say (a minor difference
of 0.00001 of x), plot the result with the x-axis as time step and Y-axis
as future value. See what happen.
(iii) Change the initial conditions of y and z and check the results.
(iv) Any interesting findings and explain why.
8.9 What is bifurcation phenomenon in chaos theory? What is the importance of
bifurcation in a complex system?
8.10 Programming exercise III—logistic equation
8.5 Conclusion 231

(i) Write a MATLAB program to implement the logistic map using Eq. (8.4).
(ii) Plot the logistic map for x between 0 and 4.0, see whether you can obtain
the result below.

(iii) Locate all the bifurcation zones.

(iv) What is the implication and importance for the discovery of chaotic
phenomena and bifurcation zones in this simple logistic map equation?
8.11 State any five applications of chaos theory in consumer electronics and health
care. Explain how they work.
8.12 State three applications of chaos theory in finance industry and explain how
they work.
8.13 What is the relationship between chaos theory and Heisenberg’s uncertainty
principle in quantum theory? And discuss how such findings are important to
quantum finance.
8.14 What are fractals? State any five examples of fractals in nature.
8.15 Programming Exercise III—fractals
(i) Write MATLAB programs to implement: (1) Koch snowflake fractals;
(2) Julia set fractals; and (3) Mandelbrot set fractals.
(ii) What are the main properties and characteristics of fractals?
(iii) Calculate the fractal dimension for these three fractals.
8.16 Programming exercise IV—fractals in MQL
(i) State and explain fractals in MQL.
(ii) In order to implement fractals in MQL, at least one other technical indi-
cator is needed. What is this technical indicator? How it works.
(iii) Write a simple trading program using MQL on three financial products:
(1) Dow Jones Index (DJI); forex product such as AUDCAD; gold or
silver (i.e., XAUUSD or XAGUSD).
(iv) Compare their trading performances.
232 8 Chaos and Fractals in Quantum Finance

(v) State and design how can we integrate quantum finance with this fractal-
based trading program.
(vi) Implement this hybrid system and compare it with fractal-based trading
system.

References

Chau, K.T. & Wang, Z. (2011) Application of Chaotic Motion in Chaos in Electric Drive Systems.
Wiley, Singapore, 283–314.
Davis, P. (1990) Application of optical chaos to temporal pattern search in a nonlinear optical
resonator. Japanese Journal of Applied Physics. 29(7): 1228–1240.
Freeman, W. J. (2001) How brains make up their mind. Columbia University Press.
Heisenberg, W. (2007) Physics and Philosophy: The Revolution in Modern Science. Harper Peren-
nial Modern Classics.
Kapitaniak, T. (2000) Chaos for engineers: theory, applications, and control. Springer.
Katayama, R. (1993) Developing tools and methods for applications incorporating neuro, fuzzy and
chaos technology. Computers and Industrial Engineering. 24(4): 579–592.
Krishnamurthy, V. (2015) Edward Norton Lorenz: Discoverer of chaos. Resonance. 20(3): 191–197.
Kazem, A. et al. (2013) Support vector regression with chaos-based firefly algorithm for stock
market price forecasting. Applied Soft Computing Journal. 13(2): 947–958.
Lee, R. S. T. (2006) LEE-Associator – A Transient Chaotic Autoassociative Network for Progressive
Memory Recalling. Neural Networks. 19(5): 644–666.
Lee, R. S. T. (2005) Advanced Paradigms in Artificial Intelligence From Neural Oscillators, Chaos
Theory to Chaotic Neural Networks. Advanced Knowledge International, Australia.
Lorenz, E. (1995) The Essence of Chaos. University of Washington Press.
Lorenz, E. (1963) Deterministic nonperiodic flow. Journal of the Atmospheric Sciences. 20:
130–141.
Mandelbrot, B. (1997) Fractals and scaling in finance: discontinuity and concentration, Springer.
Mandelbrot, B. (1982) The fractal geometry of nature. Times Books.
Mandelbrot, B. (1967) How Long Is the Coast of Britain? Statistical Self-Similarity and Fractional
Dimension. Science, New Series. 156(3775): 636–638.
Munakata, T. (1998) Fundamentals of the new artificial intelligence: beyond traditional paradigm.
Springer.
Peters, E. E. (1994) Fractal market analysis: applying chaos theory to investment and economics.
Wiley.
Sambas, A. et al. (2016) A 3-D novel jerk chaotic system and its application in secure communication
system and mobile robot navigation. Studies in Computational Intelligence. 636: 283–310.
Soriano-Sanchez et al. (2018) Synchronization and FPGA realization of complex networks with
fractional–order Liu chaotic oscillators. Applied Mathematics and Computation. 332: 250–262.
Schuste, H. G. and Just, W. (2005) Deterministic Chaos: An Introduction. Wiley-VCH.
Tuchong (2019a) Fractals in Nature - Snowflakes. http://stock.tuchong.com/image?imageId=
79052163044817970. Accessed 21 Aug 2019.
Tuchong (2019b) Fractals in Nature - Blood vessels. http://stock.tuchong.com/image?imageId=
452726017391067492. Accessed 21 Aug 2019.
Tuchong (2019c) Fractals in Nature - Landscape. http://stock.tuchong.com/image?imageId=
479820371888701550. Accessed 21 Aug 2019.
Tuchong (2019d) Fractals in Nature - Saturn ring. http://stock.tuchong.com/image?imageId=
429030579742310411. Accessed 21 Aug 2019.
References 233

Tuchong (2019e) Fractals in Nature - Trees. http://stock.tuchong.com/image?imageId=

130438294248095746. Accessed 21 Aug 2019.
Tuchong (2019f) Fractals in Nature - Sea shell. http://stock.tuchong.com/image?imageId=
459360616581955957. Accessed 21 Aug 2019.
Tuchong (2019g) Fractals in Nature - Corals. http://stock.tuchong.com/image?imageId=
411274248616935479. Accessed 21 Aug 2019.
Tuchong (2019h) Fractals in Nature - Cloud formation. http://stock.tuchong.com/image?imageId=
238389471191761069. Accessed 21 Aug 2019.
Tuchong (2019i) Fractals in Nature - Financial patterns. http://stock.tuchong.com/image?imageId=
481643508196442202. Accessed 21 Aug 2019.
Tuchong (2019j) Fractals in Nature - Lightning bolts. http://stock.tuchong.com/image?imageId=
540477009592057896. Accessed 21 Aug 2019.
Chapter 9
Chaotic Neural Networks in Quantum
Finance

If we look at the way the universe behaves, quantum mechanics

gives us fundamental, unavoidable indeterminacy, so that
alternative histories of the universe can be assigned probability.
Murray Gell-Mann (1929–2019)

What would happen if time stopped?

Would they like movies and TV shows: everything freezes when time stopped?
In 1952, Professor Erwin Schrödinger addressed at a lecture with a witty remark to
his audience what he was about to say might “seem lunatic” (Fig. 9.1).
He stated that when his equations seemed to narrate several different histories,
these were not alternatives, but all really happen simultaneously—as we are now so-
called parallel universe/multiverse theory. The author was really and truly, enticed
by this statement to take up theoretical physics especially on quantum mechanics
and general relativity courses at Hong Kong University in 1986–89.
Current cosmology (superstring and M-theory) even tells us that we live in one
of the many so-called realities, as Professor Murray Gell-Mann, Nobel Prize winner
for physics in 1969 mentioned in his famous quotation.
More importantly, is that all of these are coexisting realities and vibrating in
different frequencies.
If we are clever enough, we should realize why Professor Schrödinger said so.
If it is true. If the entire universe with everything inside it is under continuous
vibrations and oscillations in every single moment. If why we can percept the real-
ity and our surroundings is because of these continuous vibrations and oscillations
between our sensory organs and the external world—which means when time is really
stopped, everything is stopped; or should it be rather disappeared and/or cannot be
precepted instead?
In this chapter, we will study an interesting topic—chaotic neural oscillators in
quantum finance. First, we will explore the quantum world of chaotic oscillation from
the multiverse in quantum theory, chaos in the brain to our sensory of oscillations.
Next, we will review chaotic neural oscillators, neuroscience motivation, Wang-

© Springer Nature Singapore Pte Ltd. 2020 235

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_9
236 9 Chaotic Neural Networks in Quantum Finance

Fig. 9.1 What would

happen if time stopped?
(Tuchong 2019a)

oscillators, and the author’s latest research on Lee-oscillators; its neural dynamics and
applications. After that, we will study quantum finance oscillators (QFO) using Lee-
oscillators and different application of QFO in quantum finance including quantum
financial prediction using chaotic neural networks, chaotic deep neural networks,
and chaotic intelligent multiagent-based trading systems.

9.1 Quantum World of Chaotic Oscillation

9.1.1 Multiverse and Quantum Theory

Cosmologist Professor Max Tegmark has provided a taxonomy of universes beyond

the familiar observable universe (Tegmark 2014). In Tegmark’s chaotic inflation
theory—bubble universes in Fig. 9.2—every disk represents a bubble universe.
According to Tegmark’s bubble universe theory, our universe is just one of the
many disks of universes. Universes 1–6 represent bubble universes. Five of them
have different physical constants than our universe.
The four levels of the multiverse are the following:
9.1 Quantum World of Chaotic Oscillation 237

Fig. 9.2 Illustration of

Tegmark’s bubble universe The Universe we live

Universe 6
Universe 5

Universe 4
Universe 3
Universe 2

Universe 1

1. Level I: an extension of our universe

2. Level II: universes with different physical constants
3. Level III: many-worlds interpretation of quantum mechanics
4. Level IV: ultimate ensemble.
Related to the many-worlds idea were originated from Professor Richard Feyn-
man’s multiple histories theory, who used the theory to explain the existence of
infinite path integrals of realities.

9.1.2 Chaos in Brain—Brainwave

Current neuroscience tells us that oscillations can not only be found in the physical
world but also can be found at the core of our body—our brain.
If one takes a closer look at neural dynamics in our brain—the original of neural
networks: What triggers the entire neural dynamics?
The stimulus, maybe, but how?
Current neurophysiology also tells us that all neurons in our brain exist in the state
of oscillations, chaotic oscillation to be exact.
It is that such oscillations trigger the entire thinking and learning processes.
The oscillations of these neurons within our brain are collectively called brain
waves in layman’s terms as shown in Fig. 9.3.
• Awake, conscious state—beta wave
• Insight and peak focus conscious state—gamma wave
• Relaxed, calm, lucid dream (but not thinking)—alpha wave
• REM dream state, deep relaxation and mediation—theta wave
• Deep, dreamless sleep—delta wave.
238 9 Chaotic Neural Networks in Quantum Finance

Fig. 9.3 Illustration of human brain waves patterns

In fact, not only from our brain, chaotic oscillations can be easily found in the
human body including our five senses:
• eye (sight)—visual nerves;
• nose (smell)—olfactory nerves;
• ear (sound)—auditory nerves;
• tongue (taste)—gustatory nerves;
• skin (touch)—somatosensory nerves;
• our heartbeats and others.

9.2 Chaotic Neural Oscillators

9.2.1 An Introduction

We learnt from Chap. 7 that classical artificial neural networks, ANN in Fig. 9.4 are
composed of simple artificial neurons that emulate biological neural activities.
These kinds of neural models have been strongly criticized as being far simpler
than real neural models.
The latest studies in neuroscience and neurophysiology have examined such issues
as functional properties of the hippocampus, neural activities in pyloric CPG (central
pattern generator) of the lobster, and other human brain activities. For instance,
hippocampus—a major component of the brain as shown in Fig. 9.5, plays important
9.2 Chaotic Neural Oscillators 239

Fig. 9.4 Artificial neural networks (Tuchong 2019b)

Fig. 9.5 Hippocampus of human brain

240 9 Chaotic Neural Networks in Quantum Finance

roles in the consolidation of information from short-term to long-term memory, and

in spatial memory that enables navigation.
In Alzheimer’s disease, the hippocampus is one of the first regions of the brain to
suffer damage with short-term memory loss and disorientation. Studies have provided
strong evidence of chaotic neural activities in these complex neural behaviors.
Researchers have also proposed various chaotic neural models over the past
decades. Latest research include: chaotic oscillators proposed by Falcke et al. (2000)
to model pyloric CPG neurons, cortical networks proposed by Hoshino et al. (2003)
for recalling LTM (long-term memory), TCNN (transient chaotic neural network)
proposed by Chen and Aihara (1995) for handling combinational optimization prob-
lems, and Zhou’s work (Zhou and Chen 2000) based on chaotic annealing technique
for dynamic pattern retrieval.

9.2.2 Wang-Oscillator—The Structure

Most of the contemporary neural oscillators that developed theoretically from

Wilson–Cowan model are focused on the time-continuous framework, which is in
fact too complex for network modeling, let alone for actual computer implementation
on actual applications.
Wang (1991) in his Ph.D. thesis, proposed a simple time-discrete neural oscillator
model (namely, the Wang-oscillator).
Figure 9.6 shows the neural structure of Wang-oscillator.
Unlike its continuous model counterpart, the Wang-oscillator provided simple but
remarkable neural dynamics ranging from fixed points (through quasi-periodicity)
to chaos, as revealed by its bifurcation diagram.
This bifurcation diagram can be used as computational elements (called the
BTU—bifurcation transfer unit) to replace traditional sigmoid function for temporal
information processing.
The Wang-oscillator is a neural oscillatory model consisting of two neurons, one
excitatory and one inhibitory neuron.
Just like biological neurons, the excitatory neurons promote chances of neural
firing while the inhibitory neurons hinder neural activities.
Together, they form a simple but powerful oscillatory couple.
In Wang-oscillator, the time-discrete neural dynamics are given by

E(t + 1) = Sig[ωEE · E(t) − ωEI · I (t) + SE (t) − ξE ] (9.1)

I(t + 1) = Sig[ωIE · E(t) − ωII · I (t) + SI (t) − ξI ] (9.2)

W(t) = E(t) − I (t) (9.3)

9.2 Chaotic Neural Oscillators 241

Fig. 9.6 Wang-oscillator

where SE and SI are input stimulus; ω are weights; ξE and ξI are threshold values;
the sigmoid function Sig() is given by

Sig(k) = 1 1 + e−k (9.4)

One important finding of Wang-oscillator is the bifurcation behavior of Wang-

oscillator with the variation of input stimulus I.

9.2.3 Wang-Oscillator—Bifurcation Diagram

Figure 9.7 shows the bifurcation diagram of Wang-oscillator.

As shown, the bifurcation diagram of a typical Wang-oscillator consists of sigmoid
and bifurcation zones.
The original idea is that the sigmoid zone imitates the classical sigmoid function
and bifurcation zone imitates the chaotic property of a complex system during the
sigmoid-transition period.
However, there are two intrinsic problems:
1. The existence of an extra bifurcation zone I destructs the continuity of sigmoid
curve in the initial period;
2. The bifurcation zone II is “too chaotic” to model the chaotic-transition region in
real-world problems.
242 9 Chaotic Neural Networks in Quantum Finance

Fig. 9.7 Bifurcation diagram of Wang-oscillator

A more transient progressive growth in terms of neural dynamics is needed to be

achieved.

9.2.4 Wang-Oscillator—Major Contributions and Limitations

One important finding and contribution of Wang-oscillator is the property of change

in neural dynamics according to the input stimulus.
In most classical neural network models for information encoding and associa-
tion (e.g., Hopfield Network), the network model can be generalized as a nonlinear
function operator which, based on various input stimuli, alters its internal states and
fires output according to the nonlinear transfer function.
The Wang-oscillator, on the other hand, encodes information (i.e., the input stim-
ulus) and gives responses by altering the behavior of neural dynamics (from chaotic
states to sigmoid growth), which is totally consistent with the latest findings on how
the brain processes information (Freeman 2000, 2008, 2013).
9.2 Chaotic Neural Oscillators 243

However, the Wang-oscillator has several major limitations in the bifurcation

behavior of the model, preventing it from being used as an effective BTU or temporal
information processing model.
As mentioned previously, the undesirable chaotic region exists in bifurcation
zone I which not only affects the continuity of the BTU but also violates the original
function of BTU, namely that of stimulating the temporal information processing for
the brain.
So, we need to look for a total solution.

9.3 Lee-Oscillator

9.3.1 The Motivation

After over 5 years of research and testing, the author created an ideal chaotic neural
oscillator, now known as Lee-oscillator in 2004 and published in two major neural
network journals IEEE Transactions on Neural Networks (Lee 2004) for the intro-
duction of Lee-oscillator and its transient chaotic behavior; and Journal of Neural
Networks (Lee 2006) for the exploration of Lee-oscillator on progressive memory
recalling capability in visual psychology.
Different from Wang-oscillator, Lee-oscillator successfully emulates the transient
chaotic progressive growth in its neural dynamics, which helps to shed light on acting
as a perfect CTU to model complex and chaotic problems.
Figure 9.8 shows the neural structure of Lee-oscillator.
Basically, Lee-oscillator consists of 4 neural elements: E, I, , and L, which
corresponds to excitatory, inhibitory, input, and output neurons.
The neural dynamics are given by

E(t + 1) = Sig[e1 · E(t) − e2 · I (t) + S(t) − ξE ] (9.5)

I(t + 1) = Sig[i1 · E(t) − i2 · I (t) − ξI ] (9.6)

(t + 1) = Sig[S(t)] (9.7)

L(t) = [E(t) − I (t)] · e−kS (t)

2
+ (t) (9.8)

where e1 , e2 , i1, and i2 are weights; ξE and ξI are threshold values and S(t) is an
external input stimulus.
The bifurcation diagram of Lee-oscillator is illustrated in Fig. 9.9.
244 9 Chaotic Neural Networks in Quantum Finance

Fig. 9.8 Lee-oscillator

9.3.2 Lee-Oscillator—Bifurcation Diagram

Unlike Wang-oscillator, the Lee-oscillator bifurcation diagram in Fig. 9.9

is composed of three main regions, sigmoid zones I and II and single
bifurcation zone I.
From neural dynamics point of view, sigmoid zones I and II denote sigmoid-
shape region, which corresponds to nonchaotic neural activities in the oscillators;
and bifurcation zone I is hysteresis region; which corresponds to the area of chaotic
behavior that results when a weak external input stimulus is received.
Such neural dynamics can be perfectly served as chaotic transfer unit (CTU)
to model complex and chaotic systems such as severe weather prediction, complex
scene analysis, and real-time financial prediction such as quantum finance.
9.3 Lee-Oscillator 245

Fig. 9.9 Bifurcation diagram of Lee-oscillator

9.3.3 Lee-Oscillator—Potential Applications

Potential applications of Lee-oscillator include the following:

1. Basic chaotic neural elements for temporal information processing

• Fundamentally, information processing is the overall process of encoding,

recognizing, and discriminating various types of information (e.g., images,
patterns, etc.).
• Since a single Lee-oscillator can provide a two-state transient chaotic attraction
from input space, it is suitable to serve as the basic element for the information
processing unit for quantum computing.
• In addition, owing to its unique chaotic features where neural dynamics change
with variations in the input signal, the Lee-oscillator provides a good analog
for simulating chaotic and temporal information processing behavior in brain
science (Freeman 2000, 2008; Malsburg 1995; Fernandes and Malsburg 2015).

2. Transient chaotic auto-associator

246 9 Chaotic Neural Networks in Quantum Finance

• As an extension and generalization of using Lee-oscillators for information

processing, a two-dimensional (2D) layer of Lee-oscillators can be adopted as
a pattern associator (Lee 2006).
• As an analog of the classical Hopfield Network (Hopfield 1984) as an auto-
associator, a transient chaotic autoassociative network based on Lee-oscillators
as its constituting neuron elements can be used to provide an innovative pro-
gressive memory association and recalling scheme (Lee 2004; 2006) to expand
how human can recall memory and to explain an important phenomena—
Gestalt psychology.

3. Chaotic neural oscillatory units for advanced applications

• In fact, Lee-oscillators can be adopted and integrated with each other to form
a complex chaotic neural oscillatory model to tackle problems such as scene
analyses, robot vision, navigation, and of course, real-time financial prediction
in quantum finance.
• We will study this in detail in Part II of this book.

9.4 Quantum Finance Oscillators (QFO)

9.4.1 The Concept

In Chap. 2, quantum field theory, we mentioned that the quantum world can be mod-
eled as quantum oscillators, influenced and oscillating like a ball over a trampoline
as shown in Fig. 9.10a.
So, in terms of quantum finance, it is natural to analog the dynamic of every
financial market as quantum oscillations of quantum financial particles influenced
by quantum energy (price) field as shown in Fig. 9.10b.
But instead of simple quantum harmonic oscillators, we are talking about quantum
anharmonic oscillator counterparts.
Such notion, in fact, is totally equivalent and coherent with our modeling of
quantum dynamics as quantum anharmonic oscillations studied in Chap. 4.
These quantum anharmonic oscillators are known as quantum finance oscillators
(OFO).

9.4.2 The Model

One of the most direct adoptions of the chaotic oscillator is the introduction of
quantum finance oscillator—QFO by the replacement of Lee-oscillator with financial
particle.
9.4 Quantum Finance Oscillators (QFO) 247

(a) Quantum price oscillator

(b) Quantum energy field of quantum finance particles

Fig. 9.10 Quantum finance oscillators (QFO)

Figure 9.11 shows a typical USDX (USD Index) QFO model in the forex market.
The neural dynamics is given by

E(t + 1) = Sig[e1 · E(t) − e2 · I (t) + S(t) − ξE ] (9.5)

I(t + 1) = Sig[i1 · E(t) − i2 · I (t) − ξI ] (9.6)

(t + 1) = Sig[S(t)] (9.7)

QFO|USDX (t) = [E(t) − I (t)] · e−kS (t)

2
+ (t) (9.9)
248 9 Chaotic Neural Networks in Quantum Finance

Fig. 9.11 Quantum finance

oscillator model

By using the same interpretation, practically speaking we can model any financial
market using QFOs and investigate the chaotic oscillation dynamics of these financial
markets.
More importantly, such modeling technique can be integrated with AI tools and
technologies mentioned in Chap. 7 to complex and highly chaotic phenomena, finan-
cial fluctuation in particular.

9.5 Applications of QFO in Quantum Finance

9.5.1 The Concept

In Chaps. 7 and 8, we have learnt 3 basic AI tools/technologies that include the

following:
9.5 Applications of QFO in Quantum Finance 249

• artificial neural network (ANN),

• fuzzy logic and fuzzy logic systems (FLS),
• genetic algorithms (GA),
• chaos theory, and
• fractals.
In fact, QFO (quantum finance oscillators) can be integrated into all these
tools/technologies to tackle highly complex and chaotic phenomena—financial mar-
kets and forex markets in particular.
This section will study some typical QFO applications in quantum finance appli-
cations.
They are the following:
• QFO for real-time financial prediction using chaotic neural networks,
• chaotic fuzzy-neuro oscillators—quantum finance signals (QFS),
• chaotic deep neural networks in quantum finance, and
• chaotic multiagent trading system (Fig. 9.12).

Fuzzy
Logics

Chao c Neural
Networks

QUANTUM
Chaos
PRICE Fractals
Theory FIELD

Neural
Oscillators

Gene c
Algorithms

Fig. 9.12 Concentric sphere model of quantum finance

250 9 Chaotic Neural Networks in Quantum Finance

9.5.2 QFO for Real-Time Financial Prediction Using

Chaotic Neural Networks

As a direct adoption in artificial neural networks, QFO can be directly integrated with
classical neural networks such as feedforward backpropagation networks (FFBPN)
by replacing all simple neurons in input/hidden/output layers by Lee-oscillations.
By doing so, technically speaking, we convert a simple neural network system
into a multilayer chaotic neural network as shown in Fig. 9.13.
In the figure, two inputs types (time series input signals and technical index
signals) are replaced by Lee-oscillators.
From neural dynamics point of view, such changes bring (1) chaotic oscillation
dynamics during network training; (2) chaotic bifurcation transfer function (CBTF)
into forecast system to resolve the network training deadlock and trapped by local
minimum problems which usually occur during network training of highly complex
patterns (e.g., financial patterns) using classical FFPBN.
Detail implementation will be studied in Chap. 11.

Fig. 9.13 QFO for real-time financial prediction using chaotic neural networks
9.5 Applications of QFO in Quantum Finance 251

9.5.3 Chaotic Fuzzy-neuro Oscillators—Quantum Finance

Signals (QFS)

In Chap. 7, we have studied fuzzy logic (FL) and fuzzy logic system (FLS).
A simple fuzzy logic membership function (also known as type-1 fuzzy logic) of
financial technical signal (RSI) is shown in Fig. 9.14a.
Different from type-1 FL, a type-2 FL provides a higher degree of fuzziness by
the adoption of second-tier fuzziness in overlapping regions of the primary fuzzy
membership function.
Figure 9.14b shows a typical so-called interval type-2 MF f or RSI.
By QFO adoption, we technically can generate a chaotic overlapping region for
any technical financial signals—quantum finance signals (QFS).
Figure 9.14c shows a typical chaotic type-2 fuzzy logic MF (membership function)
for RSI QFO.
Detailed implementation will be studied in Chap. 12.

μ (x) μ (x)

Over-Sell Bearish Bullish Over-Buy Over-Sell Bearish Bullish Over-Buy

Type-1 FMF

0 x 100 0 x 100
Type-2 uncertainty regions

(a) Type-1 fuzzy logic MF on RSI (b) Interval type-2 fuzzy logic MF on RSI

(c) Chaotic type-2 fuzzy logic MF on RSI

Fig. 9.14 Chaotic fuzzy-neuro oscillators—quantum finance signals

252 9 Chaotic Neural Networks in Quantum Finance

Input Layer
8-Level Bifurca on Hidden Layers (BHL)

39 Technical Indicators signal Generator Output

Layer
Time Series Input Signals

Next Day [O, H, L, C]

Fig. 9.15 Chaotic deep neural networks in quantum finance

9.5.4 Chaotic Deep Neural Networks in Quantum Finance

Latest neuroscience research discovered that neural dynamics for memory recall and
learning existed as the so-called retrograding signaling (RS), a kind of delayed and
feedback mechanisms of information processing.
Such an important finding leads to the latest development of LORS (LEE-
oscillators with Retrograde Signals).
Between 2012 and 2017, the author had categorized 8 major types of LORS
bifurcation signals (LORS#0-#7), which corresponded to 8 different kinds of chaotic
bifurcations and neural dynamics.
More importantly, such findings can be adopted to DNN (deep neural network)
technology for the conversion of single hidden layer FFBPN into CDNN (chaotic
deep neural networks) with 8-level bifurcation hidden layers (BHL).
Figure 9.15 shows the system architecture of CDNN on quantum finance. Detail
implementation will be studied in Chap. 12.

9.5.5 Chaotic Multiagent Trading System

Neural networks can be used in supervised learning (SL) such as financial prediction,
it can also be used on unsupervised learning (UL) and reinforcement learning (RL)
as we learnt from Chap. 7.
9.5 Applications of QFO in Quantum Finance 253

A multiagent financial trading system is a typical reinforcement-learning problem

in which there are no well-defined output/target pairs in a typical SL system, e.g.,
financial forecast.
All we have (and need to do) is to optimize the net-profit (or returns) at (fixed)
period of time and minimize the potential risks.
In that case, we can make use of reinforcement-learning (RL) neural networks
such as actor-critic (AC) RBF-based RLN.
Again, by replacing simple neurons in the RBF-layer of the AC-RBFN, we tech-
nically convert this RLN into chaotic AC-RBFN which can significantly improve the
RL efficiency.
Figure 9.16 shows the system architecture of chaotic AC RBF-based RLN.
Detail implementation will be studied in Chap. 13.

Fig. 9.16 Chaotic multiagent-based trading system

254 9 Chaotic Neural Networks in Quantum Finance

9.6 Conclusion

In this chapter, we have learnt the basic theory of chaotic neural oscillators.
Also, we have learnt how to integrate chaotic neural oscillators (Lee-oscillator) to
effectively convert a typical neural network into chaotic neural networks to handle
highly complex and chaotic phenomena—quantum finance in particular.
As we can see, chaotic neural oscillators can be adopted into almost any AI tools
and technologies to improve efficiency and performance.
How can we do that?
As mentioned, all these various kinds of AI tools and technologies are just different
tools and methods to model and handle real-world problems, finance in our case,
which is highly complex and chaotic in nature.
More importantly, when we mentioned a (financial) market is highly chaotic or
fuzzy in nature, we are just seeing (modeling in technical perspective) the problem
with a particular perspective, the problem never changes itself.
Just like we are talking about the particle–wave duality property of financial
market in quantum finance.
The financial market always as it is, we are just using different angles and per-
spectives to observe and investigate it.
In Part II of this book, we will study four major applications of quantum finance
in financial prediction and intelligent trading systems.

Problems

9.1 Discuss and explain why Schrödinger in his 1952 lecture interpreted the quan-
tum events such as Schrödinger’s cat as different coexisting realities instead
of different alternative events, and what such notion is critical and related to
Heisenberg’s uncertainty principle.
9.2 Based on the brain waves shown in Fig. 9.3, please discuss and explain why our
perception of time is different in various brain wave states, and how it can be
related to the understanding of time in: (1) human perception; and (2) financial
markets.
9.3 If all human sensory schemes are products of sensory oscillators and their inter-
actions with neural networks in the brain, is it possible to have what we called
sensory interchange in theoretical perspective? And what is its implication?
9.4 What is the importance of converting a traditional artificial neural network
(ANN) into chaotic neural networks in terms of (1) network training; and (2)
network performance ?
9.6 Conclusion 255

9.5 What are the physical evidence and explanations for the existence of chaotic
neural dynamics in humans? And how such discoveries are vital to the future
development of AI and intelligent systems?
9.6 Based on Eqs. (9.1)–(9.4) and Eqs. (9.5)–(9.8), write two MATLAB programs to
plot the normalized bifurcation diagrams of Wang-oscillator and Lee-oscillator
(i.e., input and output in the range of [0 … 1]) and discuss how Lee-oscillator
can be used to convert ANY artificial neural networks (ANN) or recurrent neural
networks (RNN) into chaotic neural networks (CNN).
9.7 In both Wang-oscillator and Lee-oscillator, there are excitatory and inhibitory
neurons in their neural models. Discuss and explain the biological and math-
ematical meanings of these neurons in a typical chaotic neural network, as
compared with traditional ANN.
9.8 For the modeling of forex market (say AUDCAD) using quantum finance oscil-
lator (QFO):

(i) Draw the quantum finance oscillator model of QFO|AUDCAD.

(ii) Write down the formulation of QFO|AUDCAD.
(iii) What are the physical meanings of excitatory (E), inhibitory (I), and input
stimulus (S) in QFO|AUDCAD model?
(iv) Discuss how QFO|AUDCAD can be integrated with chaotic neural networks
for (1) short-term forex price prediction; (2) long-term trend pattern recog-
nition.

