Accurate prediction of the Rock Blastability Index (RBI) is essential for optimizing blasting operations, reducing excavation costs, and minimizing environmental impacts in surface mining. This study explores the application of three advanced machine learning (ML) approaches: super learner, multigene genetic programming (MGGP), and gene expression programming (GEP), to predict RBI using a comprehensive data set of rock mechanics parameters collected from the Mouteh Gold Mine in Iran. Comparative evaluations based on multiple statistical indicators confirm that MGGP consistently achieves the highest predictive accuracy, outperforming both Super Learner and GEP. Feature importance analysis using Shapley Additive Explanations (SHAP) reveals that Rock Quality Designation (RQD) is the most influential parameter in RBI prediction. Unlike black-box models, MGGP and GEP offer the added advantage of producing explicit mathematical equations, which enhance interpretability, enable sensitivity analysis, and support site-specific customization in engineering practice. Although the Super Learner model demonstrates strong predictive performance, its limited transparency restricts its practical usability. Visual tools, including scatter plots, Taylor diagrams, wind rose plots, and violin plots, further highlight the robustness and generalization capability of MGGP. Overall, the findings emphasize the potential of equation-based machine learning models to provide both accuracy and transparency in RBI prediction. Future research should explore hybrid strategies that integrate interpretability with adaptive learning to enhance decision-making in rock engineering applications.
