MAJOR PROJECT REPORT

ON

“Predicting Games Sale

Trends using ML Algorithm”

Subinitted in partial fulfillment of the requirement of

Bachelors of Computer Application (BCA)
Guru Gobind Singh Indraprastha University, Delhi

Session 2022-2025

Under Supervision of: Submitted by:

Dr Ruchi Agarwal HARSHIT
(HOD, BCA Department) SHARMA
[00225502022]

JIMS Engineering Management Technical Campus

(JEMTEC), Greater Noida
(Affiliated to Guru Gobind Singh Indraprastha University)
 SELF CERTIFICATE

This is to certify that the project report entitled “Video game sales prediction” is done by us
is an authentic work carried out for the partial fulfilment of the requirements for the award
of the degree of Bachelor of Computer Applications (2022-25) under the guidance of DR.
Ruchi Aggarwal. The matter embodied in this project work has not been submitted earlierfor
the award of any degree or diploma to the best of our knowledge and belief.

[Harshit Sharma]
[00225502022]
 ACKNOWLEDGEMENT

I would like to express my heartfelt gratitude to several individuals who played a pivotal role in
the successful completion of this project report.

I am deeply thankful to Dr. Ruchi Agarwal, our Project Mentor, for his continuous guidance, expert
knowledge, and constant motivation. His mentorship has been invaluable in steering this project
towards its goals.

Special thanks are due to our project guide, DR. Ruchi Agarwal, for her expert guidance, patience,
and dedication. Her insights and feedback were instrumental in the successful execution of this
project. I am also grateful to all my peers and colleagues who contributed to this project in various
ways.

Lastly, I would like to acknowledge the support and encouragement of my family and friends
throughout this project.

Thank you all for your unwavering support and guidance

[Harshit Sharma]
[00225502022]
 SYNOPIS

1. Name / Title of the Project:

Title: "Predicting Games Sale Trends using Machine Learning Algorithm".

2. Problem Statement:
The video game industry is an ever-evolving, highly competitive market, where developers and
publishers strive to create captivating experiences that stand out among the vast number of
annual releases. As technology advances and player demographics broaden, understanding
market trends, evolving player preferences, and key factors influencing sales has become
essential for game developers, publishers, and stakeholders alike. This project seeks to analyze
global video game sales data to uncover patterns and actionable insights that drive the industry's
success. By identifying trends and utilizing predictive modeling, we aim to equip industry
stakeholders with data-driven strategies to anticipate market demands, optimize game
development and marketing, and ultimately, improve the potential for future success in this
dynamic market.

3. Significance of the Project (Present State of the Art):

In the present state of the art, the video game industry lacks comprehensive, data-driven insights
that consider both historical trends and emerging factors, such as shifts in player preferences,
technological advancements, and evolving monetization models. Current analytics often focus
narrowly on short-term metrics without fully capturing the broader market dynamics or
predicting future trends. This project bridges that gap by leveraging advanced data science
techniques, such as predictive modeling, machine learning, and sentiment analysis, to provide
actionable recommendations for game developers, publishers, and investors. By integrating
both historical and real-time data, the project aims to deliver a holistic view of market trends,
helping stakeholders make more informed decisions that drive innovation and long-term
success.

4. Objective:
The primaryobjective of this project is to analyze historical video game sales data to:
 Develop robust predictive models capable of forecasting sales for upcoming games
across various platforms and regions.
 Identify key factors and trends influencing game performance, including genre,
platform, release timing, and marketing efforts.
 Provide actionable insights for optimizing game development, marketing strategies, and
distribution channels to maximize potential sales and improve return on investment.
5. Scope:
This project will focus on a comprehensive analysis of global video game sales data, covering
the past four decades, to understand long-term trends and shifts in the gaming industry. It will
explore sales performance across a wide range of platforms, from legacy consoles to modern-
day gaming systems, encompassing different hardware generations. The analysis will include
various game genres, highlighting popular and emerging genres, as well as regional preferences
across key markets such as North America, Europe, Asia, and others. The scope extends to the
development of predictive models for sales forecasting, utilizing historical data to provide
insights into future trends. Additionally, the project will assess factors influencing game sales,
including seasonality, marketing efforts, and technological advancements, to provide a holistic
view of the video game market dynamics.

6. H/W & S/W Specification:

Hardware:
 Processor: A Pentium® 4 CPU 2.26 GHz or higher for smooth computational
performance.
 RAM: Minimum 128 MB RAM, though 4 GB or more is recommended for
efficient handling of machine learning tasks.
 Display: A screen resolution of at least 800 x 600, though higher resolutions such
as 1024 x 768 or more are preferable for better visualization.
 Optical Drive: CD ROM Driver for external media support.

Software:
 Programming Environment: Python, with tools like Jupyter Notebooks for
development and libraries like Pandas, NumPy, Scikit-Learn, Matplotlib, and
Seaborn for data analysis, visualization, and machine learning.
 Operating System: Any Windows (95/98/2000/XP/NT and above) or Linux based
system.
 Documentation: Microsoft Word or WordPad for creating reports and
documentation.

7. Data Collection and Methodology:

 Data will be sourced from the Kaggle dataset repository, ensuring a comprehensive and
reliable basis for analysis.
 Initial data processing will include thorough data cleaning and preprocessing to handle
missing values, inconsistencies, and formatting issues, followed by exploratory data
analysis (EDA) to identify trends, patterns, and key variables.
 Advanced machine learning algorithms will be employed for predictive modeling,
leveraging various techniques to forecast video game sales across different categories,
regions, and timeframes.
8. Algorithm:
The project will primarily employ supervised machine learning algorithms, with a focus on
linear regression, to build predictive models for video game sales forecasting. Linear regression
will be used to identify relationships between variables such as genre, platform, and region,
allowing the model to predict future sales based on historical data. This approach is chosen for
its interpretability and effectiveness in capturing linear relationships within the data, making it
suitable for understanding how individual factors impact sales. Additionally, linear regression’s
simplicity enables efficient model training and testing, facilitating quicker insights during the
development phase.