9.9 Programming Exercise I (bifurcation diagrams of Lee-oscillator and Wang-

oscillator)
Below MATLAB program implements the Wang-oscillator and generates its bifur-
cation diagram.
256 9 Chaotic Neural Networks in Quantum Finance

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%
% Wang-Oscillator
%
% Created By Dr. Raymond Lee on May 2019
%
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Define parameters
%
N = 600; % n = no. of time step default is 600
s = 5; % parameter for tanh function
a1 = 5; % default is 5
a2 = 5;
b1 = 1; % default is 5
b2 = 1;
eu = 0; % u threshold default is 0
ev = 0; % v threshold default is 0
e = 0.02;

%%%%%%%%%%
% Define and initialize u(t) and v(t)
%
u = zeros(1,N);
v = zeros(1,N);
z = zeros(1,N);
z(1) = 0.2; % IC of z set to 0.2
u(1) = 0.2;

hold on;

for i=-1:0.001:1 % 1000 i values

valueOf_stimulus_i = sprintf('%0.5g',i)
sim = i + e*sign(i);
for t = 1:N-1
tempu = a1*u(t) - a2*v(t) + sim - eu;
tempv = b1*u(t) - b2*v(t) - ev;

u(t+1) = (exp(s*tempu) - exp((-1)*s*tempu))

/(exp(s*tempu) + exp((-1)*s*tempu));
v(t+1) = (exp(s*tempv) - exp((-1)*s*tempv))
/(exp(s*tempv) + exp((-1)*s*tempv));
z(t+1) = u(t+1) - v(t+1);

if (t > 500)
plot(i, z(t+1));
end
end
end
9.6 Conclusion 257

(i) Copy this program script into MATLAB and run the program, you should see
the bifurcation diagram as above.
(ii) As shown in the MATLAB code, we only plot the output z when t > 500. Why?
(iii) Modify this MATLAB to implement the Lee-oscillator using Eqs. (9.5)–(9.8).
You should obtain the below bifurcation diagram.

(iv) From the bifurcation diagram of Lee-oscillator, identify the sigmoid zone and
the bifurcation zone.
258 9 Chaotic Neural Networks in Quantum Finance

(v) As shown in the bifurcation diagram, the x-axis and y-axis are both between
[−1 … +1]. Modify the MATLAB program to convert the bifurcation diagram
of Lee-oscillator to Lee chaotic transfer function (LCTF) with normalized value
between [0 … 1], as shown in the below figure.

(vi) Modify the MATLAB program by creating a 2-D array Lee() which contains
the output values of Lee-oscillator for 1000 values of i between [1 … 1] and
100 time-step values of Lee-oscillator between 500–600. The definition of
Lee() array as shown below:
Lee = zeros(1001,100);
(vii) Store the LCTF-Lee() table array into a CSV data file for later use.
9.10 Programming exercise II
(i) Start with 7.28 with the implementation of MQL-based FFPBN for the
next-day forecast of at least 5 Forex products (e.g., AUDCAD, USD-
CAD, GBPUSD, USDJPY, EURUSD, etc.).
(ii) Implement the MATLAB program in 9.16 to obtain the CSV datafile of
LCTF (Lee chaotic transfer function).
(iii) Modify the MQL FFBPN program in (i) by: (1) Read the LCTF array in
the program initialization process; (2) Replace sigmoid function for the
FFPBN program with the LCTF function array.
(iv) Compare these two neural networks in terms of: (1) training perfor-
mance; and (2) forecast performance using five forex products mentioned
in (i) by plotting the following charts:
• Mean RMSE (root-mean-square error) in network training for 1000
iterations;
9.6 Conclusion 259

• Mean standard deviation of RMSE in network training for 1000 iter-

ations;
• Mean forecast % Error for a test run (or simulation) of 100 trading
days.
• Mean standard deviation of forecast % Error for a test run (or simula-
tion) of 100 trading days.

References

Chen, L. and Aihara, K. (1995) Chaotic stimulated annealing by a neural model with transient
chaos. Neural Networks. 8(6): 915–930.
Falcke, M. et al. (2000) Modeling observed chaotic oscillators in bursting neurons: The role of
calcium dynamics and IP3. Biol. Cybern. 82: 517–527.
Fernandes, T. and Malsburg, C. (2015) Self-organization of control circuits for invariant fiber
projections. Neural Computation. 27(5): 1005–1032.
Freeman, W. J. (2013) Imaging Brain Function With EEG: Advanced Temporal and Spatial Analysis
of Electroencephalographic Signals. Springer.
Freeman, W. J. (2008) A pseudo-equilibrium thermodynamic model of information processing in
nonlinear brain dynamics. Neural Network. 21(2–3): 257–265.
Freeman, W. J. (2000) Neurodynamics: An Exploration in Mesoscopic Brain Dynamics (Perspec-
tives in Neural Computing). Springer.
Hopfield, J. J. (1984) Neurons with Graded Response Have Collective Computational Properties
like Those of Two-State Neurons. Proceedings of the National Academy of Sciences of the United
States of America. 81(10): 3088–3092.
Hoshino, O. et al. (2003) Roles of dynamic linkage of stable attractors across cortical networks in
recalling long-term memory. Biol. Cybern., 88(3): 163–176.
Lee, R. S. T. (2006) LEE-Associator – A Transient Chaotic Autoassociative Network for Progressive
Memory Recalling. Neural Networks. 19(5): 644–666.
Lee, R. S. T. (2004) A Transient-chaotic Auto-associative Network (TCAN) based on LEE-
oscillators. IEEE Trans. Neural Networks. 15(5): 1228–1243.
Malsburg, C. (1995) Binding in models of perception and brain function. Current Opinion in
Neurobiology. 5(4): 520–526.
Tegmark, M. (2014) Our Mathematical Universe: My Quest for the Ultimate Nature of Reality.
Vintage.
Tuchong (2019a) What would happen if time stopped? http://stock.tuchong.com/image?imageId=
446677423538045208. Accessed 21 Aug 2019.
Tuchong (2019b) Artificial Neural Networks. http://stock.tuchong.com/image?imageId=
492056055110172742. Accessed 21 Aug 2019.
Wang, X. (1991) Period-doublings to chaos in a simple neural network: An analytic proof. Complex
Syst. 5: 425–441.
Zhou, C. and Chen, T. (2000) Chaotic neural networks and chaotic annealing. Neurocomputing.
30(1): 293–300.
 Part II
Quantum Finance Applications
Chapter 10
Quantum Price Levels for Worldwide
Financial Products

The markets represent the aggregate interaction of many

investors. Their attitudes, philosophies, and behavioral patterns
on many levels are predictable… and repetitive.
Rick Santelli (Born in 1956)

Mr. Rick Santelli gave a vivid description of financial markets. As an old saying:
History repeats itself.
As mentioned in Chap. 1, humans are a species of habit. Under the same condi-
tions, we will always respond in the same ways, no matter the consequences are. If
we believe financial markets are the consequences of investors’ collective conscious
for market responses in every single moment. Patterns exist will be a natural cause.
If we believe every single phenomenon (including financial market) happened
among us no matter how chaotic or unpredictable at the very basic and microscopic
level, are all governed by quantum mechanics under the influence of quantum energy
levels generated by their own existence with neighboring quantum particles; thus, the
existence of financial energy levels; or what we called quantum price levels (QPL)
should be a natural cause too.
In Part I of this book, we have studied the basic concept and theory of quantum
finance with methods for the modeling and quantum price levels (QPL) evaluation.
In Part II, this chapter will begin on how to apply the theory and formulation of QPLs
evaluation on worldwide secondary financial markets which include: cryptocurrency,
forex, commodity, and worldwide financial indices.
First, we will introduce financial market types and their importance to the world-
wide economy.
Second, we will give a brief overview of traditional concepts and beliefs of quan-
tized market price levels—the support and resistance (S&R) levels, and how they are
related to our understanding of quantum price levels (QPLs) in quantum finance.
Third, we will introduce the MetaTrader (MT) system and MQL (MetaQuotes
Language)—the real-time worldwide financial program trading and system devel-
opment platform.
After that, we will study the system architecture of QPL evaluation system and
introduce two major MT4 platforms: Forex.com (one of the biggest worldwide forex

© Springer Nature Singapore Pte Ltd. 2020 263

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_10
264 10 Quantum Price Levels for Worldwide Financial Products

trading platforms) and AvaTrade.com (one of the major cryptocurrency trading plat-
forms).
For the system implementation section, we will study in detail the MT4 program on
automatic evaluation of worldwide 120+ financial products, based on the numerical
formulations revealed in Chap. 5.
In the last section, we will study QPL system implementation results; conclude its
related works including the design; implementation of real-time financial prediction,
and multi-agent-based trading system in Chaps. 11–13.

10.1 Financial Markets

10.1.1 An Overview

A financial market is a broad term describing any marketplace where trading of

securities including equities, bonds, currencies, and derivatives occurs.
Some financial markets are small scale with little activities, while some financial
markets like the New York Stock Exchange (NYSE) trade trillions of dollars of
securities daily.
Financial market prices may not indicate the true intrinsic value of stock due to
macroeconomic forces like taxes. In addition, the prices of securities are heavily
reliant on informational transparency to ensure efficient and appropriate prices are
set by the market.
The stock market is a financial market that enables investors to buy and sell shares
of publicly traded companies.
The primary stock market is where new issues of stocks are first offered. Any
subsequent trading of stock securities occurs in the secondary market.
In terms of market pricing—primary market prices are often set beforehand,
while prices in the secondary market are determined by the basic forces of supply
and demand. If the majority of investors believe a stock will increase in value and
rush to buy it, the stock’s price rises typically. If a company loses favor with investors
or fails to post sufficient earnings, its stock price declines as demand for that security
dwindles.
In secondary markets, investors exchange with each other rather than with the
issuing entity. Through a massive series of independent yet interconnected trades, the
secondary market drives the price of securities toward their actual value. Moreover,
the secondary markets create additional economic value by allowing more beneficial
transactions to occur. The net result is that almost all market prices—interest rates,
debt, houses, and the values of businesses and entrepreneurs—are more efficiently
allocated because of secondary market activity.
10.1 Financial Markets 265

10.1.2 Primary Versus Secondary Market

It is important to understand the distinction between the primary market and the
secondary market. When a company issues stock or bonds for the first time and
sells those securities directly to investors, that transaction occurs on the primary
market. Some of the most common and well-publicized primary market transactions
are IPOs, or initial public offerings. During an IPO, a primary market transaction
occurs between the purchasing investor and the investment bank underwriting the
IPO. Any proceeds from the sale of shares of stock on the primary market go to the
company that issued the stock, after accounting for the bank’s administrative fees.
If these initial investors later decide to sell their stakes of the company, they can
do so on the secondary market. Any transactions on the secondary market occur
between investors, the proceeds of each sale go to the selling investor, not to the
company that issued the stock or to the underwriting bank. The secondary market
is where securities are traded after the company has sold its offering on the primary
market. It is also referred to as the stock market. The New York Stock Exchange
(NYSE), London Stock Exchange, and Nasdaq are secondary markets.
Small investors have a much better chance of trading securities on the secondary
market since they are excluded from IPOs. Anyone can purchase securities on the
secondary market as long as they are willing to pay the asking price per share.
A broker typically purchases the securities on behalf of an investor in the secondary
market. Unlike the primary market, where prices are set before an IPO takes place,
prices on the secondary market fluctuate with demand. Investors will also have to
pay a commission to the broker for carrying out the trade.
The volume of securities traded varies from day to day, as supply and demand for
the security fluctuates. This has a great effect on the security’s price.
When an initial offering is completed, the issuing company is no longer a party to
any sale between two investors, except in the case of a company stock buyback. For
example, after December 12, 1980, Apple’s IPO on the primary market, individual
investors were able to purchase Apple stock on the secondary market. Because Apple
is no longer involved in the issue of its stock, investors would, essentially, deal with
one another when they trade shares in the company.
The secondary market (Kuznetsov 2006; Drayer et al. 2008) has two different cat-
egories: the auction and the dealer markets. The auction market is home to the open
outcry system, where buyers and sellers congregate in one location and announce
the prices at which they are willing to buy and sell their securities. The NYSE is one
such example. In dealer markets, though, people trade through electronic networks.
Most small investors trade through dealer markets.
266 10 Quantum Price Levels for Worldwide Financial Products

10.1.3 Types of Financial Markets

Over-The-Counter Markets
The over-the-counter (OTC) market is an example of a secondary market. An OTC
market handles the exchanging of public stocks not listed on the Nasdaq, New York
Stock Exchange, or American Stock Exchange. Companies with stocks trading on
the OTC market are usually smaller organizations, as this financial market requires
less regulation and trading is less expensive.
Financial Markets for Bonds
A bond is a security in which an investor loans money for a defined period of time at
a preestablished rate of interest. Bonds are not only issued by corporations but may
also be issued by municipalities, states, and federal governments from around the
world. They are also referred to as debt, credit, or fixed-income market. For example,
the bond market sells securities such as notes and bills issued by the United States’
Treasury.
Money Markets
A money market is a portion of the financial market that trades highly liquid and
short-term maturities. The intention of the money market is for short-term borrowing
and lending of securities with maturity typically less than 1 year. This financial
market trades certificates of deposit, banker’s acceptances, certain bills, notes, and
commercial paper.
Derivatives Market
The derivatives market is a financial market that trades securities that derive its value
from its underlying asset. The value of a derivative contract is determined by the
market price of the underlying item. This financial market trades derivatives include
forward contracts, futures, options, swaps, and contracts-for-difference.
Forex Market
The forex market is a financial market, where currencies are traded. This financial
market is the most liquid market in the world, as cash is the most liquid assets. The
interbank market is the financial system that trades currency between banks.

10.1.4 OTC Market and Market Makers

Over-the-counter (OTC) or off-exchange trading is done directly between two parties,

without the supervision of an exchange (Arnett 2011; Atkeson et al. 2015; Huber
et al. 2012).
It is contrasted with exchange trading, which occurs via exchanges.
In traditional security trading such as stock trading, a stock exchange has the
benefit of facilitating liquidity, providing transparency, and maintaining the current
market price.
10.1 Financial Markets 267

In an OTC trade, the price is not necessarily published for the public.
OTC trading, like exchange trading, occurs with commodities, financial instru-
ments (including stocks), and derivatives of such products.
In 2008, approximately 16% of all U.S. stock trades were OTC trading, by April
2014 it has increased to about 40%.
A market maker or liquidity provider is a company or an individual that quotes
both a buy and a sell price in a financial instrument or commodity held in inventory,
hoping to make a profit on the bid-offer spread, or turn.
The U.S. Securities and Exchange Commission defines a market maker as a firm
that stands ready to buy and sell stock on a regular and continuous basis at a publicly
quoted price.
Most foreign exchange trading firms are market makers and so are many banks.
Market makers both sell to and also buy from its clients and is compensated by
means of price differentials for the services of providing liquidity, reducing transac-
tion costs and facilitating trade.
They make a profit by the difference between the price at which a market maker
is willing to buy a stock (the bid price) and the price that the firm is willing to sell it
(the ask price) is known as the market maker spread, or bid–ask spread.
Market makers also provide liquidity to their own firm’s clients for which they
earn a commission.

10.1.5 Forex, OTC, and Market Maker

The foreign exchange market (forex, FX, or currency market) is a global decentralized
or OTC market for currencies trading.
This market determines the foreign exchange rates.
It includes all aspects of buying, selling, and exchanging currencies at current or
determined prices.
In terms of trading volume, it is by far the largest market in the world.
In retail forex trading, most foreign exchange trading centers/firms are market
makers and so are many banks.
Unlike a stock market, the foreign exchange market is divided into levels of access.
At the top is the interbank foreign exchange market, which is made up of the
largest commercial banks and securities dealers.
Key participants include the following:
• central banks,
• commercial banks,
• foreign exchange brokers,
• investment management firms,
• retail foreign exchange traders, and
• money transfer/remittance companies.
268 10 Quantum Price Levels for Worldwide Financial Products

The main participants in this market are the larger international banks.
Financial centers around the world function as anchors of trading between a wide
range of multiple types of buyers and sellers around the clock, with the exception of
weekends.
Currencies are traded against one another in pairs.
Each currency pair thus constitutes an individual trading product and is tradi-
tionally noted XXXYYY or XXX/YYY, where XXX and YYY are the ISO 4217
international three-letter code of the currencies involved.
For instance, the quotation EURUSD (EUR/USD) 1.5465 is the price of the Euro
expressed in US dollars, meaning 1 euro = 1.5465 US dollars.
Table 10.1 shows the list of the top 35 most traded currencies in 2016.

10.1.6 Forex Trading Characteristics

Forex trading is a huge trading volume, representing the largest asset class in the
world leading to high liquidity due to its geographical dispersion.
Its continuous operation is 24 h a day except weekends, i.e., trading from 22:00
GMT on Sunday (Sydney) until 22:00 GMT Friday (New York) as Asia trading
session ends, Europe session begins, followed by North America session and then
back to the Asia session.
There are a variety of factors that can affect exchange rates.
The margins of relative profit are moderate as compared with other markets of
fixed income.
There is no unified or centrally cleared market for the majority of trades, and there
is very little cross-border regulation.
The main trading centers are London, New York, Tokyo, Hong Kong, and other
prominent trading centers throughout the world.
The top 3 most heavily traded bilateral currency pairs were the following:
1. EURUSD: 23.0%;
2. USDJPY: 17.7%;
3. GBPUSD: 9.2%.
Different from stocks which are localized to the company’s financial situation,
fluctuations in exchange rates are usually caused by actual worldwide monetary flows
as well as by expectations of changes in monetary flows.
These are caused by changes in worldwide gross domestic product (GDP) growth,
inflation (purchasing power parity theory), interest rates, budget and trade deficits or
surpluses, large cross-border M&A deals, and other macroeconomic conditions.
Major financial reports, news, and indices from major financial/government insti-
tutions such as Federal Reserve Board (USA) and Organization of the Petroleum
Exporting Countries (OPEC) are important factors, which are released publicly, often
on scheduled dates.
Political conditions including internal, regional, and international political con-
ditions and events can have a profound effect on currency markets.
Supply and demand is also a major factor.
10.1 Financial Markets 269

Table 10.1 Top 35 most traded currencies 2016

Rank Country Currency Symbol % of daily trades
1. United States dollar USD (US$) 87.60

2. Euro EUR (e) 31.40

3. Japanese yen JPY (¥) 21.60

4. Pound sterling GBP (£) 12.80

5. Australian dollar AUD (A$) 6.90

6. Canadian dollar CAD (C$) 5.10

7. Swiss franc CHF (Fr) 4.80

8. Renminbi CNY (元) 4.00

9. Swedish krona SEK (kr) 2.20

10. New Zealand dollar NZD (NZ$) 2.10

11. Mexican peso MXN ($) 1.90

12. Singapore dollar SGD (S$) 1.80

13. Hong Kong dollar HKD (HK$) 1.70

14. Norwegian krone NOK (kr) 1.70

15. South Korean won KRW (₩) 1.70

16. Turkish lira TRY () 1.40

17. Russian ruble RUB () 1.10

18. Indian rupee INR (|) 1.10

19. Brazilian real BRL (R$) 1.00

20. South African rand ZAR (R) 1.00

21. Danish krone DKK (kr) 0.80

(continued)
270 10 Quantum Price Levels for Worldwide Financial Products

Table 10.1 (continued)

Rank Country Currency Symbol % of daily trades
22. Polish złoty PLN (zł) 0.70

23. New Taiwan dollar TWD (NT$) 0.60

24. Thai baht THB (฿) 0.40

25. Malaysian ringgit MYR (RM) 0.40

26. Hungarian forint HUF (Ft) 0.30

27. Saudi riyal 0.30

28. Czech koruna CZK (Kč) 0.30

29. Israeli shekel 0.30

30. Chilean peso CLP (CLP$) 0.20

31. Indonesian rupiah IDR (Rp) 0.20

32. Colombian peso COP (COL$) 0.20

33. Philippine peso 0.10

34. Romanian leu RON (L) 0.10

35. Peruvian sol PEN (S/) 0.10

Other 2.10
Total 200.00
Note
(1) Source Triennial Central Bank Survey Foreign exchange turnover in April 2016. Triennial
Central Bank Survey. Basel, Switzerland: Bank for International Settlements. 11 December 2016.
p. 7
(2) The total sum is 200% because each currency trade always involves a currency pair; one currency
is sold (e.g., US$) and another bought (e). Therefore, each trade is counted twice, once under the
sold currency (US$) and once under the bought currency (e). The percentages above are the percent
of trades involving that currency regardless of whether it is bought or sold, e.g., the U.S. Dollar is
bought or sold in 87% of all trades, whereas the Euro is 31.4%
10.1 Financial Markets 271

10.1.7 Retail Forex Trading Platforms

Retail foreign exchange trading is a small segment of the larger foreign exchange
market, where individuals speculate on the exchange rate between different curren-
cies.
This segment has developed with the advent of dedicated electronic trading plat-
forms and the internet, which allows individuals to access the global currency mar-
kets.
In 2016, it was reported that retail foreign exchange trading represented 5.5% of
the entire foreign exchange market ($282 billion in daily trading turnover).
Prior to the development of forex trading platforms in the late 90s, forex trading
was restricted to large financial institutions.
Nowadays, traders are able to trade spot currencies with market makers on margin.
This means they need to put down only a small percentage of the trade size and can
buy and sell currencies in seconds.
As both profit and loss will be amplified by margin trading, risk management such
as stop-loss strategy must execute in every trading transaction.
MetaTrader 4 (MT4) platform is the most popular retail forex trading platform.
By using MetaQuotes Language (MQL), a trader can develop one’s own trading
program for automatic program trading (Young 2015; Walker 2018).
Details of MT4 program trading will be studied in the next section.
Figure 10.1 shows a snapshot for forex trading over MT4 provided by Forex.com
(www.Forex.com), one of the largest forex trading platforms in the world.

Fig. 10.1 Snapshot for forex trading over MT4 platform provided by Forex.com
272 10 Quantum Price Levels for Worldwide Financial Products

10.2 Supports and Resistances

10.2.1 Introduction

The concepts of support and resistance are undoubtedly two of the most highly
studied attributes of technical analysis. Part of analyzing chart patterns, these terms
are used by traders to refer price levels on charts that tend to act as barriers, preventing
the price of an asset from getting pushed in a certain direction (Bulkowski 2005;
Kirkpatrick and Dahlquist 2015; Murphy 1999; De Angelis and Peskir 2016; Gomes
and Waelbroeck 2010; Zapranis and Tsinaslanidis 2012).
In the Wall Street environment, the terms support and resistance are almost syn-
onymous with demand and supply respectively.
Support is a price level at which there is adequate demand for security to stop its
downward price movement and, normally, turn prices upward.
Typically, support occurs at reaction lows.
Resistance is a price level at which there is a significant supply of a stock causing
prices to halt an upward move and turn prices down.
Typically, resistance occurs at reaction highs.
Once an area or zone of support or resistance has been identified, it provides
valuable potential trade entry or exit points. This is because, as a price reaches a
point of support or resistance, it will do one of two things—bounce back away from
the support or resistance level or violate the price level and continue in its direction—
until it hits the next support or resistance level.
Most forms of trades are based on the belief that support and resistance zones
will not be broken. Whether price is halted by the support or resistance level, or it
breaks through, traders can bet on the direction and can quickly determine if they
are correct. If the price moves in the wrong direction, the position can be closed
at a small loss. If the price moves in the right direction, however, the loss may be
substantial. Figure 10.2 shows typical support and resistance lines.

Fig. 10.2 Typical supports and resistances

10.2 Supports and Resistances 273

Fig. 10.3 Rising and declining supports and resistances

10.2.2 Rising and Declining Supports and Resistances

In an uptrend, both support and resistance levels rise as illustrated in Fig. 10.3.
Typically, support levels hold while resistance offers temporary halts to upward
movements in prices. Resistance levels are repeatedly broken until the uptrend is
reversed. In a downtrend, both support and resistance levels move lower as illustrated
in the right figure. Typically, resistance levels hold while support levels temporarily
stop downward price movements. Support levels are repeatedly broken until the
downtrend is reversed.

10.2.3 Role Reversal on Support and Resistance

In an uptrend, both support and resistance levels rise as illustrated in Fig. 10.4.
Typically, support levels hold while resistance offers temporary halts to upward
movements in prices.
Resistance levels are repeatedly broken until the uptrend is reversed.

Fig. 10.4 Role reversals in supports and resistance

274 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.5 Trend reversals in supports and resistance

In a downtrend, both support and resistance levels move lower as illustrated in

the right figure.
Typically, resistance levels hold while support levels temporarily stop downward
price movements.
Support levels are repeatedly broken until the downtrend is reversed.

10.2.4 Trend Reversal on Support and Resistance

Trend reversal at top: In an uptrend, a trend reversal occurs when prices are held at
a resistance level. A double top or some other reversal formation develops at that
point, and the trend changes direction.
Trend reversal at the bottom: A trend reversal occurs in a downtrend when prices
are unable to penetrate a support level. In this case, a bottom reversal pattern is
formed, and the trend changes direction to the upside.
Figure 10.5 shows typical trend reversal on support and resistance.
Keep in mind that a trend reversal is not signaled by the first failure to break
through a resistance level (in an uptrend) or a support level (in a downtrend). A
reversal pattern must fully develop before one gets the signal that the trend has
changed. In other words, a trader should not rush to sell all of one’s securities or sell
short just because prices have held at a resistance level.

10.2.5 Support and Resistance Versus QPLs

As mentioned in Chap. 5, although we don’t know whether quantum price level

(QPL) whether exists or not in financial market. But for many experienced analysts
and traders, we all believe that quantized price levels definitely exist in all financial
markets such as support (S) and resistance (R).
10.2 Supports and Resistances 275

In fact, the basis of technical analysis in the past 50 years is based on this funda-
mental concept.
The difference is that, with the exponential growth of program trading, espe-
cially HFPT (high-frequency-program-trading) (Durbin 2010). The financial markets
become so complex and chaotic that we don’t believe one can observe (or locate) all
these hidden-force-levels manually.
However, different from the traditional understanding of support and resistance
(S&R) levels, there is one major difference between them.
In traditional technical analysis, S&R levels are normally totally different price
levels as they play different roles in the supply and demand forces for market move-
ments. However, in quantum finance, quantum price levels (QPLs) are interpreted as
discrete energy levels for the movements of the financial particles from one energy
state to another, so basically there is no concepts either support or resistance, but just
energy states. In other words, one QPL can be acted as a support level for a moment
and then acted as resistance level at another moment.
The truth is, such interpretation becomes more realistic and practical in nowadays
financial markets, as we normally see a particular price level can be acted as both a
support and resistance levels all the times within a day or a period of time, especially
for highly fluctuated financial markets such as forex and cryptocurrency. The main
reason is maybe due to the popularity of program trading and especially HFPT
(Durbin 2010) in which trading programs normally define the same set of price
levels as the thresholds to trigger buying or selling signals, and seldom evaluate
them as two separated price levels for automatic program trading. That’s why the
frequent interchange of S&R roles is commonly found in nowadays financial markets
include stock, forex, and cryptocurrency markets (Fig. 10.6).

Fig. 10.6 QPLs in financial markets

276 10 Quantum Price Levels for Worldwide Financial Products

10.3 The MetaTrader (MT) Platform and MQL

10.3.1 An Introduction

MetaTrader 4, also known as MT4, is an electronic trading platform widely used by

online retail foreign exchange speculative traders (Brown 2016; Walker 2018; Young
2015).
It was developed by MetaQuotes software and released in 2005.
The software is licensed to foreign exchange brokers, who provide the software
to their clients.
The software consists of both a client and server component. The server compo-
nent is run by the broker and the client software is provided to the broker’s customers,
who use it to see live streaming prices and charts; to place orders; to manage their
accounts.
The client is a Microsoft windows-based application that became popular mainly
due to the ability to end-users to write their own trading scripts and robots that could
automate trading. Figure 10.7 shows a snapshot of MT4 in PC platform.
In 2010, MetaQuotes released a successor, MetaTrader 5. However, most traders
and brokers keep on using MT4.

Fig. 10.7 Snapshot of MT4 in a PC platform

10.3 The MetaTrader (MT) Platform and MQL 277

10.3.2 MQL—MetaQuotes Language

MetaQuotes Language 4 (MQL4) is a built-in language for programming trading

strategies (Brown 2016; Walker 2018; Young 2015).
This language is developed by MetaQuotes Software Corp. based on their long
experience in the creation of online trading platforms.
Using this language, we can create our own expert advisors that make trading
management automated and are perfectly suitable for implementing our own trading
strategies.
Besides, using MQL4 we can create our own technical indicators (custom indi-
cators), scripts, and libraries.
MQL4 contains a large number of functions necessary to analyze current and
previously received quotes, along with built-in basic indicators and functions to
manage and control trade orders.
The MetaEditor that highlights different constructions of MQL4 is used for writ-
ing the program code. It helps users to orientate themselves in the expert system text
quite easily.
Although MQL5 had launched for years, most traders and brokers keep on using
MQL4 over MT4 platform. As any MQL4 program can be compiled and run in
MT4/MT5 platforms (but not the other way around). In this book, we focus on MT4
programming.
Programs written in MQL4 have different features and purposes:
1. expert advisor (EA)—is a mechanical trading system linked up to a certain chart.
An expert advisor starts to operate when an event happens that can be handled
by it: events of initialization and deinitialization, event of a new tick receipt, a
timer event, depth of market-changing event, chart event, and custom events.
2. custom indicator—is a technical indicator written independently in addition to
those already integrated into the client terminal. Like built-in indicators, they
cannot trade automatically and intend to implement analytical functions only.
3. scripts—is a program that intends to a single execution of some actions. Unlike
expert advisors, scripts do not process any actions, except for the start event.
4. library—is a set of custom functions that intend to store and distribute used blocks
of custom programs frequently. Libraries cannot start executing by themselves.
5. include File—is a source text of the most frequently used blocks of custom
programs like C++.

10.3.3 MQL—A Brief Overview

MetaQuotes Language 4 (MQL4) is a built-in language for programming trading

strategies.
MQL4 is an OO-programming language intended to write automated trading
strategies, custom technical indicators for various financial markets analyses.
278 10 Quantum Price Levels for Worldwide Financial Products

It allows not only to write a variety of expert systems, designed to operate in real
time, but also to create their own graphical tools to help users make trade decisions.
MQL4 is based on the concept of the popular programming language C++.
Like C++, MQL4 provides the OO-programming basics which include the fol-
lowing:
• Syntax,
• data types,
• operations and expressions,
• operators,
• functions,
• variables,
• preprocessor, and
• object-oriented programming.
(Details please refer to https://docs.mql4.com/basis)
In addition to all C++-style programming fundamental tools and functions, MQL4
provides predefined variables, classes, and functions tailored for program trading in
online financial markets which include the following:
• account information,
• checkup,
• market info,
• time series and indicators access
• chart operations,
• trade functions.
• trade signals,
• global variables of the terminal, and
• technical indicators.

10.3.4 MQL—Time Series and Indicators Access Functions

These are functions for working with time series and indicators.
A time series differs from usual data array by its reverse ordering—elements of
time series are indexed from the end of an array to its start (from the most recent data
to the oldest ones).
To copy the time series values and indicator data, it is recommended to use
dynamic arrays only, because copying functions are designed to allocate the nec-
essary size of arrays that receive values.
Commonly used functions include the following:
• iOpen
• iHigh
• iLow
10.3 The MetaTrader (MT) Platform and MQL 279

Fig. 10.8 Syntax of iOpen() function and program example of using time series and indicators
access functions

• iClose
• iVolume
• iTime
Figure 10.8 shows the syntax and example of using time series and indicators
access functions.

10.3.5 MQL—Trade Functions

This is the group of functions that intend to manage trading activities.

