Diversity in Ensembles for One-Class Classification
One-class classification, known also as learning in the absence of counterexamples, is one of the most challenging problems in the contemporary machine learning. The scope of the paper focuses on creating a one-class multiple classifier systems with diverse classifiers in the pool. An approach is proposed in which an ensemble of one-class classifiers, instead of a single one, is used for the target class recognition. The paper introduces diversity measures dedicated to the selection of such specific classifiers for the committee. Therefore the influence of heterogeneity assurance on the overall classification performance is examined. Experimental investigations prove that diversity measures for one-class classifiers are a promising research direction. Additionally the paper investigates the lack of benchmark datasets for one-class problems and proposes an unified approach for training and testing one-class classifiers on publicly available multi-class datasets.
Keywordsone-class classification machine learning multiple classifier system diversity classifier selection one-class benchmarks
Unable to display preview. Download preview PDF.
- 5.Frank, A., Asuncion, A.: UCI machine learning repository (2010), http://archive.ics.uci.edu/ml
- 7.Golestani, A., Azimi, J., Analoui, M., Kangavari, M.: A new efficient fuzzy diversity measure in classifier fusion. In: Proceedings of the IADIS Conference, pp. 722–726 (2007)Google Scholar
- 14.Kuncheva, L.I., Whitaker, C.J.: Measures of diversity in classifier ensembles and their relationship with the ensemble accuracy. Machine Learning 51(2) (2003)Google Scholar
- 15.R Development Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria (2008) ISBN 3-900051-07-0Google Scholar
- 16.Schölkopf, B., Smola, A.J.: Learning with kernels: support vector machines, regularization, optimization, and beyond. Adaptive computation and machine learning. MIT Press (2002)Google Scholar
- 18.Tax, D.M.J., Duin, R.P.W.: Characterizing one-class datasets. In: Proceedings of the Sixteenth Annual Symposium of the Pattern Recognition Association of South Africa, pp. 21–26 (2005)Google Scholar