Design of Linguistically Interpretable Fuzzy Rule-Based Classifiers: A Short Review and Open Questions
Hisao Ishibuchi, Yutaka Kaisho and Yusuke Nojima
Fuzzy rule-based classifier design often involves conflicting criteria: accuracy and interpretability. This paper discusses the design of linguistically interpretable fuzzy rule-based classifiers. The emphasis of our discussions is placed on the linguistic interpretability of fuzzy rule-based classifiers rather than their accuracy. First we give a short survey on the design of fuzzy rule-based classifiers with linguistic conditions. Next we discuss the relation between the complexity of fuzzy partitions and their interpretability using fuzzy grids with different granularities. Then we discuss the interpretability of fuzzy rule-based classifiers using simple numerical examples. We also discuss the interpretability of classification results by fuzzy rule-based classifiers. That is, we discuss the explanation ability of fuzzy rule-based classifiers to explain the classification result of each pattern in a human understandable manner. The main contribution of this paper is to demonstrate that we still have a number of open questions with respect to the interpretability of fuzzy rule-based classifiers.
Keywords: Linguistic rules, fuzzy systems, fuzzy rule-based classifiers, fuzzy classification, interpretability.