In this paper we propose a generalization of the standardclustering problem, which we call Structural Constrained Costarring(SCC) problem. In SC. problem, the cluster interconnections are givenby a binary relation R. If this relation is empty then SCC problemreduces to the standard classification problem. The cluster structureimposed by R may be described by the covering graph of R. For exampleif this graph is a rooted tree then a hierarchical structure isimposed to the clusters. We formulate SCC as a fuzzy clusteringproblem with prototype inter-relation. A general algorithm to solvethis problem is proposed. For three particular distance measures(squared, Euclidean and L_1 metric) we compute the prototypes bysolving a particular multifacility location problem.
展开▼
