Abstract
Clustering is a method to classify data into groups called clusters so that objects in the same cluster are similar. In this research, we focus on nominal data sets. Conventional clustering methods for nominal data use dissimilarities of clusters which are defined by aggregating dissimilarities of objects. However, even if objects in the same cluster are similar in the sense of the dissimilarity measure, it may fail to represent common features of the whole objects in clusters. To overcome such a drawback, we propose new dissimilarities using discernibilities of clusters on attribute subsets. We investigate properties of the proposed dissimilarities in the application of hierarchical clustering.
