ABSTRACT
This chapter illustrates some relevant clustering methods in a fuzzy framework useful for clustering time series. Traditional clustering arbitrarily forces the full assignment of such an object to one of the clusters, although they should almost equally belong to all of them. Fuzzy clustering relaxes the requirement that objects have to be assigned to only one cluster. The fundamental justification concerning the adoption of a fuzzy approach in a clustering framework lies in the recognition of the vague (fuzzy) nature of the cluster assignment task. The former objective is achieved by maximizing the entropy (and, therefore, the uncertainty) of the clustering of the objects into the various clusters. The latter objective is obtained by constraining the maximization process in such a way as to minimize the overall distance of the objects from the cluster prototypes Coppi and D'Urso. The chapter introduces some robust fuzzy clustering capable of neutralizing the negative effects of outliers in the clustering process.
