Visualization of players' game performance based on random forest and particle swarm models

Abstract

In this study, a model combining Particle Swarm Optimization (PSO) and Random Forest Algorithm (RF) is proposed for processing and selecting features in a sports competition dataset and optimizing the model parameters. First, the scale differences between different features were eliminated by data normalization, followed by optimizing the parameters of the random forest model using the PSO algorithm, which significantly improved the prediction accuracy of the model. The study also introduced the performance level scoring model, which assigns weights according to the importance of selected features, and the momentum scoring model, which explores the relationship between dynamic changes in the game and the outcome of the game. The validity of the models was verified through real-world data analysis, showing that this approach can accurately predict the outcome of a match. In addition, Wilcoxon rank sum test and Spearman correlation coefficient were applied to statistically validate the model, and the results showed that the selected features were significantly correlated with the match results. The application of these methods not only improves the explanatory ability of the model, but also provides a new research direction and methodological guidance for the field of sports data analysis. In summary, the integrated modeling framework developed in this study demonstrates high efficiency and accuracy in handling large-scale sports data, demonstrates its potential for application in competitive sports data analysis, and provides a solid foundation and new perspectives for subsequent research in related fields.