Predicting Football Match Results: An Analysis of Feature Selection and Machine Learning Techniques Using a Curated Dataset

The global appeal of football is fueled by its exhilarating dynamic and unpredictable nature. Football match prediction is still a challenging issue due to its growing interest. The accurate prediction of football match outcomes has been a topic of significant interest for various stakeholders, including coaches, analysts, and fans. This study explored the use of dimensionality reduction methods and different machine learning models to predict the results of football matches. A historical dataset was created from La Liga matches spanning four seasons, incorporating both technical and tactical actions for each match and both teams, comprising 297 features per match. Most important attributes are chosen from the feature set using various feature selection techniques including Pearson Correlation, Spearman's Rank Correlation, ANOVA F Test, and Random Forest. For prediction, a variety of classification techniques including Logistic Regression, K-Nearest Neighbors (KNN), Support Vector Machines (SVM), Naïve Bayes, Decision Tree, and Random Forest are used. The combination of the ANOVA F Test feature selection technique and the Random Forest classification method outperformed other models achieving an accuracy of 64.78% with 22 selected features when applied to an independent test set. This paper advances the field of match result prediction research by demonstrating how several machine learning classifiers and dimensionality reduction approaches can accurately predict the outcome of a football match. The suggested methodology can be expanded to other sports or high-dimensional data domains, and it has potential applications in sports analytics and team management.