Adapting non-hierarchical multilabel classification methods for hierarchical multilabel classification

Abstract

In most classification problems, a classifier assigns a single class to each instance and the classes form a flat (non-hierarchical) structure, without superclasses or subclasses. In hierarchical multilabel classification problems, the classes are hierarchically structured, with superclasses and subclasses, and instances can be simultaneously assigned to two or more classes at the same hierarchical level. This article proposes two new hierarchical multilabel classification methods based on the well-known local approach for hierarchical classification. The methods are compared with two global methods and one well-known local binary classification method from the literature. The proposed methods presented promising results in experiments performed with bioinformatics datasets.