Exploring momentum in tennis using LightGBM-MLR fusion modeling

Tennis Performance Prediction is a highly regarded study published daily in the 《Journal of Sports Science》. Its simplicity of prediction methodology and depth of insight have made it widely popular. In this paper we identify the factors influencing momentum in tennis, materialize the vague concept of momentum, and based on the results obtained we advise players on how to improve their performance. Initially, we meticulously scrubbed the raw data and narrowed down eight critical factors influencing player performance. Leveraging the gradient boosting tree, we formulated a LightGBM-MLR Evaluation Model to ascertain the significance of these eight factors. Post normalization, we employed a Multiple Linear Regression Model to compute a comprehensive player performance score. We then evaluated the model's accuracy, analyzed the outcomes to offer targeted recommendations for player performance enhancement, and compared the fit of five machine learning models, ultimately concluding that LightGBM offered the most optimal fit. This research has helped to improve player performance and prediction of game trends, and it can be extended to other sports.