Master of Technology in Enterprise Business Analytics page 1 of 16

NUS-ISS

MASTER OF TECHNOLOGY IN
ENTERPRISE BUSINESS ANALYTICS

Graduate Certificate Examination

Sample Examination Questions

Subject: Specialized Predictive Modeling & Forecasting

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 2 of 16

Case Study
Harrison Electric Company

Harrison Electric Company (hereafter called HEC), located in Chicago’s Old Town area, is the
largest and most profitable manufacturer of electrical products. They make a range of 300
products which are sold to wholesalers and large retailers throughout the whole world.

The company offers ex-stock delivery of all items, and dispatch within 24 hours of receipt of
the orders by their business partner – Speed Shipment Company as a third party logistic
service provider.

Demand for all the products is somewhat seasonal in nature, being heavier in spring and early
summer and tapering off through fall and winter (except for a heavy surge in the six weeks
before Christmas). A breakdown of the previous year’s business with the retail outlets usually
forms the basis for HEC’s yearly operations plan.

A growth factor (either positive or negative) is used to refine further the demand estimate by
reflecting the upcoming yearly market growth scenario. By developing a yearly plan and
updating it when appropriate, HEC can establish a reasonable basis for obtaining any
necessary financing from the bank.

To cope with the severe competition in the global market, HEC has set up a data science team
to assist the various functional areas in data-driven decision making with the following
objectives:

 Perform a set of experiments that will assess the potential customers attitude towards a
new product launch (Bladeless Fan)
 Facilitate efficient go-to-market strategy, assess the appropriate sales and marketing
channels for demand generation
 Optimise inventory planning and management, which affects overall service level &
enable efficient cost management
 Optimise production and distribution of different products by category to maximise
profit while minimizing production & distribution cost

One of the production issues that will affect the success of the marketing and sales of the
Bladeless Fan is the reliability of this fan, i.e., how long will the fan run before breaking
down? HEC would like the Bladeless Fan to have a mean time to failure at least twice that of
similar products from competitors
 Currently the highest mean time to failure claimed by a competitor is 2200 hours of
continuous use
 So, HEC must achieve the reliability target of at least twice the time to failure with
their Bladeless Fan

As it is normal with most modern electronic products, the Bladeless Fan will contain both
HEC and other sub-contractor designed components which will be assembled & sub-
assembled by all the parties involved.

One such most vital component is the Electrostatic Linear Accelerator (ELA) which is
supplied by a subcontractor.

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 3 of 16

Three subcontractors (A, B, and C) have been identified as the potential supplier of ELA and
HEC needs to decide on which one to choose
 The choice will depend on which subcontractor can supply the best ELA that will allow
the Bladeless Fan to achieve the required reliability target (as described above)

To do this, HEC has obtained two samples of ELA’s from each subcontractor and installed
them into 100 Bladeless Fans. A study was then performed where these Bladeless Fans were
subjected to continuous operation during a fixed period and either,
 The time to failure was recorded,
or
 For the Bladeless Fans that were still working at the end of the study period, the time
to end were marked as censored observation

Currently the Bladeless Fans are being produced on two pre-launch production lines;

 Production line 1
 Production line 2

The 100 samples were selected randomly from either of the two production lines to minimize
other random effects

The following Data was collected during the study.

 The ID of the Bladeless Fan (1-100)

 The time to failure of the Bladeless Fan (hours)
 Whether the time to failure was censored (1 = censored, 0 = not censored)
 In which Production Line (1 or 2) the Bladeless Fan was assembled
 Which Subcontractor (A, B or C) provided the ELA that was incorporated into the
Bladeless Fan

Appendix A shows a sample of the data collected during the study

Appendix B1 shows the results of fitting Kaplan-Meier graphs to the collected data
Appendix B2 shows the results of fitting a Proportional Hazards Model to the collected data

Appendix A Sample of Times to failure of Bladeless Fan

Time to Production
ID failure Censored Line Subcontractor
1 4350 0 2 A
2 3600 0 2 A
3 1950 0 2 C
4 1050 0 1 C
5 4050 0 1 C
6 1950 0 1 B
7 3900 0 1 C
8 1650 1 1 B
9 2400 0 1 B
10 3450 0 1 A

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 4 of 16

Q1 (7 Marks)

Please answer the following questions by evaluating the production issues articulated in pages
3 & 4 and analyzing the information provided in Appendices B1 & B2.

