Development of A System To Detect Patterns In Sports Activities Using Wearables and Machine Learning Algorithms

In recent years, wearable technology has become an essential part of sports performance enhancement, providing real-time information about biomechanical and physiological data to optimize and improve the athletes’ performance. Following this idea, this project focuses on utilizing wearables in basketball, collecting data from Inertial Measurement Units placed on the dominant arm of players to analyze their shooting technique during free throws, thereby predicting which shots are scored and which are missed. This classification was done by developing a machine learning model based on the features extracted from the measurements. In addition, a complementary study was carried out to evaluate the performance of the sensors regarding their battery consumption according to the transmission power. The results exhibit that the Random Forest model is the best for prediction, achieving an accuracy of 94.17%. Besides, the maximum elbow flexion angle is the most relevant and influential feature, suggesting that a controlled flexion of the elbow could be a key biomechanical factor to get a successful free throw.