On Constructing Fuzzy Classifiers from Interval-Valued Data in Case of Unstable Clustering Structure

Abstract

The theoretical paper deals in the preliminary way with the problem of extracting fuzzy classification rules from the interval-valued data in the case of unstable clustering structure of the training data set. The corresponding technique is based on a heuristic method of possibilistic clustering. Types of clustering structures are considered. The description of basic concepts of the heuristic method of possibilistic clustering based on the allotment concept is provided. A technique of the interval-valued data preprocessing is also given. An extended technique of constructing of fuzzy rules based on clustering results is described. An illustrative example of the method's application to the Ichino and Yaguchi's oil data set is carried out. An analysis of the experimental results is given and some conclusions are forwarded.