Clustering of fuzzy data using credibilistic expected and critical values

This paper introduces a new approach for handling fuzzy data sets in producing crisp clusters. The proposed method uses the notion of expectation of discrete fuzzy variables. The proposed method has been compared with another approach through fuzzy critical values pursued by Sampath and Kalaivani (2010). Comparative experimental study has been carried out with the help of data sets simulated from multivariate normal populations where fuzziness has been induced using a well defined procedure. In the process of comparison two partitioning clustering methods, namely, k-means and k-medoids algorithm have been considered.