Applying data mining techniques to football data from European championships

Abstract

Data Mining is the process of finding new, potentially useful and non trivial knowledge from data. Football is a popular game worldwide and a rich source of data. Gathering only part of this data we are able to collect hundreds of cases. In this paper we describe an exploratory work where we use Data Association Rules, Classification and Visualization techniques to find patterns in datasets from several European championships. For each one of these techniques, different hypotheses were stated. For Association Rules and Visualization, our hypothesis was that we would be able to find non trivial knowledge and confirm several known patterns. For Classification, our hypothesis was that we would be able to classify matches according to their results based on the available history. Our findings didnt confirm our hypotheses to their full extent as expected. Our exploratory work confirmed several well known patterns in football and highlighted borderline cases. Among the several techniques used, visualization produced the best results.