Predicting EURO Games Using an Ensemble Technique Involving Genetic Algorithms and Machine Learning

This paper summarizes our first attempt to use an ensemble technique to predict soccer games. We chose to represent the games by converting the difference of the teams' statistics into the values -1,0, and 1. This conversion was done using tolerance values that "zero-out" some of the differences. We used a genetic algorithm to find the tolerance values. 101 tolerance vectors produced by the genetic algorithm were used to create game representations on which various machine learning algorithms were trained to create an ensemble technique. Our approach had an accuracy of 70% predicting the Women's Euro 2022 and an accuracy of 67% predicting the men's Euro 2020.