Utilization of Statistical Learning Algorithms for Prediction of Elastic Modulus of Jointed Rock Mass

Abstract

This study uses two statistical learning algorithms for the prediction of elastic modulus (E_j) of jointed rock mass. The first algorithm uses support vector machine (SVM) that is firmly based on the theory of statistical learning and uses regression technique by introducing e-insensitive loss functionhas been adopted. The second algorithm uses relevance vector machine (RVM). It is based on a Bayesian formulation of a linear model with an appropriate prior that results in a sparse representation. The RVM model gives variance of predicted data. The inputs of models are joint frequency (J_n), joint inclination parameter (n), joint roughness parameter (r), confining pressure (s₃) and elastic modulus (E_i) of intact rock. Equations have been developed for the determination of E_j of jointed rock mass based on the SVM and RVM models. The results of SVM and RVM models are compared with a widely used artificial neural network (ANN) model. This study shows that the developed SVM and RVM models can be used for the prediction of E_j of jointed rock mass.

Keywords: elastic modulus, jointed rock, support vector machine, relevance vector machine, artificial neural network