Analysis of autoregressive fuzzy systems

The combination of fuzzy logic and neural networks are one of the techniques that are recently being used for modelling nonlinear system. Most of real-word systems have dynamic features, i.e. the actual output depends on the previous values. The so-called autoregressive dynamic fuzzy system with rules in the general forms has to be used. The specific problem appears when dealing with this autoregressive dynamic fuzzy system. Majority of authors treat the static and dynamic fuzzy systems in the same way. However, their properties are very different and lead to serious problems in practice. This paper demonstrates the difficulties of the usage of a fuzzy system as an autoregressive dynamic system. Completely new criteria and algorithms for the analysis of the autoregressive dynamic fuzzy systems are proposed.