Application of Support Vector Machine with Particle Swarm Optimization Algorithm in Blasting Vibration Prediction

SVM is powerful for the problem with small samples, non linear and high dimension. But such important parameters as the kernel function parameters, the insensitive parameters and the penalty coefficient are determined based on experience and cross-validation in the SVM, so it has certain blindness. In the paper, support vector machine optimized particle algorithm is used to predict the intensity of blasting vibration to solve the problems. An application indicates that support vector machine with particle swarm optimization algorithm is superiority over the empirical formula method on the prediction ability of blasting vibration intensity. Comparing the prediction results of different combinations of eight input elements such as height difference, horizontal distance, maximum charge, total charge, bench height, hole and row spacing, cast direction and batholithic resistance line, the combination of the former three elements was found to give best results, with the relative error being only at the level of 3.89%.