FUTURE SALES PREDICTION USING MACHINE

LEARNING

PROJECT BASED LEARNING REPORT

Submitted by,

SATYAM KUMAR

AARYAN B.ROY

in partial fulfillment for the award of the degree

of

BACHELOR OF TECHNOLOGY
in

ARTIFICIAL INTELLIGENCE AND DATA SCIENCE

SIKKIM MANIPAL INSTITUTE OF TECHNOLOGY, SIKKIM MANIPAL

UNIVERSITY -737136
NOVEMBER 2024

1
 BONAFIDE CERTIFICATE

 Certified that this project report “FUTURE SALES PREDICTION USING

MACHINE LEARNING” is the bonafide work of “SATYAM KUMAR
(202200196) and AARYAN B. ROY (202200401)” who carried out the project
work under your supervision. Certified further that to the best of my knowledge
the work reported herein does not form part of any other thesis or dissertation on
the basis of which a degree or award was conferred on an earlier occasion on this
or any other candidate.

 SIGNATURE SIGNATURE

 Dr. Swarup Sarkar Dr. Om Prakash

Singh
 SUPERVISOR HEAD OF THE
DEPARTMENT

 Associate Professor Head of Department

 AI&DS Department AI&DS

Department

 Sikkim Manipal Institute of Technology, Sikkim Manipal

Institute of Technology,
 SMU- 737136 SMU- 737136

 Submitted for 5 th Semester project based learning

examination held on 22 November, 2024.
2
 PROJECT REVIEW CERTIFICATE

This is to certify that the work recorded in this project report entitled “FUTURE SALES
PREDICTION USING MACHINE LEARNING” has been carried out by SATYAM
KUMAR(202200196) and AARYAN B.ROY(202200401) of Artificial Intelligence and Data Science
Department of Sikkim Manipal Institute of Technology in partial fulfilment of the requirements for the
award of Bachelor of Technology in Artificial Intelligence and Data Science. This report has been duly
reviewed by the undersigned and recommended for final submission for Project Based Learning.

Date: 22/11/2024 Supervisor Signature

Place: SMIT, Sikkim
Dr. Swarup Sarkar
Associate Professor,
Department of AI&DS,
Sikkim Manipal Institute of Technology,
Majitar, Sikkim – 737136

3
CERTIFICATE OF ACCEPTANCE
This is to certify that the below mentioned students of Artificial Intelligence and Data
Science Department of Sikkim Manipal Institute of Technology (SMIT) have worked
under the supervision of Dr. Swarup Sarkar, Associate Professor, Department of
Artificial Intelligence and Data Science from August 2023 to November 2024 on the
project entitled “FUTURE SALES PREDICTION USING MACHINE
LEARNING”.

The project is hereby accepted by the Department of Artificial Intelligence and Data
Science, SMIT in partial fulfilment of the requirements for the award of Bachelor of
Technology in Artificial Intelligence and Data Science.

Registration No Name of Student Project Venue

202200196 SATYAM KUMAR SMIT

202200401 AARYAN B.ROY SMIT

Signature of HOD
Dr. Om Prakash Singh
Associate Professor & Head,
Department of Artificial Intelligence and Data Science,
Sikkim Manipal Institute of Technology,
Majitar, Sikkim – 737136

ACKNOWLEDGMENT
4
We extend our heartfelt gratitude to Dr. Swarup Sarkar sir, our esteemed guide, for

providing invaluable guidance and unwavering supervision throughout the course. His

expertise and mentorship significantly shaped the outcome of our work.

Additionally, we would like to express our sincere thanks to Dr. Om Prakash Singh,

Head of the Department of Artificial Intelligence and Data Science at Sikkim Manipal

Institute of Technology. His encouragement and support allowed us to embark on this

project and ensured its successful completion.

We are equally appreciative of Mr. Sital Sharma, our Project based learning coordinator,

and faculty member Dr. Swarup Sarkar. Their constant elevation, inspiration, and

diligent supervision played a pivotal role in our project’s achievement.

Lastly, we extend our thanks to all other staff members of the Artificial Intelligence and

Data Science department at Sikkim Manipal Institute of Technology. Their continuous

support and guidance were instrumental in bringing our project to fruition.

