The IPL Score Predictor is an AI-powered system designed to predict the total score of a team in an IPL match. It analyzes match details such as venue, batting/bowling teams, striker, and bowler. Built using deep learning, this model leverages a neural network for accurate predictions and provides a user-friendly interface for real-time forecasting.
- Cleaned historical IPL match data.
- Handled missing values and encoded categorical variables.
- Scaled numerical features using
MinMaxScaler.
- Designed a Neural Network with multiple dense layers.
- Used ReLU activation, Huber Loss, and Adam Optimizer.
- Tuned hyperparameters for optimal performance.
- Built a live prediction interface using
ipywidgets. - Users can input match details (venue, teams, players) and get real-time score predictions.
- Mean Absolute Error (MAE): 11.71
- Mean Squared Error (MSE): 330.16
- R² Score: 60.8%
- Data Exploration: Analyzed IPL datasets for feature engineering.
- Preprocessing: Label encoding, scaling, and train-test splitting.
- Training: Developed and fine-tuned the neural network.
- Deployment: Integrated the model into an interactive tool.
- Deep Learning: Dense layers with dropout for regularization.
- Loss Function: Huber Loss to handle outliers.
- Optimization: Adam for efficient convergence.
| Category | Technologies Used |
|---|---|
| Programming | Python |
| Libraries | Pandas, NumPy, Scikit-learn, TensorFlow/Keras |
| UI | ipywidgets |
| Platform | Google Colab, VS Code |
- Accuracy Improvement: Achieved a balance between bias and variance through hyperparameter tuning.
- Real-Time Prediction: Seamlessly connected the trained model to an interactive UI for dynamic inputs.
- 🌍 Multi-League Support: Extend predictions to PSL, BBL, etc.
- 🌦️ Context-Aware Features: Incorporate weather/pitch conditions.
- 📱 Responsive UI: Develop a web/mobile app for broader accessibility.
| Role | Contributions |
|---|---|
| M Attaullah (Lead) | Designed NN, led project, evaluated metrics. |
| M Haris Nisar | Data collection, preprocessing, and EDA. |
| Abdul Rehman | Model training, tuning, and testing. |
| M Muqaddas Ali | Developed interactive prediction interface. |
Passionate about AI, cricket, and problem-solving? Let’s collaborate!
