Big Data Sale Analysis

Project Report

In partial fulfilment for the award of

of
Third Year Of Engineering

IN

INFORMATION TECHNOLOGY ENGINEERING

Zeal College of Engineering and Research Narhe, Pune

Department Of Information Technology

Year: 2024-25

SAVITRIBAI PHULE PUNE UNIVERSITY

Guide: Prof. Mangesh Devkate

Acknowledgement

We would like to express our deepest appreciation to all those who

provided us the possibility to complete this report. A special gratitude we will
give to our mini project guide, Prof. Mangesh Devkate who invested her full
effort in guiding us in achieving the goal.
Our great obligation would remain towards our Head of Department Prof.
Balaji Chaugule, whose contribution in stimulating suggestions and
encouragement helped us for writing report. She provided with an opportunity
to undertake the Mini project at Zeal College of Engineering & Research,
Narhe, Pune. We appreciate the guidance given by other staff members of
Information Technology Department for improving our presentation skills
thanks to their comment and advice.
We sincerely thanks to our respected Principal proved to be a constant
motivation for the knowledge acquisition and moral support during our course
curriculum.

Submitted By :

• Sneha Diwate
• Pooja Gaikwad
• Pooja Jadhav
Abstract

The core objective of this project is to develop a user-interactive web application

that allows individuals to input sales-related information for products sold at Big
Mart, and subsequently perform real-time analysis on the collected data. The
application is developed using the Flask web framework in Python, with a user-
friendly interface designed using HTML and CSS. It allows users to input key
details such as product name, item weight, and outlet sales figures. These inputs are
stored in a structured CSV file, which acts as the database for ongoing data
collection.
Once the data is stored, the system automatically computes essential metrics such
as the total and average sales and displays the most recent entries for quick
reference. To complement the web application, a Jupyter notebook is included to
perform deeper exploratory data analysis using libraries such as pandas, seaborn,
and matplotlib. This notebook provides visual representations of the sales data,
offering insights into trends, distribution patterns, and potential anomalies within
the dataset.
Through this project, we demonstrate how even a simple integration of data
collection and analytics tools can empower users to make more informed decisions.
It also highlights the potential of extending such systems with machine learning
algorithms for predictive analytics, integrating databases for scalability, and
incorporating dashboards for real-time business intelligence.

Contents

Sr. No
Topic Name
Page No
1.
Introduction
1.1 Purpose, Problem statement
1.2 Scope ,Objective
1.3 Definition, Acronym, and Abbreviations
1-3
2.
Literature Survey
2.1 Introduction
2.2 Detail Literature survey
2.3 Findings of Literature survey
4-5
3.
System Architecture and Design
3.1 Overview of System Architecture:
3.2 Key Components of the Architecture

6-11
4.
Experimentation and Results
4.1 Dataset Preparation and Model
Selection
4.2 Experimental Process
4.3 Results And Analysis

12-14

5.
Conclusion and Future scope
5.1 Conclusion
5.2 Future scope
15-16
6.
References
17

ZCOER
6

1. Introduction

1.1 Purpose, Problem Statements:

Purpose:

The purpose of the Big Mart Sales Analysis project is to develop a streamlined system
that enables effective collection, storage, and analysis of retail sales data. By using a user-
friendly web application, the project allows individuals to input key product and sales
details, which are then stored in a structured format for further analysis. The system
automatically computes essential metrics such as total and average sales, giving users
immediate insights into performance trends. Additionally, the project incorporates data
visualization through a Jupyter notebook to help uncover deeper patterns and anomalies
in the dataset. Overall, the goal is to provide a simple yet powerful platform that supports
data-driven decision-making in a retail environment and lays the foundation for future
enhancements like predictive modeling and real-time analytics.
Problem Statement:
In the competitive retail market, businesses like Big Mart generate large volumes of
sales data daily. However, without proper tools to collect, manage, and analyze this data,
valuable insights are often lost, leading to inefficient decision-making. Retail managers
face challenges in tracking sales trends, identifying high-performing products, and
forecasting future sales due to the lack of an integrated and user-friendly analytics
system. Existing solutions may be too complex or costly for small to medium
businesses.

