A Wavelet-Based Hybrid Prediction Method for Chaotic Time Series Combined with RBF Neural Networks and Chaos Model

Abstract

Current prediction models for chaotic time series with a lot of noise cannot accurately depict the multi-properties. To address the problem, this paper proposes a wavelet-based hybrid model called WRC model combined with Radial Basis Function neural networks (RBF neural networks) and chaos model. The chaotic time series are decomposed into an approximate time series and detailed time series with different traffic features, which are predicted by RBF neural networks and chaos model respectively. Simulation results verify the validity of the proposed WRC model and show a high prediction accuracy of the model.