A MODIFIED KRETSCHMER COMPLEXITY INDEX FOR SELECTING END PARTITIONS IN CLUSTER-ANALYSIS

An index introduced by H. KRETSCHMER [5, 6] in scientometrics for describing relations of scientific communications was modified and used to select end partitions in cluster analysis which are associated with a minimum of complexity in the system of classes. Such partitions are favoured for the interpretation of initial information given by a data matrix. Using theory of sets an algorithm was developed to compare these (and other) selected end partitions and to construct generalized classes without restriction of their number. So it is possible to obtain a high degree of information from a data matrix. The approach was applied to chemical property-structure-relations (physical properties/electronic structure, topology, state of aggregation, number of atoms).