Prediction of Football Match Results in Turkish Super League Games

Abstract

This paper presents a model that predicts the results (Home Win/Draw/Away Win) of the football matches played in the Turkish Super League, by using three machine learning classification methods, which are Support Vector Machines (SVMs), Bagging with REP Tree (BREP), and Random Forest (RF). The dataset used in this study includes the data of 70 features, which are composed of 69 input variables relating to statistical data of home and away teams, and a target variable in 1222 total football games in a 4-year period between 2009 and 2013. In connection, the most effective features of these input variables were determined as a reduced dataset by using some feature selection methods. The results showed that the match outcomes were predicted using the reduced dataset better than using the original dataset and the RF classifier produced the best result.