Major trade functions need to be used in EA are the following:
• OrderSend(),
• OrderClose(),
• OrderModify(), and
• OrderDelete().
A complete list of functions can be found at https://docs.mql4.com/trading.
Figure 10.9a shows the syntax and program example of OrderSend(), one of the
280 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.9 Syntax and program example of OrderSend() function

most important and frequently used in program trading. Besides, MT4 also provides
six types of trading orders: Buy, Sell, Buy Limit, Sell Limit, Buy Stop, and Sell Stop
(Kaufman 2013).

10.3.6 MQL—Technical Indicators

A group of functions that intend to standard and custom indicators’ calculations.

Totally 39 technical indicator functions are provided.
Commonly used technical indicators include:
iBands Bollinger Bands
iMomentum Momentum
iMA Moving Average
iMACD Moving Averages Convergence-Divergence
iRSI Relative Strength Index
iStochastic Stochastic Oscillator
10.3 The MetaTrader (MT) Platform and MQL 281

Fig. 10.10 Technical indicator functions in MQL of iMA and iRSI

iFractals Fractals
Figure 10.10 shows the syntax of two most commonly used technical indicators,
iMA (moving average), and iRSI (relative strength index). Appendix B shows the
list of 39 technical indicators functions implemented by MQL.

10.3.7 MQL Programming Tips

If the user is an experienced C++/Java programmer, it should be easy to pick up the

skill set of writing EA programs using MQL.
All we need to do is the following:
• go through MT4 documentation site,
• refresh C++/Java program skills,
• learn about MT4 program trading with technical indicator functions and how they
can be used, and
• understand file I/O techniques.
In learning new program languages, nothing is more useful than actual lab practice
(Fig. 10.11).
282 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.11 Quantum finance MQL programming workshop in QFFC.org

10.4 System Implementation—QPL Evaluation

for Worldwide Financial Products

From the implementation perspective, 129 financial products from Forex.com (one of
the biggest forex MT platform providers) and AvaTrade.com (major cryptocurrency
MT platform providers) are used for QPLs evaluation.
They include the following:
• 9 Cryptocurrency,
• 84 Forex,
• 19 Financial Indices, and
• 17 Major Commodities.
10.4 System Implementation—QPL Evaluation for Worldwide Financial Products 283

Fig. 10.12 A snapshot of MT4 platform of cryptocurrency provided by AvaTrade.com

Except cryptocurrency which has around 300–500 trading days information, all
other financial products contain 2048 time series trading day information for QPLs
evaluation.
Figure 10.12 shows a snapshot of the MT4 platform of cryptocurrency provided
by AvaTrade.com.
Appendix A shows the list of 129 financial products provided by Forex.com and
AvaTrade.com for their MT4 platforms.

10.5 MQL Program for QPLs Evaluation of Worldwide

Financial Products (QPL2019.mq4)

10.5.1 System Flowchart and Main Program Logic

The main program logic for the evaluation of the first 21 QPLs of all 120+ worldwide
financial product is shown in Fig. 10.13.
For each financial product, do the following:
• Read the daily time series and extract (Date, O, H, L, C, V)
• Calculate daily price return r(t)
• Calculate quantum price return wavefunction Q(r) (size 100)
• Evaluate λ value for the wavefunction Q(r) using F.D.M. and Eq. (5.21) & evaluate
other related parameters:
– sigma (std dev of Q)
– maxQPR (max quantum price return—for normalization).
284 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.13 Flowchart for the

determination of the first 21
QFELs and QPLs
10.5 MQL Program for QPLs Evaluation … 285

• Once λ is found, using QFSE (numerical solution) by solving the depressed cubic
equation using Cardano’s method Eq. (5.32a–c) to calculate first 21 quantum
finance energy level, QFEL(n), n = [1 … 20]
• Calculate quantum price return, QPR(n)

p = −(2n + 1)2 (10.1)

QFEL(n)
QPR(n) = (10.2)
QFEL(0)

where n = [1 … 20]
• Calculate normalized QPR(n)

NQPR(n) = 1 + 0.21 ∗ sigma ∗ QPR(n) (10.3)

where n = [1 … 20]
• Save two level of data files:
• For each financial product, save the QPL table contains QPE, QPR, NQPR for
the first 21 energy levels
• For all financial product, create a QPL summary table contains NQPR for all
FP, which will be used for financial prediction using recurrent neural networks.
Figure 10.14 shows the program header with program logic listed for the
QPL2019.mq4 program for QPLs evaluation of 120+ worldwide financial products.

Fig. 10.14 Main program logic of QPL evaluation program

286 10 Quantum Price Levels for Worldwide Financial Products

10.5.2 Program Declaration

Program Declaration consists of the following three sections:

• declaration of program directories,
• declaration of global variables,
• declaration of timing variables, and
• declaration of financial products and related variables.
According to the numerical formulation revealed in Chap. 5, daily returns of the
time series for each financial product are evaluated to calculate the wavefunction
values Qf.
The application of the finite difference method (F.D.M.) (Abreu et al. 2018; Guo
2018) is to evaluate the Lambda (λ) value.
Once the Lambda value is found, apply the quartic Schrődinger equation (Singh
et al. 2009; Tichy et al. 2012) of quantum finance to find all 21 quantum price energies
(QFEL0 .. QFEL20) and store in QFEL datafiles. Figure 10.15 shows the program
declaration of QPL evaluation program.

10.5.3 Program Initialization: init()

As MQL is a real-time program trading development platform, init() is used to

perform one-off process such as the system initialization or QPL calculation in our
case, while start() is a time-step looping process to implement real-time function
such as real time trading and hedging.
So, in our case, all QPL calculation process in this system are placed in init()
class. In other words, to implement a real-time QPL evaluation and QPL-based
trading program, one will implement QPL evaluation program in the init() class real
time trading algorithm in the start() class. Figure 10.16 shows the program, data, and
data file initialization process.

10.5.4 Process 1: Calculate All K Values

The calculation is used in accordance with quantum anharmonic oscillation function

from quantum finance revealed in Chap. 5:
 3  
E(n) E(n)
− − (K0 (n))3 λ = 0 (10.4)
2n + 1 2n + 1

where
10.5 MQL Program for QPLs Evaluation … 287

Fig. 10.15 Program declaration of QPL evaluation program

288 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.16 Program initialization of QPL evaluation program

10.5 MQL Program for QPLs Evaluation … 289

Fig. 10.17 Process 1 calculate all K values of QPL evaluation program

 1/3
1.1924 + 33.2383n + 56.2169n2
K0 (n) = (10.5)
1 + 43.6196n

First, evaluate all K0 (n) for the 20 energy levels (n).

Then open the file handles of the data files for the storage of the QPR, NQPR, QF
values. Program codes of process 1 are shown in Fig. 10.17.

10.5.5 Process 2: READ ALL Daily Time Series

For each financial product (FP), this process uses MQL time series data functions
iOpen, iHigh, iLow, iClose to read the daily time series data and calculate the daily
return by

DT_RT[d] = DT_CL[d]/DT_CL[d + 1] (10.6)

290 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.18 Process 2 read time series of QPL evaluation program

where DT_CL[d] and DT_CL[d+1] are the closing prices for d and d+1 day, respec-
tively. As mentioned, except cryptocurrencies which have 300–500 trading day
records, the other 120 financial products (forex, commodities and financial indices)
have 2048-trading day records, which are sufficient for the evaluation on distribu-
tion function (wavefunction) of the price returns (r). Program codes of process 2 are
shown in Fig. 10.18.

10.5.6 Process 3: Calculate Mean (mu) and Standard

Deviation (sigma) of Return Array

After the evaluation of daily returns DT_RT[d] of the financial time series, this
process evaluates the mean (mu, μ) and standard deviation (sigma, σ ) of the return
distribution function (ϕ).
10.5 MQL Program for QPLs Evaluation … 291

Fig. 10.19 Process 3 calculate mean (mu) and standard deviation (sigma) of return array

The purpose for mu evaluation is to locate the central point of the distribution
function, and sigma is used to calculate “dr” term—the width of each slice of returns
for the distribution function is an important factor to calculate Lambda using F.D.M.,
which is given by

sigma
dr = 3 ∗ ; (10.7)
50
For each QP distribution chart, 100 equal spacing slices with width dr are used for
to evaluate the distribution function values. Program codes of process 3 are shown
in Fig. 10.19.

10.5.7 Process 4: Generate the Returns Wavefunction

Distribution

Based on dr value calculated in process 3, this process calculates the price returns
wavefunction values Q[nQ], NQ[nQ] where nQ is the number of slice in distribution
function, Q[nQ] is the pdf (probability density function values) within the particular
slice nQ; and NQ[nQ] is the normalized pdf values of the returns wavefunction.
The NQ[nQ] is given by:
292 10 Quantum Price Levels for Worldwide Financial Products

Quantum Price Return Wavefunc on Q(r) of XAUUSD

(For the past 2048 trading-day sta s cs)
0.05

0.045

0.04

0.035

0.03
Q(r)

0.025

0.02

0.015

0.01

0.005

0 1
0.97559986
0.97657587
0.97755187
0.97852788
0.97950388
0.98047989
0.9814559
0.9824319
0.98340791
0.98438391
0.98535992
0.98633592
0.98731193
0.98828793
0.98926394
0.99023994
0.99121595
0.99219196
0.99316796
0.99414397
0.99511997
0.99609598
0.99707198
0.99804799
0.99902399

1.00097601
1.00195201
1.00292802
1.00390402
1.00488003
1.00585603
1.00683204
1.00780804
1.00878405
1.00976006
1.01073606
1.01171207
1.01268807
1.01366408
1.01464008
1.01561609
1.01659209
1.0175681
1.0185441
1.01952011
1.02049612
1.02147212
1.02244813
1.02342413
Daily Return (r)

Fig. 10.20 Process 4 quantum price return wavefunction of Q(r) of gold (XAUUSD)

no of trading days within the range of return value in nQ

NQ[nQ] = (10.8)
Total number trading day return values

Figure 10.20 shows the return wavefunction for Gold (XAUUSD) in 2048 financial
time series. Program codes of process 4 are shown in Fig. 10.21.

10.5.8 Process 5: Evaluate Lambda Value for the Returns

Wavefunction

Once all wavefunction values in each slice are evaluated, we make use of F.D.M.,
together with QFSE (Quantum Finance Schrödinger Equation) to evaluate the
Lambda value (λ), which is given by:
 2 
 r−1 ϕr−1 − r+1
2
ϕr+1 

λ= 4 (10.9)
r+1 ϕr+1 − r−1
4
ϕr−1 

where ϕr−1 and ϕr+1 are wavefunction values (normalized) for the 1st left and right-
hand side slice from the slice with max wavefunction value, normally is the central
point with returns value = 1 as in Fig. 10.22.
Program codes of process 5 are shown in Fig. 10.23.
10.5 MQL Program for QPLs Evaluation … 293

//************************************************************************************************************
//
// 4. Generate the QP Wavefunction distribution
//
//************************************************************************************************************
auxR = 0;

// Loop over all r from r - 50*dr to r + 50*dr and get the distribution function
// Reset all the Q[] first
for (nQ=0;nQ<100;nQ++)
{
Q[nQ] = 0;
}

// Loop over the maxRno to get the distribution

tQno = 0;
for (nR=0;nR<maxRno;nR++)
{
bFound = False;
nQ = 0;
auxR = 1 - (dr * 50);
while (!bFound && (nQ < 100))
{
if ((DT_RT[nR] > auxR) && (DT_RT[nR] <= (auxR + dr)))
{
Q[nQ]++;
tQno++;
bFound = True;
}else
{
nQ++;
auxR = auxR + dr;
}
}
}

// Write out the Qfile for Record

auxR = 1 - (dr * 50);
for (nQ=0;nQ<100;nQ++)
{
r[nQ] = auxR;
NQ[nQ] = Q[nQ]/tQno;
FileWrite(Qf_FileHandle,auxR,Q[nQ],NQ[nQ]);
auxR = auxR + dr;
}

// Find maxQ and maxQno

maxQ = 0;
maxQno = 0;
for (nQ=0;nQ<100;nQ++)
{
if (NQ[nQ] > maxQ)
{
maxQ = NQ[nQ];
maxQno = nQ;
}
}

// Printout the maxQ, maxQno

Print("TP",nTP+1," ",TP_Code[nTP]," MaxQ= ",maxQ," maxQno=",maxQno," Total Qno =",tQno);
FileWrite(QPLog_FileHandle,"TP",nTP+1," ",TP_Code[nTP],
" MaxQ= ",maxQ," maxQno=",maxQno," Total Qno =",tQno);

Fig. 10.21 Process 4 generate the returns wavefunction distribution

294 10 Quantum Price Levels for Worldwide Financial Products

Fig. 10.22 Illustration of F.D.M. scheme for evaluation of lambda value

//********************************************************************************************
//
// 5. Evaluate Lambda L for the QP Wavefuntion
//
//********************************************************************************************
//
// Given:
// maxQno - i.e. ground state Q[0], r[0] = r[maxQno-dr]
// We have:
// Q[+1] = NQ[maxQno+1], r[+1] = r[maxQno]+(dr/2)
// Q[-1] = NQ[maxQno-1], r[-1] = r[maxQno]-(dr*1.5)
// Apply F.D.M. into Quantum Finance Schrodinger Equation
// L = abs((r[-1]^2*Q[-1]-(r[+1]^2*Q[+1]))/(r[-1]^4*Q[-1]-(r[+1]^4*Q[+1])))

r0 = r[maxQno] - (dr/2);
r1 = r0 + dr;
rn1 = r0 - dr;
Lup = (pow(rn1,2)*NQ[maxQno-1])-(pow(r1,2)*NQ[maxQno+1]);
Ldw = (pow(rn1,4)*NQ[maxQno-1])-(pow(r1,4)*NQ[maxQno+1]);
L = MathAbs(Lup/Ldw);

// Printout r0,Q0, r1, Q1, r-1 Q-1

Print("TP",nTP+1," ",TP_Code[nTP]," r0 = ",r0," r1 = ",r1," r-1 = ",rn1," Q0 = ",
NQ[maxQno]," Q1 = ",NQ[maxQno+1]," Q-1 = ",NQ[maxQno-1]," L = Lup/Ldw = ",Lup,"/",Ldw," = ",L);

FileWrite(QPLog_FileHandle," r0 = ",r0," r1 = ",r1," r-1 = ",rn1," Q0 = ",NQ[maxQno]," Q1 = ",

NQ[maxQno+1]," Q-1 = ",NQ[maxQno-1]," L = Lup/Ldw = ",Lup,"/",Ldw," = ",L);

FileWrite(Lambda_FileHandle,TPSymbol,L);

Print(TPSymbol,L);

Fig. 10.23 Process 5 Evaluate lambda value for the returns wavefunction
10.5 MQL Program for QPLs Evaluation … 295

10.5.9 Process 6: Using QP Schrodinger Equation to FIND

the First 21 Energy Levels

Once Lambda value is evaluated, we can apply the Cardano’s formula to solve the
quantum price return Schrödinger equation, with formulation given by:
   
3 q q2 p3 3 q q2 p3
E(n) = − + + + − − + (10.10)
2 4 27 2 4 27

where:

p = −(2n + 1)2 , q = −λ(2n + 1)3 [K0 (n)]3

Please refer to Chap. 5 on detailed mathematical proof and derivations for the
numerical evaluation of quantum price level.
In this program, we calculate the first 21 quantum finance energy levels (QFEL)
of every financial product (ground state 0 and 20 Energy Levels), which is sufficient
for daily financial trading and analysis.
Figure 10.24 shows MQL codes for 21 energy levels evaluation.

10.5.10 Process 7: Program End

This process closes all datafile handles, calculate total CPU time spent and report
the total time taken to calculate QPLs of all financial products.
As shown in Fig. 10.25, MQL also provides start() class for developers to imple-
ment real time trading/hedging programs.
In other words, one can make use of the MQL learnt in this chapter to evaluate the
QPLs for any financial product, and make use of it, say, as S&R lines to implement
real time trading and hedging programs learnt in Chap. 6.
Figure 10.26 shows a snapshot of implementation result of QPL evaluation pro-
gram on Forex.com MT4 platform for QPL evaluation of 120 financial products.

10.5.11 Implementation Results

Figure 10.26 shows snapshot of the implementation results of QPL2019.mq4 MQL

program for QPLs evaluation of 120 financial products at the Forex.com MT platform.
As shown, it took only 1594 ms (i.e., less than 2 s) to evaluate the first 20 QPLs for
120 financial products, which was much faster and effective than using traditional
Path Integral Techniques.
296 10 Quantum Price Levels for Worldwide Financial Products

//*******************************************************************************************************************
// 6. Using QP Schrődinger Eqt to FIND first 21 Energy Levels
// By solving the Quartic Anharmonic Oscillator as cubic polynomial equation of the form
// a*x^3 + b*x^2 + c*x + d = 0
// Using (Dasqupta et. al. 2007) QAHO solving equation:
// (E(n)/(2n+1))^3 - (E(n)/(2n+1)) - K(n)^3*L = 0
// Solving the above Depressed Cubic Eqt using Cardano's Method
// Given t^3 + p*t + q = 0
// Let t=u+v
// Cardano's Method deduced that:
// u^3 = -q/2 + sqrt(q^2/4 + p^3/27)
// v^3 = -q/2 - sqrt(q^2/4 + p^3/27)
// The first cubic root (real root) will be:
// t=u+v
// So, combining Cardano's Method into our QF Sch Eqt.
// Substitue p = -(2n+1)^2; q = -L(2n+1)^3*(K(n)^3) into the above equations to get the real root
//
//*******************************************************************************************************************
for (eL=0;eL<21;eL++)
{
p = -1 * pow((2*eL+1),2);
q = -1 * L * pow((2*eL+1),3) * pow(K[eL],3);

// Apply Cardano's Method to find the real root of the depressed cubic equation
u = MathPow((-0.5*q + MathSqrt(((q*q/4.0) + (p*p*p/27.0)))),p3);
v = MathPow((-0.5*q - MathSqrt(((q*q/4.0) + (p*p*p/27.0)))),p3);

// Store the QFEL

QFEL[eL] = u + v;

// Printout the QF Energy Levels

Print("TP",nTP+1," ",TP_Code[nTP]," Energy Level",eL," QFEL = ",QFEL[eL]);
FileWrite(QPLog_FileHandle," Energy Level",eL," QFEL = ",QFEL[eL]);
}

// Evaluate ALL QPR values

for (eL=0;eL<21;eL++)
{
QPR[eL] = QFEL[eL]/QFEL[0];
NQPR[eL] = 1 + 0.21*sigma*QPR[eL];

// Store into ALL QFEL, QPR, NQPR into array

ALL_QFEL[nTP,eL] = QFEL[eL];
ALL_QPR[nTP,eL] = QPR[eL];
ALL_NQPR[nTP,eL] = NQPR[eL];
}

// Close Qfile
FileClose(Qf_FileHandle);

// Write out ALL_QFEL into QFEL Datafile

FileWrite(ALL_QFEL_FileHandle,TPSymbol,ALL_QFEL[nTP,0],
ALL_QFEL[nTP,1], ALL_QFEL[nTP,2], ALL_QFEL[nTP,3], ALL_QFEL[nTP,4], ALL_QFEL[nTP,5],
ALL_QFEL[nTP,6], ALL_QFEL[nTP,7], ALL_QFEL[nTP,8], ALL_QFEL[nTP,9], ALL_QFEL[nTP,10],
ALL_QFEL[nTP,11],ALL_QFEL[nTP,12],ALL_QFEL[nTP,13],ALL_QFEL[nTP,14],ALL_QFEL[nTP,15],
ALL_QFEL[nTP,16],ALL_QFEL[nTP,17],ALL_QFEL[nTP,18],ALL_QFEL[nTP,19],ALL_QFEL[nTP,20]);

// Write out ALL_QPR into QPR Datafile

FileWrite(ALL_QPR_FileHandle,TPSymbol,ALL_QPR[nTP,0],
ALL_QPR[nTP,1], ALL_QPR[nTP,2], ALL_QPR[nTP,3], ALL_QPR[nTP,4], ALL_QPR[nTP,5],
ALL_QPR[nTP,6], ALL_QPR[nTP,7], ALL_QPR[nTP,8], ALL_QPR[nTP,9], ALL_QPR[nTP,10],
ALL_QPR[nTP,11],ALL_QPR[nTP,12],ALL_QPR[nTP,13],ALL_QPR[nTP,14],ALL_QPR[nTP,15],
ALL_QPR[nTP,16],ALL_QPR[nTP,17],ALL_QPR[nTP,18],ALL_QPR[nTP,19],ALL_QPR[nTP,20]);

// Write out ALL_NQPR into NQPR Datafile

FileWrite(ALL_NQPR_FileHandle,ALL_NQPR[nTP,0],
ALL_NQPR[nTP,1], ALL_NQPR[nTP,2], ALL_NQPR[nTP,3], ALL_NQPR[nTP,4], ALL_NQPR[nTP,5],
ALL_NQPR[nTP,6], ALL_NQPR[nTP,7], ALL_NQPR[nTP,8], ALL_NQPR[nTP,9], ALL_NQPR[nTP,10],
ALL_NQPR[nTP,11],ALL_NQPR[nTP,12],ALL_NQPR[nTP,13],ALL_NQPR[nTP,14],ALL_NQPR[nTP,15],
ALL_NQPR[nTP,16],ALL_NQPR[nTP,17],ALL_NQPR[nTP,18],ALL_NQPR[nTP,19],ALL_NQPR[nTP,20]);

Fig. 10.24 Process 6 using QP Schrodinger equation to find the first 21 energy levels
10.5 MQL Program for QPLs Evaluation … 297

//*********************************************************************************************
//
// 7. Program Termination
//
//*********************************************************************************************

// Close All DataFiles

FileClose(QPLog_FileHandle);
FileClose(Lambda_FileHandle);
FileClose(ALL_NQPR_FileHandle);

// Check Global Time

Getime = GetTickCount();
Gtlapse = Getime - Gstime;

// Output time taken

Print("Total Time Taken : ",Gtlapse," msec");

return(0);
}

int start()
{

return(0);
}

Fig. 10.25 Process 7 program end

Fig. 10.26 Implementation results of QPL2019.mq4 on Forex.com MT4 platform to calculate

QPLs of 120 financial products
298 10 Quantum Price Levels for Worldwide Financial Products

10.6 Conclusion

Based on quantum finance concepts and theories in Chaps. 1–4, and numerical QPL
(quantum price levels) evaluation learnt in Chap. 5, this chapter we studied how to
implement these methods and theories to evaluate QPLs of worldwide 120+ financial
products including: worldwide forex, major cryptocurrencies, international finance
indices and major commodities.
From the implementation perspective, we demonstrated how to use MQL
(MetaQuotes Language) on MT4 Platform—one of the world’s biggest online pro-
gram trading and development platform to develop real-time QPL evaluation system.
As mentioned in Chap. 6 about quantum trading and hedging strategy, besides
using QPL as S&R lines in program trading; QPL can also be incorporated with
chaotic neural networks (CNNs), deep neural oscillatory network (DNON) for real-
time financial predictions, which will be studied in Chaps. 11 and 12, respectively.

Problems

10.1 What is the financial market? Discuss and explain why financial market prices
may not indicate the true intrinsic value, say, a stock which leads to the
popularity of programming trading nowadays.
10.2 What is the major difference between primary versus secondary financial mar-
kets? And what is the importance of Secondary markets in modern finance?
10.3 What are the two different categories of secondary market? Give one example
for each category and discuss the major characteristics.
10.4 State and explain the five major types of financial markets. Give one example
for each type of financial market and explain how it works.
10.5 What are the two major roles and functions of market makers (MM) in a
financial market, such as forex market?
10.6 State and explain the five key participants in forex market, in the descending
order of their importance and market shares.
10.7 State and explain the five key forex trading characteristics.
10.8 What are the major factors to affect the prices in forex market?
10.9 What is retail forex trading platform? Give one example and discuss it is
different from traditional forex trading platform.
10.10 What are supports and resistance (S&R) analysis in technical analysis? Give
one example in each case and discuss how it works.
10.11 Discuss and explain how S&R analysis is related to supply and demand
analysis and theory in technical analysis. Give an example by using the daily
chart of DJI (Dow Jones Index) and explain how such theory works.
10.12 What is the major difference between role reversal vs. trend reversal on S&R
analysis by using simple charts to illustrate your explanation?
10.13 How S&R analysis is related and interpreted by QPL (quantum price level)
modeling in quantum finance?
10.6 Conclusion 299

10.14 Discuss and explain why QPL (quantum price level) in quantum finance are
critical for nowadays HFPT (High-Frequency-Program-Trading).
10.15 What are the major functions and characteristics of MetaTrader system as
compared with traditional financial trading platforms?
10.16 State and explain the five categories of programs/systems generated by MQL
of MT platform.
10.17 State and explain six major components in MQL as compared with traditional
programming languages which are critical for program trading in finance.
10.18 State six types of OrderSend() in MQL and explain how they work for the
design of trading and hedging systems.
10.19 Give five commonly used technical indicators in technical analysis and
explain how they works in MQL programming.
10.20 MQL programming task 10.1 Design and implement a real-time trading pro-
gram using MQL with the implementation of classical RSI (30-70) trading
algorithm. Financial products: Using at least five forex products for trading
(e.g., USDCAD, AUDJPY, etc.) Trading criteria: Using H1 as trading time-
frame, trigger a sell order when RSI value hit 70, trigger a buy order when it
hit 30.
Using (1) MQL built-in simulator; (2) Live test for at least 10 trading days
using demo account and compared their results.
10.21 MQL Programming Task 10.2
Similar to Task 10.1, but this time using two more technical indicators to
generate automatic trading rules. Commonly used technical indicators such
as: iBand (Bollinger Bands); iMomentum (Momentum); iMA (Moving Aver-
age); iMACD (MACD) and iStochastic (Stochastic Oscillator).
Compare the trading results with the one generated by Task 10.1.
10.22 MQL Programming Task 10.3
Based on Workshop #2 (QPL2019.mq4), design and implement a real-time
MQL trading program by combining QPLs (as S&R lines) and RSI trading
program implemented in Task 10.1 to implement a simple quantum finance
trading system.
Compare trading results with trading program implemented in Task 10.1.

References

Abreu, R. et al. (2018) On the accuracy of the Complex-Step-Finite-Difference method. Journal of

Computational and Applied Mathematics 340: 390–403.
Arnett, G. W. (2011) Global Securities Markets: Navigating the World’s Exchanges and OTC
Markets. Wiley.
Atkeson, A. G. et al. (2015) Entry and Exit in OTC Derivatives Markets. Econometrica 83(6):
2231–2292.
Brown, J. (2016) MT4/MT5 High Probability Forex Trading Method. CreateSpace Independent
Publishing Platform.
Bulkowski, T. N. (2005) Encyclopedia of Chart Patterns. Wiley 2nd edition.
300 10 Quantum Price Levels for Worldwide Financial Products

De Angelis, T. and Peskir, G. (2016) Optimal prediction of resistance and support levels. Applied
Mathematical Finance 23(6): 465–483.
Drayer, J. et al. (2008) Tradition vs. Trend: A Case Study of Team Response to the Secondary Ticket
Market. Sport Marketing Quarterly 17(4): 235.
Durbin, M. (2010) All About High-Frequency Trading. McGraw-Hill Education.
Gomes, C. and Waelbroeck, H. (2010) An empirical study of liquidity dynamics and resistance and
support levels. Quantitative Finance 10(10): 1099–1107.
Guo, G. (2005) Finite Difference Methods for the BSDEs in Finance. International Journal of
Financial Studies 6(1): 26, 2018.
Huber, M. et al. (2012) Which factors drive product sales in OTC markets? International Journal
of Pharmaceutical and Healthcare Marketing 6(4): 291–309.
Kaufman, P. J. (2013) Trading Systems and Methods. Springer.
Kirkpatrick, C. D. and Dahlquist, J. R. (2015) Technical Analysis: The Complete Resource for
Financial Market Technicians. FT Press 3rd edition.
Kuznetsov, A. (2006) The Complete Guide to Capital Markets for Quantitative Professionals
(McGraw-Hill Library of Investment and Finance). McGraw-Hill Education 1st edition.
Murphy, J. J. (1999) Technical Analysis of the Financial Markets: A Comprehensive Guide to
Trading Methods and Applications. New York Institute of Finance.
Singh, R.M. et al. (2009) The solution of the Schrödinger equation for coupled quadratic and quartic
potentials in two dimensions. Pramana 72(4): 647–654.
Tichy, V. et al. (2012) Analytic energies and wave functions of the two-dimensional Schrödinger
equation: ground state of two-dimensional quartic potential and classification of solutions. Cana-
dian Journal of Physics 90(6): 503–513.
Walker, W. (2018) Expert Advisor Programming and Advanced Forex Strategies. Independently
published.
Young, A. R. (2015) Expert Advisor Programming for MetaTrader 4: Creating automated trading
systems in the MQL4 language. Edgehill Publishing.
Zapranis, A. and Tsinaslanidis, P.E. (2012) Identifying and evaluating horizontal support and resis-
tance levels: an empirical study on US stock markets. Applied Financial Economics 22(19):
1571–1585.
Chapter 11
Time Series Chaotic Neural Oscillatory
Networks for Financial Prediction

A computer would deserve to be called intelligent if it could

deceive a human into believing that it was human.
Alan Turing (1912–1954)

Financial prediction, ranging from stocks, commodities to forex forecast poses real
challenges to researchers and quantitative analysts (a.k.a. quants) due to its highly
chaotic and unpredictable nature. With the blooming of cryptocurrency trading which
provides 24 × 7 online currency trading, the financial markets, especially the world-
wide currency markets are more chaotic and fluctuating than before. An effective
and reliable financial prediction system is profoundly required.
With the integration of quantum price levels (QPL) evaluated by quantum finance
technology, a Time Series Chaotic Neural Oscillatory Network (TSCNON) for world-
wide financial market prediction is devised to effectively resolve the system over-
training and deadlock problems imposed by traditional recurrent neural networks
using classical sigmoid-based activation functions.
In terms of system implementation, TSCNON coalesces into 2048-trading day
time series financial data and 39 major financial signals as input signals for the real-
time prediction of 129 worldwide financial products which includes: 9 major cryp-
tocurrencies, 84 forex, 19 major commodities, and 17 worldwide financial indices.
In terms of system performance, past 500-trading day of average system perfor-
mance attained less than 1% forecast percentage errors.

11.1 Introduction

With the integration of quantum price levels (QPL) discussed in Chap. 10 and chaotic
neural networks learnt in Chap. 9, in this chapter, TSCNON (time series chaotic neu-
ral oscillatory networks) is devised to effectively resolve the system over-training
and deadlock problems imposed by traditional recurrent neural networks using
classical sigmoid-based activation functions (Lee 2004, 2006a; Li and Lee 2005;

© Springer Nature Singapore Pte Ltd. 2020 301

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_11
302 11 Time Series Chaotic Neural Oscillatory Networks …

Wong et al. 2008). TSCNON coalesces 2048-trading day time series financial data
with quantum finance signals (QFS) based on QPL derived in Chap. 10 into input sig-
nals for the real-time prediction of 129 worldwide financial products which includes:
9 major cryptocurrencies, 84 forex, 19 major commodities, and 17 worldwide finan-
cial indices.
The chapter is comprised of (1) TSCNON system architecture and its chaotic
neural network training algorithm in Sect. 11.2; (2) TSCNON system implementa-
tion framework for the real-time prediction of worldwide 129 financial products in
Sect. 11.3; (3) Workshop #2—quantum finance forecast system (TSCNON.mq4) will
be discussed in details in Sect. 11.4; (4) TSCNON system performance in Sect. 11.6;
(5) TSCNON—system implementation results in Sect. 11.6 and followed by (6)
conclusion in Sect. 11.7.

