Research on Match Trends in Tennis Competitions Based on Bayesian Online Change Point Detection and Random Forest Prediction Model

Abstract

The prediction of the result of the competition is helpful for the coaches and athletes to adjust their strategy and state during the competition and to summarize and study the competition mode of the opponent after the game. For a long time, the prediction of tennis has focused on the prediction of winning and losing results, and the prediction results are often low, so our work is devoted to providing more reliable data analysis and match trend identification methods for this sport. Based on the data of The Champions Wimbledon 2023, Random forest algorithm was used to predict the mutation points obtained by Bayesian online change point detection algorithm. In this process, “momentum” was taken into account, three categories of features were constructed, and the resulting mutation points were screened. Finally, a high accuracy rate (higher than 0.8) was obtained, indicating that such mutation points contained useful information and had practical significance. Finally, we use an approximate expression of Taylor expansion and the method of undetermined coefficient to correct the weights for the features according to the extreme inversion in the schedule. This mutation point provides a new observation angle and idea for the coaches and athletes during a game of daily training, which is conducive to enlightening the athletes to achieve breakthroughs.