a) Is there a significant difference in the survival times of the Bladeless Fans due to the ELA’s,
provided by the different subcontractors (in particular, observing the mean and median survival
times)? Justify your answer. (3 Marks)

b) Do any other factors, used in the model, impact on the survival times of the Bladeless Fans?
Justify your answer.
(2 Marks)

c) Which subcontractor would you select to provide the ELAs for the Bladeless Fan? Would
the Bladeless Fan achieve the desired mean time to failure? Justify your answer. (2 Marks)

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 5 of 16

Appendix B1 Kaplan Meier Analysis of Times to failure of Bladeless Fan

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 6 of 16

Appendix B2: Proportional Hazards Model of Times to failure of Bladeless Fan

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 7 of 16

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 8 of 16

Q2 (9 Marks)

HEC anticipates that with right go-to market strategy bladeless fans will eventually replace
traditional fans. If this happens then the production line of bladeless fans needs to be ramped
up and the demand curve will become similar to that of traditional fans.

There is one particular spare part that is crucial for both bladeless & traditional fans. There is
only one specific subcontractor who supplies this due to their monopolistic positioning. So
HEC wants to design a futuristic optimum procurement plan for bladeless fans production, to
ensure that there is no shortage of this particular spare part. To enable this, they developed a
demand forecast model with traditional fans data, assuming it will serve as a surrogate for
bladeless fans demand which has not been launched yet and hence there exists no demand
data.

A transfer function model was built using Spare Part as input variable and actual demand as
output (not production no.) variable. This model will help them to assess production capacity
based on the supply available for the spare part and hence assess whether they can meet the
market demand.

In the event that the market demand of bladeless fan is much higher than the production
capacity (due to limited supply of spare parts from a single supplier) an alternative supplier
needs to be sourced if possible (by searching globally) .

Various output steps in the modeling exercise has been provided below:

Fig 1

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 9 of 16

Fig 2

Fig 3

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 10 of 16

Fig 4

Fig 5

a) The daily data provided, covers the period 7/1/2018-31/12/2019. From this information and
from the relevant Figures 1-5 above, propose suitable values for the train-test split and
explain your rationale from the business standpoint. (1 Mark)

b) Using the plot pattern in Fig 5 and the model in Fig 2, guess the possible differencing
parameters and the possible model that was fitted for the input parameter “Spare part” for pre-
whitening purposes. (3 Marks)

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 11 of 16

c) From Fig 1 and referring to the course material (deck 1.3, page 26) what are the possible
values of b & s for building transfer function. From Fig 2 figure out the r value of the
denominator of the transfer function, that was used.

Also, what is wrong with the ACF, PACF graph of the white noise in Fig 1. (2 Marks)

d) With reference to the Figures 2-4, make two comments on the quality of the model.

What is the information that are missing to make a concrete decision, that it is the best model?

What is the experiment you need to do to have a more simplified solution?

(3 Marks)

Q3 (4 Marks)

As HEC wants to be a dominating player in the “Bladeless Fan” market, the management
desires to produce the product with best possible design. So they decided to run a conjoint
analysis.

a) Which type of conjoint do you think will be best suited here and what should be the
characteristics of the responders? (1 Mark)

b) You are the one who is tasked to create the conjoint plan. You have limited knowledge on
the product. Hence you took help of the advertisements of a competitor brand who are currently
the leader. The excerpts are provided in Fig 6 . The ad says these are air purifiers which doubles
as bladeless fans with latest technology:

Fig 6

Design a conjoint plan stating features with levels, so that you can run a successful conjoint
exercise using the method you have recommended in part (a) to select the best design for HEC.
Please note that there is one crucial feature information that is missing in these ads (refer to
conjoint design class exercise). Can you please identify that and include in your design?
(2 Marks)

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 12 of 16

Other than relying on the product descriptions of the competitors who else should you consult
to finalize your plan so that it is realistic for HEC product. (1 Mark)

Q4 (7 Marks)

HEC proceeded to conduct a pre-launch validation experiment to help management make the
“go/no-go” decision and achieve optimum product marketing mix for the new product. The
researchers conducted a 𝟐𝟑 X 2 replicate of the full factorial design of experiments in which all
combinations of the factor levels will be tested, with each combination tested with a random
sample of online consumers to estimate the sales conversion (Sales $). The factors are as
follows:

Response: Sales ($)

