Multi resolution wavelet least squares support vector machine network for nonlinear system modeling

Abstract

This paper proposes Multi resolution Wavelet Least Squares Support Vector Machine (MRWLS-SVM) network for the nonlinear system modeling. It composes a set of sub-wavelet least squares support vector machine networks with specified resolutions. The outputs of these sub-networks are aggregated via a set of weights to produce the network output. The structure of the MRWLS-SVM network is achieved using the fuzzy c-mean and the least squares support vector algorithm. The stability of the proposed network has been investigated using Lypunov stability theorem. The generalization of the proposed MRWLS-SVM network for the modeling purpose has investigated using two nonlinear systems.