Accurate resilience evaluation is important to help vocational college students cope with transitional stress. This study developed and validated a multidimensional resilience framework using a “dual-track” design ( N = 1,588). Psychometric analyses (Track A) revealed a robust three-factor structure with tenacity, strength, and optimism. Measurement Invariance across genders was demonstrated. Using machine learning for predictive validation (Track B), it was found that the XGBoost model performed better (AUC = 0.883) in predicting low-resilience risk than the traditional logistic regression model. Interpretability analysis through SHAP highlighted sleep quality and perceived stress as key predictors aligning with stress–resource theory. AI enhanced this by incorporating psychometrics and algorithms to give an accurate and explainable method for early identification of those in need of support in educational settings.
Xu et al. (Tue,) studied this question.