A Study of Athlete Performance Prediction Based on TOPSIS and Logistic Regression Models

Abstract

Tennis is a popular sport characterized by intense competition and demanding physical demands. To assess the performance of athletes in matches and to explore the key factors affecting their momentum and real-time scores, this paper selected eight indicators that have a significant impact on athletes' performance and processed the data using the entropy weighting method to derive the impact weights of each indicator on the athletes. The combined use of the TOPSIS method enabled the evaluation of the athletes' performance at each stage, while the real-time performance was demonstrated through visual analysis. To establish a real-time prediction model, 17 player indicators with a significant influence on training were selected, and a logistic regression model was established. The final model demonstrated an accuracy of 92%. The analysis revealed that serve percentage and serve speed exerted a significant positive influence on the player's momentum, whereas the unforced error rate exerted a significant negative influence on his performance. The prediction curve analysis yielded 2911 turning points in the players' status, providing a valuable reference for athletes' real-time performance.