Learning interpretable multi-class models by means of hierarchical decomposition: Threshold Control for Nested Dichotomies | Publicación