NEW APPROACH TO THE COMPUTATION OF THE JARDINE-SIBSON Bk CLUSTERS.

Abstract

A new approach to the computation of Jardine and Sibson's (1968a) (fine) k-cluster method B//k is described. The algorithm makes use of the fact that the set of dissimilarities which are invariant under the B//k clustering transformation can be represented as an abstract graph. The proposed algorithm for computing this graph obtains both the (weak) k-ultrametric dissimilarity matrix and the k-clusters simultaneously.