CHAOTIC TIME SERIES PREDICTION USING THE NONLINEAR PAR SYSTEMS

In this work, the nonlinear polynomial autoregressive (PAR) system has been applied to predict chaotic time series. For this purpose, different mathematical model structures based on nonlinear PAR time series have been presented to prediction of Mackey-Glass and Lorenz chaotic time series. As adaptive algorithms, Genetic algorithm (GA), differential evolution algorithm (DEA) and clonal selection algorithm (CSA) in heuristic algorithms, recursive least square algorithm (RLS) in classic algorithms have been used to determine the parameter values in the presented models and compared its performances. The simulation results have shown that both the presented mathematical models for chaotic systems and optimization works using the different algorithms to determine the parameters of these model structures have been highly successful.