Machine learning prediction of employment-related psychological distress among Chinese university students identifies a study-travel proxy as a negative predictor

Employment-related psychological distress among university students has become a significant mental health concern in China, yet modifiable predictors amenable to intervention remain insufficiently understood within predictive modeling frameworks. This investigation developed a machine learning prediction model using XGBoost algorithm combined with SHAP interpretation to examine whether a study-travel proxy is negatively associated with an employment-anxiety proxy among Chinese university students, utilizing nationally representative data from the China Family Panel Studies 2020 ( N = 1,847). Because CFPS 2020 does not contain validated direct measures of either construct, the outcome was operationalized as a composite of standardized K6 psychological distress and reverse-coded future confidence, and the key predictor as a behavioral indicator combining educational-activity participation with positive household tourism expenditure. The analytical approach demonstrated that XGBoost achieved the highest discriminative performance among all evaluated models (AUC-ROC = 0.793, 95% CI: 0.753–0.833), with SHAP interpretation identifying life satisfaction, social network size, household income, and the study-travel proxy as the most influential predictors. Study travel presence consistently yielded negative SHAP values indicating negative predictive associations with the employment-anxiety proxy, with students possessing larger social networks exhibiting stronger negative predictive contributions, consistent with theoretical expectations regarding potential synergistic relationships between experiential learning and social capital. These findings suggest that structured educational travel programs integrating experiential learning with social support enhancement may represent promising directions for addressing employment-related distress among university students facing labor market transitions, pending confirmation with validated direct measures and causal verification through longitudinal research.