SATYAM KUMAR (202200196)

AARYAN B.ROY (202200401)

5
 TABLE OF CONTENTS

SI.no. Contents Page no.

1 ABSTRACT 7
2 MOTIVATION 8
3 INTRODUCTION 9
4 LITERATURE SUREVY 10-11
5 OBJECTIVE 12

6 PROBLEM DEFINITION 13
7 SOLUTION STRATEGY 14
8 METHODOLOGY 15
9 IMPLEMENTATION 16-21
10 RESULT 22
11 CONCLUSION 23
12 REFERENCE 24

6
 ABSTRACT

By leveraging historical sales data, demographic trends, marketing efforts, and economic indicators, machine
learning models can capture complex patterns and provide more accurate forecasts than traditional methods. We
explore different machine learning algorithms such as linear regression, decision trees, support vector machines,
and deep learning approaches like neural networks, evaluating their performance on sales prediction tasks. The
study highlights the importance of In today's data-driven business environment, accurate sales forecasting is
crucial for optimizing inventory, managing resources, and making informed strategic decisions. This paper
explores the application of machine learning techniques to predict future sales for businesses across various
industries. feature engineering, model selection, and validation in improving predictive accuracy. Additionally,
we discuss the challenges of handling seasonality, outliers, and missing data, and propose solutions to address
these issues. The results demonstrate the potential of machine learning in enhancing sales forecasting accuracy,
leading to more efficient resource allocation and better decision-making in business operations. Future research
may focus on real-time prediction systems, incorporating external factors like weather, social media sentiment,
and global market trends for even more precise sales forecasting.

MOTIVATION
7
1. *Informed Decision-Making*: Machine learning models can provide more accurate forecasts than
traditional methods. With predictive insights, businesses can make data-driven decisions, reducing the
uncertainty and risks involved in sales planning.

2. *Optimization of Resources*: By predicting future sales trends, businesses can better allocate
resources (such as inventory, marketing budget, and staffing) to areas where they are most needed,
leading to higher efficiency and lower costs.

3. *Increased Profitability*: Accurate sales predictions enable businesses to optimize pricing,

promotions, and product offerings. By understanding customer demand and trends, companies can
focus on high-margin products or services and identify areas for improvement.

4. *Competitive Advantage*: Implementing machine learning for sales forecasting can give businesses
a competitive edge. It allows them to anticipate changes in the market or consumer behavior, staying
ahead of competitors who may still rely on manual methods.

8
 INTRODUCTION

In an increasingly competitive business landscape, the ability to predict future sales accurately has become a
crucial factor for success. Sales forecasting directly impacts key aspects of business operations, including
inventory management, production planning, marketing strategy, and financial forecasting. Traditionally,
businesses have relied on historical sales data and basic statistical methods such as moving averages or linear
regression to forecast future demand. However, these approaches often fall short in handling complex, non-
linear relationships within the data or accounting for external factors that influence sales.

With the rise of big data and advancements in machine learning (ML), businesses now have the opportunity to
improve their sales forecasting accuracy significantly. Machine learning algorithms can analyze vast amounts of
historical sales data, uncover hidden patterns, and make predictions with greater precision. By incorporating a
variety of factors, such as seasonal trends, promotional activities, economic conditions, and even social media
sentiment, machine learning models can generate more dynamic and adaptable forecasts compared to traditional
techniques.

This report aims to explore the application of machine learning methods in predicting future sales, discussing
various algorithms and approaches for building predictive models. It also addresses key challenges in the sales
forecasting process, such as handling missing data, seasonality, and outliers. The report provides a
comprehensive analysis of the effectiveness of different machine learning models and evaluates their
performance in forecasting sales. Ultimately, the goal is to demonstrate how machine learning can help
businesses optimize their operations, reduce risks, and make data-driven decisions for better growth and
profitability.

In the following sections, we will review the current landscape of sales prediction, explore relevant machine
learning models, and examine their application to real-world sales data. The report will also discuss the
implications of using machine learning for sales forecasting, highlighting both the opportunities and limitations
of this approach.

