Predication of multivariable chaotic time series based on maximal Lyapunov exponent

Abstract

A method for prediction of multivariable chaotic time series through selecting many neighboring reconstructed vectors is proposed with reference to the method for prediction of single-variable chaotic time series based on maximal Lyapunov exponent. The new method is used to forecast the chaotic time series of two Rssler equations coupled system, Rssler equation and Hyper Rssler equations coupled system for onestep and multistep. Results show that the algorithm can forecast multivariable chaotic time series precisely and has strong anti-chirp ability. The relation between the result and the number of neighbor points is discussed.