Study of Tennis Match Momentum based on Random Forest and BP Neural Network Models

Abstract

The aim of this study is to scientifically analyse the "momentum" of a tennis match by developing a mathematical model. Data preprocessing techniques, such as K-nearest neighbour interpolation, were used to deal with missing values in dynamic matches. By filtering key metrics through principal component analysis and constructing a random forest model, we are able to quantitatively assess player performance. Using BP neural network models and genetic algorithms to optimise weights and biases, we improve the accuracy of match trend prediction. The analyses showed the significant influence of match time and player status on match trends. These research results provide a basis for scientific and data-based tennis training, which can help coaches conduct post-match assessment and optimise training programmes to promote player performance improvement.