PREDICTING A CHAOTIC TIME-SERIES USING A FUZZY NEURAL-NETWORK

In this paper the authors present an alternative neurofuzzy architectu re for application to chaotic time series prediction. The architecture employs an approximation to the fuzzy reasoning system to considerabl y reduce the dimensions of the network as compared to similar approach es. The application considered is the chaotic Mackey-Glass differentia l equation. Simulation results for single and multi-step predictions w ere obtained using the MATLAB neural network toolbox and these are com pared with both traditional neural network implementations and other f uzzy reasoning approaches. The work not only demonstrates the advantag e of the neurofuzzy approach but it also highlights the advantages of the architecture for hardware realisations. (C) 1998 Elsevier Science Inc. All rights reserved.