Supervised Learning: Multi-Label Classification

Multi-label classification is an extension of multi-class classification where each instance under exam can be associated with a set of multiple, non-exclusive labels. Multi-label learning algorithms can be separated into two main methodological categories: Problem Transformation and Algorithm Adaptation. Problem Transformation methods turn the learning problem into multiple single-label problems, one for each label of interest, whereas Algorithm Adaptation techniques modify existing single-label learning algorithms to solve multi-label tasks directly. This chapter aims to provide a clear description of the multi-label classification problem and a detailed review of the state-of-the-art algorithms for both methodological categories, focusing on technical details and rationale for adopting each of them.