Towards hierarchical clustering (extended abstract)

Abstract

In the paper, new modified agglomerative algorithms for hierarchical clustering are suggested. The clustering process is targeted to generating a cluster hierarchy which can contain the same items in different clusters. The algorithms are based on the following additional operations: (i) building an ordinal item pair proximity ('distance') including the usage of multicriteria approaches; (ii) integration of several item pair at each stage of the algorithms; and (iii) inclusion of the same items into different integrated item pairs/clusters. The suggested modifications above are significant from the viewpoints of practice, e.g., design of systems architecture for engineering and computer systems. © Springer-Verlag Berlin Heidelberg 2007.