Find in Library

In this paper, w e present a data clustering method based on tolerance relation. The connectivity relation between objects, which is formed from tolerance sets, is a kind of equivalent relation of space partitions into several pairwise disjoint sets, called equivalent classes. Each cl ass contains objects which are considered equivalent only if they are connected. This is the basic idea of proposed clustering algorithm. Our method can discover clusters of arbitrary shapes even if they are not in a high density.