Identify Risk Characteristics of Occupational Stress Exposure in Key Industry Populations Based on Explainable Machine Learning

Objective To construct a machine learning model to identify occupational stress in key industrial populations and analyze the key influencing factors of occupational stress using the shapley additive interpretation (SHAP) method.Methods Multicenter cross-sectional design were adopted in this study. Based on entries from the 2024 Mianyang Occupational Health Literacy Study database, a total of 1,271 valid questionnaires were included in the analysis. Six machine learning algorithms—Gradient Boosting Machine (GBM), Support Vector Machine (SVM), Neural Network, Random Forest, eXtreme Gradient Boosting (XGBoost), and Logistic Regression—were employed to develop predictive models. The model with the best predictive performance was selected based on key performance metrics, including the area under the receiver operating characteristic curve (AUC), accuracy,and precision and so on, and was subsequently subjected to SHAP analysis. SHAP analysis was utilized to identify key factors influencing occupational stress and to explore the interaction effects among these factors.Results Among the six machine learning models, GBM demonstrated superior performance compared to the others,with AUC (95%CI) of 0.880 (0.853-0.907), accuracy of 0.798, sensitivity of 0.806, specificity of 0.796, precision of 0.5, and F1 score of 0.617. SHAP analysis identified the top five factors influencing the prediction of occupational stress, ranked in order of importance: depression, working hours, self-reported health status, anxiety, and age. Furthermore, significant interaction effects were observed between the following factor pairs: depression-anxiety, depression-self-reported health status, depression-sleep disorders, anxiety-self- reported health status, anxiety-sleep disorders, working hours-depression, and education level-income level, all of which contribute to occupational stress.Conclusion This study established a Gradient Boosting Machine (GBM) predictive model for occupational stress through model selection and analyzed the importance of influencing factors as well as their interaction effects. The findings contribute to a deeper understanding of the mechanisms underlying occupational stress and provide valuable insights for developing precise and effective intervention strategies.