Predictive model of psychological factors influencing university students' intentions and behavior regarding physical exercise-based on machine learning algorithms.

Insufficient physical exercise poses a substantial threat to the physical and psychological wellbeing of university students. However, participation in physical activity is shaped by a complex interplay of psychological and physiological factors. This study employed machine learning techniques to predict levels of physical exercise participation and to identify key influencing factors among university students. Questionnaire data were collected from students across multiple provinces in China. Five classification models-Multi-Layer Perceptron (MLP), Extreme Gradient Boosting (XGBoost), Random Forest, Gradient Boosting Decision Tree (GBDT), and Decision Tree-were constructed and evaluated using standard classification metrics on an independent test set. Among the five models, the GBDT exhibited the highest predictive accuracy. Feature importance analysis indicated that body weight and anxiety were the most influential predictors of physical exercise participation. Partial dependence plots revealed non-linear relationships between key psychological variables and exercise behavior, particularly in differentiating higher levels of physical activity. In addition, the decision tree model identified resistance to temptation as the primary decision node, followed by healthy habits and impulse control. By integrating the Multi-Process Action Control (M-PAC) framework with stress process theory, this study elucidates how psychological stress, self-regulatory capacity, and behavioral habits jointly influence physical exercise participation. The findings provide practical implications for the development of targeted interventions aimed at promoting sustained physical activity among university students.