Cricket Score Prediction System
An ML-powered cricket match score predictor built with Python, Scikit-learn, and Streamlit. Predicts final innings score with min/max range and upcoming overs bar chart using IPL historical data.
Interactive Showcase Gallery (3 Slides)
Predictive Engine
Project Specifications
Ready to explore this project in action?
Check out the live interactive deployment or inspect the full source code on GitHub.
Technologies Used
About The Project
The Cricket Score Prediction System is a machine learning web application that predicts the realistic final score of a cricket innings in real time. It is powered by a Gradient Boosting Regressor trained on a cleaned IPL historical dataset (ipl.csv).
The model goes beyond simple run-rate projections by engineering smart features that capture match context: current_run_rate (how fast runs are scoring), projected_score (linear extrapolation), pressure_factor (derived from wickets and overs remaining), and is_death_overs (binary flag for overs 17-20 where scoring accelerates). Teams and venues are encoded with One-Hot Encoding; numerical features are Standard Scaled.
The Streamlit app allows users to input current match state (runs, overs, wickets, batting/bowling team, venue) and instantly receive a prediction showing the average score, a minimum and maximum range, and a bar chart of projected scores for each upcoming over. Sample dataset viewing is built in.
The project is fully separated: train_model.py (data prep, feature engineering, training, saving .pkl files) and Cricket-Score-Prediction.py (Streamlit UI, model loading, inference). This separation of training and serving is a production ML best practice. Deployed live on Streamlit Cloud.
Key Features & Architecture Highlights
Gradient Boosting Regressor trained on cleaned IPL historical dataset
Smart feature engineering: run_rate, projected_score, pressure_factor, is_death_overs
One-Hot Encoding for teams/venue + Standard Scaling for numerical features
Predicts final score as average with min/max realistic range
Bar chart of projected scores for each remaining over
Clean separation: train_model.py (training) vs app file (serving) — production ML pattern
Deployed live on Streamlit Cloud with interactive web UI
In-app dataset viewer for transparency and exploration
Interested in building something extraordinary?
Whether you have an ambitious AI project, full-stack application, or collaboration in mind, I'd love to connect.