Hierarchical multi-classification

Abstract

The problem of hierarchical multi-classification is considered. In this setting a set of classes is to be assigned to a single instance, and all possible classes are structured according to a hierarchy. Example application domains are functional genomics and text classification. An algorithm is presented to solve hierarchical multi-classification tasks. It is a decision tree induction algorithm that is based on the notion of predictive clustering trees and in which a suitable distance measure is plugged in. Preliminary results with the algorithm on data sets from functional genomics and text classification are reported and discussed.