9. Limitations / Constraints of the Project:

 Limitations of this project include potential challenges related to data availability and
quality, as the dataset may have missing or inconsistent records, especially for older
games or niche regions, which could impact the accuracy of analysis.
 Predictions may also be influenced by external factors beyond the scope of the model,
such as economic downturns, unexpected global events, or major technological
advancements, which could significantly shift consumer behavior and impact sales
trends in ways the model may not predict.
 Furthermore, the project's scope is limited to data up to the knowledge cutoff year of
2020, meaning it will not account for trends or developments in the gaming industry
beyond this point, potentially affecting the relevance of long-term forecasts.

10. Conclusion, Future Scope for Modification:

The project will culminate in a comprehensive set of insights into video game sales trends,
covering factors such as platform preferences, genre popularity, and regional sales dynamics.
It will deliver predictive models that can be used to forecast future sales based on historical
data, providing valuable information for stakeholders in the gaming industry. The conclusions
drawn from the analysis will help identify key drivers of video game sales and market shifts
over time. Future enhancements could involve integrating real-time data sources, such as up-
to-date sales figures or recent release data, to make the model adaptive and responsive to
current market conditions. Additionally, incorporating social media sentiment analysis could
refine the models by reflecting consumer sentiment and trends, enhancing predictive
accuracy. Continuous model updates with new data and further exploration of advanced
algorithms like neural networks could also be considered for more complex trend
identification and robust forecasting.
 ABSTRACT
This study will help in predicting the global sales of video games based on the regional sales.

As we know the video game industry is an ever-evolving and a dynamic industry, so sales
figures play a very important role in accessing the popularity of video games. Here we have
used applied linear regression algorithm to predict the global sales.

In this project, we collected and analyzed a vast dataset comprising information on Genre, Platform
and publisher of Game with Region Sales such as North America Salse, Japan Sales, Europe Sales
and global Sales. We used this dataset to train and fine-tune a predictive model capable of
estimating Global Sales with high precision.

Keywords: video game sales prediction, machine learning, performance measures, linear
regression.
TABLE OF CONTENT
CERTIFICATE --

ACKNOWLEDGEMENT --

SYNOPSIS --

ABSTRACT --

LIST OF FIGURES --
LIST OF TABLES --

CHAPTER -1 INTRODUCTION:

1.1 Description of the topic: (Introduction the project's background and context.)

1.2 Problem Statement

1.3 Objectives

1.4 Scope of the Project

1.5 Project planning Activities

1.5.1 Team-Member wise work distribution table

1.5.2 PERT Chart

1.6 Organization of the report

CHAPTER 2 – LITERATURE REVIEW

2.1 Summary of papers studied

2.2 Findings of studied researches.

CHAPTER 3 – SYSTEM DESIGN AND METHODOLOGY:

3.1 System Design

3.1.1 Introduction to System Design

3.1.2 System Architecture

3.1.3 Data Sources

3.1.4 Data Preprocessing

3.1.5 Modelling

3.1.6 Training and Validation

3.1.7 Challenges and Trade-offs

3.1.8 Conclusion

3.2 Algorithm Used

3.2.1 About Random Forest Regressor

CHAPTER 4 – IMPLEMENTATION & RESULT:

4.1 Hardware and Software Requirement:

4.2 Implementation Details

4.2.1 Data Collection

4.2.2 Data Preparation

4.2.3 Exploratory Data Analysis (EDA)

4.2.4 Model Building/ Choosing Machine Learning Model

4.3 Results

CHAPTER 5 – CONCLUSION AND FUTURE WORK

5.1 Conclusion

5.2 Future Scope

REFERENCES
 List of Figures:

Chapters Title Page

No.
Chapter -1 Fig. 1.1 Pert Chart: Describing the work flow of the hole project. 4
Chapter -3 Fig 3.1 Formula for R2 coefficient (R2 score) 15
Fig 3.2 Formula for Mean Squared Error (MSE)
Fig 3.3 Formula for Root Mean Squared Error (RMAE)
16
Fig 3.4 Formula for Mean Absolute Error (MAE)
Fig 3.5 Figure explaining the process of Random Forest Regressor 18
Chapter -4 Fig 4.1 Heatmap showing correlation between Genre and Various 27
Sales.
Fig 4.2 Bar Chart to show what genre games have been made most. 28
Fig 4.3 Pairplot to show relations of columns. 29
Fig 4.4 Visual representation of accuracy score 33
Fig 4.5 Bar graph to show most globally sales games based on 34
genre.
Fig 4.6 Representation of global sales of games based on genre from
1990-2018
Fig 4.7 Games genre sales region wise.
Fig 4.8 Top saled games platform with genre 35
Fig 4.9 Regional Sales by genre.
Fig 4.10 Graphshowing yearly games sales.
Fig 4.11 Graph Showing Global Sales of Nintendo and Sony
Platform Games
Fig 4.12 GlobalSales bygenre
Fig 4.13 RepresentationofGame Systems Lifespan
 List of Tables:

Chapters Title Page

No.
Chapter -1 Table1.1 Responsibility wise work distribution 3
Chapter -2 Table 2.1 Showing the related work of this field. 9-12
Chapter -4 Table 4.1 Showing the Hardware and Software requirement of the 20
system developed
Table 4.2 Comparison of model at different test train size used in 33
the project.
 CHAPTER-1

INTRODUCTION
1.1 Description of the topic:

Video game sales prediction involves forecasting the future performance of a game by analyzing
historical sales data. This task is inherently challenging due to the volatility and rapid evolution
of the gaming industry, with trends shifting based on technological advancements, consumer
preferences, and external factors like global events. However, accurate predictions can provide
significant advantages for stakeholders, such as publishers, developers, and investors, enabling
informed decisions related to game development, resource allocation, marketing strategies, and
pricing models.
Various methods are available for sales prediction, ranging from traditional statistical models to
more advanced machine learning algorithms and deep learning techniques. Statistical methods
may rely on historical trends, while machine learning and deep learning models can uncover
complex patterns in data, adapt to changes over time, and improve accuracy. Machine learning
algorithms, in particular, offer flexibility in handling large, diverse datasets with features such as
game genre, platform, release time, and regional preferences. These methods enable better
forecasts, even in the face of fluctuating market conditions, making them highly effective for the
dynamic nature of the video game industry. In our project, we have specifically chosen machine
learning algorithms due to their ability to handle this complexity and provide reliable predictions.

