Chaotic time series prediction based on information entropy optimized parameters of phase space reconstruction

Abstract

This paper proposes a method of information entropy optimized parameters (IEOP) of pahse space recon struction. First, it establishes an information entropy optimum model in phase space for embedding dimension and delay time by using conditional entropy. It then solves these two parameters with genetic algorithm (GA) simultaneously. IEOP constructs an optimum phase space, which maintains independence of reconstruction coordinate and retains the dynamic characteristics of the original system. In the numerical simulations, results of the Lorenz system and Mackey-Glass system show that it not only determines two parameters at the same time, but also can obtains more information in the optimized phase space, there by improving the performance of chaotic time series prediction.