LITERATURE SURVEY

9
 Author,
Sl. Publisher, vol.
No. year Title of Paper Inference Research Gap Relevance
01 Mr. Soham FUTURE The paper discusses the • There is a • Big Data in
Patangia, Ms. SALES application of big data lack of Retail:
Rachana PREDICTION methods in retail, personnel Examines
Mohite, Mr. USING specifically focusing on with the the use of
Kevin Shah, MACHINE pricing, availability, necessary big data for
Mr. Gaurav LEARNING assortment, and layout skills to improving
Kolhe, Mrs. design. It highlights effectivel retail
Madhura challenges such as the y analyze logistics,
Mokashi, shortage of skilled and including
Mrs. Prajakta personnel, issues with IT leverage customer
Rokade integration, and the big data insights and
ability of the supply for operational
chain to respond operation planning.
promptly to data. al Challenges:
planning. Identifies obstacles
. such as skill
shortages, IT
integration issues,
and supply chain
responsiveness
02 Prajwal SALES • The paper discusses the There is a need • Business
Amrutkar, PREDICTION integration of decision for an Intelligent Dependenc
Shubhangi USING analysis with predictions Decision e on Sales
Mahadik MACHINE to enhance sales forecast Analytical Trends:
LEARNING accuracy using data System that can The
mining techniques. integrate decision research
• It evaluates various analysis with emphasizes
machine learning accurate sales the
algorithms like Gradient predictions importance
Boost Algorithm, of accurate
Decision Tree, and sales
Generalized Linear forecasting
Model for sales for
forecasting. businesses,
. as it
significantl
y impacts
their
operations
and
strategic
planning.
• Data
Mining for
Forecastin
g: It
highlights
the
effectivenes
10
 s of data
mining
techniques
in
extracting
valuable
insights
from large
datasets,
which is
crucial for
improving
sales
predictions.
.

OBJECTIVE
11
To Explore the Role of Machine Learning in Sales Prediction:
Investigate how machine learning techniques can be leveraged to improve the accuracy of sales
forecasting by analyzing historical sales data and identifying patterns that may not be captured
by traditional forecasting methods.To Compare Different Machine Learning Models for Sales
Forecasting:
Evaluate and compare the performance of various machine learning algorithms, such as linear
regression, decision trees, random forests, support vector machines, and neural networks, in
predicting future sales across different business scenarios
ToAddressKeyChallengesinSalesForecasting:
Identify and analyze challenges in sales prediction, such as seasonality, missing data, outliers,
and external factors, and propose solutions or techniques for overcoming these challenges when
applying machine learning models.
To Assess the Impact of Feature Engineering on Model Performance:
Examine the importance of data preprocessing, including feature selection and transformation,
in improving the predictive accuracy of machine learning models.
To Provide Practical Recommendations for Implementing Machine Learning in Sales
Forecasting:
Offer actionable insights for businesses on how to integrate machine learning approaches into
their sales forecasting processes, focusing on best practices, tools, and techniques for effective
implementation.
To Discuss Future Directions for Sales Prediction Models:
Explore emerging trends and research areas in sales forecasting, such as real-time prediction
systems, integration of external data (e.g., social media sentiment, economic indicators), and the
potential of advanced deep learning models for further enhancing sales prediction accuracy.

12
PROBLEM DEFINITION

1. *Lack Flexibility*: Traditional methods may fail to account for sudden shifts in
market conditions, customer preferences, or external factors such as economic
downturns or global events (e.g., pandemics).

2. *Inaccurate Forecasts*: The reliance on static historical data or basic

mathematical models can lead to poor predictions, causing businesses to
overstock or understock inventory, misallocate resources, or miss out on key sales
opportunities.

3. *Time-Consuming and Resource-Intensive*: Gathering and analyzing data

manually is not only labor-intensive but also prone to human error. As data
complexity increases, so does the difficulty of creating reliable forecasts.

4. *Limited Scope*: Traditional forecasting methods might overlook the deeper

patterns in data that machine learning can uncover, such as correlations between
product categories, customer demographics, or online behavior.