1.2 Problem Statement:

The goal of this project is to develop a predictive model capable of estimating the global
sales of video games, based on various influencing factors. This model will aim to predict
sales performance across different geographical regions, including but not limited to
North America, Europe, and Japan, based on historical data, genre, platform, release year,
and other game-specific attributes.
Key aspects of this project include building a model that is both accurate and reliable in
its predictions, capable of providing insightful sales forecasts for new games and helping
stakeholders—such as game developers, publishers, and marketers—make data-driven
decisions. The model should be able to generalize effectively, meaning it should not only
fit well on the training data but also perform accurately on new, unseen games.
The project will involve several stages, starting with data preprocessing and exploratory
data analysis to identify patterns and key predictors of sales. Various machine learning
algorithms will be tested and evaluated, potentially including linear regression, decision
trees, and ensemble methods, to find the best-performing model for this task. Finally, the
chosen model will be optimized and validated using appropriate techniques to ensure its
robustness, making it a valuable tool for forecasting sales trends in the gaming industry.
This approach will support better inventory planning, marketing strategies, and ultimately
drive business success in the highly competitive video game market.
1.3 Objectives:

The objectives of this project are to:

 Collect and clean a dataset of video game sales data.

 Develop a model for predicting the global sales of video games.

 Evaluate the performance of the model using appropriate metrics.

 Identify the factors that contribute to the sales of video games.

1.4 Scope of the Project:

In the current state of the video game industry, there is a noticeable lack of
comprehensive data-driven insights that consider both historical trends and the
influence of emerging factors. This project aims to bridge that gap by utilizing
advanced data science techniques to offer valuable, actionable recommendations
for game developers, publishers, and investors. By leveraging machine learning
models, the project provides accurate forecasts and insights that can guide decision-
making in areas such as game development, marketing strategies, and pricing
structures.

 The project will focus on a robust dataset of global video game sales spanning the
last four decades, ensuring a comprehensive analysis of market dynamics. It will
examine sales performance across various platforms, genres, and regions, enabling
a detailed understanding of how different factors contribute to success.

 The scope of the project includes developing and validating predictive models that
not only forecast future sales but also identify key factors such as platform
popularity, genre trends, and regional preferences that influence video game
success. By doing so, the project will provide stakeholders with the tools they need
to optimize their strategies in a highly competitive market.

1.5 Project planning Activities

The following project planning activities were carried out:

 A team was formed and roles were assigned.

 A timeline for the project was developed.

 A communication plan was developed.

1.5.1 Team-Member wise work distribution table

Team consists of three members Harsh Rai, Harshit Sharma, Ayush Joshi and the work is distributed
in the following manner.

Team Member Role/Responsibility Task/Contributions

Harsh Rai Project Data Collection Data Collection and
02825502022 Preprocessing
Harshit Sharma ML model selection Model Development and
00225502022 Training
Ayush Joshi Documentation Preparation of Documentation
02025502022 and Presentation on Canva

Table 1.1 Responsibility wise work distribution

The data collection was done with the agreement of both the team members and found by the
both from Kaggle website.

The preprocessing part was done with the help of both team members.

Key Highlights:
 Focused on developing the predictive model for global sales estimation.
 Utilized Python, scikit-learn for model development and training.
Gungun:
 Created presentation on the project.
 Used Canva graphic designing tool creation of presentation.
1.5.2 PERT Chart

Fig. 1.1 Pert Chart: Describing the work flow of the whole project.
1.6 Organization of the report:
The organization of this report is designed to facilitate clarity and ease of navigation, enabling
readers to access specific sections quickly and understand the logical flow of the analysis. This
structure will help the reader follow the video game sales prediction process from the initial stages
of exploration and data preparation through to the final stages of modeling and evaluation. Each
section serves a distinct purpose, contributing to a comprehensive understanding of the methods
and insights obtained. The report is divided into the following sections:

1. Introduction: This section introduces the purpose of the report, outlining the problem
statement and objectives. It explains why video game sales prediction is relevant and its
potential applications within the gaming industry.
2. Literature Review: A summary of existing work on sales prediction methodologies, especially
within the context of video game markets. This section highlights past research, commonly
used models, and existing gaps that the report aims to address.
3. Data Collection and Exploration: This section describes the data sources, attributes, and
preprocessing steps involved. It includes exploratory data analysis (EDA) to uncover trends,
correlations, and any missing or anomalous values, ensuring the dataset is suitable for
modeling.
4. Feature Engineering: Here, the report details any transformations or new features created to
improve the predictive power of the model. This might include temporal features, genre- based
factors, or platform popularity trends.
5. Model Selection and Development: This section discusses the models chosen for the task
(e.g., linear regression, decision trees, random forests, etc.), the rationale for each, and the
training process. It also includes parameter tuning and cross-validation methods to enhance
model accuracy and generalizability.
6. Evaluation Metrics: Details the performance metrics used to assess model efficacy, such as
Mean Absolute Error (MAE), Mean Squared Error (MSE), and R-squared (R²), as well as an
explanation of why these metrics are suitable for sales prediction.
7. Results and Analysis: This section presents the results of the models, compares their
performance, and interprets the outcomes. It may also include visualizations that demonstrate
model predictions against actual sales figures to highlight accuracy.
8. Conclusion and Future Work: A summary of the key findings, the report concludes by
discussing the implications of the results and offering recommendations for future work, such
as expanding the dataset or testing additional algorithms to refine the prediction model further.

This structured organization ensures that readers can systematically follow the progression of
the project, gaining insights into each phase of video game sales prediction and facilitating the
application of findings in similar contexts.
 CHAPTER 2

LITERATURE REVIEW
 2.1 Summary of papers studied

1. Joe Cox, 2013, “What makes a blockbuster video game? An empirical analysis of US
video game sales”

Joe Cox, 2013: This study investigates the factors that affect the likelihood of a video game
becoming a blockbuster. A unique dataset is collected consisting of approximately 1,800
observations relating to individual video games titles released across a variety of platforms.
Due to the long-tailed nature of the dependent variable, OLS regressions of the logarithm
ofunit sales are estimated alongside binary logistic regressions based on three different
thresholds of sales success.

