Induction in Multi-Label Domains
All the techniques discussed in the previous chapters assumed that each example is labeled with one and only one class. In realistic applications, however, this is not always the case. Quite often, an example is known to belong to two or more classes at the same time, sometimes to many classes. For machine learning, this poses certain new problems. After a brief discussion of how to deal with this issue within the framework of classical paradigms, this chapter describes the currently most popular approach: binary relevance.
- 15.Clare, A. & King, R. D. (2001). Knowledge discovery in multi-label phenotype data. In Proceedings of the 5th European conference on principles of data mining and knowledge discovery, PKDD’01, Freiburg, Germany (pp. 42–53)Google Scholar
- 32.Godbole, S. & Sarawagi, S. (2004). Discriminative methods for multi-label classification. In H. Dai, R. Srikant, & C. Zhang (Eds.), Lecture Notes in Artificial Intelligence, Berlin, Heidelberg: Springer (Vol. 3056, pp. 22–30)Google Scholar
- 47.Koller, D. & Sahami, M. (1997). Hierarchically classifying documents using very few words. In Proceedings of the 14th International conference on machine learning, ICML’07, San Francisco, USA (pp. 170–178)Google Scholar
- 57.McCallum, A. (1999). Multi-label text classification with a mixture model trained by EM. In Proceedings of the workshop on text learning (AAAI’99) (pp. 1–7).Google Scholar