Nonlinear time‐series modeling and prediction using correlation analysis
Andrzej Dydyński, Jarosław Arabas · PAMM · 2007
Abstract We present a novel way for time series prediction. The method is based on the correlation analysis and allows for handling nonlinearities of different type and character. The presented approach results in an approximation model that combines nonlinear units taken from radial basis functions (RBF) and from multilayer perceptrons (MLP). The approach leads to a low mean error of the approximation with a number of parameters significantly smaller when compared to RBF and MLP. (© 2008 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)