IPL Score Prediction using Deep Learning

Interactive deep learning project to predict IPL innings scores from match context such as teams, venue, batsman, bowler, runs, wickets, and overs.

Quick Navigation

• Project Overview
• Live Notebook Actions
• Project Highlights
• Tech Stack
• How It Works
• Run Locally
• Interactive Prediction Widget
• Model Workflow
• Future Improvements

Project Overview

This project builds a deep learning regression model that predicts final IPL scores during an innings. It includes:

• Data loading and preprocessing
• Label encoding of categorical cricket features
• Feature scaling with MinMaxScaler
• Neural network training with Keras
• Interactive score prediction using ipywidgets

Live Notebook Actions

• Open notebook locally: IPL.ipynb
• Open in GitHub: View on GitHub
• Open in Colab: Run in Colab

Project Highlights

Click to expand key features

• End-to-end notebook workflow from EDA to prediction
• Categorical encoding for venue, teams, batsman, and bowler
• Deep neural network with Huber loss for robust regression
• Interactive widget interface for real-time score predictions

Tech Stack

Area	Tools
Language	Python
Data	pandas, numpy
Visualization	matplotlib, seaborn
ML / DL	scikit-learn, keras
Notebook UI	ipywidgets

How It Works

1. Load IPL historical ball-by-ball style data.
2. Explore trends (venues, batsmen, bowlers, wickets).
3. Encode categorical inputs using LabelEncoder.
4. Select features and split into train/test sets.
5. Scale features using MinMaxScaler.
6. Train a Keras regression model.
7. Predict innings total through interactive widgets.

Run Locally

1. Clone Repository

git clone https://github.com/itsluckysharma01/IPL-Score-Prediction-using-Deep-Learning.git
cd IPL-Score-Prediction-using-Deep-Learning

2. Install Dependencies

pip install pandas numpy matplotlib seaborn scikit-learn keras ipywidgets notebook

3. Launch Notebook

jupyter notebook IPL.ipynb

Interactive Prediction Widget

The notebook includes dropdowns and numeric inputs for:

• Venue
• Batting team
• Bowling team
• Striker
• Bowler
• Runs, wickets, overs, and striker indicator

Then click Predict Score to get a predicted total runs value.

Model Workflow

Expand training setup

• Input features: teams, venue, batsman, bowler, wickets, overs, striker, runs
• Target: innings total score
• Architecture: Dense(512, relu) -> Dense(216, relu) -> Dense(1, linear)
• Loss function: Huber loss
• Optimizer: Adam

Future Improvements

• Add model comparison (XGBoost, Random Forest, LSTM)
• Add evaluation metrics dashboard (MAE, RMSE, R2)
• Deploy as a web app with Streamlit or FastAPI
• Add CI checks and reproducible environment files

Author

Lucky Sharma