Report on Mini Project

Subject: Machine Learning

AY: 2024-25

IPL Score Prediction using deep learning

Anirudh Acharya : 2103001

Manav Purswani : 2103143

Mehek Richharia : 2103148

Priya Sunariwal : 2103172

Guided By

(Prof. Darakhshan Khan)

Contents

Abstract ii

Acknowledgments iii

List of Abbreviations iv

List of Figures v

1 Introduction 1
1.1 Introduction
1.2 Motivation
1.3 Problem Statement & Objectives
1.4 Organization of the Report

2 Literature Survey 11

2.1 Survey of Existing System

2.2 Limitation Existing system or research gap
2.3 Mini Project Contribution

3 Proposed System 18

3.1 Introduction
3.2 Architecture/ Framework
3.3 Algorithm and Process Design
3.4 Details of Hardware & Software
3.4 Experiment and Results
3.5 Conclusion and Future work.

References 32
 Acknowledgement

We would like to express our gratitude and thanks to Prof. Darakhshan Khan
for her valuable guidance and help. We are indebted for her guidance and constant
supervision as well as for providing necessary information regarding the project. We
would like to express our greatest appreciation to our principal Dr. G.T. Thampi and
head of the department Dr. Tanuja Sarode for their encouragement and tremendous
support. We take this opportunity to express our gratitude to the people who have been
instrumental in the successful completion of the project.

Anirudh Acharya

Manav Purswani

Mehek Richharia

Priya Sunariwal
 Chapter 1

Introduction
This chapter explains the aim, objectives and scope of the proposed system.

1.1 Introduction
In the fast-paced world of the Indian Premier League (IPL), accurate predictions of match
outcomes are crucial for teams, analysts, and fans. This project leverages Deep Learning
to forecast IPL scores by analyzing historical data, player statistics, and real-time match
conditions. By building a neural network model using TensorFlow and Keras, we aim to
deliver precise score predictions that can enhance strategic decisions during live matches.
This project showcases the power of Deep Learning in transforming cricket analytics,
providing valuable insights and elevating the understanding of game dynamics.

1.2 Motivation
The motivation behind this project stems from the growing role of data analytics in sports,
particularly cricket. The Indian Premier League (IPL), with its fast-paced matches and
dynamic variables, offers a unique challenge for predictive modeling. Traditional
methods struggle to capture the intricate patterns in player performance, match
conditions, and team strategies. This project aims to overcome these limitations by
leveraging Deep Learning, which can process vast amounts of complex data and uncover
hidden insights. By accurately predicting IPL scores, this project not only enhances
strategic planning for teams but also enriches the fan experience, showcasing how
advanced technology can revolutionize the way we understand and engage with sports.

1.3 Problem Statement & Objectives

This project focuses on developing a Deep Learning model to predict IPL match scores
by analyzing historical data, player statistics, and match conditions, thereby improving
accuracy in score forecasting.

1. Build a Deep Learning model using TensorFlow and Keras to predict IPL scores
based on historical and real-time match data.

2. Preprocess the data by encoding categorical features and scaling to ensure optimal
model performance.

3. Train and evaluate the model using appropriate metrics to measure accuracy, such
 as mean absolute error (MAE).

4. Create an interactive tool that allows users to input match conditions and receive
real-time score predictions.

1.4 Organization of the Report

This report consists of three chapters. The first chapter deals with introduction of the
topic, problem statement, motivation behind the topic and objectives. The second chapter
is the Literature Survey. It includes all the research work done related to this topic. All
information related to study of existing systems as well as learning of new tools is
mentioned in this chapter. The third chapter is about the proposed system which is used
in this project. The block diagram, techniques used, hardware and software used
screenshots of the project are presented in this chapter. All the documents related to
development of this project are mentioned in References

Chapter 2

Literature Survey
This chapter explains the concepts used in this project, study of existing system and
contribution of this project

2.1 Survey of Existing System

Traditional cricket score prediction models, like linear regression and statistical
averages, often fall short in accounting for the complexities of the game, such as
fluctuating player performance, varying pitch conditions, and evolving team
strategies. While modern machine learning approaches—such as Random Forests,
Support Vector Machines, and Gradient Boosting—have enhanced predictive
accuracy, they still struggle to manage large volumes of real-time data and capture
intricate relationships between multiple variables. These systems, though an
improvement, lack the precision required for dynamic and fast-paced
environments like the IPL. To address these limitations, this project utilizes Deep
Learning, offering improved prediction capabilities by effectively analyzing both
historical and live match data.
2.2 Limitation of existing system
Existing cricket score prediction models exhibit several significant limitations that hinder
their effectiveness. A major drawback is their inadequate handling of complexity;
traditional approaches, such as linear regression, struggle to capture the non-linear
relationships and intricate interactions among various match-related variables, limiting
their predictive accuracy. Additionally, many systems rely heavily on static historical
datasets and fail to incorporate real-time match conditions and player form, which can
significantly influence performance.

Another limitation is their limited adaptability. While machine learning algorithms like
Random Forests and Support Vector Machines have shown promise, they often struggle
to keep pace with the fast-changing dynamics of player performance and strategies.
Moreover, existing systems may face scalability issues as the volume of available data
increases, impacting their ability to process large datasets efficiently. Lastly, many
current models fall short in their feature engineering efforts, missing crucial insights that
could enhance the understanding of factors impacting match outcomes. These limitations
highlight the need for more advanced approaches, such as Deep Learning, to improve the
accuracy and reliability of IPL score predictions.

