An Illustrative Comparison of Rough k-Means to Classical Clustering Approaches

Rough clustering has gained increasing attention in the last decade with applications in such diverse areas like bioinformatics, traffic control and retail. The relationship between rough clustering and, in particular, fuzzy and possibilistic concepts is still a topic that is raised first and foremost by practitioners who are looking for an adequate clustering algorithm. Therefore, we compare rough k-means to fuzzy c-means, possibilistic c-means and to classical k-means in our paper. We show that rough k-means is closer related to classical k-means than to fuzzy and possibilistic c-means. Besides brief theoretical evaluations we perform illustrative experiments on artificial data and the IRIS data.