Comparison of Ensemble Machine Learning Algorithms for Padel Shots Classification and Skill Level

Currently, the incorporation of technology for enhancement in the sports domain is a reality. Furthermore, the utilization of artificial intelligence for problem solving is on the rise. In this study, we merged these two aspects, leveraging artificial intelligence to enhance the performance of padel tennis players. Thus, in this research, we employ machine learning algorithms, specifically ensemble learning algorithms, to categorize various padel tennis shots and player skill levels. The data used correspond to samples collected by a gyroscope and an accelerometer on the player’s arm. For the classification, we utilize diverse algorithms including Multi-Layer Perceptron, Decision Tree, K-Nearest Neighbors, Support Vector Machine, voting classifier, Bagging, Random Forest, AdaBoost, and Gradient Boosting. According to the results achieved, the best ensemble classifier in both classification problems is the voting classifier, that obtains an accuracy higher than 92% in padel shots classification and 96% in skill levels. Nevertheless, other algorithms achieves similar results employing less training an prediction time, as extremely randomized trees or pasting classifiers.