Definition of Machine Learning Variables.

This paper aims to examine the influence mechanism of education level as a key situational factor in the relationship between risk perception and behavioral response, encompassing both behavioral intention and preparatory behavior. Utilizing non-parametric estimation techniques in machine learning, particularly the Random Forest and XGBoost algorithms, this study develops predictive models to analyze the impact of 27 influencing factors on behavioral responses following risk perception. The findings indicate that, while the model’s fit for preparatory behavior is 25.71% and its fit for behavioral intention is below 20%, the model effectively identifies key influencing factors. Further analysis employing SHAP values demonstrates that education level not only exerts a significant influence but also exhibits varying effects across different educational groups. Moreover, statistical testing corroborates the importance of education level in the relationship between risk perception and behavioral response, providing a robust scientific foundation for the development of risk management policies.