Crustal Deformation Time Series Prediction Model Based on PSO and SVM

Crustal deformation time series is a significant information source during the researches on continental deformation. In order to simulate the low frequency linear components which reflect dynamic trend as well as the high-frequency non-linear components which reflect disturbance, a prediction model based on Particle Swarm Optimization (PSO) and Support Vector Machine (SVM) is presented. With optimization of the core parameters of SVM by PSO algorithm, this model integrates the merits of fast global optimization of PSO and accurate non-linear simulation of SVM. Through comparison and analysis on traditional model as Auto Regression, BP neural network and single SVM algorithm based on measured data, results show that this model is applicable for its high precision, simple format, unnecessary of massive known data, and especially the elimination of effects from human factors.