11.2 TSCNON—System Architecture

TSCNON (time series chaotic neural oscillatory networks) is the integration of mul-
tilayer feedforward backpropagation networks (FFBPNs) (Fausett 1994; Patterson
1996) with Lee-oscillators (Lee 2004, 2005, 2006a, b) by replacing all simple neu-
rons with chaotic neural oscillators. The network training algorithm of TSCNON
is shown in Fig. 11.1. Figure 11.2 shows the system architecture of TSCNON for
financial prediction.

Fig. 11.1 System architecture of TSCNON (Source Dr. Lee’s illustration, 2019)
11.2 TSCNON—System Architecture 303

Fig. 11.2 System training algorithm of TSCNON (Source Dr. Lee’s illustration, 2019)
304 11 Time Series Chaotic Neural Oscillatory Networks …

As shown in Fig. 11.2, TSCNON consists of three neural network layers:

1. Input layer: consists of (1) 5-day time series input signal vector contains Open,
High, Low, and Closing prices; (2) quantum field signals (QFS) contain the 20
closest QPLs revealed in Chap. 10. For each input node given by Lee-oscillator,
totally we have 40 Lee-oscillators in the input layer (20 Lee-oscillators for time
series signals, and 20 Lee-oscillators of QFS).
2. Hidden layer: consists of 40 Lee-oscillators as hidden nodes.
3. Output layer: consists of 4 Lee-oscillators which model the next-day forecasts
of Open, High, Low, and Close, respectively.
Different from traditional neural networks (Fausett 1994; Patterson 1996), every
neuron in the TSCNON is replaced by Lee-oscillator with neural dynamics given by
Eqs. (11.1)–(11.4) mentioned in Chap. 9, which effectively converts the recurrent
neural network into a chaotic recurrent neural network to resolve the over-training
and deadlock problems, which is usually occurred in training of massive data such
as financial data we are conducted in this chapter.
The formulations of Lee-oscillator are given by

E(t + 1) = Sig[e1 · E(t) − e2 · I (t) + S(t) − ξ E ] (11.1)

I(t + 1) = Sig[i 1 · E(t) − i 2 · I (t) − ξ I ] (11.2)

(t + 1) = Sig[S(t)] (11.3)

L(t) = [E(t) − I (t)] · e−k S (t)

2
+ (t) (11.4)

where e1 , e2 , i1, and i2 are the weights; ξE and ξI are the threshold values and S(t) is
the external input.
Detail system performance analysis will be studied in the next section.

11.3 TSCNON—System Implementation Framework

Quantum finance forecast center (http://www.QFFC.org) is a nonprofit, self-funded

AI-fintech R&D and worldwide financial forecast center aims at the R&D and pro-
vision of a fair and open platform for worldwide traders and individual investors to
acquire free knowledge of worldwide 129 financial product forecasts based on state-
of-the-art AI, intelligent agents, chaotic neural networks, and quantum field theory
technologies.
With the adoption of TSCNON technology and the real-time data provided by
Forex.com (one of the major international forex trading platforms) and AvaTrade.
com (one of the biggest worldwide cryptocurrency trading platforms) (Narayanan
et al. 2016; Vigna and Casey 2016), QFFC launched the 129 financial products’ daily
11.3 TSCNON—System Implementation Framework 305

Fig. 11.3 Quantum finance forecast center official site for TSCNON daily financial forecast on
May 3, 2019 (Source Dr. Lee’s illustration, 2019)

and weekly forecast services from January 1, 2018 for over 1000 worldwide traders
and individual investors for testing and evaluation. Figure 11.3 shows the official site
of Quantum Finance Forecast Center.
From the system implementation perspective, real-time and historical data of the
worldwide 129 financial products provided by Forex.com and AvaTrade.com are
adopted in TSCNON for chaotic neural network training and prediction.
They include the following:
– major cryptocurrencies (9);
– major worldwide forex (84);
– major commodities (19);
– major worldwide financial indices (17).
Appendix A depicts the list of 129 financial products under these four categories.
As shown in Appendix A, owing to the short trading history of cryptocurrencies
(300-trading day records are provided by AvaTrade.com), all other financial products
consist of 2048 past trading day records for each financial product (data provided
306 11 Time Series Chaotic Neural Oscillatory Networks …

by Forex.com), which provide sufficient training and test sets for TSCNON system
testing and evaluation.
To provide a fully coherent and automation of TSCNON with both Forex.com
and AvaTrade.com trading platforms for the automatic acquisition of real-time and
historical data, the whole TSCNON system is developed in MT platform (Walker
2018; Young 2015) using MetaQuotes Language (MQL) and Expert Advisor (EA)
system for daily financial forecast. Figure 11.4 shows the system framework of
TSCNON.
As shown in Fig. 11.4 each financial product has 2048-trading day data (except
cryptocurrency which only have 300-trading day data) are automatically generated
by the MT4 engines of Forex.com and AvaTrade.com on a daily basis. Through the
QPL (quantum price level) Generator revealed in Chap. 10, 21 closed QPL signals
are generated by TSCNON together with the previous 5-day time series patterns;
they are fed into TSCNON for chaotic neural network training and testing.

Time Series
Financial Data
(1..T Trading Days)

QPL Generator
(21 QPL Signals)

Time Series
Chaotic Neural Network TSCNON
(120 Financial Products,FP1..FP120)

Financial Forecast
Results T+1
(FFR1..FFR120)

Fig. 11.4 TSCNON System Framework (Source Dr. Lee’s illustration, 2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 307

11.4 Workshop #2—Quantum Finance Forecast System

(TSCNON.Mq4)

11.4.1 Introduction

In this section, we will explore the programming details of quantum finance forecast
system (TSCNON) using MQL for real-time implementation on any MT4 platforms.
For the time series chaotic neural oscillatory network (TSCNON), we are using the
QFFC SDK developed by QFFC.org.
This section will study the complete MQL programming details for the design
and development of quantum finance forecast system by using TSCNON technology
in this chapter. Users can base on the method and program learnt in this chapter to
design and implement their own quantum finance prediction systems.
Since the entire TSCNON.mq4 consists of over 1000 program lines of coding, this
section will focus on the main module implementation process. Complete program
coding can be found in QFFC.org Workshop #2.

11.4.2 Main Program Logic

TSCNON.mq4 is a MQL4 program with the implementation of TSCNON system in

this chapter for the daily forecast of 120 financial products provided by Forex.com
MT4 platform. To use this program for any other MT4 platforms, simply replace the
product code array (TP_Code[]) with the ones adopted by users’ own MT4 system.
The entire system flowchart is shown in Fig. 11.5.
As shown in Fig. 11.6, quantum finance forecast center (a.k.a. QFFC) has imple-
mented TSCNON system training and forecast functions and packaged into QFSDK
software library. To use QFSDK software library, users should do the following:

Fig. 11.5 Main program logic of TSCNON.mq4 (Source Dr. Lee’s illustration, 2019)
308 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.6 System flowchart

of TSCNON system (Source START
Dr. Lee’s illustration, 2019)

1. PROGRAM INITIALIZZATION

2. FORECAST DAY?

3. FOR EACH FINANICAL PRODUCT IN FORECAST LIST

4. READ FC_DT/FC_HK/FC_PF DATAFILES

5. CREATE TSCNON AND INITIALIZE ALL LEE-OSCILLATORS

6. REGISTER TIME SERIES DATA

7. TSCNON SYSTEM TRAINING

8. TSCNON NEXT-DAY FORECAST

9. WITH RANGE?

10. FORECAST COMPLETED, STORE DATA AND F/C RESULT

END
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 309

1. Visit QFFC.org official site to download QFSDK;

2. Copy QFSDKv1.dll (or latest version) library file into your MT4 Libraries folder
such as “MetaQuotes\Terminal\..\MQL4\Libraries”;
3. Copy QFSDKv1.mqh (or latest version) header file into your MT4 Include direc-
tory such as “\MetaQuotes\Terminal\..\MQL4\Include”;
4. In users’ individual TSCNON.mq4 program, add #include <QFSDKv1.mqh>,
as shown in Fig. 11.5.
For the latest information about QFSDK, please visit QFFC.org official site.
In summary, TSCNON.mq4 is a real-time MQL-based multiple financial product
daily forecast using TSCNON revealed in this chapter. It integrates QPL evaluation
system studied in Chap. 10 with TSCNON network construction, training, and fore-
cast algorithm into QFSDK function library. Quantum finance developers only need
to follow this system flowchart and call these QFSDK functions to build their own
real-time quantum finance forecast systems.
Figure 11.6 shows TSCNON system flowchart. The entire TSCNON consists
of 15 steps of programming process. As the entire program consists of over 1000
program codes, we will only discuss the core concepts and program fragment of each
function process. For full and complete MQL source code, please visit QFFC.org
official site—Workshop #2 quantum finance forecast system.

11.4.3 Process #1—Program Initialization

11.4.3.1 Declaration of Data Files and Directories

The program initialization process consists of the following main modules:

1. declaration of data files and directories;
2. declaration of system parameters;
3. declaration of TSCNON forecast elements and variables;
4. creation of init() function class.
Actually, TSCNON is the first generation of quantum finance forecast system used
in QFFC.org for real-time daily forecast and forecast performance analysis system of
worldwide 129 financial products started from 2017. In other words, TSCNON not
only performs the daily forecast but also analyses the previous trading day forecast
performance, stored into the forecast performance data files for archive and upload
to QFFC.org for public access. As shown in Fig. 11.7, four directories are used for
data storage:
1. 120FC_DL stores the daily forecast results of 120 financial products that include
forecasts high/low and the closest 8 QPLs, which will also upload to QFFC.org
official site every morning. This directory contains two data files: FC_DAILY
and PF_DAILY. FC_DAILY contains the daily forecast table and PF_DAILY
contains the daily forecast performance table. They are used for daily forecast
310 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.7 Declaration of data files and directories in TSCNON (Source Dr. Lee’s illustration, 2019)

archive and upload onto the QFFC.org official site. Figure 11.8 shows the daily
forecast table for the first 20 financial products on May 3, 2019. Figure 11.9
shows the daily forecast performance table for the top 20 financial products on
May 2, 2019 (sorted by the % errors).
2. 120FC_DT stores the normalized daily time series (DT) data (i.e., Open, High,
Low, and Close) of the past 500-trading days. Totally 120 data files with each con-
taining the time series of one trading product. For example, “1 XAGUSD_DT”
contains the (normalized) time series of Silver/USD, where “1” is the product
code number.
Figure 11.10 shows the (normalized) daily time series of XAGUSD.

Fig. 11.8 FC_DAILY—TSCNON daily forecast table of first 20 financial products on May 3, 2019
(Source Dr. Lee’s illustration, 2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 311

Fig. 11.9 PF_DAILY—TSCNON daily forecast performance table of the top 20 financial products
on May 2, 2019 (Source Dr. Lee’s illustration, 2019)

Fig. 11.10 Normalized

daily time series of
XAGUSD (Source Dr. Lee’s
illustration, 2019)

3. 120FC_HL contains previous 500-trading days forecast High/Low of each finan-

cial product. They are used for forecast archive and daily performance analysis.
Figure 11.11 shows the normalized forecast H/L of XAGUSD.
4. 120FC_PF contains daily forecast performance data files for each of the 120
financial products. As mentioned, TSCNON not only performs the daily forecast,
but also performs the previous-day forecast performance analysis by calculating
the forecast errors between previous-day forecast H/L with the actual H/L, stores
into the 120FC_PF directory for the archive. Figure 11.12 shows the daily forecast
performance table of XAGUSD. In which the first three columns are year, month,
and day, columns D–G contain the actual Open, High, Low, and Close, columns
H and I contain the forecast High and Low, columns J & K contain the forecast
errors for High/Low and columns L and M contain their % errors.
312 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.11 Normalized forecast H/L of XAGUSD (Source Dr. Lee’s illustration, 2019)

Fig. 11.12 Daily forecast performance table of XAGUSD (Source Dr. Lee’s illustration, 2019)

11.4.3.2 Declaration of System Parameters

This section declares the key parameters for the TSCNON system, they include the
following:
• number of network layers;
• number of Lee-oscillators input, hidden and output layers;
• maximum training iterations (in case of deadlock);
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 313

• training and time series sizes;

• target RMSE (root-mean-square error);
• trading Product (TP) parameters include: TP code (only shown 5 for illustration
purpose); TP code number (TP_No); TP number of digits (TP_nD); TP normal-
ization value (TP_Norm); TP daily range between daily H/L (TP_range); stop loss
(TP_SL), and target_price (TP_TP).
Figure 11.13 shows the declaration section of system parameters of TSCNON
program.

Fig. 11.13 Declaration of system parameters (Source Dr. Lee’s illustration, 2019)
314 11 Time Series Chaotic Neural Oscillatory Networks …

11.4.3.3 Declaration of TSCNON Forecast Elements and Variables

This section declares the forecast elements and variables include: previous-day record
elements for Open, High, Low, and Close (OHLC); today forecast elements for OHLC
and QPL variables for the 8 closest QPLs as in Fig. 11.14.

11.4.3.4 Declaration of Variables in Init() Function Class

Declaration of variables in init() include the creation of init() and the declaration of
all variables and arrays used in the program. As mentioned previously, init() instead
of start() class is used as the forecast system is a “one-off” real-time system being

Fig. 11.14 Declaration of forecast elements and variables (Source Dr. Lee’s illustration, 2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 315

used only once every morning of trading day; instead of trading program, which will
be operated continuously throughout the entire trading day. Figure 11.15 shows the
variables and arrays declaration section of the init() class.

11.4.4 Process #2—Check for Forecast Day

As daily forecast will only perform during trading days from Monday to Friday (even
though there are half-day trading in Saturday, but normally will not perform daily
forecast as the half-day time period is too short for actual prediction).
This process will check whether today is a weekday (i.e., Mon–Fri) or not. If it is a
weekday, b_Forecast will be set to TRUE and the system will read FX_NQPR.csv data
file. Actually, this file is the implementation results of the QPL evaluation program
(QPL2019.mq4) revealed in Chap. 10 for the QPLs evaluation of worldwide financial
products, and NQPR.csv contains 21 NQPRs of 120 financial products.
By multiplying the opening price for any trading day, users can calculate the 21
closest QPLs on both sides of the opening price for TSCNON network training and
the closest 8 QPLs evaluation for daily information update in the QFFC.org official
site.
In addition, this section also prepares the file header for daily forecast file
(FC_DAILY) and daily performance file (PF_DAILY), respectively. Figure 11.16
shows the program section of the checking for forecast day.

11.4.5 Process #3—Loop over ALL Trading Products

This section will start the looping of ALL trading product data acquisition and fore-
casting. For every financial market, two cases exist. Either today day bar exists or
not. If it does not exist, we set yesterday trading day closing as today opening to cal-
culate the 8 closest QPLs and forecast result adjustment. In either case, the system
will read and store the previous-day time series (i.e., O, H, L, C) for today forecast,
archive, and calculation of yesterday performance (Fig. 11.17).

11.4.6 Process #4—Read FC_DT/HL/PF Data Files

This section will read three data source of previous time series, which are as follows:
• FC_DT—to read the previous 500-trading day normalized time series data,
together with the previous-day OHLC for today forecast;
316 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.15 Declaration of variables in the init() (Source Dr. Lee’s illustration, 2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 317

Fig. 11.16 Process #2—Check for forecast day (Source Dr. Lee’s illustration, 2019)
318 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.17 Process #3—Loop over all trading products (Source Dr. Lee’s illustration, 2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 319

• FC_HL—to read the previous 500-trading day forecast H/L data, together with
the previous-day actual H/L results to calculate yesterday forecast performance
and stored into FC_PF file. The system will also store today’s forecast results after
daily forecast;
• FC_PF—to open the previous 500-trading day daily performance data, and store
yesterday forecast performance results.

11.4.6.1 Process #4.1—Read FC_DT Data Files

This process will check whether the first data record is the current D1 Bar. If yes,
there is no need to store it into the data array. In either case, the system reads all 500
previous-day time series into the DT data array for network training (Fig. 11.18).

11.4.6.2 Process #4.2—Write FC_DT Data File if b_PF Is TRUE

This section checks whether b_PF is true or not. If it is true, record the previous-day
time series into the FC_DT data file. Figure 11.19 shows the program code of Process
#4.2.

11.4.6.3 Process #4.3—Read FC_HL Data Files

This section will read FC_HL data file, which stores the daily forecast H/L of the
previous 500-trading day (Fig. 11.20).

11.4.6.4 Process #4.4—Read FC_PF Data Files

This section will read the FC_PF data file. This data file contains the previous 500-
trading day daily forecast performance results (Fig. 11.21).

11.4.6.5 Process #4.5—Write FC_PF Data File if bWrite Is TRUE

This section will check whether bWrite is TRUE. If it is true, calculate the yester-
day forecast H/L errors, overall % errors, and store them into the FC_PF data file
(Fig.11.22).
320 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.18 Process #4.1—read FC_DT file (Source Dr. Lee’s illustration, 2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 321

Fig. 11.19 Process #4.2 write FC_DT data file if b_PF is TRUE (Source Dr. Lee’s illustration,
2019)

11.4.7 Process #5—Create TSCNON and Initialize ALL

LEE-Oscillators

By using QFSDK, this process creates the TSCNON network and initial all the
Lee-oscillator in different network layers (Fig. 11.23).

11.4.8 Process #6—Register Time Series Data

This process calls prepareData() function to register the 500 time series input data
one-by-one. Also, it makes use of NQPR array to generate all 21 QPLs for network
inputs (Fig. 11.24).

11.4.9 Process #7—TSCNON Training

This process performs network training with target RMSE defined in the declaration
section (Fig. 11.25).
322 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.20 Process #4.3 read FC_HL data file (Source Dr. Lee’s illustration, 2019)

11.4.10 Process #8—TSCNON Forecast

This section registers the past 5-trading day time series and 21 QPLs to for next-day
forecast (Fig. 11.26).
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 323

Fig. 11.21 Process #4.4 read FC_PF data file (Source Dr. Lee’s illustration, 2019)
324 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.21 (continued)

Fig. 11.22 Process #4.5 write FC_PF data file if bWrite is TRUE (Source Dr. Lee’s illustration,
2019)
11.4 Workshop #2—Quantum Finance Forecast System (TSCNON.Mq4) 325

Fig. 11.23 Process #5 Create TSCNON and Initialize ALL LEE-oscillators (Source Dr. Lee’s
illustration, 2019)

11.4.11 Process #9—Check Forecast Within Range or NOT

After TSCNON daily forecast, this process evaluates the range between forecast
H and L can check whether it is within the acceptable range defined in the declaration
section. It is to check whether the forecast OHLC are in logical values to decide
whether it needs to forecast again (Fig. 11.27).

11.4.12 Process #10.1—Forecast Completed, Store Forecast

Results and Total Time Taken

After the daily forecast is completed and within an acceptable range, this process
calculates the total time taken, calculates the closest 8 QPLs and records the forecast
results into FC_DAILY data file, for data archive and upload to QFFC.org official
site (Fig. 11.28).

11.4.13 Process #10.2—Write FC_HL Data File and Close

All Files

This process writes the Forecast H/L into FC_HL data file and closes all data files
(Fig. 11.29).
326 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.24 Process #6 register time series data (Source Dr. Lee’s illustration, 2019) Note Training
data file “Data_4_mix.csv” contains 744 records and for each record includes 40 data, the first 36
data presenting 9-days of O, H, L, C, and the last 4 data presenting next day of O, H, L, C

11.5 TSCNON—Implementation Results

Figure 11.30 shows a snapshot of TSCNON system training and forecast process of
120 financial products of Forex.com on May 6, 2019 in the server farm of quantum
finance forecast center using Intel i5 CPU 2.39 GHz 32 MB RAM Dell Server.
11.5 TSCNON—Implementation Results 327

Fig. 11.25 Process #7 TSCNON training (Source Dr. Lee’s illustration, 2019)

Fig. 11.26 Process #8 TSCNON forecast (Source Dr. Lee’s illustration, 2019)
328 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.27 Process #9—check forecast within range or NOT (Source Dr. Lee’s illustration, 2019)

As shown in Fig. 11.30, in a typical daily forecast of 120 financial products on

Forex.com MT4 platform, the TSCNON system only took 91,547 ms (91.547 s) to
finish the training and forecast of 120 financial products.
On average, it took 0.763 s (less than 1 s) to complete the network training and
forecast process of a single financial product. Figure 11.31 shows the snapshot of
TSCNON system for the system training and forecast of 9 major cryptocurrencies
over AvaTrade.com MT platform on May 6, 2019.
As shown in Fig. 11.31, in a typical forecast day, TSCNON took 56,578 ms
(56.578 s) to finish the training and forecast of the 9 cryptocurrencies. That is, on
average it took 6.275 s to train and forecast a single cryptocurrency.
As compared with all those 120 non-cryptocurrency products, TSCNON took
8.22 times to predict cryptocurrency, even though cryptocurrency only have 300-
trading day records while the other 120 financial products each have 2048-trading
day records for system training. It may due to the fact that cryptocurrencies, in
general, are much more chaotic and fluctuant in nature, which takes more time and
iterations for TSCNON to learn the market pattern.
11.6 TSCNON—System Performance 329

Fig. 11.28 Process #10.1—forecast completed, store forecast results, and total time taken (Source
Dr. Lee’s illustration, 2019)

11.6 TSCNON—System Performance

From the system performance and evaluation perspective, TSCNON system evalu-
ated the daily forecast performance of the 129 financial products in four timeframes:
daily, weekly average, monthly average, and past 500-day average. Figure 11.32
presents the past 500-day performance ranking list of the top 20 financial products.
As shown, the 500-day average forecast % error of the top 20 financial products
ranging from 0.011% to 0.227%, respectively, which is somewhat promising and sig-
nificant as reflected by over 1000+ members of QFFC which consist of professional
forex traders, quants, and investors.
330 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.29 Process #10.2—write FC_HL data file and close all files (Source Dr. Lee’s illustration,
2019)

To differentiate system performance with traditional FFBPN (feedforward back-

propagation network), four categories of worldwide 129 financial products are tested
using FFBPN and TSCNON systems with network training error rates ranging from
1 × 10−5 to 1 × 10−6 , respectively.
Table 11.1 presents the system performance comparison chart of both systems.
As shown in Table 11.1, system simulations using error rate 1 × 10−5 (Case
I); the average total time taken for 129 financial products training using FFBPN
and TSCNON were 2,915,709 ms (48.6 min) and 40,049 ms (40.05 s), respectively.
TSCNON outperformed FFBPN by 72.8 times in terms of network training time
to attain the same error rate. Also, system simulations in Case II using error rate
1 × 10−6 , TSCNON completed the system training task in 147,940 ms (147.9 s)
while FFBPN encounters “deadlocks” during system training for cryptocurrency
and forex products; which clearly demonstrated the resolution of over-training and
deadlock problems with sufficient improvement in system training efficiency by
using TSCNON over traditional neural networks.
11.6 TSCNON—System Performance 331

Fig. 11.30 Snapshot of TSCNON (Forex.com) for system training and forecast of 120 financial
products for Forex.com MT4 platform on May 6, 2019 (Source Dr. Lee’s illustration, 2019)

In terms of system performance across different financial products, the simulation

results clearly showed that both cryptocurrency and forex are more chaotic and
difficult for network training than other financial products as expected, which will
be further explored in the future research of QFFC.

11.7 Conclusion

TSCNON (time series chaotic neural oscillatory network) is for real-time financial
prediction incorporated into quantum finance forecast center—a nonprofit making
AI-fintech R&D and worldwide financial forecast center aims at R&D and provision
of a fair and open platform for worldwide traders and individual investors to acquire
free knowledge of worldwide 129 financial product forecasts based on state-of-the
-art AI, chaotic neural networks and quantum field theory technologies.
If TSCNON is so effective to forecast worldwide financial markets, is it the end
of story?
332 11 Time Series Chaotic Neural Oscillatory Networks …

Fig. 11.31 Snapshot of TSCNON (AvaTrade.com) for system training and forecast of 9 major
cryptocurrencies for AvaTrade.com MT4 platform on May 6, 2019 (Source Dr. Lee’s illustration,
2019)

The truth is: For a professional trader and investor, a reliable and effective financial
forecast system is only the beginning of the story. A good financial investment also
needs: (1) good, effective trading and hedging strategies; (2) stable, logical with
rational investment psychology, which are also the future R&D directions of quantum
finance forecast center.
Current research of QFFC include the following:
(1) Integration of TSCNON with fractal technology for market trends/patterns min-
ing and prediction;
(2) R&D on quantum entanglement of quantum finance system on severe financial
event modeling and prediction;
(3) Design and implementation of chaotic type-2 transient-fuzzy deep neural oscil-
latory network for real-time financial prediction;
(4) Design and develop intelligent agent-based hedging and trading systems as
discussed in Chap. 13 (Lee 2019).
11.7 Conclusion 333

Fig. 11.32 Past 500-day system performance ranking chart of top 20 financial products
(Source Dr. Lee’s illustration, 2019) Note 1. High (Error) = Abs(HighForecast − HighActual ) 2. Low
(Error) = Abs(LowForecast − LowActual ) 3. Average (Error) = Average(High(Error), Low(Error))
4. % Error = Average(Error)/CloseActual

Problems

11.1 What is a recurrent neural network (RNN)? How it is different from the
traditional neural network in terms of network architecture.
11.2 Explain how to integrate Lee-oscillator(s) into a recurrent neural network
(RNN) to convert RNN into chaotic RNN.
11.3 In TSCNON, we use the closest, say, 20 quantum price levels (QPLs) as
quantum finance signals (QFSs) for network training.
334 11 Time Series Chaotic Neural Oscillatory Networks …

Table 11.1 System performance comparison chart (FFBPN vs. TSCNON)

Product No. of FFBPNa TSCNONa
category products Total STTb Average STTc Total STTb Average STTc
Case I (Error rate = 1 × 10−5 )
Cryptocurrency 9 1,460,000 162222.22 15,783 1753.67
Forex 84 1,235,543 14708.85 19,126 227.69
Financial 19 111,024 5843.37 2040 107.37
index
Commodity 17 109,142 6420.12 3100 182.35
Overall 129 2,915,709 22602.40 40,049 310.46
Case II (Error rate = 1 × 10−6 )
Cryptocurrency 9 DLd – 58,400 6488.89
Forex 84 DLd – 72,679 865.23
Financial 19 577,324 30385.47 6939 365.21
index
Commodity 17 687,595 40446.76 9922 583.65
Overall 129 – – 147,940 1146.82
Note
a Results are generated by 500 simulations of both FFBPN and TSCNON system training (in ms)
b “Total STT” denotes the total average system training time for 500 simulations of network training

using FFBPN and TSCNON system for each category of financial products
c “Average STT” denotes the average system training time for a single financial product of each

category
d “DL” denotes deadlock during system training

(i) What is the ground state (i.e., QPL0 ) of QPL? Why?

(ii) Using numerical computational method learnt in Chap. 10, work out the
formulation of the closest 21 QPLs.
(i.e., QPL+10 … QPL0 … QPL-10 )
11.4 In TSCNON, we use normalized quantum price levels (QPLs) as input vectors
of QFS, can we use the normalized quantum price return (QPRs) instead of
QPLs as input signals? Why?
11.5 The following figure shows the bifurcation diagram of Lee-oscillator dis-
cussed in Chap. 9.
11.7 Conclusion 335

(i) Based on this bifurcation diagram and Eqs. (11.1)–(11.4), write a MAT-
LAB program to generate the TCAF (transient chaotic activation func-
tion) with external input stimulus S(t) and L(t) in the normalization
region [0, 1].
(ii) From the TCAF, identify the active bifurcation region.
(iii) During the normalization process of all input signals in the TSCNON,
what is the optimal normalization region? Why?
11.6 In TSCNON discussed in this Sect. 11.2, 5-day timeseries are used as input
signal vectors with QPLs to forecast next-day high/low Prices. Why? Try to
use more trading days, says 21 days to do next-day forecast and check the
difference.
11.7 In this chapter, we used TSCNON for daily financial prediction. Discuss and
explain how we can use TSCNON for weekly and monthly prediction in
terms of (1) time series input vectors; (2) time horizon for network training;
and (3) QPLs being used. Do we need to recalculate all the QPLs? Why?
11.8 In Chap. 8, we learnt how initial conditions are critical for the prediction
of complex and highly chaotic systems like weather and financial markets.
Discuss why we can still use chaotic neural networks to perform financial
prediction? What is the basic concept and logic behind?
11.9 In QFFC.org, four timeframes of daily system performance analysis are gen-
erated: daily, weekly, monthly, and 500-Day average daily performance. Dis-
cuss and explain physical meaning and importance of each type of perfor-
mance results in terms of (1) system stability and (2) significance on different
timeframes of trading. Which financial products are the best in terms of fore-
cast performance? Why?
336 11 Time Series Chaotic Neural Oscillatory Networks …

11.10 Programming task 11.1

(i) Follow workshop #2 and implement the TSCNON system for at least 10
different Forex products, using 5-day time series and 21 closest QPLs
to predict next-day high/low–system A.
(ii) Implement the same forecast system but this time without using QPLs
as input QFS–system B.
(iii) Compare systems A and B in terms of (1) system learning performance
in 1000 iterations; (2) next-day forecast performance.
11.11 Programming task 11.2
(i) As an extension of task 11.1, this time change the number of closest
QPLs from 10 to 30, with an increment of 2 QPLs (i.e.,one more QPL
on each side). So totally we have 11 sets of systems for testing.
(ii) Modify TSCNON.eq4 program to automate system testing process of
these 11 experiments.
(iii) Compare these 11 systems in terms of (1) System learning performance
in 1000 iterations; (2) next-day forecast performance, and locate the
optimal number of QPL values.

Acknowledgements The author wishes to thank Forex.com and AvaTrade.com for the provision
of historical and real-time financial data. The author also wishes to thank Quantum Finance Forecast
Center of UIC for the R&D supports and the provision of the channel and platform QFFC.org for
worldwide system testing and evaluation.

