A general theory of classificatory sorting strategies: II. Clustering systems

Abstract

Current clustering programs (i.e., non-hierarchical classificatory programs) are examined, with particular reference to the internal consistency of the methods used for initiation, allocation and reallocation. It is shown that almost all existing methods are open to serious objection, and that no method fully exploits the potentialities of such systems. The desirable properties of a clustering program are examined de novo, and suggestions made for optimum lines of further development.