Monotone Fuzzy Rule Relabeling for the Zero-Order TSK Fuzzy Inference System
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
UNIVERSITI MALAYSIA SARAWAK
Abstract
Description
A monotonically-ordered fuzzy rule base is an essential building block for a monotone fuzzy inference system (FIS). However, fuzzy rules solicited from human experts do not always constitute a monotonically-ordered fuzzy rule base, i.e., noise exists due to errors in human judgments. In this thesis, the idea of monotone fuzzy rule relabeling (MFRR) is proposed to handle the non-monotone fuzzy rules in a fuzzy rule base, for zero-order Takagi-Sugeno-Kang (TSK) FIS modelling. The aim of MFRR is to restore the monotonicity property of the fuzzy rule base in an optimal way, for practical FIS modelling scenarios. MFRR searches for a set of solutions based on the following criteria: 1) satisfying the monotonicity property as the first priority; 2) having the minimum number of fuzzy rules changed as the second priority; and 3) having the minimum loss measure as the third priority. Apart from that, a formal and systematic way is proposed to model and analyze non-monotone fuzzy rules in a fuzzy rule base. A MFRR framework is further devised as a three-stage procedure, for practical implementation. The usefulness of the proposed method is demonstrated using real world applications, and promising results were obtained.