Designing fuzzy rule-based classifiers that can visually explain their classification results to human users

In various application areas of fuzzy rule-based systems, human users want to know why a particular reasoning result is obtained. That is, fuzzy rule-based systems are required to have high explanation ability. In this paper, we propose an approach to the design of fuzzy rule-based classifiers that can visually explain their classification results to human users. That is, our fuzzy rule-based classifiers can explain to human users why an input pattern is classified as a particular class in an understandable manner. The proposed approach consists of a rule selection method and a visualization interface. Our idea is to design fuzzy rule-based classifiers using fuzzy rules with only two antecedent conditions. A genetic algorithm is employed to construct a compact fuzzy rule-based classifier by choosing only a small number of fuzzy rules. In the classification phase, we use a single winner rule-based method for classifying an input pattern. The classification result of the input pattern is visually explained in a two-dimensional space where the two antecedent conditions of the winner rule are defined. Our approach is compared with feature selection by computational experiments.