2. TM Geethanjali, Ranjan D, Swaraj HY, Thejaskumar MV, Chandana HP, 2020,

“Video Games Sales Analysis: A Data Science Approach”

TM Geethanjali, Ranjan D, Swaraj HY, Thejaskumar MV, Chandana HP, 2020: This
paperaims to predict the top-selling video game sales in North America between 1983 and
2016 The dataset is collected from an internet platform known as Kaggle.com. Exploitation
the dataset, the RStudio IDE tool and R-programming language are used for data cleaning,
analysis, and representation. The machine learning algorithm used in this project is linear
regression. They point the results of this Analysis in Associate in nursing intuitive
methodology by visualizing outcome victimization ggplot2 in R.

3. Bodduru Keerthana, Dr. K.Venkata Rao, 2019, “Sales Prediction On Video Games
Using Machine Learning.”

Bodduru Keerthana, Dr. K.Venkata Rao, 2019: In this paper, they briefly analysed the
concept of gaming sales data and sales predictions. The various machine learning
techniques and measures used for this sales prediction. On the basis of a performance
evaluation, best suited predictive models like linear regression, support vector regression,
random forest anddecision trees etc. are used for the sales trend predictions. The results
summarized in performance measures are root mean square error, r-square, and mean
absolute error. The studies found that the best fit model is Random forest algorithm, which
shows maximum accuracy in future sales prediction.

4. Alice Yufa, Jonathan L. Yu, Henry Chan, Paul D. Berger, 2019, “Predicting Global
Video-Game Sales”

Alice Yufa, Jonathan L. Yu, Henry Chan, Paul D. Berger, 2019: In this paper they examine
video games, and consider selected independent variables and explore their relationship
to global sales. Key variables that are identified include the number of critics that rate the
video game, and the average score that the critics give to the game. they also find a non-
intuitive result concerning user ratings of the game.
5. Amar Aziz, Shuhaida Ismail, Muhammad Fakri Othman, Aida Mustapha, 2017,
“Empirical Analysis on Sales of Video Games: A Data Mining Approach”

Amar Aziz, Shuhaida Ismail, Muhammad Fakri Othman, Aida Mustapha, 2017: This
paper studies factors that make the sales of video games becomes a blockbuster. The
dataset used is collected from an online database maintained by VGChartz.com. Using
the dataset, the Rapid Miner tool is used to select the features or factors and produce
efficient estimation of the data. The techniques used in this project included the k-
Nearest Neighbour (k-NN), Random Forest and Decision Tree.

6. John Sacranie, 2010, “Consumer Perceptions & Video Game Sales: A Meeting of the
Minds”

John Sacranie, 2010: This paper examines the determinants of video game software sales.
This paper incorporates several of variables that previous literature have, but also add new
one that is quality. To account for quality, the model incorporates the average review score
agame receives from professional critics. The results indicate that indeed, quality does play
a major role in consumers’ purchase decisions.

7. Jianbin Li, Yufan Zheng, Haoran Hu, Junhui Lu, Choujun Zhan, 2021, "Predicting
Video Game Sales Based on Machine Learning and Hybrid Feature Selection Method"

Jianbin Li, Yufan Zheng, Haoran Hu, Junhui Lu, Choujun Zhan, 2021: In this paper, we
proposed a new hybrid feature selection method Pearson correlation coefficient - Random
Forest Feature Selection(PCC-RFFS), and utilized 9 machine learning methods combined
with PCC-RFFS to predict the sales of video games. For the filter-based stage, we used the
absolute value of Pearson correlation coefficient and feature ranking technique, while the
wrapper-based stage is based on Random Forest to measure the importance between
featuresand target.

8. Luis Alberto Mendieta Caballero, 2015, “The Impact of Online Ratings on Video Game
Sales”

T Luis Alberto Mendieta Caballero, 2015: his study intends to quantify the impact online
ratings have over video game sales by conducting a linear regression analysis on 300 titles
forthe previous console generation (PlayStation® 3 and Xbox® 360) using a data from the
video game industry to understand the existing influence on this particular market.
Afterwards, we compare results to previous ones and discuss the managerial implications for
upcoming gaming generations.
9. Wesley J. Marshall and Andrew L. Weaver, 2012, "The Effect of Video Game Platform on
Game Sales: A Comparative Analysis"

Wesley J. Marshall and Andrew L. Weaver, 2012: This study explores the impact of the platform
on which a video game is released on its overall sales performance. Using data from multiple
platforms, the authors analyze how platform exclusivity, cross-platform availability, and console
type affect the game’s popularity and sales. Through a combination of regression analysis and data
visualization, they find that platform choice plays a significant role in sales, with console-specific
games often seeing different market trajectories compared to those released across multiple
platforms.

10. Marie Müller, 2018, "Analyzing the Impact of Social Media on Video Game Sales"

Marie Müller, 2018: This paper examines the correlation between social media presence and video
game sales performance. By analyzing Twitter mentions, YouTube reviews, and Reddit discussions
for various titles, the study assesses how social media hype and user-generated content impact sales.
The authors use sentiment analysis and time-series forecasting to track trends, finding that games
with high engagement levels on social media platforms tend to experience higher sales. They
conclude that social media has become a significant predictor of a game's commercial success,
especially for independent titles and smaller studios leveraging community support.
2.2 Findings of studied researches

Title of Research Paper Author/Author(s) detail Findings

with year
What makes a blockbuster Author: The dataset used in this study
video game? An empirical represents a snapshot of the
analysis of US video game Joe Cox US video game market during
sales September 2011. The
Department of Economics, dependent variable used in
University of Portsmouth, the empirical analysis is the
Portsmouth, Hampshire, UK lifetime unit sales of games
released in the US after
Year of publish: 2013 September 2010, which were
collected from an online
database maintained by
VGChartz.com

Using OLS regression output

(dependent variable = lnsales)
the R2 Score for General
Model & Preferred Model are
0.450 & 0.443 respectively.

