The Football Matches Outcome Prediction for English Premier League (EPL): A Comparative Analysis of Multi-class Models

Football match outcome prediction has evolved into a dynamic field of research, driven by the integration of machine learning models to achieve precise forecasts. This paper embarks on a comprehensive exploration, presenting a comparative analysis of four prominent multiclass machine learning (ML) models-Multiclass Logistic Regression, Multiclass Neural Network, Multiclass Decision Forest, and Multiclass Decision Jungle. Furthermore, it explores the application of Azure Machine Learning for predicting English Premier League (EPL) match results in terms of win, draw, and lose, underscoring the league’s significance in the global football landscape through multiclassification ML models. The paper aims to contribute valuable insights to the field, emphasizing the importance of ML model selection for enhanced predictive performance. As the result, Multiclass Logistic Regression emerges as the most accurate model with a 57.08% success rate, followed closely by the multiclass decision forest at 56.71%. The multiclass neural network exhibits a respectable accuracy of 55.59%, while the multiclass decision jungle performs similarly with an identical accuracy of 55.59%. These findings show critical role of a choosing suitable and well-performing ML model in achieving accurate football match outcome predictions.