Predicting Soccer Match Results in the English Premier League

In this report, we predict the results of soccer matches in the English Premier League (EPL) using artificial intelligence and machine learning algorithms. From historical data we created a feature set that includes gameday data and current team performance (form). Using our feature data we created five different classifiers: Linear from stochastic gradient descent, Naive Bayes, Hidden Markov Model, Support Vector Machine (SVM), and Random Forest. Our prediction is in one of three classes for each game: win, draw, or loss. Our best error rates were with our Linear classifier (.48), Random Forest (.50), and SVM (.50). Our error analysis focused on improving hyperparameters and class imbalance, which we approached using grid searches and ROC curve analysis respectively.