Skip to main content

A Self-Adaptive Fuzzy System that Improves its Performance Over Time

A Mathematics Review, Editor: S Cawley,Blackhall Publishing, Dublin, 1997 (printed in 1998), 113 - 132, ISBN: 1-901657-01-9

Abstract

This paper presents three new techniques that allow a fuzzy system to make decisions in a changing application domain. The salient characteristic of such a domain is that situations and facts can change over time. This means that the domain cannot be modelled by a fixed set of fuzzy rules. Instead, the rules must change over time and these changes must model actual changes that occur in the application domain. Three new techniques have been developed for altering a set of fuzzy rules: altering the credibility weight of an expert and using connection matrices to shift the focus of attention between different sets of rules; fine-tuning and changing the membership functions of fuzzy premise variables and thereby altering the meaning of rules; and generating new fuzzy rules by inductive learning from examples.