Research on the predicted height of water-conducting fracture zones based on the BO-RFR model and SHAP analysis.

Significant differences exist in the development height of water-conducting fracture zones (WCFZ) and their relationships with influencing factors between Carboniferous-Permian coal seams in central-eastern China and Jurassic coal seams in the western region. Utilizing 147 measured WCFZ height samples from Carboniferous-Permian seams and 111 from Jurassic seams, five influencing factors were selected: mining height, mining depth, working face slope length, hard rock ratio coefficient, and mining method. A Bayesian-optimized random forest regression model (BO-RFR) was then constructed to predict WCFZ heights, with the contributions of individual factors assessed through SHAP (SHapley Additive exPlanations) values. SHAP analysis indicated that, for Carboniferous-Permian seams, the factors influence WCFZ height in descending order as mining height, hard rock ratio coefficient, mining depth, working face slope length, and mining method. Conversely, for Jurassic seams, the ranking is mining height, mining depth, hard rock ratio coefficient, mining method, and working face slope length. This study not only validates the superior predictive capability of the BO-RFR model for WCFZ height but also systematically elucidates the differing mechanisms by which these factors impact WCFZ development across coal seams of two distinct depositional eras. The findings provide targeted theoretical support and practical decision-making guidance for the safety assessment of water-bearing coal mining in central-eastern and western mining regions of China.