Multi-Label Classification via Manipulating Labels

Unlike traditional classification problem, multi-label learning task is to predict a label set with unknown size for an example. While the exponential number of possible label sets challenges the task of multi-label learning. Many approaches by manipulating labels have been proposed. In this paper, we propose a new method via manipulating labels for multi-Label Learning: adding a virtual label to the original label set, appending the label subset selected by mutual information for each pairwise labels to the original feature set, and finally learning a binary classifier for each pairwise labels. Extensive experiments show that, compared with advanced multi-label methods, the proposed method induces models with significantly better performance..

[1]  G. G. Stokes "J." , 1890, The New Yale Book of Quotations.

[2]  Ian Witten,et al.  Data Mining , 2000 .

[3]  Aidong Zhang,et al.  Mining of Attribute Interactions Using Information Theoretic Metrics , 2009, 2009 IEEE International Conference on Data Mining Workshops.

[4]  Eyke Hüllermeier,et al.  A Unified Model for Multilabel Classification and Ranking , 2006, ECAI.

[5]  Grigorios Tsoumakas,et al.  Mining Multi-label Data , 2010, Data Mining and Knowledge Discovery Handbook.

[6]  Geoff Holmes,et al.  Classifier chains for multi-label classification , 2009, Machine Learning.

[7]  J. Ross Quinlan,et al.  C4.5: Programs for Machine Learning , 1992 .

[8]  Ian H. Witten,et al.  Data mining: practical machine learning tools and techniques, 3rd Edition , 1999 .

[9]  Zhi-Hua Zhou,et al.  A k-nearest neighbor based algorithm for multi-label classification , 2005, 2005 IEEE International Conference on Granular Computing.

[10]  Harry Zhang,et al.  Full Bayesian network classifiers , 2006, ICML.

[11]  Saso Dzeroski,et al.  Decision trees for hierarchical multi-label classification , 2008, Machine Learning.

[12]  Eyke Hüllermeier,et al.  Multilabel classification via calibrated label ranking , 2008, Machine Learning.