Comparing Local and Global Hierarchical Multilabel Classification Methods Using Decision Trees

Ricardo CerriAndré Carlos Ponce de Leon Ferreira de Carvalho

Hierarchical multilabel classification is a classification problem where an example can belong to more than one class simultaneously and these classes are structured as a hierarchy. This paper describes and evaluates two methods of hierarchical multilabel classification based on decision trees, fol­lowing the standard local and global approaches, and a new variation method of the local approach, named HMC­Label­Powerset. Ten biological datasets with gene functions of the Yeast organism, organized as a tree structure, were used in the experiments. The results show that the local approach can lead to better results, although a more complex set of rules is produced, reducing the interpretability of the induced model.

Caso o link acima esteja inválido, faça uma busca pelo texto completo na Web: Buscar na Web

Biblioteca Digital Brasileira de Computação - Contato:
     Mantida por: