Parameter determination based on composite evolutionary algorithm for reconstructing phase-space in chaos time series

Abstract

To improve the prediction accuracy of the chaotic time series prediction model, a composite optimization method of the differential evolution (DE) algorithm that is based on the phase space reconstruction and least square supported vector machine (LSSVM), is proposed. The phase space parameters and LSSVM model parameters are taken as differential evolution algorithm individuals while the prediction accuracy of the chaotic time series is used as the evaluation function of DE algorithm. The optimal parameters are obtained by mutation, crossover, and selection operators of DE algorithm. Several numerical simulation results show that not only four parameters are determined at the same time, but also the performance of chaotic time series prediction is improved.