References

Fausett, L. Fundamentals of Neural Networks Architectures Algorithms and Applications. Prentice

Hall, 1994.
Lee, R. S. T. LEE-Associator—A Transient Chaotic Autoassociative Network for Progressive Mem-
ory Recalling. Neural Networks. 19(5): 644–666, 2006a.
Lee, R. S. T. Fuzzy-Neuro Approach to Agent Applications (From the AI Perspective to Modern
Ontology). Springer-Verlag, Germany, 2006b.
Lee, R. S. T. Advanced Paradigms in Artificial Intelligence From Neural Oscillators, Chaos Theory
to Chaotic Neural Networks. Advanced Knowledge International, Australia, 2005.
Lee, R. S. T. A Transient-chaotic Auto-associative Network (TCAN) based on LEE-oscillators.
IEEE Trans. Neural Networks. 15(5): 1228–1243, 2004.
Lee, R. S. T. (2019) Chaotic Type-2 Transient-Fuzzy Deep Neuro-Oscillatory Network
(CT2TFDNN) for Worldwide Financial Prediction. IEEE Transactions on Fuzzy System. https://
doi.org/10.1109/tfuzz.2019.2914642.
Li, G. C. L. and Lee, R. S. T. A Real-Time Scene Segmentation System Using Solely Excitatory
Oscillator Networks (SEON). Journal of Intelligent Manufacturing. 16(6): 669–678, 2005.
Narayanan A. et al. Bitcoin and Cryptocurrency Technologies: A Comprehensive Introduction.
Princeton University Press, 2016.
Patterson, D. W. Artificial Neural Networks. Prentice Hall, 1996.
References 337

Vigna, P. and Casey, M. J. The Age of Cryptocurrency: How Bitcoin and the Blockchain Are Chal-
lenging the Global Economic Order. Picador, 2016.
Walker, W. Expert Advisor Programming and Advanced Forex Strategies. Independently published,
2018.
Wong et al. Wind Shear Forecasting by Chaotic Oscillatory-based Neural Networks (CONN) with
Lee-oscillator (Retrograde Signaling) Model. International Joint Conference on Neural Networks
(IJCNN), 2040–2047, 2008.
Young, A. R. Expert Advisor Programming for MetaTrader 4: Creating automated trading systems
in the MQL4 language. Edgehill Publishing, 2015.
Chapter 12
Chaotic Type-2 Transient-Fuzzy Deep
Neuro-Oscillatory Network
(CT2TFDNN) for Worldwide Financial
Prediction

In this chapter, the author proposed a chaotic type-2 transient-fuzzy deep neuro-
oscillatory network with retrograde signaling (aka CT2TFDNN) for worldwide finan-
cial prediction. With the extension of the author’s original work on Lee-oscillator—
a chaotic discrete-time neural oscillator with profound transient-chaotic property,
CT2TFDNN provides: (1) effective modeling of Interval type-2 fuzzy logic with
chaotic type-2 transient-fuzzy membership function (CT2TFMF); (2) effective time
series network training and prediction using chaotic deep neuro-oscillatory network
with retrograde signaling (CDNONRS). CT2TFDNN not only provides a fast chaotic
fuzzy-neuro deep learning and forecast solution, but also it successfully resolves the
massive data over-training and deadlock problems, which are usually imposed by tra-
ditional recurrent neural networks using classical sigmoid-based activation functions.
From the implementation perspective, CT2TFDNN is integrated with 2048-trading
day time series financial data and top-10 major financial signals as fuzzy financial
signals (FFS) for the real-time prediction of 129 worldwide financial products which
consists of: 9 major cryptocurrencies, 84 worldwide forex, 19 major commodities,
and 17 worldwide financial indices.
Over the years, financial engineering ranging from the study of financial signals
to the modeling of financial prediction is one of the most exciting topics for both
academia and financial community. With the flourishing of AI technology in the past
20 years, various hybrid intelligent financial prediction systems with the integration
of neural networks, chaos theory, fuzzy logic, and genetic algorithms were proposed.
Interval type-2 fuzzy logic system (IT2FLS) with its remarkable capability for the
modeling of highly uncertain events and attributes, provides a perfect tool to interpret
various financial phenomena and patterns.
In this chapter, the author proposed a chaotic type-2 transient-fuzzy deep neuro-
oscillatory network with retrograde signaling (aka CT2TFDNN) for worldwide finan-
cial prediction (Lee 2019a, b). With the extension of author’s original work on Lee-
oscillator—a chaotic discrete-time neural oscillator with profound transient-chaotic

© Portions of this chapter are reprinted from Lee (2019a), with permission of IEEE.

© Springer Nature Singapore Pte Ltd. 2020 339

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_12
340 12 Chaotic Type-2 Transient-Fuzzy Deep …

property, CT2TFDNN provides: (1) effective modeling of Interval type-2 fuzzy logic
with chaotic type-2 transient-fuzzy membership function (CT2TFMF); (2) effective
time series network training and prediction using chaotic deep neuro-oscillatory net-
work with retrograde signaling (CDNONRS). CT2TFDNN not only provides a fast
chaotic fuzzy-neuro deep learning and forecast solution but also more prominently
successfully resolves the massive data over-training and deadlock problems, which
are usually imposed by traditional recurrent neural networks using classical sigmoid-
based activation functions. From the implementation perspective, CT2TFDNN is
integrated with 2048-trading day time series financial data and top-10 major financial
signals as fuzzy financial signals (FFS) for the real-time prediction of 129 worldwide
financial products which consists of: 9 major cryptocurrencies, 84 worldwide forex,
19 major commodities, and 17 worldwide financial indices.

12.1 Introduction

This chapter introduces an innovative chaotic type-2 transient-fuzzy deep neuro-

oscillatory network with retrograde signaling (aka CT2TFDNN) for worldwide finan-
cial time series prediction.
With the extension of author’s original work on Lee-oscillator—a discrete-time
neural oscillator with profound transient-chaotic property on temporal information
progressing (Lee 2004a, 2005, 2006a, b), together with the author’s latest works
on the exploration of Lee-oscillator with retrograde signaling (LORS), which gen-
eralizes into eight categories of bifurcation transfer functions for deep learning,
CT2TFDNN successfully integrates discrete-time chaotic neural oscillators (for
chaotic neural activity modeling), chaotic deep neuro-oscillatory network with 8-
hidden layers of neuro-oscillators with various degree of retrograde signaling (for
chaotic time series deep learning and prediction) and chaotic type-2 transient-fuzzy
logic (for high-order uncertainty modeling of financial signals) into: (1) effective
modeling of type-2 fuzzy logic with chaotic type-2 transient-fuzzy membership
function (CT2TFMF); (2) effective time series network training and prediction using
chaotic deep neuro-oscillatory network with retrograde signaling (CDNONRS).
CT2TFDNN not only provides a fast chaotic fuzzy-neuro deep learning and
forecast solution, more prominently it successfully resolves the massive data over-
training and deadlock problems, which are usually imposed by traditional recurrent
neural networks and fuzzy neural networks using IT2 fuzzy membership functions
and/or hybrid neural networks using classical sigmoid-based activation functions.
From the implementation perspective, CT2TFDNN is integrated with 2048-
trading day time series financial data and top-10 major financial signals as input
signals for the real-time prediction and agent-based trading of 129 worldwide finan-
12.1 Introduction 341

cial products which consists of: 9 major cryptocurrencies, 84 worldwide forex, 19

major commodities, and 17 worldwide financial indices.
The main contributions and originality of this system include:
1. The introduction of an innovative type-2 fuzzy logic system (T2FLS), which
is inspired by the author’s original work on Lee-oscillators (Lee 2004). Dif-
ferent from the contemporary research on T2FLS which mainly focus on the
R&D of the interval type-2 logic system (IT2FLS)—a simplified version of
T2FLS due to its computational complexity, this chapter proposed a chaotic
type-2 transient-fuzzy logic system (CT2TFLS)—a truly IT2LS with remark-
able chaotic transient-fuzzy property to resolve the computational complexity
problem.
2. The introduction of CT2TFDNN—an innovative fuzzy-neuro deep neural net-
work, which is not only the integration of two AI technologies as two sepa-
rate functional modules proposed by numerous hybrid fuzzy-neuro systems, the
CT2TFDNN introduced in this chapter is constructed by LORS—chaotic neural
oscillators invented by the author which serves as transient-fuzzy input neurons
of the DNN and effectively converts it into CT2TFDNN system. In other words,
the chaotic transient-fuzzification process is actually part of the neural model of
CT2TFDNN.
3. The successful resolution of the massive data over-training and deadlock prob-
lems, which are usually imposed by traditional recurrent neural networks using
classical sigmoid-based activation functions.
4. The successful design and implementation of real-time financial forecast system
of worldwide financial products with state-of-the-art CT2TFDNN system.
This chapter is comprised of (1) literature review on (i) type-1 and interval type-2
fuzzy logic systems (FLS) on financial prediction; (ii) chaotic neural networks, Lee-
oscillator, and the author’s latest work on Lee-oscillator with retrograde signaling
(LORS) in Sect. 12.2; (2) system framework and methodology of CT2TFDNN for
financial time series prediction which include: chaotic type-2 transient-fuzzification
scheme, GA-based top-10 fuzzy financial signals (FFS) selection scheme and the
chaotic deep neuro-oscillatory network learning algorithm in Sect. 12.3; (3) system
implementation of CT2TFDNN for real-time prediction of worldwide 129 finan-
cial products in Sect. 12.4; (4) performance analysis of proposed CTSTFDNN
and comparison with five forecast systems: (i) traditional time series feedforward
backpropagation networks (FFBPN), (ii) support vector machine (SVM) provided
by R Project—one of the most popular financial forecasting tools used in finance
industry, (iii) deep neural network (DNN) with PCA (principal component analysis)
model (Singh and Srivastava 2017), (iv) classical interval type-2 fuzzy-neuro net-
work (IT2FNN), (v) chaotic type-1 fuzzy-neuro-oscillatory network (CT1FNON) in
Sect. 12.5; and (5) conclusion in Sect. 12.6.
342 12 Chaotic Type-2 Transient-Fuzzy Deep …

12.2 Literature Review

In a typical financial prediction problem, there are two basic problems we must tackle
with: (1) massive and highly uncertain financial patterns and signals; (2) massive time
series data and information for system training and machine learning. fuzzy logic
system (FLS) (Zadeh 1975; Ross 2016; Siler and Buckley 2004; Zadeh and Aliev
2019) with its intrinsic property of handling attributes of uncertainty, provides a
viable solution for system modeling and data representations.
This section gives a general overview of type-1 and type-2 FLSs, their systems’
characteristics and limitations, and the current research of FLS in financial predic-
tions. To tackle with highly chaotic and massive time series data such as severe
weather information and financial data, data scientists, and AI researchers are now
exploring the possibility of adopting chaotic neural oscillators into traditional recur-
rent neural networks—chaotic neuro-oscillatory networks (CNON, or chaotic neural
networks, CNN in short) for chaotic time series modeling and forecast. This section
gives a general overview of chaotic neural oscillator and the author’s original work
on Lee-oscillator, its neural architecture, and the latest works on Lee-oscillator with
retrograde signaling (LORS).

12.2.1 An Overview on Type-1 and Interval Type-2 Fuzzy

Logic Systems on Financial Prediction

Like many real-world phenomena, the linguistic variables used to describe financial
markets (or market patterns), ranging from worldwide financial indices and stocks,
to highly volatile markets such as worldwide forex and cryptocurrencies are typical
fuzzy logic systems (FLS) with high degree of uncertainty and fuzziness (Ross 2016;
Siler 2004; Zadeh 1975; Zadeh and Aliev 2019).
For example, to describe a typical stock market (e.g., Nasdaq Index), financial
professionals are usually using fuzzy linguistic variables such as: bearish, bullish,
over-buy, or over-sell. More importantly, these fuzzy linguistic variables are usually
coexisting and overlapping in nature. For instance, when a stock market (e.g., Dow
Jones Index) appeared to be bearish (means short-sell), it is also having a certain
degree of over-sell (means rebound—with just the opposite trading signal) simul-
taneously. The same case occurs in a bullish market. Besides, all these linguistic
variables are fuzzy in the sense that the meaning of bearish or bullish are highly
subjective and different for different people, even for the same person the meanings
of these linguistic variables are usually different at different times and scenarios.
In order to model these highly uncertain and fuzzy linguistic variables into a
computer system (such as neural networks) for system modeling, FLS is one of the
most viable solutions. A typical FLS consists of three main modules:
12.2 Literature Review 343

(1) fuzzification module,

(2) inference module, and
(3) defuzzification module.
Figure 12.1 shows a typical type-1 fuzzy membership function (T1-FMF) of
RSI (relative strength index)—one of the most commonly used financial indexes in
financial engineering (Lee 2006b).
Zadeh (1975) in his influential paper The Concept of a Linguistic Variable and
its Application to Approximate Reasoning introduced the concept of higher order
type-n fuzzy logic system (Tn-FLS) as
A fuzzy set is of type n, n = 2, 3, …, if its membership function ranges over fuzzy sets of type
n-l. The membership function of a fuzzy set of type 1 ranges over the interval [0, 1].

For example, in a typical type-2 FLS, the membership function of each attribute
of the system itself is also a fuzzy set with a fuzzy range over the interval [0, 1]. In
other words, a higher order (n > 1) type-n FLS will increase the degree of uncertainty
of each fuzzy variable substantially.
But it has an intrinsic problem of computational complexity. For a simple situation
of a T2-FLS with k fuzzy variables, each fuzzy variable has m membership functions,
and each membership function is quantized into N different states. The total number
of possible fuzzy rules/states will be (N m )k , which will be an astronomical number
when such method is applied to real-world financial time series prediction problems
which usually have over 50 fuzzy attribute signals (from different financial indicators
and oscillators) to consider and each attribute has 4 membership functions, each can
span up to 100 different states, which means (1004 )50 possible states/fuzzy rules!
In order to solve this computational problem, interval type-2 fuzzy logic sys-
tem (IT2FLS) is introduced. In a typical IT2FLS, only certain interval(s) of the T2
membership function (usually the intervals of the state-transition-region, STR) are
fuzzified to a second-order FMF. Figure 12.2 shows a typical interval type-2 fuzzy
membership function (IT2-FMF) of relative strength index (RSI) for financial engi-
neering.

μ (x)

Over-Sell Bearish Bullish Over-Buy

0 x 100

Fig. 12.1 Type-1 fuzzy membership function of relative strength index (RSI)
344 12 Chaotic Type-2 Transient-Fuzzy Deep …

μ (x)

Over-Sell Bearish Bullish Over-Buy

Type-1 FMF

0 x 100
Type-2 uncertainty regions

Fig. 12.2 Interval type-2 fuzzy membership function (IT2-FMF) of RSI

A typical interval type-2 fuzzy set (IT2-FS), denoted by à is formulated by

 
μà (x) = 1/u (12.1)
 
xX u∈ μ̄ (x),μ (x)
à Ã

where μ̄à (x) and μà (x) denote the upper and lower membership functions for the
IT2-FS (Ã).
As shown in Fig. 12.2, the IT2-FMF is characterized by the second-order FMF
(the grey areas) in the fuzzy membership function, which are also known as FOU
(footprints of uncertainty). Owing to this high uncertainty modeling capability of
IT2FLS as compared with its T1-FLS counterpart, extensive researches have been
done in the past 20 years, especially in the fields of fuzzy financial modeling and
time series prediction.
Latest works of type-2 FLS in financial time series modeling and prediction
include: Castillo et al. have done extensive research of IT2FLS and fuzzy-neuro
systems on time series financial predictions including Mexican Stock Exchange
and Taiwan Stock Exchange (Castillo et al. 2014; Melin et al. 2012; Gaxiola et al.
2017; Ontiveros et al. 2018; Pulido et al. 2018); Bhattacharya et al. (Bhattacharya
et al. 2016; Konar and Bhattacharya 2017; Bhattacharya and Konar 2018) using
self-adaptive IT2FLS for stock prediction; IT2FLS for chaotic time series predic-
tion proposed by Lee et al. (2014); Multi-order fuzzy time series forecasting model
proposed by Ye et al. (2016) and Yolcu et al. (2016); Compact Evolutionary Interval-
Valued Fuzzy Rule-based system for real-time financial prediction proposed by Sanz
et al. (2015); Interval Type-2 Mutual Subsethood Fuzzy Neural Inference System
proposed by Sumati and Patvardhan (2018) for time series prediction and function
approximation; Fuzzy time series forecast of Taiwan stock market based on two-
factor second-order fuzzy-trend logical relationship groups (TSFTLRGs) with par-
ticle swarm optimization technique proposed by Chen and Jian (2017); IT2FLS for
stock index forecast based on fuzzy logical relationship map proposed by Jiang et al.
12.2 Literature Review 345

(2018); and type-2 neuro-fuzzy for stock price prediction based on self-constructing
clustering method of fuzzy rules and particle swarm optimization technique by Liu
et al. (2012).
Although IT2FLS provides a viable and computational feasible solution to model
highly uncertain phenomena, the determination and formulation of FOU is still an
area of interest in which the de facto usage of triangular-shape formulation as FOU
is also queried to be a bit artificial in nature, which motivates the author to search
for an integrated mathematical framework to model both interval type-2 fuzzy and
type-1 fuzzy regions in IT2-FMF as a single continuous transient-fuzzy membership
function.
By the extension of the author’s original work on Lee-oscillators (Lee 2004a), this
chapter proposed an innovative solution for the modeling of chaotic type-2 transient-
fuzzy logic (CT2TFL) by using multiple composite Lee-oscillators.
The motivation for introducing CT2TFL is in three aspects, which are as follows:
1. Typical financial signals and indicators such as RSI and MACD coexist with
multiple linguistic variables (over-sell, bearish, bullish, and over-buy), FLS is a
viable solution for technical signal modeling;
2. These linguistic variables themselves exist with different degrees of uncertainty
and fuzziness in nature, T2FLS is the best tool for handling such high-order
degree of fuzziness;
3. If we can construct a chaotic neural oscillator model in which its transient-chaotic
bifurcation diagram itself can directly model (simulate) a typical T2FLS with a
transient-fuzzy property only appears in FOU, the computational complexity
problem of T2FLS will be effectively resolved. More importantly, if such com-
posite chaotic neural oscillators themselves are exactly input neurons of a deep
neural network (DNN), the type-2 fuzzification process will become the intrin-
sic property of the DNN itself; instead of external fuzzification process in many
contemporary fuzzy-neuro systems.
In the next section, we will study an overview of discrete-time neural oscillators
and chaotic neural networks; and how it can be used to implement a type-2 chaotic
fuzzy neural network for time series prediction. After that, we will explore how
composite Lee-oscillators can be used to model T2FMF for financial time series
financial prediction.

12.2.2 Lee-Oscillators with Retrograde Signaling (LORS)

Latest research in neuroscience revealed that retrograde signaling (RS) in neurons

plays a vital role in temporal information processing and memory recalling in human
brain (Faghihi and Moustafa 2017; Korkut et al. 2013; Naeem et al. 2015; Yoshihara
et al. 2005), while abnormal retrograde signaling may lead to various types of memory
disorders such as dementia, memory loss, Alzheimer disease, and Down syndrome
(Sidoryk et al. 2011; Steinert et al. 2010; Verkhratsky et al. 2010; Volman et al. 2007).
346 12 Chaotic Type-2 Transient-Fuzzy Deep …

Inspired by retrograde signaling in neurons, chaotic Lee-oscillators with retro-

grade signaling (aka LORS) is the extension of the author’s previous works on Lee-
oscillator (Lee 2004a; 2006a) with the integration of retrograde feedback signals
in the neural dynamics for the improvement of temporal information processing in
neural oscillatory networks. Figure 12.3 shows the neural architecture of LORS.

E(t + 1) = Sig  [a1 LORS(t) + a2 E(t) − a3 I (t) + a4 S(t) − ξE ] (12.2)

I (t + 1) = Sig  [b1 LORS(t) − b2 E(t) − b3 I (t) + b4 S(t) − ξI ] (12.3)

(t + 1) = Sig  [S(t)] (12.4)

LORS(t) = [E(t) − I (t)]e−kS (t)

2
+ (t) (12.5)

where E, I, , and LORS denote excitatory, inhibitory, input, and output neurons
of LORS; ai and bi are the weights for excitatory and inhibitory neurons; ξE and ξI
are the threshold values and S(t) is the external input stimulus.
After extensive experiments and research (2008–2018) on the bifurcation behav-
iors of LORS with different parameter settings and conditions in Table 12.1, we cate-
gorize LORS into 8 major categories (LORS#0–LORS#7), with LORS#0 corresponds
to the original Lee-oscillator without retrograde signals and LORS#1–LORS#7 corre-
spond to different degrees of bifurcations, from single to multiple bifurcation regions.
Figure 12.4 illustrates the bifurcation diagrams for the 8 major categories of LORS.

12.3 CT2TFDNN—The System

12.3.1 CT2TFDNN—System Framework

The chaotic type-2 transient-fuzzy deep neuro-oscillatory network with retrograde

signaling (aka CT2TFDNN) proposed in this chapter is a hybrid intelligent worldwide
financial prediction system with the integration of the following:
1. chaotic neural oscillators (Lee-oscillators) as basic neural structure;
2. chaotic type-2 transient-fuzzification of the financial signals (FS) for signal mod-
eling;
3. genetic algorithms (GA) for the selection of top-10 significant Fuzzy financial
signals (FFS); and
4. chaotic transient-fuzzy deep neuro-oscillatory networks with retrograde signal-
ing for the chaotic financial time series deep learning and prediction.
Figure 12.5 depicts the system framework of CT2TFDNN for worldwide financial
prediction.
12.3 CT2TFDNN—The System 347

Fig. 12.3 Neural architecture of LORS

 348

Table 12.1 Parameter settings of 8 major categories of LORS

Parameters Lee-oscillator with retrograde signaling (LORS)
LORS #0 LORS #1 LORS #2 LORS #3 LORS #4 LORS #5 LORS #6 LORS #7
a1 0.00 −0.50 −0.50 0.50 0.90 0.90 5.00 5.00
a2 5.00 0.55 0.55 0.55 0.90 0.90 5.00 5.00
a3 5.00 0.55 0.55 0.55 0.90 0.90 5.00 5.00
a4 1.00 0.50 0.50 0.50 0.90 0.90 5.00 5.00
b1 0.00 −0.50 0.50 0.50 −0.90 −0.90 −1.00 −1.00
b2 −1.00 −0.55 −0.55 −0.55 −0.90 −0.90 −1.00 −1.00
b3 1.00 −0.55 −0.55 −0.55 −0.90 −0.90 −1.00 −1.00
b4 0.00 0.50 −0.50 −0.50 −0.90 −0.90 −1.00 −1.00
k 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00
ξE 500 50 50 50 50 300 50 300
ξI 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
s 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
12 Chaotic Type-2 Transient-Fuzzy Deep …
12.3 CT2TFDNN—The System 349

LORS #0 Bifurcation LORS #1 Bifurcation

LORS #2 Bifurcation LORS #3 Bifurcation

LORS #4 Bifurcation LORS #5 Bifurcation

LORS #6 Bifurcation LORS #7 Bifurcation

Fig. 12.4 Eight major categories of bifurcations in LORS

350 12 Chaotic Type-2 Transient-Fuzzy Deep …

Chao c
Financial GA
Type-2 Chao c T2 Forecast
Signal TOP10 FS
Transient- Transient- DataBank
Generator Selec on
Fuzzifica on Fuzzy Deep
Financial Module Module
Module Neural
Time FSGM GAT10FSSM
CT2TFM Network
Series
Module CT2TFDNN
CIT2TFDNNM DataBank

Fig. 12.5 System framework of CT2TFDNN for worldwide financial prediction

12.3.2 Financial Signal Generator Module (FSGM)

In the financial signal generator module (FSGM), 39 worldwide commonly used

financial indicators/oscillators are generated. They include moving averages (MA),
relative strength index (RSI), Bollinger bands (BB), MACD, Stochastic oscillators,
etc. Appendix B shows the list of 39 financial trading signals generated in the pro-
posed system.
All the daily financial time series databank will be used to generate these 39
trading signals, which are used for type-2 transient-fuzzification process performs
in the next module. Also, for the ease of fuzzification and network training, all the
technical indicators/oscillators being generated in this module will be converted into
the technical oscillator and normalize with values between 0 and 1. For instance, the
MA (Moving Average) technical indicator will be converted into MA oscillator by
evaluating the signal crossing between two MA signals, such as MA5 and MA13, the
two most commonly used Fibonacci numbers commonly used for technical signal
crossing in financial engineering (Murphy 1999).

12.3.3 Chaotic Type-2 Transient-Fuzzification Module

(CT2TFM)

(1) Chaotic Type-2 Transient-Fuzzy Membership Function (CT2TFMF)

CT2TFMF is the modeling of interval type-2 fuzzy logic by using the author’s devised
time-discrete chaotic neural oscillator—LORS. The formulation of CT2TFMF (nor-
malized) is given by

LORS(2x, t), x ≤ 0.5
CT 2TFMF(x, t) = (12.6)
LORS(2 − 2x, t), x > 0.5 ≥ 1

where x is the fuzzy variable value (e.g., RSI value) which corresponds to the input
stimulus S(t) in the LORS formulations Eqs. (12.2)—(12.5); t is the type-2 fuzzy logic
12.3 CT2TFDNN—The System 351

Fig. 12.6 Chaotic type-2 transient-fuzzy membership function (CT2TFMF)

index; LORS(x, t) is the normalized LORS function given by Eq. (14). Figure 12.6
shows the CT2TFMF by using LORS#0 with MATLAB simulation result of 1000
× 200 time-steps.
As shown, the CT2TFMF generated by LORS has certain special features:
(i) The membership function generated is a typical type-2 fuzzy membership func-
tion with remarkable transient-fuzzy FOU, which corresponds to the bifurcation
regions of the chaotic oscillators.
(ii) The whole type-2 fuzzy membership function can be modeled and generated
by a simple time-discrete chaotic neural oscillator, which can be easily imple-
mented to model various Interval type-2 fuzzy logic systems (IT2FLS).
(iii) As the neural dynamics of FOUs generated by CT2TFMF is inherited from
the Lee-oscillators, so it naturally converges to the two stable states at both
ends, while exhibits progressive and controlled fuzziness (bifurcation in chaos
theory).
(iv) By regulating various parameters (e.g., weights of inhibitory and excitatory
neurons, threshold values ξ E and ξ I , decay constant k) in the Lee-oscillator, the
FOUs of CT2TFMF can be adjusted and transformed to other chaotic patterns
to tackle with different complex IT2FLS problems.

(2) Chaotic Type-2 Composite Transient-Fuzzy Membership Function

(CT2CTFMF)

For the ease of modeling type-2 fuzzy logic for the 39 financial signals, chaotic type-2
composite transient-fuzzy membership function (aka CT2CTFMF) is constructed to
model the following four type-2 fuzzy financial linguistic variables: over-sell (OSell),
bearish (Bear), bullish (Bull), and over-buy (OBuy).
Figure 12.7 shows the MATLAB simulation of normalized CT2CTFMF for finan-
352 12 Chaotic Type-2 Transient-Fuzzy Deep …

cial signal RSI (relative strength index) using 4 composite Lee-oscillators (LORS#0)
with MATLAB simulation results of 1000 × 200 time-steps.
The CT2CTFMF of these four linguistic variables are the following:

CT 2TFMF(2x + 0.5, t), x ≤ 0.25
CT 2CTFMFOsell (x, t) = (12.7)
0, x > 0.25 ≥ 1

CT 2TFMF(1.6x, t), x ≤ 0.625
CT 2CTFMFBear (x, t) = (12.8)
0, x > 0.625 ≥ 1

0, x ≤ 0.375
CT 2CTFMFBull (x, t) = (12.9)
CT 2TFMF(1.6x − 0.6, t), x > 0.375 ≥ 1

0, x ≤ 0.75
CT 2CTFMFOBuy (x, t) = (12.10)
CT 2TFMF(2x − 1.5, t), x > 0.75 ≥ 1

From the fuzzy-neuro perspective, Fig. 12.7 is the bifurcation diagram of the
composite Lee-oscillators with the superposition of four LORS#0 Lee-oscillators
which model the membership function CT2CTFMF of the four type-2 fuzzy vari-
ables: over-sell, bearish, bullish, and over-buy with their neural dynamics governed
by (12.7)–(12.10), respectively.
As shown in the above equations, the formulations of all type-2 fuzzy linguistic
variables are continuous state functions with transient-chaotic (named as transient-
fuzzy in Type-n FLS) properties within the state-transition regions, which correspond
to FOUs in the type-2 FLS. Besides, since CT2CTFMF is generated by Lee-oscillator

Fig. 12.7 Chaotic T2 composite transient-fuzzy membership function (CT2CTFMF) of RSI

12.3 CT2TFDNN—The System 353

which is a discrete-time chaotic oscillator, all type-2 fuzzy states can be easily modi-
fied, implemented, and evaluated by adjusting the total number of first-order (x) and
second-order (t) CTFMV (composite transient-fuzzy membership values) to fit for
different requirements of granularity for any IT2FLS problems.
From the theoretical perspective, the chaotic bifurcation diagram of the compos-
ite Lee-oscillators is the transient-chaotic output states of these chaotic oscillators
subjected to the input stimulus. In terms of type-2 fuzzy logic, the transient-fuzzy
bifurcation at type-2 transient-fuzzy regions shown in Fig. 12.7 are theoretically ana-
log to the interval type-2 concept in fuzzy logic. The difference is that this chaotic
transient-fuzzy membership function is not an additional or external fuzzification
engine that usually is applied by typical IT2 FLS, but rather it is an intrinsic property
of the fuzzy-neuro network itself. In other words, the type-2 fuzzification process
proposed in this system is just part of the neural dynamics of the input neurons by sim-
ply replaced each of the input neuron (financial signal) by 4 LORS#0 Lee-oscillators
and automatically inherited the transient-fuzzy capability during the system training
and forecast process.
(3) Chaotic T2 Transient-Fuzzification Module (CT2TFM)
In this module, all 39 normalized trading signals (oscillators) will undergo the
chaotic interval type-2 fuzzification process by applying the CT2CTFMF described
in (12.7)–(12.10). The resulted 39 (each fuzzy financial signal has 4 second-order
type-2 TFMV) fuzzified financial signals will enter to the GA selection module for
top-10 financial fuzzy signals selection.

12.3.4 GA-Based TOP-10 Financial Signals Selection

Module (GAT10FSSM)

GA-based top-10 financial signals selection module (GAT10FSSM) is a vital process

in this system to screen-out noncritical signals/data, as usually there are too many
signals and time series data appeared in many financial engineering problems. If all
related signals and time series are used for financial prediction, it will not only slow
down the whole system training and prediction process but more importantly, it will
cause serious system over-training and deadlock problems by being trapped in the
local minima.
Top10-FFSSM is a GA (genetic algorithms) based module to select the best (top
10) fuzzy financial signals (FFS) generated in the CT2TFM. It consists of six func-
tional modules shown in Fig. 12.8.
(1) Fuzzy Financial Signals Population Generation Module (FFSPGM)
In this module, a population of 1000 fuzzy financial signal vectors FFSV
(s, w) are generated, where s and w are the signal vectors of the 39 FFS and
their weights with values between 0 and 1. Noted that s is the composite fuzzy
354 12 Chaotic Type-2 Transient-Fuzzy Deep …

TOP-10 Fitness FFS Selection Module

(T10FFSSM)

New Generation Population Selection Module

(NGPSM)

Fuzzy Financial Signals Fitness Evaluation Module

(FFSFEM)

Fuzzy Financial Signals MutationModule

(FFSMM)

Fuzzy Financial Signals CrossoverModule

(FFSCM)

Fuzzy Financial Signals Population GenerationModule

(FFSPGM)

Fig. 12.8 GA-based top-10 fuzzy financial signals selection scheme

financial signal vectors with 4 dimensions correspond to the four type-2 fuzzy
variables.
In the initialization stage, a random number generator is used to generate all the
weights for these 1000 FFSVs. Note that these weights are also used as the initial
weights used for the CT2TFDNN for fitness evaluation. In fact, the objective of
this GA-based signal selection scheme is to evaluate the best (fitness) FFSV and
retrieve the top-10 FFS with highest weights for system training and prediction
in CT2TFDNN.
(2) Fuzzy Financial Signals Crossover Module (FFSCM)
The whole FFSCM consists of the following steps:
12.3 CT2TFDNN—The System 355

Step 1. Randomly selects two FFSVs from the population.

