Sport action mining: Dribbling recognition in soccer

Recent advances in Computer Vision and Machine Learning empowered the use of image positional data several high-level analyses Sports Science, such as player action classification, recognition complex human movements, tactical analysis team sports. In context sports analysis, allows new developments opportunities by taking into account players’ positions over time. Exploiting its sequence a systematic way, we proposed framework that bridges association rule mining recognition. The Action Mining (SAM) is grounded on usage for recognising actions, e.g., dribbling. We hypothesise different actions could be modelled using confidence levels computed from previous locations. method takes advantage an algorithm (e.g., FPGrowth) to generate displacement sequences modelling soccer. this context, transactions are traces representing displacements, while itemsets coordinates pitch. experimental results pointed out Random Forest classifier achieved balanced accuracy value 93.3% detecting dribbling which considered events Additionally, provides insights skills player’s roles based small amount data.