Using Logistic regression

output the Chi-Squared Score
is 312.98.
Video Games Sales Authors: Video Game Sales Dataset
Analysis: A Data Science TM Geethanjali: Assistant was used for the study.
Approach Professor, Department of
Information Science and Using Linear regression
Engineering. output the R2 and Adjusted
Ranjan D: Student R2 Score are 0.05864 &
Swaraj HY: Student 0.03425 respectively.
Thejaskumar MV: Student
Chandana HP: Student

Year of publish: 2020

Sales Prediction On Video Authors: Dataset consists of 11

Games Using Machine Bodduru Keerthana: variables and 500 samples
Learning. M.Tech Scholar with a combination of
Dr. K.Venkata Rao: categorical and numeric
Professor variables. They are Rank,
Name of video game,
Department Of Computer Platform, Year, Genre,
Science & Systems Publisher, North American
Engineering, Andhra Sales, Europe Sales, Japan
University, Visakhapatnam Sales, Other Sales and Global
Sales.
Year of publish: 2019
 Different supervised ML
models were used to analyze
data and the results are as
follows:

1 Random Forest:
RMSE: 1.4648
Accuracy: 0.9605
2 Support vector regression
RMSE: 1.9773
Accuracy: 0.8154
3 Decision Tree
RMSE: 2.8762
Accuracy: 0.8036
4 Linear regression
RMSE: 2.4830
Accuracy: 0.5734

Predicting Global Video- Authors: The dataset was titled Video

Game Sales Alice Yufa Game Sales with Ratings, and
Jonathan L. Yu was released in 2016 and
Henry Chan used data from the review
Paul D. Berger website, Metacritic.

Bentley University Stepwise regression analysis

Waltham, MA 02452 USA is done using SPSS statistical
software, to identify key
Year of publish: 2019 factors that contribute to the
final model.

The R2 value for the final

model was a modest .127
(adjusted R2 = .126),
indicating that our final
model is estimated to explain
about 12.7% of the variability
in Global Sales.
Empirical Analysis on Author: The dataset was collected
Sales of Video Games: A Amar Aziz: from two Portuguese
Data Mining Approach Faculty of Computer secondary schools during
Science and Information 2005-2006 by using two
Technology different methods which are
mark reports and
Shuhaida Ismail: questionnaire
Faculty of Science and
Technology Different supervised ML
models were used to analyze
Muhammad Fakri data and the results are as
Othman: follows:
 Faculty of Computer 1 Naïve Bayes
Science and Information Accuracy: 81.58%
Technology Recall: 43.00%
Precision: 44.12%
Aida Mustapha: 2 Decision Tree
Faculty of Computer Accuracy: 99.55%
Science and Information Recall: 86.61%
Technology Precision: 86.20%
3 K-NN
Universiti Tun Hussein Onn Accuracy: 24.86%
Malaysia Recall: 15.07%
Precision: 13.97%
Year of publish: 2017 4 Random Forest
Accuracy: 26.89%
Recall: 3.45%
Precision: 0.93%

Consumer Perceptions & Author: The data set includes

Video Game Sales: A John Sacranie information on 200 randomly
Meeting of the Minds selected titles released
Illinois Wesleyan University between 2007 and 2009. The
dependent variable will be
Year of publish: 2010 sales over the first ten weeks
of a game’s release.

The regression had an R2 of

.197 and an adjusted R2 of
.122, suggesting that this
model is a moderately
capable predictor of sales.

Predicting Video Game Author: Data is collected from

Sales Based on Machine Jianbin Li: VGChartz.VGChartz is a
Learning and Hybrid School of Electrical and video game sales tracking
Feature Selection Method Computer Engineering website, which was launched
Nanfang College, in 2005 by Brett Walton.
Guangzhou Guangzhou,
China. The dataset with 53 features
into a training set (79.8%)
Yufan Zheng: from February 2006 to
School of Electrical and December 2015 and a testing
Computer Engineering set(20.2%) from January
Nanfang College , 2016 to December 2018.
Guangzhou Guangzhou,
China It is observed that in the
establishment of a model for
Haoran Hu: sales prediction, the use of
School of Electrical and PCCRFFS for feature
Computer Engineering selection and LGB for data
Nanfang College, modeling in Group 2 has the
 Guangzhou Guangzhou, best prediction results, R2 can
China reach 0.585, and RMSE is
81,173.
Junhui Lu:
School of Electrical and
Computer Engineering
Nanfang College,
Guangzhou Guangzhou,
China

Choujun Zhan:
School of Electrical and
Computer Engineering
Nanfang College,
Guangzhou Guangzhou,
China

Year of publish: 2021

The Impact of Online Author: This study was conducted
Ratings on Video Game Luis Alberto Mendieta: with secondary data collected
Sales Caballero, Student Number from two existing and
1895 publicly available datasets.

Year of publish: 2015 a) Sales figures were

extracted from
website
www.vgchartz.com
which includes
statistics from
different regions per
game
b) Online ratings were
gathered from the
website
www.ign.com.
R-squared or the
determination coefficient
which explains how close is
the data to the fitted
regression (a linear model in
this case) resulted in a value
of 13.6%
the R-squared value
explaining that 13.6% of
video game sales movements
can be explained solely on the
rating factor

Table 2.1 Showing the related work of this field.

 CHAPTER-3

SYSTEM DESIGN AND METHODOLOGY

3.1 System Design:

3.1.1 Introduction to System Design

In the highly competitive world of video games, developers, publishers, and industry analysts
rely on accurate and comprehensive data to make informed decisions that can impact a game’s
success. Understanding market trends, optimizing marketing strategies, and predicting sales
performance are crucial steps in launching a game globally. To achieve these goals, an efficient
system design for video game sales analysis must be established, providing a robust framework
for collecting, processing, and analyzing large amounts of data on a global scale.

