Understanding rock deformation is crucial for various engineering and geological applications, including mining, tunneling, and earthquake prediction. In this study, we propose a novel approach to estimate rock deformation under uniaxial compression using extreme gradient boosting (XGB), Extra trees regression (ETR), and K-Nearest Neighbours (KNN) algorithms. The proposed methodology involves three main steps. First, a comprehensive dataset of rock samples is collected, including various positions of the strain gauge, stress, and corresponding deformation measurements under uniaxial compression. These properties serve as input and output features for the machine learning models. Second, the XGB, ETR, and KNN algorithms are trained and tested using the collected dataset. These algorithms are known for their ability to handle complex relationships and nonlinearities, making them suitable for modeling the intricate behavior of rock deformation under compression. To ensure accurate predictions, a cross-validation technique is employed to optimize the hyperparameters of each algorithm. The trained models are then evaluated using various performance evaluations like performance parameters, Actual and predicted curves, Rank analysis, Sensitivity Analysis, Error matrix, and OBJ criteria. All models perform better (i.e., coefficient of determination greater than 0.9), however, XGB is a more robust model when compared to other models. Overall, this study presents a novel and promising approach to estimating rock deformation under uniaxial compression, offering a valuable tool for engineers and geologists working in the field of rock mechanics.
