Pruning Ensembles with Cost Constraints
The paper presents a cost-sensitive classifier ensemble pruning method, which employs a genetic algorithm to choose the most promising ensemble. In this study the pruning algorithm considers constraints put on the cost of selected features, which is the one of the key-problems in the real-life decision support systems, especially dedicated medical support systems. The proposed method takes into consideration both the overall classification accuracy and the cost constraints, returning balanced solution for the problem at hand. Additionally, also to boost the value of the exploitation cost, we propose to use cost-sensitive decision trees as the base classifiers. The pruning algorithm was evaluated on the basis of the comprehensive computer experiments run on cost-sensitive medical benchmark datasets.
KeywordsPattern classification Machine learning Classifier ensemble Classifier ensemble pruning Cost-sensitive classification
Unable to display preview. Download preview PDF.
This work was supported by the Polish National Science Center under the grant no. DEC-2013/09/B/ST6/02264.
- 4.Frank, A., Asuncion, A.: UCI machine learning repository (2010). http://archive.ics.uci.edu/ml
- 13.Núñez, M.: The use of background knowledge in decision tree induction. Mach. Learn. 6(3), 231–250 (1991)Google Scholar
- 14.Núñez, M.: Economic induction: A case study. In: EWSL, pp. 139–145 (1988)Google Scholar
- 17.Tan, M., Schlimmer, J.C.: Cost-sensitive concept learning of sensor use in approach and recognition. In: Proceedings of the Sixth International Workshop on Machine Learning, pp. 392–395. Morgan Kaufmann Publishers Inc., San Francisco (1989)Google Scholar
- 18.Turney, P.D.: Cost-sensitive classification: empirical evaluation of a hybrid genetic decision tree induction algorithm. J. Artif. Int. Res. 2(1), 369–409 (1995)Google Scholar
- 19.Verdenius, F.: A method for inductive cost optimization. In: Proceedings of the European Working Session on Learning on Machine Learning, EWSL 1991, pp. 179–191. Springer-Verlag New York Inc., New York (1991)Google Scholar