Particle Swarm Optimization-Least Squares Support Vector Regression with Multi-scale Wavelet Kernel

A novel regression model combining least squares support vector regression (LS-SVR) with multi-scale wavelet kernel and particle swarm optimization (PSO) was presented in this paper, and applied to the approximation of non-stationary dataset and those continuous functions polluted by strong noise. Support vector kernel function with the multi-resolution characteristics was employed, such that LS-SVR with multi-scale wavelet kernel can estimate each details of target function accurately. The experimental results show that the proposed method is effective and feasible.