Wavelet Based Nonparametric Additive Models for Nonlinear System Identification and Prediction

Wavelet based nonparametric additive models are considered for nonlinear system identification. Additive functional component representations are an important class of models for describing nonlinear input-output relationships and eavelets, which have excellent approximation capabilities, can be chosen as the functional components in the additive models. Wavelet based additive models, combined with model order determination and variable selection, are capable of handling problems of high dimensionality. Examples are given to demonstrate the efficiency of this new modelling approach.