Predicting the outcomes of football matches using machine learning

Abstract

" Football is the most watched and played sport globally. Every year billions of dollars are risked to bet on match outcomes. These conclude in either home wins, draw or away wins. Football is regulated by FIFA organization, under which several continental and international leagues exist. Countries in each continent have a domestic league or two or more in case of England. Which this research will be focusing on, specifically English Premier League which is the most popular domestic league globally. Prediction of football matches have been attempted in Turkish League as well as FIFA world cups and the EPL. Existing research has focused on the accuracy by using techniques such as training on video game data or reducing the number of considered features. This research project provides a prediction model where attendance of the game is considered to make predictions alongside the outcome of the past 3 matches for each team. It also uses a novel, untested approach by using ensemble methods in the application & testing of the Voting Classifier. "