PERFORMANCE COMPARISON OF MACHINE LEARNING METHODS IN TURKISH SUPER LEAGUE MATCH RESULT PREDICTIONS

Abstract

The aim of this study is to determine, examine, interpret and compare the performances of the models formed by the most effective variables in predicting the results of the matches played in the Turkish Super League, using machine learning methods. For this purpose, 743 matches of 23 teams in the Turkish Football Super League were examined using data from the 2018-2021 seasons. The winning and losing situations of the teams were modeled using machine learning methods such as logistic regression, decision trees and random forest. The performances of the models were compared according to sensitivity, specificity, accuracy and F-score criteria. When the machine learning methods and models were compared, it was determined that the best model with 67.4% accuracy was the classification and regression trees (CART) with the variables "pozitive passing percentage of the opponent team", "offensive power of the home team" and "defensive power of the opponent team".