Step 2. Perform a two-point crossover operation.
Step 3. Two new FFSVs offspring are generated.
Step 4. Perform steps 1–3 for 500 times until 1000 FFSV offspring are created.
3) Fuzzy Financial Signals Mutation Module (FFSMM)
In FFSMM, a 5% (i.e., 0.05) mutation rate is adopted.
The whole FFSMM consists of the following steps:
Step 1: For each 1000 FFSV offspring, generate a random number (between 0
and 100) for each of the 39 FFS.
Step 2: If the random generated for a particular FFS is less than 5, then the
weight (w) for that FFS will be replaced by a new random number between 0
and 1.
Step 3: Repeat steps 1–2 until all 1000 FFSV offspring have to go through the
FFS mutation operation.
(4) Fuzzy Financial Signals Fitness Evaluation Module (FFSFEM)
In this module, each of the 1000 FFSV offspring will be used (in turn) with
the time series financial data as input vectors for CT2TFDNN deep training.
Detailed training algorithm will be studied in the next section.
After 1000 iterations, the overall RMSE (Root-Mean-Square Errors) between
the target next-day close and the actual-day close will be evaluated and stored
as the fitness value (FV) for that particular FFSV offspring.
(5) New Generation Population Selection Module (NGPSM)
For 1000 FFSVs of the current population and their 1000 FFSV offspring,
evaluate their FVs and choose the first 1000 best FFSV with the lowest RMSE
as members for the new generation.
(6) TOP-10 Fitness FFS Selection Module (T10FFSSM)
Repeat the above GA modules for 1000 generations. For the last generation,
select the best (fit) FFSV with the highest FV (i.e., the lowest RMSE). Inspect
the weights (w) of its 39 FFS and choose the top-10 FFS with the highest weights
as the target top-10 FFS and use these weights as the initial for network training
in the CT2TFDNN module.
Table 12.2 shows the list of top-10 FFS after 1000 generations.
As shown, it is not surprising to find out that those popular financial signals such
as: MACD, RSI, MA (OsMA), and Stochastic signals appeared in the top-10 list,
mainly because these financial signals are so popular that many traders and investors
are using them to design their trading strategies so that are more correlated with the
forecast results than other financial signals.
356 12 Chaotic Type-2 Transient-Fuzzy Deep …

Table 12.2 List of TOP−10

No FFS code Fuzzy financial signals
fuzzy financial signals
1 AO Awesome oscillator FFS
2 Bands Bollinger bands FFS
3 Gator Gator oscillator FFS
4 Ichimoku Ichimoku Kinko Hyo FFS
5 Momentum Momentum FFS
6 OsMA Moving average oscillator FFS
7 MACD MACD FFS
8 RSI Relative strength index FFS
9 RVI Relative vigor index FFS
10 Stochastic Stochastic oscillator FFS

12.3.5 Chaotic T2 Transient-Fuzzy Deep Neural Network

Module (CT2TFDNNM)

In short, CT2TFDNN is a multilayer deep supervised-learning neural oscillatory

network with the replacement of all neurons all layers of the networks with LORS with
a chaotic bifurcation transfer function (CBTF) as activation functions. Besides, eight
chaotic bifurcation hidden layers (CBHL) which correspond to the 8 different modes
of bifurcation in LORS are used to facilitate chaotic deep learning in CT2TFDNN.
Figures 12.9 and 12.10 show the system architecture and the chaotic deep learning
algorithm of CT2TFDNN for worldwide financial prediction.

Fig. 12.9 System architecture of CT2TFDNNM

12.3 CT2TFDNN—The System 357

Fig. 12.10 CT2TFDNN chaotic transient-fuzzy deep learning algorithm

358 12 Chaotic Type-2 Transient-Fuzzy Deep …

As shown in Fig. 12.9, the input vectors of CT2TFDNN include the following:
(1) Top-10 T2 fuzzy financial signals (FFS), each consists of 4 fuzzy attributes
indicate the T2 FMV of: over-sell (OS), bearish (BR), bullish (BU), and over-
buy (OB);
(2) Time series financial data, each set consists of the open (O), high (H), low (L),
close (C), and volume (V) of 10-trading days normalized time series data, each
is multiplied by time-step decay (forgetting) function fg() given by

fg(D) = e−kD (12.11)

where D is the time-horizon for time series and k = 0.5 is the decay factor.
Figure 12.11 shows the decay chart for financial time series data (Lee 2004b).
As shown in Fig. 12.11 the decay function almost converges to 0 when time
horizon (TH) close to day-10. So, we set the time horizon for time series financial
data to 10. As reflected from previous works (Lee 2004b), the introduction of decay
(forgetting) factor is vital not only because it simulates time series relationships of
the historical data, but also it can avoid over-training by effectively control the data
size of time series input vectors and speedup the whole training process. Also, by
replacing all neurons of CT2TFDNN with Lee-oscillators, the traditional FFPBN
effectively transforms into a multilayer neuro-oscillatory network with inherited
transient-chaotic neuron activation properties from Lee-oscillators, which not only
significantly speedup the whole network training process, but more importantly it

Fig. 12.11 Decay (forgetting) factor for the financial timeseries data
12.3 CT2TFDNN—The System 359

resolves the over-training and deadlock problems which commonly occurs in most
recurrent neural networks with massive financial time series data. Detailed system
performance analysis will be studied in Sect. 12.5.

12.4 CT2TFDNN—System Implementation

12.4.1 Quantum Finance Forecast Center

As shown in Fig. 12.12, CT2TFDNN forecaster is a server-side intelligent forecast

agent located at the MT4 server farm of QFFC using Dell precision T5820 tower
workstation with Intel 6-Core 3.6 GHz Xeon processor. For each financial product,
2048-trading day time series (except cryptocurrency which only has 300-trading
day data) include: open (O), high (H), low (L), close (C), and volume (V) are auto-
matically generated by MT4 engines of Forex.com and AvaTrade.com. Through
four functional processes studied in the previous section, the next-day forecasts of
these 129 financial products are stored at the CT2TFDNN forecast database for
CT2TFDNN trading agents to access and release for public access.

12.4.2 129 Worldwide Financial Products

With the successful cooperation with AvaTrade.com (major cryptocurrency MT4 ser-
vice provider) and Forex.com (major forex MT4 service provider), QFFC launched

Fig. 12.12 System architecture of quantum finance forecast center (QFFC)

360 12 Chaotic Type-2 Transient-Fuzzy Deep …

the 129 financial products free daily and weekly forecast services from January, 2018
for worldwide traders and individual investors. They include the following:
(1) 9 major cryptocurrencies;
(2) 84 worldwide forex;
(3) 19 major commodities;
(4) 17 worldwide financial indices.
Appendix A shows the list of 129 financial products under these four categories.
As shown in Appendix A, owing to the short trading history of cryptocurrency,
300-trading day of historical time series are provided by AvaTrade.com, while all
other financial products consist of a 2048-trading day of historical time series for
each financial product provided by Forex.com which provide sufficient training and
test sets for system implementation and performance analysis. Also, each time series
record consists of daily information: open (O), high (H), low (L), close (C), and
volume (V).

12.4.3 CT2TFDNN Network Parameter Settings

Table 12.3 shows parameter settings of CT2TFDNN for network training and fore-
cast. As shown, 10 trading day time series are input into CT2TFDNN for network
training, so totally we have 50 input nodes (LORS#0) for time series inputs and 100
fuzzy financial signals (FFS) inputs with totally 400 FFS input nodes (LORS#0).
For each hidden layer, 400 LORS (LORS#0–LORS#7) hidden nodes are used for
network training. For financial products with 2048-trading day time series data, 1228
(60%) time series dataset is used for training, 401 (20%) time series dataset is used
for network testing and validation. For cryptocurrency with only 300 trading day time

Table 12.3 Parameter settings of CT2FDNN

Parameters Values
No. of trading days for network input 10
No. of timeseries LORS in input layer 50
No. of input Fuzzy Financial Signals (FFS) 100
No. of FFS LORS in input layer 400
No. of LORS in each hidden layer 400
No. of output LORS 4
Network training stop criteria (RMSE) 1 × 10−6
Deadlock criteria 10,000 epochs
LORS parameter settings in input and output layers LORS#0 settings
(Table I)
LORS parameter settings in 8 hidden layers LORS#0-LORS#7
(Table I)
12.4 CT2TFDNN—System Implementation 361

series data, same proportion of datasets are employed for the next training, testing,
and validation.

12.4.4 System Implementation and Pilot Run

From the implementation perspective, the whole CT2TFDNN system are imple-
mented over MT4 (MetaTrader4) platform (Walker 2018; Young 2015)—the world’s
biggest online financial program trading and development platform with over hun-
dreds of participating financial institutions for the provision of online program trading
and development services of all common worldwide financial products.
Since January 1, 2018, QFFC released daily financial forecast using CT1FNON
(type-1 fuzzy chaotic neuro-oscillatory networks) as pilot run and open testing for
worldwide traders and investors.
From October 1, 2018, after completing the system implementation of
CT2TFDNN, QFFC daily worldwide financial predictions are conducted by
CT2TFDNN system.

12.5 CT2TFDNN—Performance Analysis

12.5.1 Introduction

From the system performance perspective, three types of system performance anal-
ysis were conducted. They were the following:
1. System training performance analysis;
2. System forecast simulation performance analysis;
3. Actual daily forecast performance analysis.
In all these performance analyses, the system performance of CT2TFDNN was
compared with five forecast models:
1. Traditional time series feedforward backpropagation network (FFBPN);
2. Support vector machine (SVM) forecasting tool provided by R Project—one of
the most popular financial forecasting tools used in the finance industry;
3. Deep neural network (DNN) with PCA (principal component analysis) model
(Singh and Srivastava 2017);
4. Classical interval type-2 fuzzy-neuro network (IT2FNN);
5. Chaotic type-1 fuzzy-neuro-oscillatory network (CT1FNON).
362 12 Chaotic Type-2 Transient-Fuzzy Deep …

12.5.2 System Training Performance Analysis

In CT2TFDNN training performance analysis, 60% of time series data of 129 finan-
cial products are employed for system training in two aspects. Figure 12.13 shows
system performances of the six forecast models over 500 epochs of network training

Fig. 12.13 System training performance (over 500 epochs) of six financial forecast models for 129
financial products. (TOP) Mean of RMSE. (BOTTOM) Standard deviation of RMSE
12.5 CT2TFDNN—Performance Analysis 363

of the 129 financial products in terms of mean and standard deviations of RMSE
(root-mean-square errors).
As shown in Fig. 12.13, two observations can be found: (1) CT2FDNN outper-
formed the other five models in terms of both mean and standard deviation of RMSE;
(2) CT2FDNN attained the promisingly low RMSE in 70 epochs while the RMSE
of other networks especially FFBPN, SVM, and IT2FNN were still half-way of their
lowest RMSE levels.

12.5.3 System Forecast Simulation Performance Analysis

In the system forecast simulation performance analysis, four categories of worldwide

129 financial products were tested with target RMSE (Root-Mean-Square Error) of
the forecast next-day closing price ranging from 1 × 10−4 to 1 × 10−7 , respectively.
The test was done by applying 500 forecast simulations for each system. Table 12.4
presents the system forecast simulation performance test of these six systems. Certain
interesting findings are revealed in Table 12.4:
(1) For case 1 simulation (RMSE 1 × 10−4 ), CT2TFDNN forecaster outperformed
FFBPN (453.94), SVM (298.92), DNN-PCA (227.40), IT2FNN (246.81), and
CT1FNON (5.58) times. Similar findings can be found in case II simulation
results. It clearly reflected the improvement of network learning rate achieved
by the hybrid type-2 transient-fuzzy with deep chaotic neuro-oscillatory network
system and GA-based top-10 FFS selection scheme.
(2) Across the 3 cases with decreasing RMSE from 1 × 10−4 (case 1), 1 × 10−5
(case 2), 1 × 10−6 (case 3) to 1 × 10−7 (case 4). All forecast systems achieved
the target RMSE in cases 1 and 2. However, for cases 3 and 4 simulations using
target RMSE 1 × 10−6 and 1 × 10−7 , both FFBPN and IT2FNN (which were
using sigmoid-based FFBPN for machine learning) encountered deadlock prob-
lems during the network training of cryptocurrency and forex products; while
CT2TFDNN forecaster could still complete the network training with promising
training speeds. It clearly demonstrated the resolution of over-training and dead-
lock problems and improvement of system training efficiency by CT2TFDNN
forecaster over traditional recurrent neural networks using classical sigmoid-
based activation function.
(3) Comparing CT2TFDNN against CT1FNON across the three cases, it was inter-
ested to reveal that CT2TFDNN outperformed its counterpart by 5.58 times
(case 1), 5.65 times (case 2), 17.48 (case 3), and 28.50 (case 4) respectively. As
shown in Table 12.4, the overall performance of the CT1FNON forecast sys-
tem deteriorated substantially when the target RMSE was set to 1 × 10−6 and
beyond, especially on cryptocurrency and forex, while CT2TFDNN still per-
formed stable result. It clearly reflected the merits for the integration of type-2
transient-fuzzification scheme with chaotic neural oscillator technology with
retrograde signaling for time series chaotic deep learning.
Table 12.4 System forecast simulation performance analysis
364

Product category FFBPN SVM DNN-PCA IT2-FNN CT1-FNON CT2TFDNN

Total Av. STT Total STT Av. STT Total STT Av. STT Total Av. STT Total Av. STT Total Av. STT
STT STT STT STT
Case 1 (RMSE = 1 × 10−4 )
Cryptocurrency 557,251 61,916.78 372,244 41,360.41 244,633 27,181.47 305,240 33,915.56 5720 635.56 1023 113.67
Forex 508,453 6053.01 331,511 3946.56 291,344 3468.38 275,154 3275.64 6323 75.27 1128 13.43
Financial Index 41,120 2164.21 26,152 1376.44 19,738 1038.82 21,183 1114.89 923 48.58 159 8.37
Commodity 46,641 2743.59 29,384 1728.46 22,108 1300.46 25,564 1503.76 1223 71.94 231 13.59
Overall 1,153,465 8941.59 759,291 5885.98 577,823 4479.25 627,141 4861.56 14,189 109.99 2541 19.70
Case 2 (RMSE = 1 × 10−5 )
Cryptocurrency 1,460,000 162,222.22 937,320 104,146.67 823,440 91,493.33 862,334 95,814.89 15,121 1680.11 2673 297.00
Forex 1,235,543 14,708.85 841,405 10,016.72 720,322 8575.26 782,507 9315.56 17,231 205.13 2894 34.45
Financial Index 111,024 5843.37 68,391 3599.51 50,627 2664.58 62,236 3275.58 2312 121.68 481 25.32
Commodity 109,142 6420.12 71,925 4230.86 44,967 2645.09 64,733 3807.82 2640 155.29 551 32.41
Overall 2,915,709 22,602.40 1,919,041 14,876.29 1,639,356 12,708.19 1,771,810 13,734.96 37,304 289.18 6599 51.16
Case 3 (RMSE = 1 × 10−6 )
Cryptocurrency DL – 6,102,255 678,028.38 3,601,307 400,145.17 DL – 234,663 26,073.67 10,324 1147.11
Forex DL – 5,477,816 65,212.10 4,184,833 49,819.44 DL – 218,661 2603.11 13,234 157.55
Financial Index 577,324 30,385.47 373,529 19,659.40 318,683 16,772.78 459,441 24,181.11 6623 348.58 1292 68.00
Commodity 687,595 40,446.76 468,252 27,544.25 385,741 22,690.64 622,775 36,633.82 9126 536.82 1983 116.65
Overall – – 12,421,852 96,293.43 8,490,564 65,818.33 – – 469,073 3636.22 26,833 208.01
Case 4 (RMSE = 1 × 10−7 )
Cryptocurrency DL – 25,141,291 2,793,476.73 14,405,228 1,600,580.89 DL – 1,928,929 214,325.44 51,213 5690.33
Forex DL – 26,896,077 320191.39 17,785,540 211732.62 DL – 1,600,598 19054.74 62,256 741.14
Financial Index DL – 1,714,498 90,236.74 1,446,821 76,148.46 2,451,732 129,038.53 37,088 1952.00 6231 327.95
(continued)
12 Chaotic Type-2 Transient-Fuzzy Deep …
Table 12.4 (continued)
Product category FFBPN SVM DNN-PCA IT2-FNN CT1-FNON CT2TFDNN
Total Av. STT Total STT Av. STT Total STT Av. STT Total Av. STT Total Av. STT Total Av. STT
STT STT STT STT
Commodity DL – 2,088,404 122,847.29 1,917,133 112,772.52 2,756,693 162,158.41 48,641 2861.24 7132 419.53
Overall – – 55,840,269 432,870.30 35,554,721.84 275,618.00 – – 3,615,256 28,025.24 126,832 983.19

Note
1. Results are generated by 500 simulations of each neural network system (in msec)
2. “Total STT” denotes the total average system training time for 500 simulations
3. “Av. STT” denotes the average system training time for a single financial product
4. “DL” denotes deadlock during system training
12.5 CT2TFDNN—Performance Analysis
365
366 12 Chaotic Type-2 Transient-Fuzzy Deep …

(4) In terms of the system performance across different financial products, the
simulation results clearly reflected that both cryptocurrency and forex were
more chaotic and difficult for network training than other financial products as
expected, which will be one of the future R&D directions of quantum finance
forecast center.

12.5.4 Actual Daily Forecast Performance Analysis

Started from October 1, 2018, CT2TFDNN provides an official daily forecast of

these 129 financial products via QFFC official site on every trading day at 08:00
HKT (00:00 UTC) and reports the previous day forecast performance. Figure 12.14
shows the actual daily forecast performance chart of the six forecast models between
October 1, 2018 and March 15, 2019 (120 trading days).
The daily forecast % error (DFPE) is defined by

Average(DFErr(H ), DFErr(L))
DFPE = × 100% (12.12)
DClose
where DFErr(H ), DFErr(L) are the daily errors of daily forecast high (low) with
the daily actual high (low), respectively, and DClose is the daily closing price.
As shown in Fig. 12.14, the 120-trading day actual forecast performance of six
forecast models were rather consistent with their system training results shown in
previous tests, in which FFPBN performed the worst and CT2TFDNN performed
the best.
But one interesting difference is that although the ranking of their performances
were the same between system training performance against actual forecast perfor-
mance, the overall forecast performance gaps between these six forecast systems
were more distinct in which CT2FDNN outperformed the other five forecast models
significantly and it took less than 20-trading days to attain its stable forecast per-
formance state while other forecast systems took 40–60 trading days to attain their
stable forecast performance states.
Figure 12.15 shows the daily forecast performance of CT2TFDNN across the four
different categories of financial products. As shown, owing to difference levels of
chaotic behavior and disturbance between different categories of financial products,
there were two interesting findings: (1) the forecast performance of cryptocurrency
was significantly poorer than all other categories even though their differences in
training performance were not so significant. This might due to the highly chaotic
property of cryptocurrency and insufficient time series history data; (2) the forecast
performance of forex was better than its performance in system training, which
was rather consistent with the observations from the professional traders using the
12.5 CT2TFDNN—Performance Analysis 367

Fig. 12.14 Actual daily forecast performance of six financial forecast models for 129 financial
products between 10/1/2018 and 03/15/2019. (TOP) Mean of forecast % error. (BOTTOM) Standard
deviation of forecast % error
368 12 Chaotic Type-2 Transient-Fuzzy Deep …

Fig. 12.15 Actual daily forecast performance of CT2TFDNN of four different categories of finan-
cial products between 10/1/2018 and 3/15/2019. (TOP) Mean of RMSE. (BOTTOM) Standard
deviation of RMSE
12.5 CT2TFDNN—Performance Analysis 369

CT2TFDNN forecast service during this period of time also concluded that although
forex products overall were highly chaotic in nature, CT2TFDNN provided a reliable
daily forecast with rather a stable degree of accuracy.

12.6 Conclusion

The dawn of big data epoch has driven us to face new challenges from over-flooding
data and information. This chapter devised an innovative chaotic type-2 transient-
fuzzy deep neuro-oscillatory network (CT2TFDNN) to address the over-training
and deadlock problems, which are commonly occurred during network training of
massive data such as financial, weather, and biomedical big data.
For the system architecture perspective, CT2TFDNN provided integration of four
different innovative AI technologies:
(1) chaotic neural oscillator for the modeling of neural dynamics;
(2) type-2 transient-fuzzy logic for the modeling of fuzzy financial signals;
(3) genetic algorithm for the selection of the best fuzzy financial signals;
(4) deep chaotic neural network for network training and prediction.
From the implementation perspective, CT2TFDNN has been adopted for the real-
time prediction of 129 worldwide financial products. In comparison with five fore-
cast systems ranging from traditional FFBPN, SVM, DNN with PCA, IT2FNN to
CT1FNON, CT2TFDNN produced promising results in terms of system training
performance and actual daily forecast performance.
The future and related works include the following:
1. Further research and study of CT2TFLS for the modeling, analysis, and data
mining of other big data problems such as weather, biometric and biomedical
engineering.
2. Further study of CT2TFDNN for the categorization of various CT2TFMF using
neural retrograde signaling techniques.
3. R&D of intelligent agent-based hedging and trading systems based on CT2TFLS.
4. Integration of CT2TFDNN with quantum price level (QPL) study using quantum
anharmonic oscillatory model (QAOM) for financial trend prediction and long-
term investment.

Problems

12.1 What is fuzzy logic? Discuss and explain the difference between type-1 vs.
type-2 fuzzy logic. Give two examples and explain how type-2 fuzzy logic
can be applied in finance engineering and quantum finance.
12.2 What are the major advantages and shortcomings of type-1 fuzzy logic? Use
two applications of type-1 fuzzy logic on technical indicators in the financial
market for illustration.
370 12 Chaotic Type-2 Transient-Fuzzy Deep …

12.3 What is interval type-2 fuzzy logic (IT2FL)? State and explain the major
difference between type-2 fuzzy logic and IT2FL. Use two examples in real-
world situation for illustration.
12.4 The following figure shows the type-1 fuzzy membership function of RSI

(i) Use the same methodology, draw the type-1 fuzzy membership function
to describe the fuzzy variable of temperature.
(ii) Compare these two fuzzy membership functions, discuss and explain
why the fuzzy membership of RSI for the description of financial market
is more meaningful and critical for fuzzification than traditional fuzzy
variable such as temperature and humidity for the description of weather
situation.
(iii) Based on the given type-1 membership function of RSI:
• Draw the corresponding type-2 membership function and write the
corresponding membership function formulation;
• Draw the corresponding internal type-2 membership function and
write the corresponding membership function formulation;
• Contract their difference (i.e., pros and cons) and describe how they
work;
• Why they are more meaningful and useful than type-1 fuzzy mem-
bership?
12.5 State and describe the formulation and logic behind the type-n fuzzy logic.
Use two real-time examples (weather and finance) to illustrate your explana-
tion.
12.6 What are the major advantages and shortcomings of type-n fuzzy logic?
Justify your explanation by using real-world example in finance engineering.
12.7 What is retrograde signaling in biological neuroscience? State and discuss
the latest research and findings on symptoms and illness in human memory
related to this effect.
12.8 What is Lee-oscillators with retrograde signaling? What is the mathematical
formulation? How can it relate to biological neuroscience?
12.9 State and explain the formulation of Lee-oscillator with retrograde signaling.
Discuss how it works in real-world application on weather prediction and
financial engineering.
12.6 Conclusion 371

12.10 Discuss and explain the major differences between Lee-oscillator and Lee-
oscillator with retrograde signaling in terms of (1) physical meaning in biolog-
ical neuroscience; (2) mathematical formulation; physical meaning in chaotic
neural network modeling.
12.11 Fig. 12.4 shows the bifurcation diagrams of 8 major Lee-oscillators with
retrograde signaling. Discuss and explain their major characteristic in terms
of (1) bifurcation; (2) chaotic transfer function capability.
12.12 Below figure shows the system framework of CT2TFDNN (chaotic type-2
transient-fuzzy deep neural network)

(i) Discuss and explain the major roles and functions of each functional
modules of CT2TFDNN.
(ii) What is deep neural network (DNN)? What is different between DNN
and ANN?
(iii) What are the major shortcomings of DNN for modeling a complex
system such as real-time financial prediction systems?
(iv) What are the major advantages and characteristics of CT2TFDNN ver-
sus traditional DNN (deep neural networks)?
12.13 Below figure shows the chaotic type-2 composite transient-fuzzy membership
functions (CT2CTFMF) of RSI.

(i) Write down the formulation of CT2CTFMF of RSI.

(ii) In addition to RSI, assume we also need to model MA crossing of two
MA signal lines (e.g., MA5 and MA21):
372 12 Chaotic Type-2 Transient-Fuzzy Deep …

• Draw the corresponding CT2CTFMF of MA-crossing indicator.

• Write down the corresponding formulation for CT2CTFMF of MA-
crossing indicator.
• What is the advantage of using CT2CTFMF of MA-crossing indicator
values instead of the MA-crossing signals as an input node in the DNN
for system training?

12.14 Discuss and explain why GA-based top-10 financial signals section module
is critical in the CT2TFDNN system.
12.15 Programming exercise I
(i) Draw the system flowchart for the implementation of CT2TFDNN
system for real-time financial prediction.
(ii) Write the system training algorithms of CT2TFDNN.
(iii) Based on the MQL program skill learnt in QFFC.org, implement
CT2TFDNN for at least five forex products using MQL on MT4 plat-
form.
(iv) Compare their system training performance in terms of (1) RMSE
(root-mean-square error) and (2) standard deviations in 1000 iterations
using error rate 1 × 10−6 .
(v) Compare their next-day forecast performance results in terms of (1) %
Error and (2) Standard deviation of % error for at least 20 trading days
(or 3 months MT4 simulation results).
(vi) Implement and perform the same performance test for three other cate-
gories of financial products: (1) financial indices; (2) major commodity
and (3) cryptocurrency.
(vii) Compare their results in terms of different categories of financial prod-
ucts. Discuss and explain the experimental results.
12.16 Figure 12.15 shows the actual daily forecast performance of CT2FDNN for
four different categories of financial products: cryptocurrency, forex, financial
indices, and major commodity.
(i) Discuss and explain why cryptocurrency is always the worst in terms
of forecast performance. How can we improve it?
(ii) As compared with system training performance shown in Fig. 12.4,
the forecast performance of forex products is significantly improved.
Why?
(iii) What is a suggestion and conclusion in terms of a financial investment
perspective?

Acknowledgements The author wishes to thank Forex.com and AvaTrade.com for the provision
of historical and real-time financial data. The author also wishes to thank Quantum Finance Forecast
Center of UIC for the R&D supports and the provision of the channel and platform qffc.org for
worldwide system testing and evaluation.
References 373

References

Bhattacharya et al. (2016) Secondary Factor Induced Stock Index Timeseries Prediction using
Self-Adaptive Interval Type-2 Fuzzy Sets. Neurocomputing 171:551–568.
Bhattacharya, D. and Konar, A. (2018) Self-adaptive type-1/type-2 hybrid fuzzy reasoning tech-
niques for two-factored stock index timeseries prediction. Soft Computing 22(18): 6229–6246.
Castillo et al. (2014) Application of Interval Type-2 Fuzzy Neural Networks in Non-Linear Identi-
fication and Timeseries Prediction. Soft Computing 18(6): 1213–1224.
Chen, S. M. and Jian, W. S. (2017) Fuzzy Forecasting Based on Two-Factors Second-Order Fuzzy-
Trend Logical Relationship Groups, Similarity Measures and PSO Techniques. Information Sci-
ences 391–392: 65–79.
Faghihi, F. and Moustafa, A. A. (2017) Sparse and burst spiking in artificial neural networks inspired
by synaptic retrograde signaling. Journal of Information Science 421: 40–42.
Gaxiola et al. (2017) Comparison of T-Norms and S-Norms for interval type-2 fuzzy numbers in
weight adjustment for neural networks. Information 8(3): 114.
Jiang et al. (2018) An Interval Type-2 Fuzzy Logic System for Stock Index Forecasting Based on
Fuzzy Timeseries and a Fuzzy Logical Relationship Map. IEEE Access 6: 69107–69119.
Konar, A. and Bhattacharya, D. (2017) Handling main and secondary factors in the antecedent for
type-2 fuzzy stock prediction. Intelligent Systems Reference Library 127: 105–132.
Korkut et al. (2013) Regulation of Postsynaptic Retrograde Signaling by Presynaptic Exosome
Release. Neuron 77: 1039–1046.
Lee, R. S. T. (2019a) Chaotic Type-2 Transient-Fuzzy Deep Neuro-Oscillatory Network
(CT2TFDNN) for Worldwide Financial Prediction. IEEE Transactions on Fuzzy System. https://
doi.org/10.1109/tfuzz.2019.2914642.
Lee, R. S. T. (2019b) Chaotic Interval Type-2 Fuzzy Neuro-oscillatory Network (CIT2-FNON) for
Financial Prediction. International Journal of Fuzzy Systems. https://doi.org/10.1007/s40815-
019-00688-w
Lee, R. S. T. (2006a) LEE-Associator – A Transient Chaotic Autoassociative Network for Progres-
sive Memory Recalling. Neural Networks 19(5): 644–666.
Lee, R. S. T. (2006b) Fuzzy-Neuro Approach to Agent Applications (From the AI Perspective to
Modern Ontology). Springer-Verlag, Germany.
Lee, R. S. T. (2005) Advanced Paradigms in Artificial Intelligence From Neural Oscillators, Chaos
Theory to Chaotic Neural Networks. Advanced Knowledge International, Australia.
Lee, R. S. T. A (2004a.) Transient-chaotic Auto-associative Network (TCAN) based on Lee-
oscillators. IEEE Trans. Neural Networks 15(5): 1228–1243,
Lee, R. S. T. (2004b) iJADE Stock Advisor: An Intelligent Agent based Stock Prediction System
using Hybrid RBF Recurrent Networks. IEEE Trans. on SMC A 34(3): 421–428.
Lee et al. (2014) An Efficient Interval Type-2 Fuzzy CMAC for Chaos Timeseries Prediction and
Synchronization. IEEE Transactions on Cybernetics 44(3) 29–341.
Liu et al. (2012) Application of Type-2 Neuro-Fuzzy Modeling in Stock Price Prediction. Applied
Soft Computing Journal 12(4): 1348–1358.
Melin et. al. (2012) A new approach for timeseries prediction using ensembles of ANFIS models.
Expert Systems with Applications 39(3): 3494–3506.
Murphy, J. J. (1999) Technical Analysis of the Financial Markets: A Comprehensive Guide to
Trading Methods and Application. New York Institute of Finance, New York.
Naeem et al. (2015) On the Role of Astroglial Syncytia in Self-Repairing Spiking Neural Networks.
IEEE Trans. On Neural Networks and Learning Systems 26(10): 2370–2380.
Ontiveros et al. (2018) High order α-planes integration: A new approach to computational cost
reduction of General Type-2 Fuzzy Systems. Engineering Applications of Artificial Intelligence
74: 186–197.
Pulido, M. and Melin, P. (2018) Optimization of Ensemble Neural Networks with Type-1 and Type-
2 Fuzzy Integration for Prediction of the Taiwan Stock Exchange. Studies in Fuzziness and Soft
Computing 361:151–164.
374 12 Chaotic Type-2 Transient-Fuzzy Deep …

Ross, T. J. (2016) Fuzzy Logic with Engineering Applications. Wiley.