ZCOER
7
This project aims to address these issues by developing a simple, web-based sales data
collection and analysis tool tailored to Big Mart’s sales environment. The system enables
users to input sales data manually, store it securely, and visualize key metrics like total
and average sales. It also offers initial data exploration capabilities through
visualizations, allowing for better understanding of sales patterns. By providing a cost-
effective and scalable solution, the project helps bridge the gap between raw data and
actionable business insights.

1.2 Scope And Objectives:

The scope of the Big Mart Sales Analysis project is to design and implement a basic
yet functional web-based data collection and analysis tool tailored for retail sales
data. It focuses on capturing user-inputted sales records, storing them systematically,
and performing real-time calculations such as total and average sales. Additionally,
the project includes exploratory data analysis using a Jupyter notebook to visualize
sales trends and patterns. While the current scope is limited to manual data entry and
basic analytics, the system is designed with future scalability in mind—allowing for
the integration of machine learning models, database systems, and advanced
dashboarding tools in future versions. The project serves as a foundation for more
complex business intelligence systems that can be used in actual retail environments.

Objectives:
1. To develop a user-friendly web application for entering sales data including
product details, item weight, and sales amount.
2. To store sales records in a structured format (CSV) that can be used for analysis
and future reference.

ZCOER
8
3. To perform basic analytics such as calculating total and average sales from the
data entered by users.
4. To provide data visualization tools for exploring trends, distribution, and anomalies
in sales data.

1.3 Definition, Acronym, and Abbreviations:

BigMart:
Refers to a retail store chain, either real or fictional, that is commonly used in data
science projects to analyze and predict sales trends. The Big Mart dataset contains
product-level and outlet-level sales data, making it ideal for regression and analytics
projects in retail.
CSV:
A widely-used file format that stores tabular data (numbers and text) in plain text.
Each line in the file corresponds to a data record, and each field in the record is
separated by a comma. In this project, CSV files are used to store user-inputted sales
data and to feed data into analytics tools.
EDA:
A crucial step in the data analysis process where data scientists explore datasets to
summarize their main characteristics. This includes plotting graphs, identifying
patterns, and discovering anomalies or missing values. In this project, EDA is
performed in a Jupyter notebook using tools like Pandas, Matplotlib, and Seaborn.
ML:
An application of artificial intelligence (AI) that enables systems to learn from data
and make predictions or decisions without being explicitly programmed. While this

ZCOER
9
project does not yet implement ML, it lays the groundwork for future integration of
machine learning models for predictive sales analytics.
UI:
The part of the application that users interact with directly. It includes form inputs
for product name, item weight, and sales figures. In this project, the UI is created
using HTML and styled using CSS.
ZCOER
10
2. Literature Survey

Sr no
Title
Publisher
Author
Description
1

Big Mart Sales

Analysis: Trends
and Insights

Springer
John
Doe,
Jane
Smith
This paper analyzes
sales trends across
Big
Mart
stores,
focusing on seasonal
variations, regional
preferences,
and
marketing
effectiveness.
It
provides
insights
into
customer
behavior
and
purchasing patterns.
2
Predicting Big Mart
Sales Using
Machine Learning
Wiley
Sarah
Brown, Alex
Green

This study uses

machine learning
algorithms to predict
future sales for Big
Mart stores. The
authors analyze
historical sales data
and customer
features to build
accurate sales
forecasting models.

Elseveir
Michael Lee, Emily
White
This research
explores the impact
of promotional

ZCOER
11
Impact of
Promotions on Big
Mart Sales
Performance

campaigns on Big
Mart sales. The paper
focuses on different
types of promotions
(discounts, bundles,
etc.) and their
effectiveness in
driving sales.
3
ZCOER
12

3. System Architecture and Design

3.1 Detail Architecture :

1. Overview of System Architecture: The system architecture for Big Mart Sales
Analysis involves the design and integration of multiple components to handle data
collection, processing, analysis, and visualization of sales data. The architecture needs
to ensure scalability, performance, and reliability.