Purpose and Importance of System Design for Video Game Sales Analysis
The global video game industry generates billions of dollars in revenue each year, with hundreds
of new titles competing for market share. As games are distributed across various platforms,
regions, and formats, having a reliable system for tracking and analyzing sales data is essential
for gaining a competitive edge. The insights drawn from such a system enable companies to
tailor their marketing strategies, identify profitable regions, and forecast future demand based
on current and past sales trends. This data-driven approach not only helps game developers
prioritize resources but also supports publishers in making strategic distribution decisions.

3.1.2 System Architecture:

The high-level architecture of the video games global sales analysis system is designed to
streamline the process of data extraction, preprocessing, visualization, and predictive modeling,
providing valuable insights into video game sales trends.

 Components:
o Data Source: The primary data source is a CSV file containing comprehensive sales
data, including fields such as game title, genre, platform, release year, and sales figures
across various regions. This data serves as the foundation for all subsequent analysis.
 Data Flow:
o Input: The CSV file containing the raw sales data is imported using the Pandas
library.
o Preprocessing: Data cleaning, handling missing values, and preparing the dataset for
modeling.
o Modeling and Analysis: The clean data is split into training and test sets, and then
passed through machine learning algorithms to create predictive models.
o Output: Predictions, along with visual insights, are produced, showcasing key metrics
like genre popularity, sales trends by region, and projected future sales.
 Technology Stack:
o Python Libraries: The project uses a suite of Python libraries. Pandas handles data
import and manipulation, Seaborn and Matplotlib create visualizations, and Scikit-
Learn performs machine learning tasks, including data splitting, model training, and
evaluation. Together, these tools create a seamless pipeline for analyzing and predicting
video game sales trends.
3.1.3 Data Sources:

The data source consists of a dataset CSV file downloaded from Kaggle website. Kaggle is a
data science competition platform and online community of data scientists and machine
learning practitioners under Google LLC. Kaggle enables users to find and publish datasets,
explore and build models in a web-based data science environment, work with other data
scientists and machine learning engineers, and enter competitions to solve data science
challenges.

The reference is given as below.

Link: https://www.kaggle.com/datasets/gregorut/videogamesales

3.1.4 Data Preprocessing:

Data preprocessing is essential for preparing raw data for modeling, ensuring the dataset is
clean, relevant, and free of inconsistencies. Here are the steps involved:
 Data Cleaning: Removed irrelevant data, outliers, and noise to improve data quality.
Cleaned incorrect or inconsistent entries, like spelling errors, and standardized values
where necessary.
 Feature Engineering: Created new features that better capture underlying patterns in
the data. This included aggregating, transforming, or combining existing features to add
predictive power.
 Feature Transformation: Applied techniques like normalization, scaling, and
encoding to convert categorical data into numerical form. Transformation ensures model
compatibility and consistency, especially for algorithms sensitive to data scales.
 Handling Missing, Null, and Duplicate Values: Imputed missing values using
strategies like mean, median, or mode imputation, or advanced techniques if needed.
Removed or corrected null entries and dropped duplicate rows to avoid bias or
redundancy in modeling.
3.1.5 Modelling:

The machine learning models we employed for estimating global Sales was Linear regression.
The main theme of this algorithm is to find a mathematical equation for continuous variables
Y when we have one or more X variables. This algorithm establishes a relation between two
variables one variable is predicted variable and another one is resulting variable whose value
is derived from the predictive variable.

Function: model = lm (formula, data)

Y=aX+b

Where: Y is result variable

X is predicted variable

a and b are coefficients

Since we were using python for machine learning algorithm we used sci-kit learn library for
model selection.

3.1.6 Training and Validation:

The training dataset taken was 85% of the data and 15% was used for testing. We also used
cross validation technique and performance metrics used was R2_score and for error detection
we used MSE, RMSE, MAE.

Where R2_Score is given by:

Fig 3.1 Formula for R2 coefficient (R2 score)

MSE is Mean Squared Error given by:

Fig 3.2 Formula for Mean Squared Error (MSE)

RMSE is Root Mean Squared Error given by:

Fig 3.3 Formula for Root Mean Squared Error (RMAE)

MAE is Mean Absolute Error given by:

Fig 3.4 Formula for Mean Absolute Error (MAE)

3.1.7 Challenges and Trade-offs:

Challenges and trade-offs in data preprocessing included dealing with missing values and outliers.
Our dataset had some null values, which we addressed by filling these with the median of each
respective column, ensuring that the central tendency was preserved without introducing biases from
extreme values. For rows with extensive missing data, we opted to remove them to maintain data
quality. After this, we encountered outliers, particularly in our target column. Outliers can skew
results, so we identified and removed these extreme values to improve the reliability of our analysis.
Each step involved careful consideration, as filling null values and removing outliers can sometimes
lead to loss of valuable information, but ultimately these trade-offs were necessary to prepare a
cleaner dataset for analysis.
3.1.8 Conclusion:
In conclusion, estimating or predicting accurate global sales, which involves numerical data, requires
the application of a reliable regression algorithm. Among the various regression algorithms available,
we selected Simple Linear Regression due to its effectiveness and simplicity in handling regression
tasks. This approach not only models the relationship between variables efficiently but also minimizes
prediction errors significantly, leading to a more accurate and dependable forecast. By implementing
Simple Linear Regression, we achieved a reduction in error margins, thereby enhancing the precision
of our predictions and ultimately delivering the best possible results for salesestimation.
3.2 Algorithm Used
In machine learning, there are numerous algorithms available, each with strengths suited to particular
types of problems, whether they involve classification, regression, clustering, or dimensionality
reduction. The selection of an appropriate algorithm is a critical aspect of any data science project, as
the model's success hinges on its alignment with the nature of the problem and the characteristics of
the dataset.
For our project, we conducted a thorough analysis of the dataset's structure, correlations, and target
variables, ultimately determining that a regression approach would be the most suitable. Among
various regression techniques, we opted for the Linear Regression model due to its simplicity,
interpretability, and efficiency with continuous target values. Using Linear Regression allowed us to
achieve a strong predictive accuracy, yielding a maximum score of 97.1% after fine-tuning parameters
such as test size and random state. This high score demonstrates the model's effectiveness in capturing
relationships within the data, offering valuable insights and reliable predictions.

