[Analysis of occupational stress risk factors and risk prediction research for female coal miners]

Objective: To analyze the risk factors of occupational stress among female coal miners and construct a risk prediction model. Methods: In July 2023, a cluster sampling method was conducted to select all female workers on duty at a large coal mine in Gansu Province during a specific period (totaling 1437 individuals) as research subjects. Occupational stress levels were assessed through questionnaire surveys. A total of 1435 valid questionnaires were collected, with an effective recovery rate of 99.86%. Single-factor analysis was conducted using the rank sum test, while multiple linear regression models was used for multivariate analysis. Deep learning models, including XGBoost and LightGBM, were constructed to analyze factors influencing occupational stress among female coal miners. The performance of the optimal XGBoost model was further explained using the Shapley additive explanatory (SHAP) algorithm. Results: Among the 1435 female coal miners, 89 cases (6.2%) had no occupational stress (score ≤20 points); 930 cases (64.8%) had mild occupational stress (scores 21-30 points); 405 cases (28.2%) had moderate occupational stress (scores 31-40 points); and 11 cases (0.8%) had severe occupational stress (scores 41-50 points). Educational level, marital status, annual personal income, physical exercise, job type, average weekly working time, work form, work position, exposure to noise and noise intensity, heavy labor, coal dust exposure, exposure to occupational hazards and exposure time, injury or infringement during work, and whether to wear protective equipment were influencing factors affecting the occupational stress of female coal miners (P<0.05). For the XGBoost model on the test set, the coefficient of determination (R(2)), mean squared error (MSE), and root mean squared error (RMSE) were 0.994, 0.152, and 0.390, respectively, which were better than LightGBM's 0.933, 1.692, 1.301 and the multiple linear regression model's 0.268, 18.408, and 4.290. The SHAP results showed that work position and daily household chores were the main factors affecting the occupational stress of female coal miners. In addition, physical exercise, work-related injuries or violations, and coal dust exposure also had a certain impact on occupational stress. Conclusion: The occupational stress of coal miners is the result of the combined effects of occupational and lifestyle factors. The XGBoost model demonstrates superior predictive ability and proves the feasibility of the SHAP algorithm in analyzing the occupational stress of female coal miners using machine learning models.