Optimizing Military Sports Training Effectiveness: Predictive Analysis Using the XGBoost Model

In this study, we apply the XGBoost algorithm to predict the results of military sports training while conducting a detailed analysis of the relevant data. Variables were assembled from a range of military athletic competitions—including measures of athlete performance, lengths of training periods, weather-related influences, attributes of the competition sites, and levels of participant expertise—to build a predictive model. The outcomes from experiments reveal that, in shooting events, the model obtained an average RMSE of 1.80 and MAE of 1.30, while in the intricate military pentathlon, the RMSE was 2.45 and MAE was 1.98. Such results underscore a solid predictive strength across various event categories, with accuracy surpassing 90% in applications to high-level athletes. To improve the model's effectiveness, we adjusted hyperparameters through grid search and Bayesian optimization methods, which yielded greater predictive accuracy and affirmed the model's reliability across different circumstances. This research not only offers a trustworthy mechanism for projecting training results in military sports but also sets the stage for its expansion into wider areas of athletic preparation.