Cluster analysis of multivariate data : efficiency versus interpretability of classifications