Research on Tennis Match Momentum Prediction Based on Logistic Regression and GBDT Model

Tennis is a complex sport where shifts in match momentum can dramatically alter the outcome of games. This paper aims to predict changes in momentum during tennis matches using Logistic Regression and Gradient Boosting Decision Tree (GBDT) models. By analyzing and establishing relevant momentum models, we can address challenges within the sport. We initially preprocess the raw data to remove outliers and perform a correlation analysis to identify variables associated with momentum. Subsequently, a logistic regression model is employed to quantify momentum and predict turning points in the match. Additionally, a scoring system is constructed that integrates player performance, time differentials, and pressure factors, enabling the visualization of momentum fluctuations. The scoring system is then refined using the GBDT algorithm for accurate prediction of momentum shifts. The experimental results demonstrate that our models effectively capture the fluctuations in momentum throughout the match and provide crucial insights for coaches and players to devise more effective game strategies.