Reduction Methods on Multi-label Data based on Granular Computing
Multi-label classification (MLC) is one specific type of classification. In a multilabel dataset, an object can belong to several classes at the same time, that is, more than one class label can be associated with the same object. In the context of multi-label classification, most data reduction methods have been addressed to the selection and extraction of features, and discretization. For this reason, we propuse three methods of instance selection, and three
methods of instance generation (prototypes). The former are relying on the Rough Set Theory (RST), which is one of the most representative theories within Granular Computing. On the other hand, the prototype generation methods are capable of generating a set of new objects in the application domain from the initial objects. Our proposal is based on two different ways of the granularity of the information. The methods proposed in these researches are independent of the learning algorithm to be used.
We explore the global performance of our methods when coupled with the ML-kNN algorithm. We leaned upon several multi-label training sets taken from the MULAN and RUMDR repositories.
Rafael Bello received his Bachelor degree in Mathematics and Computer Science (1982) at Universidad Central ¨Marta Abreu¨ de Las Villas (UCLV), Santa Clara, Cuba and his PhD in Mathematics at UCLV in 1988. He has been a visiting scholar at some universities in Spain, Germany and Belgium. He is a Full Professor at Computer Science Department, UCLV, Cuba, and exhibits a long record of academic exchange with many universities in Latin America and Europe. He has authored/edited 12 books, published over 200 papers in conference proceedings and scientific journals. His research interests comprise Metaheuristics, Soft Computing (Rough and Fuzzy Set theories), Knowledge discovery and Decision making.