Divisive hierarchical bisecting min-max clustering algorithm

Abstract

The inspiration for the Divisive Hierarchical Bisecting Min-Max Clustering Algorithm came from the Bisecting K-Means clustering Algorithm. To obtain K clusters, bifurcate the set of input values into two clusters, select one of these clusters to split further (each time bisect the selected cluster using the Min-Max Clustering Algorithm), and so on, until K clusters have been produced. The Min-Max Clustering Algorithm initially computes the minimum of the input set and then finds a point which is at the greatest distance from the minimum. The remaining values from the set of data items are then accumulated into twoclusters formed by the maximally disjoint min and max values.