Section: New Results

ClustGeo: Ascendant Hierarchical Clustering (AHC) with geographical constraints

The following result has been obtained by Marie Chavent (CQFD member), Vanessa Kuentz-Simonet, Amaury Labenne and Jerome Saracco (CQFD member).

Hierarchical Ascendant Clustering (HAC) is a well-known method of individual clustering. This method aims to bring together individuals who are similar regarding to variables which describe them. But when individuals are geographical units, the user may wish geographically close individuals to be put in same clusters and that, without too much deteriorating the quality of the partition. The proposed ClustGeo method allows geographical constraints of proximity to be taken into account within the HAC. For that purpose, a new Ward homogeneity criterion based on two different matrices of distances is proposed.