Prediction of chaotic time series based on fuzzy model

Abstract

For dynamic systems with complex, illconditioned, or nonlinear characteristics, the fuzzy model based on fuzzy sets is very useful to describe the properties of the dynamic systems using fuzzy inference rules. Modeling and prediction of nonlinear systems using fuzzy modeling is discussed in this paper. First, the fuzzy space of input variables is partitioned by means of online fuzzy competitive learning. Further, the parameters of fuzzy model are estimated by means of Kalman filtering algorithm. To illustrate the performance of the proposed method, simulations on the chaotic MackeyGlass time series prediction are performed. Combining either offline or online learning with the proposed method, we can show that the chaotic MackeyGlass time series are accurately predicted, and demonstrate the effectiveness.