Tennis Match Trend Prediction Based on LSTM

The focus of this research is on the data processing, including data mining, data cleaning, feature extraction, feature engineering and other aspects. Whether the data can be used to fit the game trend well is the key to the final effect of the model. At the same time, there are many reasons (i.e., characteristics) that affect the trend of the game. The purpose of this study is to find out the most drastic variables and make corresponding tactical changes according to them. The model based on LSTM network constructed in this paper has a good ability to handle variable-length inputs, and uses the attention mechanism to enhance the "memory" of long-distance inputs, and uses the Dropout layer to avoid model overfitting to a certain extent. Finally, the random forest method was used to extract several variables with the highest sensitivity for the sensitivity analysis of the model.