Tennis Match Momentum Prediction Based on Spearman Correlation and LightGBM Model

In this paper, a new method for predicting tennis match momentum based on Spearman Correlation and LightGBM model is proposed. In the correlation analysis, we use Spearman correlation coefficient to measure the monotonic relationship between the momentum of athletes and the outcome of the game. We found that the serve score rate and unforced errors in the primary indicator had a significant impact on the momentum difference of the players. In addition, we construct a LighTGBM-based predictive model, which uses the gradient raised decision tree (GBDT) algorithm to support efficient parallel training, and has the advantages of fast training speed, low memory consumption, high precision, and support for large-scale data processing. Through 50% cross-validation, LightGBM model has shown good predictive performance on the training set, cross-validation set and test set, and has excellent performance on the mean square error (MSE), root mean square error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and determination coefficient (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>). The feature importance analysis shows that serve score rate and unforced error are the two key factors affecting the momentum difference of athletes, accounting for 21.12% and 13.19% of the feature importance respectively.