Intelligent Prediction of Compressive Strength for Cemented Tailings Backfill Using Bayesian Optimized Machine Learning

With the increasing demand for sustainable mining, cemented tailings backfill (CTB) realizes the resource recycling of massive mine tailings solid waste and has become a key material for underground mine backfill and support [1]. Rational CTB preparation can drastically reduce tailings stockpiling hazards and alleviate environmental pollution triggered by piled mine solid residues. Unconfined compressive strength (UCS) is a core index to evaluate the performance of backfill, but traditional laboratory tests are time-consuming and inefficient [2]. To achieve efficient and accurate prediction, this study established 10 machine learning models based on 745 experimental data points, and adopted Bayesian optimization (BO) as the unified hyperparameter tuning method. The models include: Gaussian Process Regression (GPR), Linear Regression (LR), Random Forest (RF), Gradient Boosting (GB), XGBoost, Multilayer Perceptron (MLP-ANN), Radial Basis Function-Support Vector Regression (RBF-SVR), Stacked-ensemble, Local Cascade Ensemble (LCE), and LSTM-XGBoost. A comprehensive evaluation was conducted using R², RMSE, MAE, MSE, and MAPE, while ANOVA and residual analysis were used to verify the statistical validity. The results show that Bayesian optimized Gaussian Process Regression (BO-GPR) achieves the best prediction performance with a test R² of 0.9614, RMSE of 0.1969 MPa, MAE of 0.0989 MPa, MSE of 0.0388, and MAPE of 17.46%. SHAP analysis quantifies the top-five influencing factors with corresponding mean absolute SHAP values: cement-tailings ratio (CTR, 0.0008), curvature coefficient (Cc, 0.0006), temperature (TEM, 0.0005), mass concentration (MC, 0.0004), and Al₂O₃ (0.0003). This study provides an interpretable and efficient intelligent method for strength prediction and mix proportion optimization of mine cemented tailings backfill materials.

References:

1. Wang, J.; Chen, G.; Chen, Y.; Ye, Z.; Lin, M.; Su, R.; Hu, N. Intelligent Mixture Optimization for Stabilized Soil Containing Solid Waste Based on Machine Learning and Evolutionary Algorithms. Constr. Build. Mater.2024, 445, 137794, doi:10.1016/j.conbuildmat.2024.137794.
2. Yuan, Z.; Zheng, W.; Qiao, H. Machine Learning Based Optimization for Mix Design of Manufactured Sand Concrete. Constr. Build. Mater.2025, 467, 140256, doi:10.1016/j.conbuildmat.2025.140256.