E2CM: An Evolutionary Version of Evidential C-Means Clustering Algorithm

This paper aims to propose an Evolutionary version of Evidential C-Mean (E2CM) clustering method based on a Variable string length Artificial Bee Colony (VABC) algorithm. In the E2CM, the centers of clusters are encoded in form of a population of strings with variable length to search optimal number of clusters as well as locations of centers based on the VABC, by minimizing objective function non-specificity, in which the assignment of objects to the population of cluster centers are performed by the ECM. One significant merit of the E2CM is that it can automatically create a credal partition without requiring the number of clusters as a priority. A numerical example is used to intuitively verify our conclusions.