Prediction and treatment of joint injuries in basketball training based on improved regression algorithm from the perspective of sports biomechanics

With the increasing popularity of basketball, especially in collegiate competitions like the University Basketball Super League, the sport has become a significant part of student life. The intensity of basketball training and competition has risen, necessitating athletes to have enhanced physical capabilities to meet modern demands. This heightened physical confrontation often leads to various injuries, with joint injuries being particularly common and impactful. This study integrates sports biomechanics with machine learning to address the prediction and treatment of joint injuries in basketball training. By employing an improved regression algorithm and leveraging high-performance computing, we have experimentally analyzed the prediction of joint injuries and proposed effective solutions. Our results indicate that the difference between the highest and lowest predicted residual values for the Back Propagation (BP) model was 0.92, and for the Extreme Learning Machine (ELM) regression model was 0.87. Notably, the improved ELM regression model demonstrated a reduced residual difference of 0.43. This improvement suggests that the enhanced ELM regression model offers superior prediction accuracy for joint injuries in basketball training and provides more comprehensive monitoring of athletes’ physical health, thereby supporting the advancement of basketball training programs.