IPL Win Predictor

The IPL Win Predictor is a machine learning model that predicts the probability of a team winning an Indian Premier League (IPL) match based on current match statistics. This project includes a logistic regression model implemented in Python and an interactive web application built with Streamlit, as well as containerized deployment on AWS ECR and SageMaker.

Table of Contents

• Overview
• Installation
• Deployment
• Usage
• Files
• Model Details
• Streamlit Application
• Results
• Requirements
• License

Overview

The IPL Win Predictor leverages historical match data to train a logistic regression model that provides real-time win probabilities based on the current state of the game. The model can be accessed through a Streamlit web app, allowing users to interactively input match details and receive win predictions.

Installation

To run this project locally, follow these steps:

1. Clone the repository:

   git clone https://github.com/shivangsingh26/IPL-WIN-PREDICTOR.git
   cd IPL-WIN-PREDICTOR

2. Install the required packages:

   pip install -r requirements.txt
   

3. Download the dataset:

   • Place matches.csv and deliveries.csv in the data directory.

   • The dataset can be found here.

4. Run the app locally or use run the docker container from AWS ECR:

   `streamlit run app.py`
   

Deployment

This project is deployed using Docker containers on AWS ECR and AWS SageMaker.

Docker Image Creation and Deployment to AWS ECR

1. Build Docker Image:

   docker build -t registry_ipl_win_pred .
   

2. Push Image to AWS ECR:

   Make sure your ECR repository is set up (e.g., registry_ipl_win_pred), then push your Docker image:

   docker tag registry_ipl_win_pred:latest 992382843941.dkr.ecr.us-east-1.amazonaws.com/registry_ipl_win_pred:latest
   docker push 992382843941.dkr.ecr.us-east-1.amazonaws.com/registry_ipl_win_pred:latest
   

Deploying on AWS SageMaker

1. Create a SageMaker Notebook Instance and open a terminal.

2. Execute Deployment Code in the notebook to pull the Docker image from ECR and deploy it using a SageMaker model and endpoint.

3. Access Endpoint: Once the deployment is successful, the endpoint can be accessed for predictions.

Usage

To use the web application:

1. Select the batting and bowling teams.

2. Select the host city.

3. Enter the target score.

4. Enter the current score, overs completed, and wickets out.

5. Click on the Predict Probability button to get the win probabilities for both teams.

Files

• notebook.ipynb: Jupyter notebook containing the data analysis and model training code.

• app.py: Streamlit app script for the interactive web interface.

• data/matches.csv: Historical match data.

• data/deliveries.csv: Ball-by-ball delivery data.

Model Details

The model training includes:

• Loading and merging match and delivery data.
• Feature engineering to create useful features like current_score, runs_left, balls_left, wickets, etc.
• Using logistic regression to predict the win probability.

The main steps in the notebook.ipynb include:

1. Loading and preprocessing data.
2. Feature engineering.
3. Model training and evaluation.
4. Saving the trained model using pickle.

Streamlit Application

• The app.py script loads the trained model and provides an interactive UI for users to input match details and get win probabilities.
• The application is built using Streamlit, a powerful framework for creating web applications in Python.

Results

• The model predicts win probabilities based on the current state of the match.
• The Streamlit app displays the win probabilities for both the batting and bowling teams.

Requirements

• streamlit

• pandas

• numpy

• scikit-learn

• matplotlib

License

• This project is licensed under the MIT License. See the LICENSE file for more details.