2.3 Mini Project Contribution

Write about your own work. If you have attempted to solve any of the limitations
mentioned in 2.2 then write about that or else you can write about how your system is
different from the existing ones (difference can be in terms of technology used or any
additional features included etc)
 Chapter 3

Proposed System

3.1 Introduction

This project utilizes a combination of programming languages, tools, and

technologies to develop a robust model for predicting IPL scores using Deep
Learning. The primary programming language employed is Python, chosen for its
extensive libraries and frameworks that facilitate data analysis and machine
learning.

For data manipulation and analysis, libraries such as Pandas and NumPy are used
to efficiently handle and preprocess the dataset. Visualization is accomplished
using Matplotlib and Seaborn, allowing for insightful explorations of the data and
results.

The Deep Learning model is built using TensorFlow and Keras, which provide
powerful capabilities for designing, training, and evaluating neural networks. This
combination enables the creation of complex architectures that can learn from vast
amounts of historical and real-time data, enhancing predictive accuracy.

Additionally, tools like Jupyter Notebook and Google Colab are utilized for
interactive coding, making it easier to visualize results and share the project with
collaborators. The entire setup fosters an efficient environment for
experimentation and iteration, ultimately leading to improved score prediction for
IPL matches.

3.2 Architecture

Fig 3.2.1 IPL score prediction Architecture

3.3 Algorithm and Process Design
This detailed step-by-step methodology encapsulates the systematic approach taken in
developing the IPL score prediction project, emphasizing the integration of data
processing, machine learning, and user interaction to deliver predictive analytics.
1. Dataset Acquisition: The dataset was sourced from Kaggle, containing IPL match
data from 2008 to 2017. It included features such as match date, venue, batting
and bowling teams, individual player performances, and the total score.
2. Data Loading: The dataset was loaded into a Pandas DataFrame using the
pd.read_csv() function. This allowed for easy access and manipulation of the data.
3. Data Cleaning: The data was cleaned to address missing values and duplicates:
Used isnull() and dropna() methods to handle missing values. Applied
drop_duplicates() to remove any duplicate entries.
4. Feature Selection: Features were selected based on their relevance to score
prediction. Unimportant columns were dropped to streamline the dataset.
5. Data Preprocessing: Categorical variables were transformed into numerical
formats using LabelEncoder from Scikit-learn. The dataset was split into
independent variables (X) and the dependent variable (y), with y representing the
total score. The data was divided into training and testing sets using the
train_test_split() function, with a 70:30 ratio. Min-Max scaling was applied to
normalize the features, ensuring all inputs were within the same range for
improved model performance.
6. Model Design: A neural network architecture was defined using TensorFlow and
Keras: The input layer was set to match the number of features. Two hidden layers
were included, the first with 512 neurons and the second with 216 neurons, both
using ReLU activation functions. The output layer consisted of a single neuron
with a linear activation function for regression.
7. Model Compilation: The model was compiled using the Adam optimizer for
efficient training and Huber loss as the loss function, which is robust against
outliers.
8. Model Training: The model was trained using the .fit() method: Specified the
number of epochs (e.g., 100) and batch size (e.g., 32). A validation split was used
to monitor performance and prevent overfitting. Training and validation loss were
recorded for analysis.
9. Model Evaluation: After training, the model was evaluated using the Mean
Absolute Error (MAE) and Mean Squared Error (MSE) metrics. Predictions were
 made on the test set, and these predictions were compared against actual values to
assess accuracy.
10. Interactive Widget Development: An interactive widget was created using
ipywidgets to allow users to input match conditions: Dropdown menus were
created for users to select the venue, batting team, bowling team, striker, and
bowler. A function was defined to encode and scale user inputs, matching the
format used during model training. The model was then used to predict the score
based on user inputs, and the predicted score was displayed.
11. Final Implementation: The final setup enabled real-time score prediction for
specific match scenarios, demonstrating the model's capabilities in providing
insights into IPL match outcomes.

3.4 Details of Hardware & Software

System Configuration:
The project was developed and executed on a system with the following specifications:

• Processor: Intel Core i7-10700K (8 cores, 3.8 GHz base clock speed)

• RAM: 16 GB DDR4

• Graphics Card: NVIDIA GeForce GTX 1660 (for GPU acceleration during model
training, if applicable)

• Operating System: Windows 10 (64-bit)

3.5 Results

Fig 3.5.1 Selecting input for performing process

Fig 3.5.2 Processing of input using ML algorithm

3.6 Conclusion and Future Work
In conclusion, the application of deep learning in IPL score prediction represents a
transformative approach to cricket analytics. By harnessing the power of advanced
algorithms and historical data, teams and analysts can forecast match outcomes with
greater accuracy than ever before. This not only enhances strategic decision-making
during live matches but also enriches the fan experience by providing real-time insights
and predictions. As technology continues to evolve, the future of cricket analytics
promises to be increasingly data-driven, offering new opportunities to unravel the
complexities of the game and elevate its competitive edge.