A Study of Momentum Change in Tennis Matches Based on LightGBM and Support Vector Machine Algorithms

Aiming at the key factors of tennis match scoring, this paper discusses the relationship between the player's "momentum" change and various factors, establishes and solves an accurate model. First, 16 possible influencing factors are constructed through logistic regression, and five machine learning models, including LGBM, XGB, SVC, MLP and LR, are used to quantify and visualize the progress of a tennis match. Second, Pearson correlation analysis was performed to verify the correlation between "momentum" and match scores. Finally, different datasets are used to remove uncertainties and the SVC model is predicted using roc-auc visualized random game data. It is found that the model constructed in this paper has good generalization and accuracy.