On knowledge-based fuzzy classifiers: A medical case study

Abstract This article discusses the role of knowledge in the problem of classification and presents a knowledge-oriented fuzzy classification system (KOFC) suitable for use in fields in which classification criteria, though numerically imprecise, can be formulated in natural language, and in which it is important to retain the expert's conceptual descriptions. KOFC extends previous fuzzy K nearest neighbor techniques in that it generalizes the concept of a design set, in order to allow both reference sets and their labelling to be defined in fuzzy terms by an expert, expressing items of his knowledge in production rule format. The results of applying KOFC to the study of fetal behavior are presented and discussed.