Detecting homogeneous groups in clustering using the Euclidean distance

The use of the Euclidean distance as dissimilarity measure to detect groups in a sample implies that the features space is homogeneous. The clustering process can be realised under this condition taking the sample as one grou p, but when the groups are detected we must assure that each one is homogen eous, in order to apply the Euclidean distance. To detect homogeneous group s we propose a criterion, which treats each group as a fuzzy set in the sam e universe of discourse of the sample. (C) 2001 Elsevier Science B.V. All r ights reserved.