Multi Label Classification using Label Clustering
Pranav Gupta, Ashish Prabhu Anand · 2013
Multi Label Classifier classifies a given instance into one or more labels from a set of labels. In literature, such classifiers have been classified into two categories: Problem Transformation (PT) and Algorithm Adaptation (AA). Because of simplicity and competitive performance of PT-methods, we explore a PT method, namely Label Powerset (LP). Existing LP methods are either too slow or tend to underutilize multi label information. We propose a novel LP approach, achieving competitive performance with respect to Hamming Loss and F1-measure, in relatively less time.