Selective Neural Network Ensemble Based on Clustering
To improve the generalization ability of neural network ensemble, a selective method based on clustering is proposed. The method follows the overproduce and choose paradigm. It first produces a large number of individual networks, and then clusters these networks according to their diversity. Networks with the highest classification accuracies in each cluster are selected for the final integration. Experiments on ten UCI data sets showed the superiority of the proposed algorithm to the other two similiar ensemble learning algorithms.
KeywordsBase Classifier Generalization Performance Ensemble Size Final Integration High Classification Accuracy
Unable to display preview. Download preview PDF.
- 1.Krogh, A., Vedelsby, J.: Neural Network Ensembles, Cross Validation, and Active Learning. In: Tesauro, G., Touretzky, D., Leen, T. (eds.) Advances In Neural Information Processing Systems, vol. 7, pp. 231–238. MIT Press, Cambridge (1995)Google Scholar
- 4.Brodley, C., Lane, T.: Creating and Exploiting Coverage and Diversity. In: Proc. AAAI 1996 Workshop on Integrating Multiple Learned Models, pp. 8–14 (1996)Google Scholar
- 5.Giacinto, G., Roli, F., Fumera, G.: Design of Effective Multiple Classifier Systems by Clustering of Classifiers. In: Proc. of ICPR 2000, 15th Int’l Conf. on Pattern Recognition, Barcelona, Spain, pp. 3–8 (2000)Google Scholar