A heuristic approach for the hypervolume method in cluster analysis

We propose a heuristic approach for the hypervolume method in cluster analysis. The algorithm consists of two parts. The first one is a hierarchic divisive procedure which yields a partition of the set of objects into k clusters. The aim of the second part of the algorithm is to improve that partition according to the hypervolume criterion. We describe the new algorithm and we apply it to data sets test; we analyse the results obtained and compare them with the exact solution.