Abstract: Constructing an intelligent classification model for rock mass blastability is of great significance for the safety and economy of mining blasting engineering. Four key indicators, namely rock bulk density (ρ), uniaxial compressive strength (σ_c), rock mass integrity coefficient (η), and unit explosive consumption (q), were selected to establish a classification indicator system. After data cleaning and processing using the Synthetic Minority Oversampling Technique (SMOTE), 195 balanced samples were obtained. An intelligent classification model for rock mass blastability, named BO-XGBoost, was constructed by optimizing the hyperparameters of the Extreme Gradient Boosting (XGBoost) algorithm using Bayesian Optimization (BO). The SHapley Additive Explanations (SHAP) method was further employed to analyze the contribution of each indicator. The proposed BO-XGBoost model was then applied to classify the blastability levels of relevant engineering cases both in China and abroad. The results show that the BO-XGBoost model exhibits excellent classification performance on the test set, achieving an accuracy of 96.67%, precision of 97.50%, recall of 96.92%, and an F₁-score of 0.97. SHAP analysis reveals that the importance ranking of the indicators is as follows: rock bulk density, unit explosive consumption, rock mass integrity coefficient, and uniaxial compressive strength. The classification results are in full agreement with the actual conditions, demonstrating the model's engineering practicality and accuracy.