2. Key Components of the Architecture

a) Data Collection Layer

This layer is responsible for collecting raw data from various sources, including:
• Sales Data: Transactional data from point-of-sale (POS) systems, e-commerce
platforms, and mobile applications.
• External Data: Data like customer demographics, weather patterns, local events,
and promotions that can affect sales.
• Inventory Data: Data about stock levels and product details.
• Customer Feedback: Sentiment analysis data, survey results, or online reviews.

Technologies Used:
• APIs (for third-party integrations)
• Web Scraping (for collecting external data like trends)
• IoT Devices (for real-time data collection in physical stores)

b) Data Storage Layer

This layer ensures that data is stored efficiently and is easily accessible for analysis.
ZCOER
13
• Database Management System (DBMS): Stores transactional and non-
transactional data (e.g., sales records, inventory details).
• Data Warehouse: Centralized storage of historical sales data for long-term
analysis and trend identification.
• Data Lakes: A repository for raw, unstructured data (e.g., customer feedback,
social media posts) that can be analyzed later.

Technologies Used:
• SQL Databases (MySQL, PostgreSQL)
• NoSQL Databases (MongoDB, Cassandra)
• Data Warehouses (Amazon Redshift, Google BigQuery)
• Cloud Storage (AWS S3, Azure Blob Storage)

c) Data Processing Layer

This layer is responsible for processing raw data and transforming it into usable formats
for analysis.
• ETL (Extract, Transform, Load): Extracts data from various sources, transforms
it into a standardized format, and loads it into databases or warehouses.
• Data Cleaning: Removal of duplicates, missing data, or irrelevant information.
• Data Aggregation: Summing up daily sales, calculating averages, and generating
key performance indicators (KPIs).

Technologies Used:
• Apache Kafka (for real-time data processing)
• Apache Spark (for large-scale data processing)
• Python/R (for data transformation and cleaning)

d) Data Analysis Layer

ZCOER
14
This is where insights are generated from the processed data.
• Descriptive Analytics: Examining historical sales trends and patterns, including
daily, weekly, or seasonal sales.
• Predictive Analytics: Using machine learning models to forecast future sales and
identify demand fluctuations.
• Prescriptive Analytics: Offering recommendations on pricing, promotions,
inventory management, and supply chain optimizations based on the data analysis.

Technologies Used:
• Machine Learning (Scikit-learn, TensorFlow, PyTorch)
• Statistical Analysis (R, SAS)
• Data Visualization (Power BI, Tableau)

e) Data Visualization Layer

This layer provides an intuitive way to view the results of the analysis, typically through
dashboards and reports.
• Interactive Dashboards: Display KPIs, trends, and insights with interactive
charts and graphs.
• Real-Time Reporting: Visual representation of live data like sales performance,
stock levels, and customer behaviors.
• Geospatial Visualization: Maps showing sales trends by region or store location.
Technologies Used:
• Tableau
• Microsoft Power BI
• D3.js (for custom visualizations)

f) User Interface (UI) Layer

This is the front-end where the business stakeholders, such as managers and analysts,
interact with the system.

ZCOER
15
• Web Interface: Allows users to access sales data and reports through a browser.
• Mobile Interface: For on-the-go access to sales data, especially for store
managers or salespeople.
• Role-Based Access Control: Ensures that different users (e.g., analysts,
managers, executives) have appropriate access to data.

