AN INCREMENTAL CLUSTERING APPROACH WITHIN BELIEF FUNCTION FRAMEWORK

摘要

The Belief K-modes Method (BKM) is a recently developed clustering approach handling uncertainty encountered in the attribute values of categorical dataset objects. This uncertainty is represented and managed under the belief function framework. This proposed method, as generally existing clustering ones, starts with a known dataset of objects characterized by a given set of attributes. However, there are numerous applications where this attribute set evolves. So, we propose in this paper an Incremental Belief K-modes Method (IBKM), that is able to cluster uncertain data within such dynamic environment. The main objective is to efficiently maintain clusters as new attributes are inserted without frequently performing complete recluster-ing.