Predicting and Analysing Match Momentum in Tennis Matches Using Entropy Weight Method-based TOPSIS and XGBoost Model

Abstract

Tennis players' round performance scores and the prediction and analysis of match momentum in tennis matches are important for players, coaches and the media. Accurate and appropriate predictions and analyses are important for players and coaches to adjust their tactics in time, for media to analyse and report, and for spectators to experience and interact. Firstly, in order to quantify the performance of players, a comprehensive evaluation model of TOPSIS based on entropy weight method was constructed based on the theory of TOPSIS model and the advantage of entropy weight method in reducing subjectivity in the evaluation process. Secondly, in order to accurately predict the match momentum of tennis matches, XGBoost (Extreme Gradient Boosting) model was constructed. The model not only has extremely high prediction accuracy, but also improves the generalisation ability of the model, reduces overfitting, and has the advantages of automation and robustness. Finally, the constructed machine learning model is reasonably interpreted using the shap model to provide a scientific basis for players and coaches to change tactics and adjust training methods in a timely manner.