Time delay dynamic fuzzy networks for time series prediction

Creative Commons License

Oysal Y.

5th International Conference on Computational Science - ICCS 2005, Atlanta, GA, United States Of America, 22 - 25 May 2005, vol.3514, pp.775-782 identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 3514
  • Doi Number: 10.1007/11428831_96
  • City: Atlanta, GA
  • Country: United States Of America
  • Page Numbers: pp.775-782
  • Anadolu University Affiliated: Yes


This paper proposes a Time Delay Dynamic Fuzzy Network (TDDFN) that can be used for tracking and prediction of chaotic time series. TDDFN considered here has unconstrained connectivity and dynamical elements in its fuzzy processing units with time delay state feedbacks. The minimization of a quadratic performance index is considered for trajectory tracking applications. Gradient with respect to model parameters are calculated based on adjoint sensitivity analysis. The computational complexity is significantly less than direct method, but it requires a backward integration capability. For updating model parameters, Broyden-Fletcher-Golfarb-Shanno (BFGS) algorithm that is one of the approximate second order algorithms is used. The TDDFN network is able to predict the Mackey-Glass chaotic time series and gives good results for the nonlinear system identification. © Springer-Verlag Berlin Heidelberg 2005.