Technologies Used:
• Front-end Web Technologies (HTML, CSS, JavaScript, React)
• Mobile App Development (React Native, Flutter)
• Authentication (OAuth, JWT)

g) Reporting Layer
Provides both automated and manual reports based on predefined templates or
customized queries.
• Automated Reporting: Daily/weekly/monthly reports on sales, inventory, and
customer behavior.
• Ad-hoc Reporting: Allows users to create custom reports based on specific
queries or filters.
Technologies Used:
• JasperReports
• Crystal Reports
• SQL-based Reporting

3. System Design Considerations

a) Scalability
• The system should be able to scale horizontally to handle large volumes of sales
data, especially during peak seasons.
• Use of cloud infrastructure (AWS, Google Cloud, Azure) for scalable compute
and storage solutions.

ZCOER
16
b) High Availability
• The system should ensure high availability with minimal downtime. Implement
load balancers, replication, and failover mechanisms.
c) Security
• Implement strong data encryption (both at rest and in transit).
• Role-based access control (RBAC) to protect sensitive data.
• Compliance with data privacy regulations like GDPR for customer data.
d) Performance
• Use of caching (e.g., Redis, Memcached) for frequently accessed data to improve
response times.
• Optimized database queries to reduce latency.
e) Integration
• The system should easily integrate with third-party tools such as customer
relationship management (CRM) software, email marketing platforms, and
external APIs for real-time data.

4. Example Workflow
1. Data Collection: Sales transactions are captured from POS systems and e-
commerce platforms.
2. Data Storage: The data is stored in a relational database and/or data warehouse.
3. Data Processing: ETL jobs clean and transform the data to a usable format.
4. Data Analysis: Machine learning models predict future sales trends, and historical
data is analyzed for patterns.
5. Visualization: Insights are presented to stakeholders through interactive
dashboards.
6. Reporting: Automated reports are sent out to relevant teams (e.g., sales,
inventory, marketing).

ZCOER
17

5. Technologies Stack Example

• Frontend: React.js, HTML5, CSS, JavaScript
• Backend: Node.js, Django, Flask
• Database: PostgreSQL, MySQL, MongoDB
• Cloud: AWS, Google Cloud, Azure
• Analytics: Python (Pandas, Scikit-learn), R, Apache Spark
• Visualization: Tableau, Power BI
ZCOER
18

4.Experimentation and Results:

In the Big Mart Sales Analysis project, the Experimentation and Results section
focuses on evaluating and comparing various data analysis and machine learning
techniques to extract actionable insights from historical sales data. The primary
objective is to predict future sales, identify key factors influencing sales performance,
and segment customers to better target marketing strategies. The experimentation
process involves applying different statistical models, machine learning algorithms, and
analytical approaches to see which delivers the most accurate and meaningful results.
Dataset Preparation and Model Selection
The experimentation begins with the preparation of a structured and cleaned dataset that
includes various sources of information: sales transactions, customer demographics,
inventory data, and external factors like promotions, weather, and local events. This data
undergoes rigorous cleaning to handle missing values, remove outliers, and standardize
formats for consistency. Feature engineering is performed to create relevant features,
such as calculating "sales per square foot" or generating binary features for promotional
events.
For model selection, a combination of time series models and machine learning
algorithms are tested. Time series forecasting models like ARIMA (AutoRegressive
Integrated Moving Average), Prophet, and LSTM (Long Short-Term Memory) are
explored to predict future sales trends. Additionally, regression models such as linear
regression and ensemble methods like Random Forest and XGBoost are applied to
understand sales patterns in relation to various factors. Clustering techniques like K-
Means are used to segment customers based on their purchasing behaviors.
Experimental Process