Sanz et al. (2015) Compact Evolutionary Interval-Valued Fuzzy Rule-Based Classification System
for the Modeling and Prediction of Real-World Financial Applications with Imbalanced Data.
IEEE Trans. on Fuzzy Systems 23(4): 973–990.
Sidoryk et al. (2011) Role of astrocytes in brain function and disease. Toxicol. Pathol. 39(1):
115–123.
Siler, W. and Buckley, J. J. (2004) Fuzzy Expert Systems and Fuzzy Reasoning. Wiley-Interscience.
Singh, R. and Srivastava, S. (2017) Stock prediction using deep learning. Multimedia Tools and
Applications 76(18): 18569–18584.
Steinert et al. (2010) Nitric oxide signaling in brain function, dysfunction, and dementia. Neurosci-
entist 16(4): 435–452.
Sumati, V. and Patvardhan, C. (2018) Interval Type-2 Mutual Subsethood Fuzzy Neural Inference
System (IT2MSFuNIS). IEEE Trans. on Fuzzy Systems 26(1): 203–215.
Verkhratsky et al. (2010) Astrocytes in Alzheimer’s disease. Neurotherapeutics 7(4): 399–412.
Volman et al. (2007) The astrocyte as a gatekeeper of synaptic information transfer. Neural Com-
putation 19(2): 303–326.
Walker, W. (2018) Expert Advisor Programming and Advanced Forex Strategies. Independently
published.
Ye et al. (2016) A Novel Forecasting Method Based on Multi-Order Fuzzy Timeseries and Technical
Analysis. Information Sciences 367–368: 41–57.
Yolcu et al. (2016) High Order Fuzzy Timeseries Forecasting Method Based on an Intersection
Operation. Applied Mathematical Modelling 40: 8750–8765.
Yoshihara et al. (2005) Retrograde signaling by Syt 4 induces presynaptic release and synapse-
specific growth. Science 310: 858–863.
Young, A. R. (2015) Expert Advisor Programming for MetaTrader 4: Creating automated trading
systems in the MQL4 language. Edgehill Publishing.
Zadeh, L. A. (1975) The Concept of a Linguistic Variable and Its Application to Approximate
Reasoning-1. Information Sciences 8: 199–249.
Zadeh, L. A. and Aliev, R. A. (2019) Fuzzy Logic Theory and Applications: Part I and Part II.
World Scientific Publishing Company.
Chapter 13
Quantum Trader—A Multiagent-Based
Quantum Financial Forecast
and Trading System

Over the years, financial engineering ranging from the study of financial signals to
the modeling of financial prediction is one of the most stimulating topics for both
academia and financial community. Not only because of its importance in terms of
financial and commercial values, but also it vitally poses a real challenge to world-
wide researchers and quants owing to its highly chaotic and almost unpredictable
nature. This chapter devises an innovative multiagent-based quantum financial fore-
cast and trading system (a.k.a. quantum trader) for worldwide financial prediction
and intelligent trading. With the adoption of author’s theoretical works on Lee-
oscillator with profound transient-chaotic property, quantum trader effectively inte-
grates quantum field signals (QFS) and quantum field oscillators (OFS) studied in
Part I for neural network training and prediction into: (1) quantum forecaster—
chaotic FFBP-based time series supervised-learning agent for worldwide financial
forecast and; (2) quantum trader—chaotic RBF-based actor-critic reinforcement-
learning agents for the optimization of trading strategies. Quantum trader not only
provides a fast reinforcement-learning and forecast solution, more prominently it
successfully resolves the massive data over-training and deadlock problems which
is usually imposed by traditional recurrent neural networks and RBF networks using
classical sigmoid or Gaussian-based activation functions.
From the implementation perspective, the quantum trader is integrated with 2048-
trading day time series financial data and 39 major financial signals as input signals
for the real-time prediction and intelligent agent trading of 129 worldwide financial
products which consists of: 9 major cryptocurrencies, 84 forex, 19 major commodi-
ties, and 17 worldwide financial indices. In terms of system performance, past 500-
day average daily forecast performance of quantum trader attained less 1% forecast
percentage errors and with promising results of 8–13% monthly average returns.

© Portions of this chapter are reprinted from Lee (2019), with permission of KeAi Publishing
Communications Ltd.

© Springer Nature Singapore Pte Ltd. 2020 375

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_13
376 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

13.1 Introduction

Conventional technical analysis and chart analysis methods probe various trading
signals (e.g., K-lines, MAs, Bollinger-band index, KDJ and RSI indices) and chart
patterns (e.g., major reversal and trend patterns, golden-ratio patterns, Fibonacci
patterns, Elliot-wave patterns) that believe to affect the market price trends, trad-
ing patterns and individual reasonings to determine the best time to trigger buy or
sell decisions (Borden 2018; Brown 2012; Bulkowski 2005; Murphy 1999). How-
ever, these numerous trading signals and chart-patterns are usually self-contradictory
between different timeframes, let’s alone with the fact that they are highly subjective
to the traders’ own judgment and psychological condition.
With the advancement of computational capacity in past decades, the modeling of
complex financial prediction systems ranging from artificial neural networks (ANNs)
(Chang et al. 2012; Dai et al. 2012) to fractal-based financial forecast systems (Ling
2013) using ordinary desktop personal computers and workstations is no more a
reverie.
Current research on financial prediction include: stock prediction using deep neu-
ral network (DNN) with principal component analysis (PCA) by Singh and Srivas-
tava (2017); fuzzy models by Hwang and Oh (2010); Postfix-GP (Genetic Program-
ming) models by Dabhi and Chaudhary (2016); SVM (Support Vector Machine) and
hybrid SVR (Support Vector Regression) models by Henrique et al. (2018), Nahil
and Lyhyaoui (2018) and Ouahilal et al. (2017).
Moreover, the popularity of free and open quantitative financial system develop-
ment platforms such as MetaTrader (MT) platform provides an ideal environment
for worldwide researchers and quants to test their trading algorithms, strategies and
financial signals with real-time financial data streams which prosper the popular-
ity of program trading, especially the high-frequency algorithmic program trading
(HFAPT) in the past 10 years (Durbin 2010; Walker 2018; Young 2015).
From the financial market perspective, the blooming of cryptocurrency origi-
nated from bitcoin in 2009 had increased to more than 4000 cryptocurrencies in
the worldwide financial market (Narayanan 2016; Vigna and Casey 2016). More
importantly, major international fund houses and forex trading platforms integrate
24 × 7 electronic trading of cryptocurrency into their forex trading platforms since
2013. Together with the flourishment of HFAPT in the past 10 years, the world-
wide financial markets; especially the international currency market becomes more
volatile and unpredictable. An effective, open, and reliable worldwide financial pre-
diction, trading and advisory system is profoundly required for worldwide traders
and investors, especially independent investors than ever before.
This chapter devises an innovative multiagent-based quantum financial forecast
and trading system (a.k.a. Quantum Trader) for worldwide financial prediction and
intelligent trading. With the adoption of author’s theoretical works on Lee-oscillator
with profound transient-chaotic property (Lee 2004), quantum trader effectively inte-
grates quantum field signals (QFS) and quantum field oscillators (OFS) studied in
Part I for neural network training and prediction into: (1) quantum finance forecaster
13.1 Introduction 377

(or quantum forecaster in short)—chaotic FFBP-based supervised-learning agent for

worldwide financial forecast and; (2) quantum finance trader (or quantum trader in
short)—chaotic RBF-based actor-critic reinforcement-learning agents for the opti-
mization of trading strategies. Quantum trader not only provides a fast reinforcement-
learning and forecast solution, more importantly, it successfully resolves the mas-
sive data over-training and deadlock problems, which usually imposed by traditional
recurrent neural networks and RBF networks using classical sigmoid or Gaussian-
based activation functions.

13.2 Quantum Forecaster—Chaotic FFBP-Based Time

Series Supervised-Learning Neural Networks
for Financial Prediction

13.2.1 Time Series Prediction Using

Supervised-Learning-Based Artificial Neural
Networks

Time series prediction, ranging from weather forecast to stock prediction has been
studied for over 50 years. With the improvement of computational speed, nonlinear
models such as artificial neural networks (ANN) have proven success to tackle these
problems. Traditional ANN such as feedforward backpropagation neural network
(FFBPN) is a typical kind of supervised-learning (SL) neural network to tackle these
problems with certain success.
However, when it comes to massive input data such as financial prediction with
over thousands time series financial data and signals as input vectors, these conven-
tional ANNs using classical sigmoid-based activation function are usually “trapped”
in local minima during neural network training; which not only affects the efficiency
(time cost), but also the accuracy of the forecast results (Lee 2006).

13.2.2 Quantum Finance CSL Network—Chaotic

FFBP-Based SL Network for Time Series Financial
Prediction with Quantum Field Signals (QFF)

Quantum finance chaotic supervised-learning (CSL) network is the integration of

Lee-oscillator with classical FFBPN by replacing all neurons with Lee-oscillators.
Besides time series signals, the most important characteristics of quantum finance
CSL network is the adoption of all related quantum field signals (QFS) for the finan-
cial market studied in Part I as quantum field oscillators (QFO) for network training
and financial prediction. Figures 13.1 and 13.2 show the system architecture and
378 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Fig. 13.1 System architecture of quantum finance CSL network for time series financial prediction

chaotic FFBP-based time series supervised-learning algorithm of quantum finance

CSL network for financial prediction.

13.3 Quantum Trader—Chaotic RBF-Based Actor-Critic

Reinforcement-Learning Networks—Optimization
of Trading Strategy

13.3.1 An Overview of Reinforcement Learning (RL)

Different from the supervised-learning model with well-defined input/target-output

pairs to train the network, there are many situations in which input/target-output pairs
do not exist. For example, in stock investment, even though we have the best stock
forecast to tell us when to set the buy/sell bid, how much we should invest? or when
to close the bid? in order to get the highest returns is a typical optimization problem
without an exact solution. In that case, we can make use of the reinforcement-learning
(RL) method.
Reinforcement-learning (RL) theory is originated from behavior psychology. Its
main concept is to train the neural network with the adoption of feedback signals,
namely reinforcement signal (RS). For the right behavior, the network will respond
with a positive RS to award the RL network; while for the wrong behavior, the
13.3 Quantum Trader—Chaotic RBF-Based Actor-Critic Reinforcement … 379

Fig. 13.2 Quantum finance CSL network learning algorithm

380 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

network will respond with a negative RS to punish the RL network. As we can see,
RL networks don’t need well-defined input/target-output pairs, all they need to do is
to search for a set of optimal weights to minimize the negative reinforcement signals.
Classical RL model such as Markov decision process (MDP) using
stochastic-based reinforcement-learning algorithms such as Q-Learning, dynamic-
programming (DP) and TD-learning (TDL) with certain success (Kaelbling et al.
1996). However, when it comes with time series financial optimization problems
with over thousands of input signals and possible trading strategies, these classical
stochastic RL methods are either too computationally intensive or difficult to adopt
for actual implementation.
With the flourishing of recurrent neural networks in the past decades, researchers
began to explore how recurrent neural networks can be applied for RL on vari-
ous optimization problems (Li et al. 2009; Liu et al. 2005); financial engineering
include Deng et al. (2017) using deep direct reinforcement-learning network for
financial signal representation and trading; Pendharkar and Cusatis (2018) applied
reinforcement-learning agents for trading financial indices; Tan et al. (2011) used
reinforcement-learning system for stocking trading with cycles; Chang et al. (2017)
used asymmetric reinforcement-learning and conditioned responses technique to
analyze global financial crisis; Carapuco et al. (2018) using reinforcement-learning
system for short-term speculation in the foreign exchange market and Jeong and Kim
(2019) using deep Q-Learning to improve financial trading decisions.

13.3.2 Discrete-Time Actor-Critic RL Model

DTAC-RLM is a multiagent system (Lee 2006) consists of four components: envi-

ronment space (E), action space (A), actor agent (Actor), and critic agent (Critic).
Different from its continuous-time AC model counterpart (such as MDP), DTAC-
RLM visualizes its world as a collection of discrete-time-step states and actions.
More vitally, all these states are related to time series events. The role of the actor is
based on input signals provided at current state at time t (st ) to respond with the best
available actions (at ); whereas the role of the critic is to evaluate the reinforcement
signal (RS, or reward, r t in short) based on action(s) taken by the actor and signals
provided by the new states (st+1 ). Then the reward signal rt feedbacks as input to the
actor to decide the next action a at time t + 1 (Fig. 13.3).
The formulations are given by

s ∈ {S}, a ∈ { A}, r ∈ {R} (13.9)

= {a t , t = 1 . . . T } (13.10)

T
V = r t (s t , a t ) (13.11)
t=1
13.3 Quantum Trader—Chaotic RBF-Based Actor-Critic Reinforcement … 381

Environment Space (E)

Critic
Actor Agent Action Space (A)
Agent
(Actor)
(Critic)

Fig. 13.3 Discrete-time actor-critic RL model

V∗ = max V  (13.12)


As shown in the above formulation: S and A denote the state space and action
space, s and a denote the state and action vectors; r denotes the reward vector; 
denotes the action policy from time t = 1 to t = T (dimension of policy space); V
denotes the value function which represents the total returns for a particular policy .
For DRL with maxI iterations, there will be totally maxI policies  being generated,
and V* is the optimal returns obtained by the calculation of all possible return values
V
k (where k = 1… maxI) from these policies. So, the whole optimization problem
is to find the best policy * via DRL in order to attain the optimal returns V* . Next
section we will explore how to adopt CRBFN (chaotic radial basis function neural
network) for DRL.

13.3.3 Radial Basis Function Neural Network (RBFN)

for RL

Like FFBPN, a typical multilayer radial basis function neural network (RBFN) also
consists of input, hidden, and output layers. Although network architectures between
FFBPN and RBFN are highly similar, the network activations in RBF networks
are localized RBF functions such as Gaussian functions, resulting in a much faster
training rate (Markopoulos et al. 2016).
Unlike FFBPN, input vectors in RBFN distribute values to the hidden layer neu-
rons uniformly, without multiplying them with weights. The neurons in the hidden
layer are presented by RBF neurons using radial basis function such as Gaussian
function as activation function, which is given by
382 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …
 r 
G(r) = ex p − 2 (13.13)
2σ
Different from FFBPN which is only tailored for supervised learning, experimen-
tal results revealed that RBFNs can basically approximate any functions. As a result,
they are usually known as “universal-approximators”. In other words, RBFN can be
used for both SL and RL by using RBFN to approximate the V-value function.
However, like FFBPNs, when it comes with handling massive input data and/or
highly chaotic nature such as time series financial forecast using over thousands of
input values and financial signals as network input vectors, RBFNs also encounter
“over-training” and “deadlock” problems, which hinder further improvement of net-
work accuracy and resulted in an expensive time costs from the computational per-
spective.

13.3.4 Quantum Finance CRL Network—Chaotic

RBF-Based Actor-Critic RL Network
for the Optimization of Trading Strategy

Quantum finance chaotic reinforcement-learning (CRL) network integrates the Lee-

oscillator technology with conventional RBFN by: (1) replacing all Gaussian-based
radial basis functions in the hidden layer of RBFN with Lee-oscillatory RBF functions
(namely, LRbf ); and (2) replacing output neurons with Lee-oscillator to facilitate
transient-chaotic neural activations. Figure 13.4 shows LRbf() which served as the
chaotic RBF activation function in the quantum trader.

Fig. 13.4 Chaotic RBF activation function using Lee-oscillators

13.3 Quantum Trader—Chaotic RBF-Based Actor-Critic Reinforcement … 383

The formulation of LRbf is given by


L(x), x < c
L Rb f (x) = (13.14)
L(2c − x), x ≥ c

where L(x) is the Lee-oscillator function as mentioned in Chap. 9 and c is the

center of LRbf().
In contrast with the Gaussian function counterpart, the LRbf exhibits a
progressive-transient-chaotic property in two RBF activation regions which provide
some sort of controlled-hysteresis to sort out the over-training and deadlock problems
during RL with massive input vectors.

13.3.5 Quantum Finance CRL Network System Architecture

for the Optimization of Trading Strategy

As depicted in Fig. 13.5, quantum finance CRL network is a three-layer chaotic

RBF-based neural network with input, hidden, and output layers. To accomplish
actor-critic-based RL, the input layer consists of two classes of input vectors: the
reward signal vector r and input signal vector x. The input signal vector x composes

Fig. 13.5 Quantum finance CRL network system architecture

384 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

of all time series input signals together with the corresponding quantum price levels
(QPLs) as quantum field signal (QFS) for reinforcement training. Reward signal
vector r contains four reinforcement feedback signals which are generated by the
critic agent in the previous time-step. These four reward signals rT,B , rT,S , rP,B , rP,S
correspond to four actions aT,B , aT,S , aP,B , and aP,S which emulate de facto short-term
trading strategies:
(1) Time-driven-buy strategy—trigger buy action aT,B when current price reaches
forecast low, day-end harvest;
(2) Time-driven-sell strategy—trigger sell action aT,S when current price reaches
forecast high, day-end harvest;
(3) Price-driven-buy strategy—trigger buy action aP,B when current price reaches
forecast low, closing when reaching target price/stop-loss price;
(4) Price-driven-sell strategy—trigger sell action aP,S when current price reaches
forecast high, closing when reaching target price/stop-loss price.
The hidden layer of quantum finance CRL network consists of H LRbf nodes,
which facilitate transient-chaotic RRB-based RL.
The output layer consists of four possible actions: aT,B , aT,S , aP,B , and aP,S as
described above. Their scalar values are normalized values between 0 and 1 which
represent the size (lot) of investment (“0” means not invest; “1” means to invest one
complete lot).

13.3.6 Quantum Finance CRL Network Learning Algorithm

The overall quantum finance CRL network learning algorithm is presented in

Fig. 13.6.
As shown in Fig. 13.6, the Lee-oscillators are adopted into the algorithm in two
aspects: (1) replacement of the Gaussian RBF with chaotic RBF (LRbf) in the hid-
den layer; (2) replacement of sigmoid-based activation function with Lee-transient-
chaotic activation function L in output layer for the determination of action a. From
the neural dynamic perspective, the adoption of chaotic RBF and activation function
can be considered as the emulation of chaotic trading behaviors in the time series
events and actions. While from the financial perspective, such adoption of transient-
chaotic activation and RBF can be considered as the imitation of transient-hysteresis
investors’ behavior during real-time trading.
For the value function V, the physical meaning of its optimal value V* in terms
of financial investment is even more profound: while rt is total returns (rewards) at
time-step t which represents the short-term returns in day trade; its value function V
plainly represents the long-term returns of the whole investment policy ; and its
optimal V* directly represents the optimal long-term returns by using the optimal
quantum trader network configuration and hence the optimal policy *.
13.4 System Implementation 385

Fig. 13.6 Quantum finance CRL network learning algorithm

13.4 System Implementation

13.4.1 Quantum Finance—TEQNA 5-Tier Architecture

From the implementation perspective, quantum finance system model adopts a 5-tier
system implementation architecture, namely, TEQNA as shown in Fig. 13.7.
Quantum finance TEQNA system architecture consists of the following:
1. Technology layer—supports MT4/MT5 platforms and related programming
technologies such as expert advisors (EAs), quantum trader agent protocol
(CAP).
2. Encryption layer—supports critical cryptographical technologies include both
encryption and blockchain technologies.
386 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Quantum Finance
5-Tier Model ARCHITECTURE

Quantum Quantum Quantum Quantum

Autoassociator Forecaster Trader Cri cs

AGENTS LAYER
Quantum Finance CULN Quantum Finance CSLN Quantum Finance CRLN
Chao c UnsupervIsed-Learning Networks Chao c Supervised-Learning Networks Chao c Reinforcement-Learning Networks

NEURAL NETWORKS LAYER

QUANTUM PRICE LEVEL QUANTUM FIELD SIGNALS QUANTUM FIELD OSCILLATORS
(QPL) (QFS) (QFO)

QUANTUM FIELD LAYER

Quantum Finance CRYPTOGRAPHIC TECHNOLGY Quantum Finance BLOCKCHAIN TECHNOLOGY
(ECDSA, RSA, SHA-256, MD5) (Hyperledger, Mul chain, Ethereum)

ENCRYPTION LAYER
MT4 MT5 CAP
(MetaTrader4 Pla orm) (MetaTrader5 Pla orm) (Chao c Agent Protocol)

TECHNOLOGY LAYER

Fig. 13.7 Quantum trader—TEQNA 5-layer system architecture

3. Quantum field layer—supports three basic components of quantum field mod-

eling of financial markets: quantum price levels (QPL), quantum field signals
(QFS), and quantum field oscillators (QFO) studied in Part I.
4. Neural networks layer—supports quantum finance chaotic supervised-learning
(CSL) networks, quantum finance unsupervised-learning (CUSL) networks, and
quantum finance chaotic reinforcement-learning (CRL) networks.
5. Agents layer—supports intelligent chaotic agent applications include quantum
auto-associator, forecaster, trading and critic agents.

13.4.2 Quantum Trader System Implementation Using 129

Worldwide Financial Products

From the application perspective, real-time and historical data of worldwide 129
financial products provided by Forex.com (major online Forex trading platform) and
AvaTrade.com (the biggest cryptocurrency trading platform) are adopted for quantum
trader system implementation. They include: major cryptocurrencies (9), worldwide
13.4 System Implementation 387

forex (84), major commodities (19), worldwide financial indices (17). Appendix A
shows the list of financial products under these four categories.

13.4.3 Quantum Trader—System Implementation

For the ease of fully integration and automation of quantum trader system with both
real-time and historical financial data provided by Forex.com and AvaTrade.com, the
whole intelligent agent-based system fully integrated with MT platforms. Figure 13.8
shows the system implementation framework of quantum trader.
The whole quantum trader system consists of two main modules: (1) quantum
forecaster using quantum finance CSL network for time series worldwide financial
prediction; and (2) quantum trader and critic agents using quantum finance CRL
network for the optimization of trading strategy.
The agent activities of these quantum finance agents are described as follows:
Quantum Forecaster Agent
Quantum forecaster is a server-side forecast agent located at the server farm of
quantum finance forecast center (QFFC) using Intel i5 CPU 2.39 GHz 32 MB ram
Dell servers.
For each financial product, 2048-trading day data (except cryptocurrency which
only have 300-trading day data) include: open (O), high (H), low (L), close (C), and
volume (V) are automatically generated by MT4 engines of Forex.com and AvaTrade.
com. Through the trading signal generator, 39 most common trading signals are
generated, together with the 2048-trading day data, they are fed into the forecast
system for training and forecast of next-day open (O), high (H), low (L), and close
(C). The predicted forecasts of these 129 financial products are stored at the quantum
forecast database for the quantum trader trading agents to access. Appendix B shows
the 39 trading signals generated by trading signal generator.
Besides input time series data and trading signals, one of the major characteristics
of a quantum forecaster is to incorporate with quantum field signals (QFS) generated
by quantum field signal generator (QFSG) as shown in Fig. 13.8 that was studied in
Part I. These QFS are stored in QFS databank which will also be used by quantum
traders and critic agents for reinforcement learning.
Quantum Trader Agents
The quantum trader receives inputs from four sources: (1) input signals from markets
(stored at the input signal databanks of the quantum finance server farm) of current
and past records; (2) quantum field signals (QFS) from QFS databank; (3) current
forecasts stored at quantum forecast database of the quantum finance server farm;
and (4) reward signals from previous time-step generated by quantum finance critic
agent. By using the quantum finance CRL algorithm, it evaluates four possible actions
a. Updates its trading policy and keeps track of the market movements in order to
trigger the trading actions.
388 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Fig. 13.8 System implementation framework of quantum trader

13.4 System Implementation 389

Quantum Finance Critic Agent

Quantum finance critic agent receives inputs from two sources: (1) input signals
from the markets of current and past records; (2) quantum field signals (QFS) from
QFS databank; (3) latest action a taken by the quantum trader. By using the reward
evaluation algorithm, quantum finance critic agent evaluates the latest rewards r of
four actions and feedback to the quantum trader as reinforcement signals for the
determination of investment actions in the next time-step.

13.4.4 Quantum Finance Daily Forecast at QFFC

Figure 13.9 shows a snapshot of quantum finance forecaster (or quantum forecaster
in short) for the training and forecast of 120 financial products of Forex.com on
November 09, 2018. As shown, in a typical daily forecast, the quantum forecaster
only took 62,341 ms (62.341 s) to finish the training forecast of 120 financial products.
On average, it took 0.519 s (less than 1 s) to finish the network training and forecast
process of a single financial product.
Figure 13.10 shows the snapshot of quantum forecaster for the system training
and forecast of 9 major cryptocurrencies over AvaTrade.com MT platform on the
same trading day.
As shown, quantum forecaster took 44,976 ms (44.976 s) to finish the training
and forecast of 9 cryptocurrencies. On average, it took 4.99 s to train and forecast a
single cryptocurrency.

Fig. 13.9 Snapshot of quantum forecaster for the training and forecast of 120 financial products
for Forex.com MT4 platform on November 09, 2018
390 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Fig. 13.10 Snapshot of quantum forecaster for training and forecast of 9 major cryptocurrencies
for AvaTrade.com MT4 platform on November 09, 2018

As compared with all those 120 non-cryptocurrency products, quantum forecaster

took 9.62 times of computer time to predict cryptocurrency, even though cryptocur-
rency only have 300-trading day records while the other 120 financial products each
have 2048-trading day records for system training. It might due to cryptocurrencies,
in general, are much more chaotic and fluctuant in nature, which takes more time
and iterations for the quantum forecaster to learn the market patterns.

13.4.5 Quantum Forecaster Performance Test

The forecast performance of quantum forecaster is compared with traditional FFBPN

by applying 500 forecast simulations for each system. For the ease of comparison,
four categories of worldwide 129 financial products are tested network MSE (mean-
square-error) ranging from 1 × 10-4 to 1 × 10-6, respectively.
Besides traditional FFPBN, two state-of-the-art financial forecasting methods:
(i) SVM (support vector machine) forecasting tool provided by R project, one of
the most popular financial forecasting tools used in finance industry and (ii) deep
neural network (DNN) with principal component analysis (PCA) model (Singh and
Srivastava 2017) are contrasted with quantum forecaster for forecast performance
comparison. Table 13.1 presents quantum finance forecast performance comparison
chart of these four systems.
Table 13.1 Quantum forecaster forecast performance comparison chart
Product category No. of PRD FFBPN SVM DNN-PCA Quantum forecaster
Total STT Av. STT Total STT Av. STT Total STT Av. STT Total STT Av. STT
Case 1 (MSE = 1 × 10-4)
Cryptocurrency 9 557,251 61916.78 372,244 41360.41 244,633 27181.47 5720 635.56
Forex 84 508,453 605. 01 331,511 3948.56 291,344 3488.38 6323 75.27
Financial index 19 41,120 2184.21 26,152 1378.44 19,738 1038.82 923 48.58
13.4 System Implementation

Commodity 17 46,641 2743.59 29,384 1728.46 22,108 1300.46 1223 71.94

Overall 129 1,153,465 8941.59 759,291 5885.98 577,823 4479.25 14,189 109.99
Case 2 (MSE = 1 × 10-5)
Cryptocurrency 9 1,460,000 162222.22 937,320 104148.67 823,440 91493.33 15,121 1680.11
Forex 84 1,235,543 14708.85 841,405 10016.72 720,322 8575.26 17,231 205.13
Financial index 19 111,024 5843.37 68,391 3599.51 50,627 2664.58 2312 121.68
Commodity 17 109,142 6420.12 71,925 4230.86 44,967 2645.09 2640 155.29
Overall 129 2,915,709 22602.40 1,919,041 14878.29 1,639,356 12708.19 37,304 289.18
Case 3 (MSE = 1 × 10-6)
Cryptocurrency 9 DL 6,102,255 678028.38 3,601,307 400145.17 57,235 6359.44
Forex 84 DL 5,477,816 65212.10 4,184,833 49819.44 71,243 848.13
Financial index 19 577,324 30385.47 373,529 19659.40 318,683 16772.78 6623 348.58
Commodity 17 687,595 40446.76 468,252 27544.25 385,741 22690.64 9126 536.82
Overall 129 – 12,421,852 96293.43 8,490,564 65818.33 144,227 1118.04
Note
1. Results are generated by 500 simulations of each neural network system (measured in ms)
2. “Total STT” denotes the total average system training time for 500 simulations of network training
3. “Av. STT” denotes the average system training lime for a single financial product of each category
4. “DL” denotes deadlock during system training
391
392 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Certain interesting findings were revealed:

(i) For the case I simulation using MSE 1 × 10-4, quantum forecaster outper-
formed FFBPN, SVM, and DNN-PCA by 81.29, 53.51, and 40.72 times,
respectively. Similar findings can be found in case II simulation results. It
clearly reflected the improvement of network learning rate by the adoption of
chaotic neural oscillators into traditional FFBPN.
(ii) Across the 3 cases with decreasing MSE from 1 × 10-4 (case I), 1 × 10-5
(case II) to 1 × 10-6 (case III). All forecast systems achieved MSE in case I
and case II. However, case III simulations using MSE 1 × 10-6, FFBPN encoun-
tered “deadlock” problems during the network training of cryptocurrency and
forex products; while quantum forecaster could still finish the network training
with promising training speeds. It clearly demonstrated the resolution of over-
training and deadlock problems and sufficient improvement of system training
efficiency by quantum forecaster over traditional neural networks using clas-
sical sigmoid-based activation function.
(iii) In terms of system performance across different financial products, the simula-
tion results clearly reflected that both cryptocurrency and forex were more
chaotic and difficult for network training than other financial products as
expected, which will be one of the future R&D directions of quantum finance
forecast center.

13.4.6 Quantum Trader Trading Strategy Performance Test

In order to conduct a systematic test of quantum trader on intelligent trading, two

more benchmark agents were implemented for comparison purposes:
(1) time-driven trading agent (TmAgent);
(2) price-driven trading agent (PrAgent).
These two agents simulated real-world traders by using quantum finance forecasts
together with simple short-term trading strategies of:
(1) day-end harvest strategy (time-driven strategy); or
(2) target-profit and stop-loss strategy (price-driven strategy) with ratio 2:1 (i.e.,
target-profit = 2 × stop-loss).
For the 9 cryptocurrency products, 300-trading day simulations were performed
for each trading agent; while 500-trading day simulations were performed for 120
non-cryptocurrency financial products. Table 13.2 summarizes the trading perfor-
mance results for three trading strategies.
 13.4 System Implementation

Table 13.2 Quantum trader trading strategy performance comparison chart

Product category Time-driven strategy Price-driven strategy Quantum trader strategy
BUY (%) SELL (%) Overall (%) BUY (%) SELL (%) Overall (%) BUY (%) SELL (%) Overall (%)
Cryptocurrency 2.66 3.17 2.92 3.12 3.12 3.47 6.71 9.37 8.04
Fores 5.52 7.22 6.37 6.32 1.16 7.04 12.41 13.83 13.12
Financial index 5.72 6.11 5.92 5.49 6.71 6.10 9.56 10.21 9.89
Commodity 4.39 6.01 5.20 5.12 6.43 5.7S 10.43 11.88 11.16
Average 4.57 5.63 5.10 5.01 6.18 5.60 9.78 11.32 10.55
Note
1. Results (Monthly Returns %) are generated by 300 trading day simulations of the 9 Cryptocurrencies and 500 trading day simulations of the 120 non-
cyptocurrency products
393
394 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Major findings can be concluded in three aspects:

(i) Across the three trading strategies, overall speaking, quantum trader trad-
ing strategy outperformed time-driven strategy and price-driven strategy by
106.86% and 88.52%, respectively. It clearly reflected the effectiveness of
the combined strategy in quantum trader by using the chaotic reinforcement-
learning technique. Comparing time-driven strategy with price-driven strategy,
it was interested to reveal that price-driven strategy was consistently outper-
formed its time-driven counterpart by around 9.73%. It can be explained by
the fact that although the time-driven strategy was rather typical in automatic
program trading for day trade, the closing-price normally wouldn’t be the opti-
mal price for bid-closing, as compared with the 2:1 target-profit and stop-loss
price-driven strategy which was more sensible in terms of short-term trading.
(ii) Across the four different categories of financial products, the overall returns of
cryptocurrency products were the worst as expected, since the forecast perfor-
mance of cryptocurrency products were also the lowest as compared with the
other three categories. However, for forex products, it was interested to find
out that although the forecast performance of forex products was not as good
as financial indices and commodities, their overall returns outperformed these
two categories and rank no. 1. As reflected by experienced traders, it might be
owing to the fact that although the daily patterns of forex products were highly
chaotic, their price patterns were mostly (over 80% of the time) oscillations
between the predicted high/low without trigger the stop-loss thresholds.
(iii) As compared with the buy versus sell strategy, it was interested to find out
that the overall returns from sell strategy were consistently higher than its
buy strategy counterpart by around 16%, which was quite consistent with the
comments from experienced traders that automatic program trading usually
favors sell versus buy strategy.

