Abstract
Four types of discussions as to usefulness of fuzzy clustering are shown. First, fuzzy graphs play the central role in agglomerative hierarchical clustering algorithms. Second, Entropy-based methods in fuzzy c-means clustering show generalizations of Gaussian mixture models and connect the hard c-means with a statistical model. Third, fuzzy classification functions show theoretical properties of fuzzy classification rules induced from fuzzy c-means clustering. Lastly, fuzzy cluster validity measures provide a good approach to determine the number of clusters.
