Construction of Dynamic Momentum Model based on Neural Network

Tennis matches are usually highly dynamic and multi-faceted, and the winners and losers of the matches are affected by a variety of factors. Therefore, in order to deeply analyze the game trend and identify the key factors, and help the athletes scientifically establish a good game strategy, this paper introduces a series of complex models. In this paper, a BP neural network model is used to systematically identify the key factors that determine the game trend, and the TOPSIS model is used to calculate the athlete's comprehensive score. In addition, the LSTM model is used to predict the flow and fluctuation of the race, and the eigenvalues are calculated by introducing small perturbations and comparing the variance. In order to further optimize the model, additional factors such as world ranking and age were also considered. It was found that a high percentage of aces, serves and break points were crucial to the match outcome. Through model calculations, we identified the six most important metrics, with momentum ranking first. Based on these findings, we recommend that athletes focus on training in areas such as serving. Improving the model with additional factors resulted in further improvements in prediction and analysis. In this paper, a multi-model fusion approach was used to reveal the key factors affecting the tennis game and provide feasible training suggestions.