Prediction of Peak Shear Strength of Joints Based on Machine Learning Algorithms

The peak shear strength of rock joints is closely related to the normal stress, the roughness of the joint surface, and the mechanical properties of the material. Based on the published data, three machine learning (ML) algorithms, including Support Vector Machine (SVM), BP Neural Network (BPNN), and Random Forest (RF) are used to established regression models for predicting the peak shear strength of joints. 10-fold cross-validation is adopted to evaluate the performance of three ML models. In addition, three empirical peak shear strength models are selected for comparison with the ML models. Compared to the empirical models, we find that the ML models predict the peak shear strength with higher accuracy and lower error, indicating that the ML models perform better. Finally, the parameter analysis is performed to evaluate the influence of input parameters and the robustness of the ML models.