3.2.5 About Linear Regression:

Linear regression is a type of supervised machine learning algorithm that computes the linear
relationship between a dependent variable and one or more independent features. Here’s how
it works:

3.2.5.1 Data Collection:

The first step in building any predictive model, including linear regression, is to gather a
dataset that contains both the independent variables, also known as features, and the dependent
variable, also referred to as the target. Independent variables are the predictors that provide
information relevant to the prediction, while the dependent variable is the outcome we aim to
predict. The quality and relevance of this data are crucial, as they directly impact the model's
performance. In our case, we carefully collected and prepared a dataset that captures all
necessary aspects to ensure our linear regression model could be accurately trained and
evaluated.

3.2.5.2 Model Representation:

Linear regression operates under the assumption of a linear relationship between the
independent variables and the dependent variable. This relationship suggests that any change
in the independent variable(s) will result in a proportional change in the dependent variable.
The general form of a simple linear regression model, which includes only one independent
variable, can be represented by the following equation:

y=β0+β1x+ϵ

.
3.2.5.3 Model Training:
The goal is to find the best values for a and b that minimize the difference between the
predicted values and the actual target values. This is typically done using a method called
"least squares optimization." The model iteratively adjustsa and b to minimize the sum of the
squared differences between predicted and actual values.

3.2.5.4 Predictions:
Once the model is trained, you can use it to make predictions. Givennew values of the
independent variable(s), you can calculate the corresponding predicted values of the
dependent variable using the linear equation.

3.2.5.5 Model Evaluation:

To assess the quality of the model's predictions, you can use various metrics such as
Mean Squared Error (MSE), Root Mean Squared Error (RMSE), or R-squared (R2),
depending on your specific goals.

Fig 3.5 Figure explaining the Linear regression

 CHAPTER-4

IMPLEMENTATION AND RESUTLS

4.1 Hardware and Software Requirements

Hardware Software
 Pentium® 4 CPU 2.26GHz, 128  Any window-based operating
MB RAM system (Windows
95/98/2000/XP/NT or above).
 Screen resolution of at least 800
x 600 required for proper and  Word pad or Microsoft Word
complete viewing of screens.
Higher resolution would not be a
problem.

 CD ROM Driver.

Table 4.1 Showing the Hardware and Software requirement of the system developed

4.2 Implementation Details:

4.2.1 Data Collection

Data Collection is the systematic process of gathering information from multiple sources, including
surveys, experiments, public databases, web scraping, sensors, and social media. The quality and
relevance of the collected data are crucial for the success of the model, as they significantly impact
its performance and reliability. Here are the steps we followed to collect and prepare our dataset:

 Identify the data needed:

To begin, we identified the data requirements based on our project objectives. This involved
thoroughly understanding the problem we aimed to solve—in this case, estimating global video
game sales—and determining the specific features needed to make accurate predictions. Key
features included regional sales, platform type, genre, and other relevant factors that could
influence global sales.

 Collect data from appropriate sources:

After defining our data needs, we gathered the dataset from reliable sources. In this project, we
chose to use a public dataset of video game sales available on the Kaggle platform. This dataset
provided comprehensive information on various video games, including sales by region, genre,
and platform, all of which were pertinent to our analysis. Using a reputable source like Kaggle
ensured that the data was credible, widely used, and well-maintained.
 Assess data quality:
Following data collection, we assessed the dataset for quality and completeness. This step
involved checking for any errors, inconsistencies, and missing values, as well as ensuring that
each entry met the accuracy requirements for our analysis. Data quality assessments are
essential to avoid any potential issues that could impact model performance.

 Import necessary modules:

To proceed with the data preparation and model development, we began by importing all the
necessary Python libraries and modules. These included libraries such as pandas for data
manipulation, numpy for numerical computations, and scikit-learn for implementing and
evaluating the machine learning model. Each of these tools played a crucial role in processing
the data and building the model efficiently.

Following these steps enabled us to gather, validate, and prepare the dataset, ensuring that we
had a reliable foundation for our linear regression model
Now, we will follow the steps given below:

Following this, we have imported the data into a dataframe named dataset,
Detailed information of data,

Overview of our dataset,

4.2.2 Data Preparation

The process of cleaning, transforming, and enriching data to make it suitable for machine
learning. This may involve cleaning the data, handling missing values, and feature
engineering.

 Cleaning the data: This involves removing errors and inconsistencies from the data.
For example, you may need to remove duplicate entries, correct typos, or convert data
types.
 Handling missing values: This involves deciding how to deal with missing values in
the data. There are a variety of methods for handling missing values, such as imputing
them with the mean or median value, or deleting them from the dataset.
 Feature engineering: This involves creating new features from the existing data. This
can be done to improve the performance of the machine learning model. For example,
you may create a new feature that represents the difference between two existing
features.
In this step we first started with checking the no. of null value in each columns, as follows

Before proceeding to data cleaning process, first we have to create a copy of dataframe to
avoid the loss of original data, as shown in
Now, we found that year column has the most null value so we filled it with its Median, as
shown in
Now, we will remove the N/A value present in any other column (i.e. Publisher’s column),

Checking for any duplicate values,

Removing outliers
Our final dataset,

4.2.3 Exploratory Data Analysis (EDA)

EDA, or Exploratory Data Analysis, is a crucial initial step in the data science process. It
involves the analysis and visualization of data to understand its main characteristics, detect
patterns, spot anomalies, and generate hypotheses. It consists of data visualization, handling
missing data, outlier detection etc.
Heatmap to see the correlation between Genre and Various Sales:

Fig 4.1 Heatmap showing correlation between Genre and Various Sales.
Bar Chart to show what genre games have been made most.

Fig 4.2 Bar Chart to show what genre games have been made most.
pairplot to show relations of columns.

Fig 4.3 Pairplot to show relations of columns.

