Incremental possibilistic k-modes

Abstract

This paper proposes an incremental version of a soft clustering approach under uncertainty. The possibility theory and the k-modes algorithm are combined together in an incremental way to deal with two aspects of uncertainty. On one hand, the possibility theory deals with uncertain values of attributes of instances using possibility distributions and handles the belonging of objects to different clusters based on possibilistic membership degrees. On the other hand, the incremental aspect is studied in this new method by adding clusters without re-clustering initial instances. Experimental results clearly demonstrate the advantages of our proposal in a variety of databases using different evaluation criteria. © 2013 Springer-Verlag.