A Machine Learning Model for Analysis and Prediction of Football Match Outcomes in the English Premier League

Abstract

Football stands as the world's most popular sport, captivating billions globally. The English Premier League, in particular, is widely regarded as the pinnacle of professional football, boasting immense global viewership and attracting widespread interest. Its dynamic and unpredictable nature fuels a massive industry built around match analysis, reflecting the deep desire to anticipate match outcomes. Early attempts at football match prediction often relied on static historical data, assumed independence among events, failed to adapt quickly to football's rapid evolution, and lacked the capacity to capture complex nonlinear interactions among multiple features. This study develops a machine learning model for football match analysis in the English Premier League to predict match outcomes, addressing gaps in previous models by using ensemble machine learning algorithms to provide timely, accurate, and real-time analysis. The study utilised Random Forest (RF), XGBoost, and LightGBM. Performance evaluation using standard classification metrics, including Accuracy, Precision, Recall, F1-Score, and ROC-AUC, showed that Random Forest achieved the best overall performance, with an accuracy of 87.14% and an ROC-AUC of 99.00%. The ensemble model further enhanced prediction consistency by combining the strengths of the three machine learning models. This study demonstrates the effectiveness of machine learning for match predictions and, from an industry perspective, offers practical recommendations for football to enhance retention, efficiency, and competitiveness.