USE OF THE FIRST AND SECOND HALVES RESULTS TO CLASSIFY THE FINAL OUTCOME OF ENGLISH PREMIER LEAGUE MATCHES

English premier league (EPL) is one of the top leagues in Europe and any analysis of data generated from the league is highly sought after by fans, betters, coach, managers and scouters. The paper applied four machine learning models in classifying the outcome of five seasons using the results of the first and the second halves. Each half and the final outcome were made up of just three data points, namely, home win (HW), draw (DR) and away win (AW). Home win is the most frequent followed by AW and DR in descending order. There is no significant relationship between the results of the two halves. On the other hand, there are significant relationships between the first half and the outcome and also, between the second half and the outcome. Random forests (RF), gradient boosting (GB) and adaptive boosting (AB) yielded better results than the logistic regression (LR). Generally, the accuracy averaged over 90 percent with few misclassifications. Implementation of the research in a decision support system is highly recommended