Abstract
Water conducting fracture zone (WCFZ) is a common damage phenomenon in coal mining. Accurately predicting the height of WCFZ is of great scientific significance for protecting the mining area environment and ensuring safe production. This study introduces a novel intelligent prediction model for the height of WCFZ in coal mines based on Categorical Boosting (CatBoost). The CatBoost model, utilized to capture and explore the complex relationships between coal mining factors and WCFZ height, incorporates four cutting-edge intelligent algorithms for adaptive optimization of model performance. The experimental results demonstrate that the proposed model exhibits excellent performance and strong generalization ability, with the coefficient of determination achieving 0.969 on the test set. In addition, the importance and global contribution mechanism of mining factors to WCFZ height were analyzed and revealed using Shapley Additive exPlanations (SHAP). The model established in this study exhibits excellent generalization ability and interpretability, providing a reliable method for the accurate prediction of WCFZ height and can offer auxiliary support for decisions regarding mine safety production and environmental protection.