13.4.7 Quantum Trader Worldwide Trading Strategy

Evaluation

Started from January 1, 2018, 342 members of QFFC which consists of professional
traders and quants from major fund houses are invited to join the quantum trader
system evaluation focus group for a 1-year worldwide evaluation program of quantum
trader trading system as compared with their own trading strategies.
They were all provided with free quantum finance daily forecasts for the 129
financial products and the two de facto agent trading programs (i.e., the time-driven
and price-driven trading agents). Based on quantum finance daily forecasts, they are
free to use either the de facto trading agents or their own trading strategies to do
trading.
13.4 System Implementation 395

Table 13.3 Quantum trader trading performance comparison chart

Product FFBPN SVM DNN- ALL Top Top 5% Quantum
(%) (%) PCA traders 15% traders trader
(%) (%) traders (%) (%)
(%)
Cryptocurrency 2.03 5.08 5.32 2.76 6.54 8.76 7.62
Forex 2.72 7.91 8.29 5.32 12.31 16.71 13.16
Financial index 3.65 7.59 7.89 4.24 11.22 15.23 11.37
Commodity 3.87 9.52 10.01 4.12 11.81 16.71 12.67
Average 3.07 7.53 7.88 4.11 10.47 14.35 10.55

Table 13.3 presents the nine months (1.1.2018–9.30.2018) trading performance

comparison chart for the worldwide evaluation scheme.
For the ease of comparison, quantum trader trading performance (overall 10.55%)
was compared with the following:
(1) Traditional FFBPN with day-trade algorithm (overall 3.07%);
(2) SVM forecasting tool with day-trade algorithm (overall 7.53%);
(3) DNN-PCA forecasting model with day-trade algorithm (overall 7.88%);
(4) Overall average performance of the focus group (overall 4.11%);
(5) Top 15% traders’ performance (overall 10.47%);
(6) Top 5% traders’ performance (overall 14.35%).
As revealed from the performance analysis, it was clear that the overall trading
performance of quantum trader outperformed FFBPN, SVM, and DNN-PCA by
around 7.49%, 3.03%, and 2.67%, respectively (in terms of overall returns). On the
other hand, as compared with professional traders, quantum trader trading perfor-
mance was close to a typical top 15% trader, outperformed the average traders by
over 6.44%, which was rather promising as feedback from professional traders.

13.5 Conclusion

This chapter devised an innovative multiagent-based quantum finance forecast and

trading system—quantum trader. From the implementation perspective, the quan-
tum trader was integrated with quantum finance forecast center for the provision of
worldwide 129 financial product forecasts and intelligent trading systems.
In fact, for a professional trader and investor, a reliable and effective financial
forecast system is only the beginning of the story. A good financial investment also
needs: (1) good and effective trading and hedging strategies; (2) stable, logical, and
rational investment psychology.
396 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Problems

13.1 What is an intelligent agent? What are the major differences between tradi-
tional computer programs and intelligent agents? Give two real-world exam-
ples on how to apply intelligent agent technology in finance engineering.
13.2 Discuss and explain the major advantages of using intelligent agent technology
for the implementation of intelligent trading systems.
13.3 Discuss and explain why program trading become so popular in nowadays
financial community. What are the major differences between traditional finan-
cial trading versus program trading? Give two examples of program trading
and describe how they work.
13.4 Describe and explain what is high-frequency algorithmic program trading
(HFAPT). Give two examples of HFAPT in forex markets and describe how
they work.
13.5 Discuss and explain why nowadays AI become so popular in building program
trading systems. Give two examples of financial product trading for illustration.
13.6 Financial prediction nowadays becomes one of the hottest topics in financial
engineering. Why now as AI prediction already exists for over half a century?
13.7 State and discuss three contemporary AI-based financial prediction technolo-
gies and briefly explain how they work.
13.8 As mentioned in Chap. 8, chaos theory told us that the forecast of complex
systems such as weather and financial markets are bounded by the initial
condition. How can AI such as artificial neural networks or recurrent neural
networks get around with this intrinsic problem? Use forex prediction system
as an example for explanation.

Acknowledgements The author wishes to thank Forex.com and AvaTrade.com for the provision
of historical and real-time financial data. The author also wishes to thank Quantum Finance Forecast
Center of UIC for the R&D supports and the provision of the channel and platform Qffc.org for
worldwide system testing and evaluation.

References

Borden, C. (2018) Fibonacci Trading: How to Master the Time and Price Advantage. McGraw-Hill
Education, UK.
Brown, C. (2012) Elliot Wave Principle: Elementary Concepts, Wave Patterns, and Practice Exer-
cises. Bloomberg Press, New York.
Bulkowski, T. N. (2005) Encyclopedia of Chart Patterns. Wiley, UK.
Carapuço, J. et al. (2018) “Reinforcement learning applied to Forex trading,” Applied Soft Computing
Journal 73: 783–794.
References 397

Chang, P. C. et al. (2012) A novel model by evolving partially connected neural network for stock
price trend forecasting. Expert Systems with Applications 39(1): 611–620.
Chang, C. et al. (2017) Asymmetric Reinforcement Learning and Conditioned Responses During
the 2007–2009 Global Financial Crisis: Evidence from Taiwan. Review of Pacific Basin Financial
Markets and Policies 20(2): 1750010.
Dabhi, V. K. and Chaudhary, S. (2016) Financial time series modeling and prediction using postfix-
GP. Computational Economics 47(2): 219–253.
Dai, W. et al. (2012) Combining nonlinear independent component analysis and neural network for
the prediction of Asian stock market indexes. Expert Systems with Applications 39(4): 4444–4452.
Deng Y. et al. (2017) Deep Direct Reinforcement Learning for Financial Signal Representation and
Trading. IEEE Transactions on Neural Networks and Learning Systems 28(3): 653–664.
Durbin, M. (2010) All About High-Frequency Trading. McGraw-Hill Education.
Henrique, B. M. et al. (2018) Stock price prediction using support vector regression on daily and
up to the minute prices. The Journal of Finance and Data Science 4(3): 183–201.
Hwang, H. and Oh, J. (2010) Fuzzy models for predicting time series stock price index. International
Journal of Control, Automation, and Systems 8(3): 702–706.
Jeong G. and Kim, H. Y. (2019) Improving financial trading decisions using deep Q-learning:
Predicting the number of shares, action strategies, and transfer learning. Expert Systems with
Applications 117: 125–138.
Kaelbling, L. P. et al. (1996) Reinforcement Learning: A Survey. Journal of Artificial Intelligence
Research 4: 237–285.
Lee, R. S. T. (2004) A Transient-chaotic Auto-associative Network (TCAN) based on LEE-
oscillators. IEEE Trans. Neural Networks 15(5): 1228–1243.
Lee, R. S. T. (2006) Fuzzy-Neuro Approach to Agent Applications (From the AI Perspective to
Modern Ontology). Springer-Verlag, Germany.
Lee, R. S. T. (2019) COSMOS Trader – Chaotic Neuro-oscillatory Multiagent Financial Prediction
and Trading System,” The Journal of Finance and Data Science 5: 61–82.
Li, C. G. et al. (2009) An Actor-Critic reinforcement learning algorithm based on adaptive RBF
network. 2009 IEEE International Conference on Machine Learning and Cybernetics 984–988.
Ling, P. W. (2013) The stock price forecasting comparative research of the use of fractal theory
at Taiwan traditional industry and technology industry. Applied Mechanics and Materials 274:
53–56.
Liu, F. et al. (2005) Neural network model for time series prediction by reinforcement learning.
2005 IEEE International Joint Conference on Neural Networks 2: 809–814.
Markopoulos, A. P. et al. (2016) On the use of back propagation and radial basis function neural
networks in surface roughness prediction. Journal of Industrial Engineering International 12(3):
389–400.
Murphy, J. J. (1999) Technical Analysis of the Financial Markets: A Comprehensive Guide to
Trading Methods and Application. New York Institute of Finance, New York.
Nahil, A. and Lyhyaoui, A. (2018) Short-term stock price forecasting using kernel principal com-
ponent analysis and support vector machines: the case of Casablanca stock exchange. Procedia
Computer Science 127: 161–169.
Narayanan A. et al. (2016) Bitcoin and Cryptocurrency Technologies: A Comprehensive Introduc-
tion. Princeton University Press.
Ouahilal, M. et al. (2017) A novel hybrid model based on Hodrick–Prescott filter and support vector
regression algorithm for optimizing stock market price prediction. Journal of Big Data 4(1): 1–22.
Pendharkar P. C. and Cusatis, P. (2018) Trading financial indices with reinforcement learning agents.
Expert Systems with Applications 103: 1–13.
Singh, R., & Srivastava, S. (2017) Stock prediction using deep learning. Multimedia Tools and
Applications 76(18): 18569–18584.
Tan, Z. et al. (2011) Stock trading with cycles: A financial application of ANFIS and reinforcement
learning. Expert Systems with Applications 38(5): 4741–4755.
398 13 Quantum Trader—A Multiagent-Based Quantum Financial Forecast …

Vigna, P. and Casey, M. J. (2016.) The Age of Cryptocurrency: How Bitcoin and the Blockchain
Are Challenging the Global Economic Order. Picador,
Walker, W. (2018) Expert Advisor Programming and Advanced Forex Strategies. Independently
published.
Young, A. R. (2015) Expert Advisor Programming for MetaTrader 4: Creating automated trading
systems in the MQL4 language. Edgehill Publishing.
Chapter 14
Future Trends in Quantum Finance

The most important application of quantum computing in the

future is likely to be a computer simulation of quantum systems,
because that’s an application where we know for sure that
quantum systems in general cannot be efficiently simulated on a
classical computer.
David Deutsch (Born in 1953)

Professor David Deutsch in his famous quotation pointed out an important role and
function of a quantum computer—the simulation of quantum systems and related
applications, which cannot be efficiently simulated by classical computers.
Is it really true?
The answer is yes and no. “Yes” in the sense that the major objective of the quantum
computer is to model quantum dynamics and simulate quantum applications with
quantum computing technology—so-called hard quantum computing. “No” in the
sense that without the usage of quantum computers, technically we can still model
quantum dynamics and quantum applications such as quantum finance model we
have studied in this book—so-called soft quantum computing.
In this final chapter, we will explore the future of quantum finance and the new era
of quantum computing. Similar to the history of artificial intelligence (AI) that has
both hard AI and soft AI. Each of them focuses on different aspects and applications,
but after over half a century; the boundary between them becomes obscure. In this
chapter, we will study two disciplines of quantum computing and how they interact
to shape the new age of intelligent computing.
As mentioned in previous chapters, quantum finance is a cross-discipline subject to
challenge the utmost question of finance engineering—the exploration of the nature
and dynamics of financial market activities in the quantum realm.
Quantum finance is one of the major research areas of quantum finance forecast
center (QFFC.org). In this chapter, we will explore and introduce other aspects and
active research of QFFC in terms of (1) AI-fintech—the integration of AI technology
with state-of-the-art finance technology such as blockchains and digital ledgers;
(2) quantum computing—the R&D of potential applications of quantum computing
technology in various areas such as quantum entanglements in quantum finance,
quantum cryptography, quantum holographic systems, etc.

© Springer Nature Singapore Pte Ltd. 2020 399

R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8_14
400 14 Future Trends in Quantum Finance

14.1 Introduction—The Dawn of Quantum Computing

Since the birth of AI and computing technology from over half a century ago, we are
now coming to the historical moment of a new era on computing technology—quan-
tum computing. With the official launch of IBM’s System Q—the first commercially
available quantum computer at CES 2019 Las Vegas USA in January 2019 signify
the dawn of quantum computing technology (Moran 2019; Silva 2018).
Different from traditional computers which use bits of 0s and 1s as the fun-
damental binary representation in data storage and program calculation, quantum
computers use qubits (stands for quantum bits)—quantum superpositions of the two
states which allow for far greater flexibility than the traditional binary system in data
representation and programming (Hirvensalo 2010; Scherer 2019; Zygelman 2018).
In general, a quantum computer technically would be able to perform calculations
on a far greater order of magnitude than traditional computers. As mentioned by
Professor David Deutsch in his famous quotation stated at the beginning of this
chapter, quantum computing is the best solution tailored for the modeling of quantum
systems with related highly complex and sophisticate quantum phenomena such as
quantum cryptography and quantum finance in our case (Assche 2006; Bernhardt
2019; Jaeger 2018).

14.2 Two Sides of the Same Coin—Hard Quantum

Computing Versus Soft Quantum Computing

The origin of quantum computing can be traced back to 1959 a speech by Professor
Richard Feynman in which he spoke about the effects of miniaturization, including
the idea of exploiting quantum effects to create more powerful computers. However,
before the quantum effects of computing could be realized, quantum computing is still
a concept and idea in the scientific community. It was not until 1985, Professor David
Deutsch proposed a new idea of quantum logic gates in his paper Quantum theory
as a universal physical theory as a means of harnessing the quantum realm inside a
computer. In fact, his paper on the subject showed that any physical process could
be modeled by a physical quantum computer—so-called hard quantum computing
(Deutsch 1985; Steane 1998).
The road of hard quantum computing was not an easy path. Owing to technical
complexity and difficulty, after almost decades, Professor Peter Shor devised an algo-
rithm in 1994 that could use only 6 qubits to perform basic factorization operations.
Since then, a handful of quantum computers were built. The first, a 2-qubit quantum
computer in 1998, could perform trivial calculations before losing decoherence after
a few nanoseconds. In 2000, quantum computing teams including IBM and AT&T
successfully built both a 4-qubit and a 7-qubit quantum computers.
IBM has showcased its IBM Q System One at CES 2019, which was claimed as
the world’s first integrated quantum computing system designed for scientific and
14.2 Two Sides of the Same Coin … 401

commercial use. The 20-qubit system by IBM, which was seen as one of the leaders
in the field of quantum computing (Silva 2018; Moran 2019).
Just like hard AI versus soft AI in artificial intelligence (AI) realm, soft quantum
computing refers to the modeling of quantum theory and its mathematical models by
means of computing systems and programs, similar to the analog of soft AI with the
focus of building computer models using artificial neural networks (ANN) to model
human intelligence and learning process.
In other words, all quantum theory concepts and models we have studied in this
book, ranging from Feynman’s path integral for the modeling of forward interest
rate to quantum anharmonic oscillator model for the modeling of quantum price
levels; along with quantum financial signals are components and frameworks of soft
quantum computing.
If we can model quantum theory and related phenomena by means of traditional
computers, why do we need to build quantum computers?
The answer is simple. If we can model quantum theory effectively using tra-
ditional computers with binary formulation and representations, integrating these
quantum models with quantum computers will become more direct and effective
while the fundamental construction of quantum computer itself totally conforms to
quantum model and architecture. In other words, by implementing a quantum model,
say quantum finance model into a traditional computer; technically speaking, we are
converting a traditional financial computation system into a quantum financial com-
putation machine!

14.3 From Quantum Finance to AI Fintech

As mentioned earlier in this chapter, quantum finance is one of the major R&D
projects of quantum finance forecast center (QFFC.org)—a worldwide financial pre-
diction and R&D center for the design and implementation of next generation of
AI-fintech standards, toolkits and applications (Nicoletti 2017).
Figure 14.1 depicts the AI-fintech 5-layer architecture of QFFC.
(1) AI-applications Layer
• This layer focuses on AI-fintech-related systems and applications in four
major areas: forecasts such as quantum finance forecast, deep learn-
ing/reinforcement learning such as multiagent-based trading agents; data
mining and NLP (natural language processing) such as robo-advisor (Lim
et al. 2011; Sironi 2016).
(2) Business Layer
• This layer focuses on the business applications and fintech systems includ-
ing cryptocurrency, blockchain systems on financial trading, payment, clear-
ing and hedging transactions (Antonopoulos 2017; Gaur 2018; Sironi 2016;
Nielsen 2011, Swan 2015).
402 14 Future Trends in Quantum Finance

AI
Forecasting
Deep Learning
Data Mining
Natural Language Processing

Business
Cryptocurrency &
Blockchain System on Trading,
Payment, Clearing, Hedging

Security
Blockchain 1.0/2.0/3.0 Technology
Quantum Cryptography

Technology
Cloud, Bit Torrent, Intelligent Agents, Neural
Networks, Deep Learning, Biometrics, Data Mining

Internet

Fig. 14.1 AI-Fintech 5-layer architecture

(3) Security Layer

• This layer focuses on the implementation of fintech security systems and
technology include Blockchain 1.0/2.0/3.0 systems and quantum cryptosys-
tems on crypto-payment and transactions (Assis 2012; Balygin 2017; Elliott
2004; Fehr 2010).
(4) Technology
• This layer focuses on all the fundamental and building blocks technology and
standards for the implementation of AI-fintech applications.
(5) Internet Layer
• This layer provides the basic framework and backbone for AI-fintech intelli-
gent agent communication technology, protocols, and technology.

14.4 Quantum Finance System Development Platform

Besides the provision of various AI-fintech forecasts and applications, QFFC also
actively developing a next-generation AI-fintech development kit—quantum finance
development kit (QFDK), an easy to use C/C++ library with the implementation of
all state-of-the-art AI-fintech functions and libraries; so that AI-fintech developers
14.4 Quantum Finance System Development Platform 403

Fig. 14.2 QFSDK 5-layer model

and data scientists can base on these functional libraries to design and develop their
own quantum finance and AI-fintech systems and applications.
Figure 14.2 depicts the QFSDK 5-layers model of version 1.0, which will be
launched officially in late 2019 at QFFC.org official site.

14.5 Conclusion

In this final chapter, we studied the future trend in quantum finance—a cross-
discipline subject that comprises quantum theory as a theoretical foundation;
numerical computational theory as computer model; artificial intelligence as machine
404 14 Future Trends in Quantum Finance

learning; prediction and trading optimization models; and finally quantum comput-
ing as the ultimate intelligence machine. In other words, the future R&D and growth
of quantum finance technology not only depends on the development of quantum
finance theory itself but also depends on future development and growth of related
technologies such as AI and quantum computing technologies.
As mentioned in the introductory chapter, since Professor Heisenberg introduced
the ground-breaking quantum mechanics’ idea to nowadays, we have concrete mathe-
matical and computational models to model quantum dynamics and QPL of financial
markets, along with the interpretation of path integrals of the forward interest rate.
It is both a challenging and exciting path. The contribution is not only the credit of
an individual but also is the collective contribution of intelligence, innovation, and
R&D from many centuries.
In view of the advancements of technology and AI in the past decades, quantum
finance, together with AI and quantum computing technology should work together to
provide more scientific and intelligent applications, systems and services for world-
wide investors and financial professionals.

Problems

14.1 What is a quantum computer? State and explain the major difference between
a quantum computer and the traditional computer systems.
14.2 State and explain the major challenges in quantum computing. Give two exam-
ples of illustrations.
14.3 State and contrasts the major differences between hard quantum computing
and soft quantum computing. Give two examples in each case for illustration.
14.4 Discuss whether quantum computer in the future will or will not replace tra-
ditional computers. Why?
14.5 What is finance technology (fintech)? State three typical fintech-related tech-
nologies and briefly explain how they work.
14.6 What is a robo-advisor in fintech? State and explain the basic underlying
technology. Give two examples of robo-advisor in fintech and explain briefly
how they work.
14.7 What is blockchain technology? State and explain the difference between
Blockchain 1.0, 2.0, and 3.0 technologies. Give one example in each case
for illustration.
14.8 What is quantum cryptography? State and explain the major difference
between traditional cryptosystem and quantum cryptosystem. What are the
major challenges in quantum cryptography?
References 405

References

Antonopoulos, A. M. (2017) Mastering Bitcoin: Programming the Open Blockchain, O’Reilly.

Assche, G. (2006) Quantum Cryptography and Secret-Key Distillation. Cambridge University Press.
Assis et al. (2012) Improving classical authentication over a quantum channel. Entropy 14(12):
2531–2549.
Balygin et al. A. (2017) Practical quantum cryptography. JETP Letters 105(9): 606–612.
Bernhardt, C. (2019) Quantum Computing for Everyone. MIT Press.
Deutsch, D. (1985) Quantum theory as a universal physical theory. International Journal of Theo-
retical Physics 24(1): 1–41.
Elliott, C. (2004) Quantum cryptography,” IEEE Security & Privacy, vol. 2, (4), pp. 57–61.
Fehr, S. (2010) Quantum cryptography. Foundations of Physics 40(5): 494–531.
Gaur, N. (2018) Hands-On Blockchain with Hyperledger: Building decentralized applications with
Hyperledger Fabric and Composer, Packt Publishing.
Hirvensalo, M. (2010.) Quantum Computing. Springer.
Jaeger, L. (2018) The Second Quantum Revolution: From Entanglement to Quantum Computing
and Other Super-Technologies. Copernicus.
Lim, E. H. Y., Liu, J. N. K. and Lee, R. S. T. (2011) Knowledge Seeker – Ontology Modelling for
Information Search and Management, Springer-Verlag, Germany.
Moran, C. C. (2019) Mastering Quantum Computing with IBM QX: Explore the world of quantum
computing using the Quantum Composer and Qiskit. Packt Publishing.
Nicoletti, B. (2017) The Future of Fintech: Integrating Finance and Technology in Financial Ser-
vices, Palgrave Macmillan.
Nielsen, M. A. and Chuang, I. L. (2011) Quantum Computation and Quantum Information: 10th
Anniversary Edition. Cambridge University Press.
Scherer, W. (2019) Mathematics of Quantum Computing: An Introduction. Springer.
Silva V. (2018) Practical Quantum Computing for Developers: Programming Quantum Rigs in the
Cloud using Python, Quantum Assembly Language and IBM QExperience, APress.
Sironi, P. (2016) Fintech Innovation: From Robo-Advisors to Goal Based Investing and Gamifica-
tion, Wiley.
Steane, A. (1998) Quantum computing. Reports on Progress in Physics 61(2): 117–173.
Swan, M. (2015.) Blockchain: Blueprint for a New Economy, O’Reilly.
Zygelman, B. (2018) A First Introduction to Quantum Computing and Information. Springer.
Appendix A
List of 139 Financial Products

Code Product description Code Product description

Cryptocurrencies (Data provided by AvaTrade.com)
BCHUSD BitCoin Cash versus US Dollar EOSUSD EOS versus US Dollar
BTCEUR BitCoin versus Euro ETH Ethereum
BTCJPY BitCoin versus Japanese Yen LTC Litecoin
BTCUSD BitCoin versus US Dollar XRP XRP
BTGUSD Bitcoin Gold versus US Dollar
Financial index (Data provided by Forex.com)
AUS200 AUSSIE 200 N25 Netherlands 25 Index
CHINAA50 China A50 Index NAS100 Nasdaq Index
ESP35 Spain 35 Index SIGI Singapore Index
ESTX50 EURO STOXX 50 Index SPX500 SP500 Index
FRA40 CAC 40 Index SWISS20 Switzerland 20 Index
GER30 DAX 30 Index UK100 FTSE 100 Index
HK50 Hang Seng Index US2000 US Small Cap 2000
IT40 Italy 40 Index US30 Dow Jones Index
JPN225 Nikkei Index
Commodity (Data provided by Forex.com)
COPPER Copper US_OIL WTI Crude Oil
CORN Corn WHEAT Wheat
COTTON Cotton XAGUSD Silver versus US Dollar
HTG_OIL HTG Oil XAUAUD Gold versus Australian Dollar
PALLAD Palladium XAUCHF Gold versus Swiss Franc
PLAT Platinum XAUEUR Gold versus Euro
SOYBEAN Soybean XAUGBP Gold versus British Pound
(continued)
© Springer Nature Singapore Pte Ltd. 2020 407
R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8
408 Appendix A: List of 139 Financial Products

(continued)
Code Product description Code Product description
SUGAR Sugar XAUJPY Gold versus Japanese Yen
UK_OIL Brent Crude Oil XAUUSD Gold versus US Dollar
US_NATG US Natural Gas
Forex (Data provided by Forex.com)
AUDCAD Australian Dollar versus GBPDKK British Pound versus Danish
Canadian Dollar Krone
AUDCHF Australian Dollar versus Swiss GBPHKD British Pound versus Hong
Franc Kong Dollar
AUDCNH Australian Dollar versus GBPJPY British Pound versus Japanese
Chinese Yuan Yen
AUDJPY Australian Dollar versus GBPMXN British Pound versus Mexican
Japanese Yen Peso
AUDNOK Australian Dollar versus GBPNOK British Pound versus
Norwegian Krone Norwegian Krone
AUDNZD Australian versus New Zealand GBPNZD British Pound versus New
Dollar Zealand Dollar
AUDPLN Australian Dollar versus Polish GBPPLN British Pound versus Polish
Zloty Zloty
AUDSGD Australian Dollar versus GBPSEK British Pound versus Swedish
Singapore Dollar Krona
AUDUSD Australian Dollar versus US GBPSGD British Pound versus Singapore
Dollar Dollar
CADCHF Canadian Dollar versus Swiss GBPUSD British Pound versus US Dollar
Franc
CADJPY Canadian Dollar versus GBPZAR British Pound versus South
Japanese Yen African Rand
CADNOK Canadian Dollar versus HKDJPY Hong Kong Dollar versus
Norwegian Krone Japanese Yen
CADPLN Canadian Dollar versus Polish NOKDKK Norwegian Krone versus
Zloty Danish Krone
CHFHUF Swiss Franc versus Hungarian NOKJPY Norwegian Krone versus
Forint Japanese Yen
CHFJPY Swiss Franc versus Japanese NOKSEK Norwegian Krone versus
Yen Swedish Krona
CHFNOK Swiss Franc versus Norwegian NZDCAD New Zealand versus Canadian
Krone Dollar
CHFPLN Swiss Franc versus Polish Zloty NZDCHF New Zealand Dollar versus
Swiss Franc
CNHJPY Chinese Yuan versus Japanese NZDJPY New Zealand Dollar versus
Yen Japanese Yen
EURAUD Euro versus Australian Dollar NZDUSD New Zealand Dollar versus US
Dollar
(continued)
Appendix A: List of 139 Financial Products 409

(continued)
Code Product description Code Product description
EURCAD Euro versus Canadian Dollar SGDHKD Singapore versus Hong Kong
Dollar
EURCHF Euro versus Swiss Franc SGDJPY Singapore Dollar versus
Japanese Yen
EURCNH Euro versus Chinese Yuan TRYJPY Turkish Lira versus Japanese
Yen
EURCZK Euro versus Czech Koruna USDCAD US Dollar versus Canadian
Dollar
EURDKK Euro versus Danish Krone USDCHF US Dollar versus Swiss Franc
EURGBP Euro versus British Pound USDCNH US Dollar versus Chinese Yuan
EURHKD Euro versus Hong Kong Dollar USDCZK US Dollar versus Czech Koruna
EURHUF Euro versus Hungarian Forint USDDKK US Dollar versus Danish Krone
EURJPY Euro versus Japanese Yen USDHKD US Dollar versus Hong Kong
Dollar
EURMXN Euro versus Mexican Peso USDHUF US Dollar versus Hungarian
Forint
EURNOK Euro versus Norwegian Krone USDILS US Dollar versus Israeli Shekel
EURNZD Euro versus New Zealand USDJPY US Dollar versus Japanese Yen
Dollar
EURPLN Euro versus Polish Zloty USDMXN US Dollar versus Mexican Peso
EURRON Euro versus Romanian Leu USDNOK US Dollar versus Norwegian
Krone
EURRUB Euro versus Russian Ruble USDPLN US Dollar versus Polish Zloty
EURSEK Euro versus Swedish Krona USDRON US Dollar versus Romanian
Leu
EURSGD Euro versus Singapore Dollar USDRUB US Dollar versus Russian
Ruble
EURTRY Euro versus Turkish Lira USDSEK US Dollar versus Swedish
Krona
EURUSD Euro versus US Dollar USDSGD US Dollar versus Singapore
Dollar
EURZAR Euro versus South African USDTHB US Dollar versus Thai Baht
Rand
GBPAUD British Pound versus Australian USDTRY US Dollar versus Turkish Lira
Dollar
GBPCAD British Pound versus Canadian USDZAR US Dollar versus South African
Dollar Rand
GBPCHF British Pound versus Swiss ZARJPY South African Rand versus Jap
Franc Yen
Appendix B
List of 39 Financial Indicators

No Function name Signal name

1 iAC Accelerator Oscillator
2 iAD Accumulation/Distribution
3 iADX Average Directional Index
4 ADXWilder Av Dir by Welles Wilder
5 iAlligator Alligator
6 iAO Awesome Oscillator
7 iATR Average True Range
8 iBearsPower Bears Power
9 iBands Bollinger Bands
10 iBandsOnArray Bollinger Bands indicator, stored in array
11 iBullsPower Bulls Power
12 iCCI Commodity Channel Index
13 Chaikin Chaikin Oscillator
14 iCustom Custom indicator
15 DEMA Double Exponential MA
16 iDeMarker DeMarker
17 iEnvelopes Envelopes
18 iEnvelopesOnArray Calculation of Envelopes, stored in array
19 iForce Force Index
20 iFractals Fractals
21 iGator Gator Oscillator
22 iIchimoku Ichimoku Kinko Hyo
23 iBWMFI Market Facilitation Index by Bill Williams
24 iMomentum Momentum
(continued)
© Springer Nature Singapore Pte Ltd. 2020 411
R. S. T. Lee, Quantum Finance,
https://doi.org/10.1007/978-981-32-9796-8
412 Appendix B: List of 39 Financial Indicators

(continued)
No Function name Signal name
25 iMomentumOnArray Calculation of Momentum, stored in array
26 iMFI Money Flow Index
27 iMA Moving Average
28 iMAOnArray Calculation of Moving Average indicator on data
29 iOsMA Moving Average of Oscillator
30 iMACD MACD indicator
31 iOBV On Balance Volume
32 iSAR Parabolic Stop And Reverse System
33 iRSI Relative Strength Index
34 iRSIOnArray Calculation of Momentum, stored in array
35 iRVI Relative Vigor Index
36 iStdDev Standard Deviation
37 iStdDevOnArray Standard Deviation indicator, stored in array
38 iStochastic Stochastic Oscillator
39 iWPR Williams’ Percent Range