A clustering procedure that begins with all the objects in a single cluster. It is opposite of agglomerative procedures. The procedure begins with a single large cluster that is divided into separate clusters based on the the most dissimilar objects.