4.2.4 Model Building/ Choosing Machine Learning Model

Model selection is the process of choosing a machine learning model that is appropriate for
the problem you are trying to solve. There are a variety of different machine learning
algorithms available, each with its own strengths and weaknesses like, Linear regression,
Logistic regression, Decision trees, Random forests, Support vector machines, Neural
networks

When choosing a machine learning model, we considered the following factors:

 The type of problem we are trying to solve (e.g., classification, regression, clustering,
etc.)
 The characteristics of our data (e.g., the number of features, the presence of missing
values, etc.)
 The desired level of accuracy and explainability.
so considering the above factors we chose Linear Regression model

Linear regression is a machine learning algorithm that is used to model and predict continuous
(numerical) target variables. It is one of the simplest and most widely used machine learning
algorithms. Linear regression works by fitting a linear line to the data, where the slope and
intercept of the line represent the relationship between the input features and the target
variable.

Linear regression is a good choice for your project because the target value is numerical.
Linear regression is specifically designed to predict numerical target variables.

Linear Eqaution= y=m1 x1 + m2 x2 + m3 x3 + m4 x4 +b

Our project Linear Eqaution,

Global_sales=m1* NA_Sales + m2 * EU_Sales + m3 *JP_Sales + m4* Other_Sales +b

Here,

Independent variable=Global_sales

Dependent variable= NA_Sales,, EU_Sales, JP_Sales, Other_Sales.

Coeffient= m1, m2, m3, m4

Intercept=b

Training the Model

Model training is the process of fitting a machine learning model to a dataset. This involves
feeding the model the data and allowing it to learn the relationships between the features and
the target variable.
Now, splitting our dataset into training and testing sets.

Dependent values,

Target value,
Checking the error values in the model:

Fitting our training data to MLR model:

Predicting the test set result,

Evaluating the Model

This is important to do to ensure that the model does not overfit the training data.

There are a variety of different model evaluation metrics that you can use, depending on the
type of problem you are trying to solve. For example, if you are trying to solve a classification
problem, you may use metrics such as accuracy, precision, recall, and F1 score.

Checking accuracy score:

4.3 Results:

Visual representation of the model accuracy

Fig 4.4 Visual representation of accuracy score

Result Scores:
Comparison tables of the models that we used in our project:

Table. 4.2 Comparison of model at different test train size used in the project.
Visual representation of all the findings:

Fig 4.5 Bar graph to show most globally sales games based on genre.

Fig 4.6 Representation of global sales of games based on genre from 1990-2018

Fig 4.7 Games genre sales region wise.

Fig 4.8 Top saled games platform with genre.

Fig 4.9 Regional Sales by genre.

Fig 4.10 Graph showing yearly games sales.

Fig 4.11 Graph Showing Global Sales of Nintendo and Sony Platform Games

Fig 4.12 GlobalSales bygenre

Fig 4.13 RepresentationofGame Systems Lifespan
CONNECTION TO FROTEND VIA LOCAL HOST

SNAPSHOT OF THE OUTPUT SCREEN

PREDICTION SCREENSHOT

ENTERING VALUES FOR SALES ANALYSIS

 CHAPTER-5

CONCLUSION AND FUTURE WORK

5.1 Conclusion:
The model developed in this project demonstrates a high level of accuracy in estimating global sales
for video games. Through careful analysis, the model provides valuable insights into predicting
global sales based on regional sales data, allowing for a clearer understanding of the impact that
regional popularity has on a game's overall success.
By experimenting with various test sizes and random state configurations, we were able to maximize
the model's performance, ultimately achieving a remarkable accuracy of 97.2%. This level of
precision, coupled with minimal training errors, highlights the model’s robustness and reliability in
making accurate sales predictions.
Furthermore, the model effectively considers essential aspects such as the genre and platform of the
games, which play a critical role in shaping sales trends. Achieving this objective of accurately
predicting global sales signifies the model’s potential value to stakeholders in the gaming industry.

5.2 Future Work:

In the future, this model has the potential to serve as a powerful tool for forecasting the global sales
of video games by taking into account a range of influential factors. As it leverages data on regional
sales and platforms, it could greatly aid in strategic planning for game releases, enabling developers
and publishers to optimize distribution strategies across different regions. Additionally, this model
could be incorporated into industry service systems, providing gaming companies with a predictive
resource that could inform decisions on game development and marketing, ultimately driving higher
revenue.
By incorporating further refinements and expanding the feature set, including aspects such as
seasonality, genre trends, or even specific in-game features, the model could offer even deeper
insights. The platform type and game genre, identified as significant predictors in this project, will
continue to be focal points in refining future versions of this model, supporting the industry in creating
games that align with market demands and player preferences.
 REFERENCES

 https://www.kaggle.com/datasets/gregorut/videogamesales

 https://pure.port.ac.uk/ws/portalfiles/portal/1090821/COXj_2013_cright_MDE_What
_makes_a_blockbuster_video_game.pdf

 https://ijcrt.org/papers/IJCRT2005182.pdf

 https://www.jetir.org/papers/JETIR1907H50.pdf

 https://www.questjournals.org/jrbm/papers/vol7-issue3/I07036064.pdf

 https://www.researchgate.net/publication/326510277_Empirical_Analysis_on_Sales_
of_Video_Games_A_Data_Mining_Approach

 https://www.researchgate.net/profile/Choujun-
Zhan/publication/360045887_Predicting_Video_Game_Sales_Based_on_Machine_Le
arning_and_Hybrid_Feature_Selection_Method/links/62b83dc11010dc02cc5eb5b6/Pr
edicting-Video-Game-Sales-Based-on-Machine-Learning-and-Hybrid-Feature-
Selection-Method.pdf

 https://digitalcommons.iwu.edu/cgi/viewcontent.cgi?article=1107&context=econ_hon
proj

 https://run.unl.pt/bitstream/10362/15686/1/Mendieta_2015.pdf

 ResearchGate. (2022). Predicting Video Game Sales Based on Machine Learning and
Hybrid Feature Selection Method. Available at:

https://www.researchgate.net/publication/360045887_Predicting_Video_Game_Sales_Base
d_on_Machine_Learning_and_Hybrid_Feature_Selection_Method

 Journal of Applied Data Science and Machine Learning. (2023). Enhancing Video Game Sales
Prediction with Advanced Regression Models. Available at: https://jadsml.org/papers/vol10-
issue2/Enhancing-Video-Game-Sales.pdf