Fast Unfolding of Credal Partitions in Evidential Clustering

International audienceEvidential clustering, based on the notion of credal partition, has been successfully applied in many fields, reflecting its broad appeal and usefulness as one of the steps in exploratory data analysis. However, it is time-consuming due to the introduction of meta-cluster, which is regarded as a new cluster and defined by the disjunction (union) of several special (singleton) clusters. In this paper, a simple and fast method is proposed to extract the credal partition structure in evidential clustering based on modifying the iteration rule. By doing so, the invalid computation associated with meta-clusters is effectively eliminated. It is superior to known methods in terms of execution time. The results show the potential of the proposed method, especially in large data