In the world of professional soccer, performance analytics about the skill level of a player and the overall tactics of a match are supportive for the success of a team. These analytics are based on positional data on the one hand and events about the game (e.g. pass, shot on target) on the other hand. The positional data of the ball and players is tracked automatically by cameras or via sensors. However, the events are still captured manually by human, which is time-consuming and error-prone.
In this research project, we introduce a novel approach to detect events in soccer matches by utilizing different machine learning. As input for the machine learning techniques (e.g. neural networks, random forest, support vector machines, k-nearest neighbor, we used several time-dependent features, which were calculated on the basis of the positional data. The evaluation of the results showed that it is possible to recognize soccer events in spatio-temporal data with a high accuracy. Apart of that, we discovered that the size of the used model and the data granularity have a strong influence on the quality of the predicted results.