Prediction of Chaotic Time Series Based on Gauss Kernel Function and Spatial Constraints Fuzzy Clustering Algorithm

Abstract

The prediction of chaotic time series based on Gauss kernel function and spatial constrains fuzzy clustering algorithm enhanced the robust ability to noise and outlier of the system model by adding Gauss kernel function. First of all, started from an initial fuzzy partition of input space by a simplified nearest-neighbor clustering method to get the number of fuzzy rules and the initial clustering center vector, then smoothed the input sample points in form of vector by using the Gauss kernel function. And then computed and optimized the fuzzy membership and the clustering center vector with the spatial constrains fuzzy clustering algorithm At last identified the conclusion parameters by the least square method. The proposed method was applied to simulations on Mackey-Glass chaotic time series modeling and prediction. The results demonstrated the robust, effectiveness and practicability of the method of this paper, and it had the better simulation result than support vector machine and the neural networks methods.