184 Int. J. Innovative Computing and Applications, Vol. 4, Nos. 3/4, 2012 Copyright © 2012 Inderscience Enterprises Ltd. Improving a dynamic ensemble selection method based on oracle information Leila Maria Vriesmann* and Alceu de Souza Britto Jr. State University of Ponta Grossa (UEPG), Av. General Carlos Cavalcanti, 4748, Ponta Grossa (PR), 84030-900, Brazil and Pontifical Catholic University of Parana (PUCPR), R. Imaculada Conceição, 1155, Curitiba (PR), 80215-901, Brazil E-mail: lmvriesmann@uepg.br E-mail: alceu@ppgia.pucpr.br *Corresponding author Luiz Eduardo Soares de Oliveira Federal University of Parana (UFPR), Av. Cel. Francisco H. dos Santos, s/n, Curitiba (PR), 81530-900, Brazil E-mail: lesoliveira@inf.ufpr.br Robert Sabourin and Albert Houng-Ren Ko École de Technologie Supérieure (ÈTS) – University of Quebec, 1100, Notre-Dame West, Montréal (QC), H3C1K3, Canada E-mail: robert.sabourin@etsmtl.ca E-mail: albert@livia.etsmtl.ca Abstract: This work evaluates some strategies to approximate the performance of a dynamic ensemble selection method to the oracle performance of its pool of weak classifiers. For this purpose, we evaluated different distance metrics in the K-nearest-oracles (KNORA) method, the use of statistics related to the class accuracy of each classifier in the pool and some additional information calculated by using a clustering process in the validation dataset. Moreover, different strategies are also evaluated to combine the results of the KNORA dynamic ensemble selection method with the results of its built-in K-nearest neighbour (KNN) used to define the neighbourhood of a test pattern during the ensemble creation. A strong experimental protocol based on more than 60,000 samples of handwriting digits extracted from NIST-SD19 was used to evaluate each strategy. The experiments have shown that the fusion of the KNORA results with the results of its built-in KNN is a very promising strategy. Keywords: dynamic ensemble selection; oracle; K-nearest-oracles; KNORA. Reference to this paper should be made as follows: Vriesmann, L.M., de Souza Britto, A., Jr., de Oliveira, L.E.S., Sabourin, R. and Ko, A.H-R. (2012) ‘Improving a dynamic ensemble selection method based on oracle information’, Int. J. Innovative Computing and Applications, Vol. 4, Nos. 3/4, pp.184–200. Biographical notes: Leila Maria Vriesmann received her MSc in Informatics from the Federal University of Parana, Brazil in 2006. In 2007, she started her PhD in Informatics in Pontifical Catholic University of Parana (PUC-PR), Brazil. Her research interests are ensemble classification methods and pattern recognition. Alceu de Souza Britto Jr. received his MSc in Industrial Informatics from the Federal Centre for Technological Education of Parana, Brazil in 1996, and PhD in Computer Science from Pontifical Catholic University of Parana (PUC-PR), Brazil in 2001. In 1989, he joined the Computer Science Department of the Ponta Grossa University, Brazil. In 1995, he also joined the Computer Science Department of the PUC-PR. His research interests are in the areas of pattern recognition, document analysis and handwriting recognition.