Dutch football prediction using machine learning classifiers

Sports betting is becoming more popular every year and more people are betting now than ever. With the growth of the betting market comes the growth of research done on match prediction. Research done in the 1950s has been the basis for match predictions up until the 1980s. Since then prediction techniques have shifted from distribution prediction towards a more modern data mining predicting. Using machine learning methods has been proved to be good analyzing methods for tournaments and league matches. In this research paper multiple data mining techniques are analyzed and prediction results are compared to come to a good model for predicting matches of the Dutch football team (soccer team, in American English). Based on the prediction results of a random tree model, Naïve Bayes model and a k-nearest neighbor model one single model is chosen and results are looked at more in-depth. By using the best prediction model, one can see which variables of the dataset have little predictive power and which variables have a lot of predictive power. From the random tree model, which was the most predictive power, it is surprising to see that tactics of the coach hold little predictive value for predicting the end results of a match. To support this claim more research has to be done the prediction of the Dutch football team.