In the part, we present a factorial approach for clustering following the least squares criterion, for every choice of the metrics in the individual space. We deduce that the between-clusters inertia has an upper bound that depends on the number of clusters and the results of a Principal Component Analysis; this enables us to generalize a coefficient that measures the quality of the approximation of an optimal partition.In the second part, we demonstrate that the inertia induces a strict ordering of the set of optimal partitions. Finally, we propose a heuristic for choosing the number of clusters.
