Classifier Selection Approaches for Multi-label Problems

TitleClassifier Selection Approaches for Multi-label Problems
Publication TypeConference Paper
Year of Publication2011
AuthorsPillai, I, Fumera, G, Roli, F
Conference Name10th Int. Workshop on Multiple Classifier Systems (MCS 2011)
Date Published15/06/2011
Conference LocationNaples
Keywordsdoc00, doc01, mcs00, mcs01, multi-label

While it is known that multiple classifier systems can be effective also in multi-label problems, only the classifier fusion approach has been considered so far. In this paper we focus on the classifier selection approach instead. We discuss a specific selection strategy for ensembles of multi-label classifiers, based on selecting one or more two-class classifiers for each class, possibly coming from different multi-label classifiers. We then derive static selection criteria based on the F measure, which is widely used in multi-label problems. Preliminary experimental results show that the considered selection strategy can effectively exploit the complementarity between the multi-label classifiers on the different classes. They also show that the derived selection criteria can improve the trade-off between the macro- and micro-averaged F measure, despite it is known that an increase in either of them is usually attained at the expense of the other one.

Citation Key 1091
pillai_MCS2011.pdf282.8 KB