Factors Low Level (-1) High Level (+1)
A : Message Easy to clean Energy saving
B : Product Traditional Fan Bladeless Fan
C : Promotion Discount on current purchase Discount on next purchase

a) Interpret the ANOVA table results below and write down the regression model for predicting
online sales conversion based on the results of this experiment. (3 marks)

b) To understand the effect of the Brand, the researchers assigned the two replicates as two
blocks for Brand, i.e. branded (+1) vs unbranded (-1). Based on the regression output below,
is there any brand effect? (1 mark)
MTech EBAC Sample Exam Paper:
Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 13 of 16

c) Based on an analysis of interaction plots and regression output below, provide the factor
levels and business implications to maximize sales. (3 marks)

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 14 of 16

Q5 (4 Marks)

HEC has been doing direct marketing to its customer for the past 5 years. It has regularly sent
out marketing materials to their registered customers through email every 2 weeks. When a
registered customer clicked links in email and visited HEC website, response is recorded in
HEC system. Occasionally, for some customers, discount coupon will be attached in the
email.

The data science team in HEC is requested to build a deep learning model that can predict if a
registered customer is going to respond to the upcoming material and make a purchase.
Propose a model that can achieve the above business objective. Specify the input(s) and the
output(s). Use a diagram to illustrate the solution if necessary.

Q6 (13 Marks)

Two of their popular products are (i) bladeless fan and (ii) traditional fan.

Both bladeless fan and traditional fan require a two-step production process involving wiring
and assembly.

It takes about 2 hours to wire each bladeless fan and 3 hours to wire a traditional fan.

Final assembly of the bladeless fan and traditional fan requires 6 and 5 hours, respectively.

The production capability is such that only 12 hours of wiring time and 30 hours of assembly
time are available.

Each bladeless fan and traditional fan produces $7 and $6 profit respectively.

a) The HEC’s management is interested to find the optimal profit by solving the product-mix
problem. Assuming you are a member of the data science team, you are tasked to conduct the
prescriptive analytics via optimization.

Formulate the optimization model for optimizing profit using integer programming technique.

Prove that linear programming is not appropriate to be used as a substitute for integer
programming. (hint: to use graphical linear programming approach). (6 marks)

b) During strategic planning for the post Covid-19 business strategy, the HEC’s management
is exploring to relocate the manufacturing plant to a new location, which may affect the
production during that period of relocation. Hence, the management agreed that maximizing
profit is not a realistic goal.

Management set a profit level of $31, which would be reasonably acceptable during the
adjustment period.

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 15 of 16

As a data scientist, you propose the goal programming technique to be carried out, which is
meant to find the production mix that achieves this goal as closely as possible, given the
production time constraints.

After discussion with management, the finalized 4 goals are stated as follows:

Goal 1: to produce as much profit above $31 as possible during the relocation period
Goal 2: to use the available assembly plant hours as much as possible
Goal 3: to avoid overtime in the wiring department
Goal 4: to meet a contract requirement to produce at least 8 bladeless fans

It is assumed that the priority of the indicated goals is in sequence from highest priority (Goal
1) to lowest priority (Goal 4).

Formulate the multiple objective problem using Goal Programming approach.

(4 marks)
c) Due to the impact of Covid-19, the HEC’s top management decided to invest in online
sales channel supported by real time customer service helpline.

To arrange for the appropriate capacity from the strategic point of view, the management has
performed a high-level human resource planning for the 24/7 helpline. However, the
members of the operations team are concerned and interested to carry out detailed manpower
planning and optimization.

As a data scientist, you have proposed to adopt Queuing Model approach. This will help to
identify the exact number of customer service staffs to be deployed.

Queuing models are developed as shown in the following figures.

Queuing Model – M/M/c (8/9/1)

Queuing Model – M/M/c (8/9/2)

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting
Master of Technology in Enterprise Business Analytics page 16 of 16

Queuing Model – M/M/c (8/9/3)

Note for M/M/c Queuing Model in R:

- lambda: arrival rate
- mu: server service rate
- c: number of servers
- n: number of customers in the system from which you want to obtain its probabilities. Put
n=0 for an idle probability (no customer present in the system or system idle). With n=-1,
no probabilities are computed

Recommend the best model to be selected as guideline to be carried out in the actual 24/7
online helpline in production.

Justify the recommendation with optimization for optimal resource utilization and quality of
service delivery.
(3 marks)

END

MTech EBAC Sample Exam Paper:

Specialized Predictive Modeling & Forecasting