A Study on Visualization of Tennis Players' Game Performance Based on Wavelet Analysis and ARIMA-LightGBM Model

Abstract

With the continuous progress of science and technology, scientific prediction methods gradually penetrate into the field of competitive sports, and the performance of athletes on the field can be visualized by a series of indicators to further predict the outcome of the match. In order to visualize player performance, this paper first establishes a performance scoring model and quantifies the fluctuation of the matching situation and the player's performance, and the quantitative results are consistent with the trend of the actual matching. Considering the data characteristics of its time series, we carried out multi-scale decomposition on the basis of discrete wavelet transform, and finally reconstructed and synthesized the original sequence using multi-scale wavelet to obtain the corresponding feature matching trend graph. In order to verify whether there is a relationship between consecutive scores and players winning matches, this paper uses Spearman correlation analysis and Kruskul-Walis test to verify that there is a significant correlation between the two. At the same time, by synthesizing the relevant data indicators of each tennis match, this paper combines the advantages of ARIMA and LightGBM to construct a more accurate prediction model.