A Study on Predicting Tennis Player's Competition Direction Based on Logistic Regression and LSTM Model

A tennis player's performance during a match affects the outcome of the match. The study of the issue can provide advice to tennis players on how to prepare for the tournament, and to coaches and their teams on how to develop training programs that will help athletes get better results in the tournament. Initially, according to the entropy weight method to assess the player's performance on the field. Two factor layers were used to finally get a score measuring the performance at each point. Next, based on the score and making use of conditional probabilities in logistic regression models to predict the winning rate in each point and then in each game. The result shows high accuracy in the logistic regression in both points and games. Ultimately, by assuming the existence of average effect of the players’ performance in all points of one game, deep learning with long short-term memory is used to predict game fluctuations and identify factors that contribute to them.