Improving ODI Cricket Predictions: A Novel Approach Using Key Features and Advanced Regression Models

The outcomes of cricket matches, a sport followed passionately by millions worldwide, are influenced by many factors. This study presents a robust machine learning model to predict scores and win probabilities in One Day International (ODI) cricket matches. The model leverages key features such as toss outcomes, venue-specific attributes, team performance metrics, match locations, and detailed match statistics to enhance predictive accuracy. The proposed methodology employs several machine learning algorithms to forecast ODI cricket scores, including XGBRe-gressor, Lasso Regression, Ridge Regression, CatBoost, Gradient Boosting, LGBM, and Linear Regression. To evaluate model performance, established metrics such as the coefficient of determination (R2), mean absolute error (MAE), root mean squared error (RMSE), and mean squared error (MSE) are utilized. This study highlights not only the predictive accuracy and generalizability of the model but also its practical application through integration into a mobile application. This integration enhances user interaction and serves as a valuable tool for cricket enthusiasts and analysts, enabling the identification of trends in ODI matches while demonstrating the practical application of machine learning in sports analytics.