Self-generating hierarchical fuzzy systems

In this paper; two types of self-generating hierarchical fuzzy system (SG-HFS) are proposed. Based on the ability of FuNe I Neuro-Fuzzy System to identify the rule-relevant nodes, a hierarchical fuzzy system (HFS) can be automatically generated. In a computationally complex environment, in which a large number of inputs are present, a hierarchically structured fuzzy system becomes more desirable as the total number of rules in the rule bases can be reduced dramatically. This paper describes a novel method of generating HFS from numerical data for classification-type problems. The generated hierarchical fuzzy system can further be optimized through iterative training, rule reduction, and hierarchical level reduction.