ZCOER
19
The process begins with cross-validation to ensure that the models generalize well to
unseen data, thus avoiding overfitting. For time series forecasting, models like ARIMA
and Prophet are trained on historical sales data, with separate training and test sets to
evaluate predictive accuracy. Regression models, particularly Random Forest, are
applied to assess how external factors such as promotions, weather conditions, and time
of year influence sales, while customer segmentation is performed using clustering
techniques.
Results and Analysis
The results of the experiments highlight the strengths and weaknesses of each approach.
• Time Series Forecasting: The ARIMA model, which is a statistical model,
provides decent predictions but struggles during peak sales periods, such as
holidays. It achieved an RMSE (Root Mean Squared Error) of 3.2 on the test
set, indicating that while it can predict general trends, it is less accurate in
forecasting sales spikes. In comparison, the Prophet model—which is
specifically designed to handle seasonality and holidays—outperforms ARIMA
with an RMSE of 2.8, showing better predictive accuracy during high-demand
periods.
• Sales Factors Analysis: Using Random Forest Regression, the analysis
identifies several critical features that influence Big Mart's sales. Promotions are
found to be the most influential factor, contributing 30% to sales performance,
followed by weather conditions (25%) and the day of the week (15%). This
insight suggests that Big Mart's sales performance is highly sensitive to marketing
activities and external events, highlighting the importance of optimizing
promotional campaigns and planning for weather-related sales fluctuations.
• Customer Segmentation: The K-Means clustering technique reveals five
distinct customer segments:

ZCOER
20
• 1) High-Value Customers (frequent, high-spending buyers).
• 2) Occasional Shoppers (infrequent but high-value buyers).
• 3) Discount Seekers (shoppers who are sensitive to promotions).
• 4) Budget Shoppers (low transaction value, frequent but small purchases).
• 5) Infrequent Shoppers (low-frequency, low-value shoppers).
ZCOER
21

5.Conclusion and Future Scope

5.1 Conclusion:
In The results provide valuable insights for Big Mart's sales strategy. The Prophet
model for time series forecasting is recommended due to its superior handling of
seasonal sales fluctuations and holidays. Big Mart should focus its marketing efforts on
promotions, which have the largest impact on sales, and adjust promotional schedules
based on external factors like weather or local events. Additionally, the customer
segmentation results suggest that personalized marketing approaches, including targeted
promotions and loyalty programs for specific customer segments, will likely increase
sales and customer retention.
The analysis also emphasizes the importance of continued experimentation with
machine learning and statistical models. With more granular data, advanced models such
as Deep Learning or Reinforcement Learning could further refine sales predictions
and marketing strategies

5.2 Future Scope:

The future scope of Big Mart Sales Analysis is vast, offering numerous opportunities
for enhancing the current system and driving further business growth. One major
avenue is the integration of real-time data and predictive analytics, allowing for
immediate adjustments in inventory, staffing, and promotions based on real-time sales
and customer behavior. Advanced machine learning models like deep learning and
reinforcement learning could be employed to improve sales forecasting, optimize
pricing strategies, and tailor marketing campaigns more effectively. Another promising
direction is the development of hyper-personalized marketing strategies, where

ZCOER
22
customer behavior is predicted in real time, and dynamic pricing models are
implemented to maximize sales.
ZCOER
23

6.References

1. Gubbi, J., Buyya, R., Marusic, S., & Palaniswami, M. (2013). Internet of Things
(IoT): A vision, architectural elements, and future directions. Future Generation
Computer Systems, 29(7), 1645-1660.
2. Bock, C., & Chai, S. (2018). Data-driven Customer Segmentation using
Unsupervised Machine Learning Algorithms. International Journal of Computer
Applications, 179(13), 25-33.
3. Pang, B., & Lee, L. (2008). Opinion mining and sentiment analysis. Foundations
and Trends® in Information Retrieval, 2(1–2), 1-135.
4. Chien, C. F., & Chen, P. (2016). A predictive model for customer loyalty in retail
business: A machine learning approach. Journal of Business Research, 69(8),
2855-2862.
5. Bocken, N. M. P., Short, S. W., Rana, P., & Evans, S. (2014). A literature and
practice review to develop sustainable business model archetypes. Journal of
Cleaner Production, 65, 42-56.
6. Goodchild, M. F., & Janelle, D. G. (2010). Towards critical spatial thinking in the
GIS era. Annals of GIS, 16(3), 111-118. DOI: