Research on Tennis Match Momentum Prediction Based on BSA-XGBoost Algorithm

In recent years, with the rapid development of machine learning and deep learning technology in sports analysis, the evaluation of athlete performance and status has been more accurate. The data provided by sports analytics help coaches to gain a better understanding of an athlete’ s condition at current times and strategize game plans accordingly to provide a better winning chance. We were inspired by some analytical research in badminton and refocused on tennis and examined the shot-by-shot data of "2023 Wimbledon Championships." Initially, we used unique one hot encoding for the data. Then we apply the XGBoost and its novel Bird Swarm Algorithm (BSA-XGBoost) extension to predict momentum shifts throughout the game. Also, we utilized MAPE and R² to compare the performance of the models, tested included Decision Trees, Random Forests, AdaBoost, Gradient Boosting, XGBoost and BSA-XGBoost. These comparative results demonstrate that BSA-XGBoost hyperparameter optimization achieves notable improvements in the prediction accuracy and general performance of the model.