Least Squares Wavelet Support Vector Machines for Nonlinear System Identification

Abstract

A novel admissible support vector kernel, namely the wavelet kernel satisfying wavelet frames, is presented based on the wavelet theory. The wavelet kernel can approximate arbitrary functions, and is especially suitable for local signal analysis, hence the generalization ability of the support vector machines (SVM) is improved. Based on the wavelet kernel and the least squares support vector machines, the least squares wavelet support vector machines (LS-WSVM) are constructed. In order to validate the performance of the wavelet kernel, LS-WSVM is applied to a nonlinear system identification problem, and the computational process is compared with that of the Gaussian kernel. The results show that the wavelet kernel is more efficient than the Gaussian kernel.