Bayesian optimization-enhanced ensemble learning for the uniaxial compressive strength prediction of natural rock and its application

• DOI: 10.1016/j.ghm.2024.05.002
• Publication Date: 2024-05-22T16:03:20Z
• Authors: Chukwuemeka Daniel; Xin Yin; Xing Huang; Jamiu Ajibola Busari; Amos Izuchukwu Da...; Honggan Yu; Yucong Pan
• Keywords: Boosting; AdaBoost; Overfitting; Ensemble Learning; Gradient boosting; Matthews correlation coefficient

Engineering disasters, such as rockburst and collapse, are closely related to structural instability caused by insufficient bearing capacity of geological materials. Uniaxial compressive strength (UCS) holds considerable significance in rock engineering projects. Consequently, this study endeavors devise efficient models for the expeditious economical estimation UCS. Using a dataset 729 samples, including Schmidt hammer rebound number, P-wave velocity, point load index data, we evaluated six algorithms, namely Adaptive Boosting (AdaBoost), Gradient Decision Tree (GBDT), Extreme (XGBoost), Light Machine (LightGBM), Random Forest (RF), Extra Trees (ET) utilized Bayesian Optimization (BO) optimize aforementioned algorithms. Moreover, applied model evaluation metrics Root Mean Squared Error (RMSE), Absolute (MAE), Variance Accounted For (VAF), Nash-Sutcliffe Efficiency (NSE), Weighted Percentage (WMAPE), Coefficient Correlation (R), Determination (R2). Among models, BO-ET emerged most optimal performer during training (RMSE = 4.5042, MAE 3.2328, VAF 0.9898, NSE WMAPE 0.0538, R 0.9955, R2 0.9898) testing 4.8234, 3.9737, 0.9881, 0.9875, 0.2515, 0.9940, 0.9875) phases. Additionally, conducted systematic comparison between ensemble traditional single machine learning decision tree, support vector machine, K-Nearest Neighbors, thus highlighting advantages learning. Furthermore, enhancement effect BO on generalization performance was assessed. Finally, BO-ET-based Graphical User Interface (GUI) system developed validated Tunnel Boring Machine-excavated tunnel.