A new clustering criterior based on the distance matrix

Abstract

A new clustering method of a distance matrix is proposed here. The algorithm is based on the arrangement of the original entities in increasing order along a line of minimum distance and on permuting the elements of the distance matrix. The clustering is made with the constraint that the clusters must be formed by contiguous entities of the arranged series. Some applications to real problems prove the good properties of the new method both for the provided final partitions and for the strong reduction of calculations.