The primary goal of blasting in open-pit mines is to achieve the desired rock fragmentation while minimizing adverse effects such as flyrock, backbreak (BB), airblast, and ground vibration. Among these, BB is particularly critical as it reduces operational efficiency and poses considerable safety hazards. In this study, support vector regression (SVR) was employed to predict blast-induced BB in the Mouteh gold mine. To enhance the model’s generalization capability and determine optimal hyperparameters, grid search (GS) technique combined with 5-fold cross-validation was applied during model development. A total of 80 datasets from the Mouteh gold mine, consisting of five key input parameters, rock quality designation (RQD), uniaxial compressive strength (UCS), density (D), cohesion (C), and blastability index (BI), were collected and randomly divided into training and testing subsets. For comparative analysis, a linear multivariable regression (LMR) model was also developed. The predictive performance of both models was evaluated using three statistical metrics: coefficient of determination (R²), root mean square error (RMSE), and mean absolute error (MAE). During training, the SVR model achieved R² = 0.978, RMSE = 0.062, and MAE = 0.069, outperforming the LMR model (R² = 0.940, RMSE = 0.087, MAE = 0.124). In the testing phase, SVR similarly outperformed LMR, yielding R² = 0.967, RMSE = 0.081, and MAE = 0.094 compared to R² = 0.886, RMSE = 0.167, and MAE = 0.207 for LMR. Furthermore, sensitivity analysis, incorporating Shapley Additive Explanation (SHAP) method and partial dependence plots (PDPs), revealed that RQD and BI were the most influential parameters affecting BB, whereas density had the least impact. These findings confirm that the optimized SVR model can be used as a robust and reliable tool for accurately predicting blast-induced BB and for optimizing blasting practices in the mining operations of the Mouteh gold mine.