SOLUTON STRATEGY
13
Automating Data Collection and Preparation: Collect the sales data and external factors
into a centralized database and clean the data for accuracy purposes while highlighting
the critical variables. Data Analysis and Modeling: Perform EDA to identify patterns;
train machine learning models, such as random forest, regression, and evaluate the
performance by using metrics like MAE and RMSE. Deploy a Automatically Update
Forecasting: Deploy trained models generating sales forecasts and establish automated
processes which are periodically upgraded. . Link to Inventory Management: Use the
forecasts to adjust stock levels, minimize stockouts/ overstock, and set alerts for
potential inventory problems. Monitor and Improve Performance: Continuously
evaluate and retrain the model with new data, refining strategies based on the outcome
of inventory to improve forecasting and business results.

METHODOLOGY
14
1. 1.Define the Objective: - Clearly outline the goals of the sales prediction
model, including specific metrics to evaluate (e.q., accuracy).

2. Data Collection: - Gather relevant historical sales data, including features

such as date, product type, pricing, promotions, seasonality, and external
factors (e.g., economic indicators).

3. Data Preprocessing: Cleaning: Remove duplicates, handle missing values,

and correct any inaccuracies in the dataset. - Feature Engineering: Create
new features that may improve model performance, such as lagged sales,
moving averages, or promotional flags.

4. 4.Exploratory Data Analysis (EDA): Analyze the data to uncover patterns,

trends, and correlations. Use visualizations to understand distributions and
relationships among variables.

5. Split the Data: Divide the dataset into training and testing subsets (e.g., 80%
for training, 20% for testing) to evaluate model performance effectively.

6. Select Machine Learning Models: - Choose suitable models for the prediction
task (e.g. Random forest regression, Decision Tree regressor, KNN
regressor).

7. Train the Model: - Fit the selected models to the training data, adjusting
hyperparameters as necessary using techniques like cross-validation.

8. Evaluate Model Performance: Assess the models on the test set using metrics
such as Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), or
R-squared. Compare the performance of different models to
select the best one

15
IMPLEMENTATION

16
17
18
19
20
21
RESULTS

MODEL MAPE RMPSE

Random forest regeressor 14.15% 21.33%
Decision tree regresor 14.39% 21.33%
K nearest neighobour 12.62% 20.55%
Linear regresion 10.31% 15.77%

So, after analyzing all the model we have consider that the linear regression model is
performing well with less error

22
CONCLUSION

In conclusion, as businesses face increasing pressure in a competitive market, the ability to accurately
predict future sales has become an essential tool for driving operational success. Traditional methods
of sales forecasting, while useful, often fall short in capturing the complexities and non-linear
relationships within the data, as well as accounting for external influences. However, the advent of
machine learning presents a transformative opportunity to overcome these limitations.
Machine learning algorithms have demonstrated their ability to analyze vast quantities of data, uncover
hidden patterns, and generate more accurate and adaptable sales forecasts by incorporating a broader
range of variables—such as seasonality, economic conditions, and social media sentiment. These
models not only offer enhanced predictive accuracy but also enable businesses to make more informed,
data-driven decisions.

REFERENCES
23
1. Trnka., "Market Basket Analysis with Data Mining Methods", International Conference on
Networking and Information Technology (ICNIT),2010..

2. W Yanthy, T. Sekiya, K. Yamaguchi, "Mining Interesting Rules by association and

Classification Algorithms", FCST 09.

3. Chiu, K.S.Y., Luk, R.W.P, Chan, K.C.C., and Chung, K.F.L, "Marketbasket Analysis with
Principal Component Analysis: An Exploration", IEEE International Conference on Systems,
Man and Cybernetics, Vol.3, 2002.

4. Huang, Q., & Zhou, F. (2017, March). Research on retailer data clustering algorithm based on
spark. In AIP Conference Proceedings (Vol. 1820, No. 1, p. 080022). AIP Publishing.

5. Saylı, A., Ozturk, I., & Ustunel, M. (2020). Brand loyalty analysis system using KMeans
algorithm. Journal. of Engineering Technology and Applied Sciences, 1(3).

6. Maingi, M. N. A Survey on the Clustering Algorithms Sales Data Mining. For dataset:
https://www.kaqgle.com/datasets/tevecsystems/retail-sales-forecasting

24