A visual explanation system for explaining fuzzy reasoning results by fuzzy rule-based classifiers

In this paper, we develop a visual explanation system for explaining fuzzy reasoning results (i.e., classification results of input patterns) by fuzzy rule-based classifiers in an understandable manner to human users. Our explanation system can clearly explain why an input pattern is classified as a specific class. We use fuzzy rules with only two antecedent conditions. That is, the antecedent part of each fuzzy rule is defined on only two attributes. We assume the use of a single winner rule-based fuzzy reasoning method for pattern classification. Thus a single fuzzy rule is responsible for the classification of an input pattern. Our visual explanation system depicts the input pattern to be classified, the given training patterns, and the winner rule in a two-dimensional pattern space with the same two attributes as in the antecedent part of the winner rule.