Modeling time series volatility using fuzzy rule systems

Summary form only given. A special class of volatility models based on Takagi-Sugeno fuzzy rules is presented in the paper. The corresponding identification problem is stated and techniques for constructing and estimating fuzzy rules by observation data are thoroughly described. A comparison between models of proposed class and traditional ARCH-/GARCH-models is made for a sample data produced by a non-linear model. A question of using fuzzy rule systems for modeling time-varying volatility of real-life time series is briefly discussed.