Pairwise fusion matrix for combining classifiers

Pairwise fusion matrix for combining classifiers

Ko, Albert H.R. and Sabourin, Robert and Britto, Alceu de Souza and Oliveira, Luiz

Pattern Recognition 2007

Abstract : Various fusion functions for classifier combination have been designed to optimize the results of ensembles of classifiers (EoC). We propose a pairwise fusion matrix (PFM) transformation, which produces reliable probabilities for the use of classifier combination and can be amalgamated with most existent fusion functions for combining classifiers. The PFM requires only crisp class label outputs from classifiers, and is suitable for high-class problems or problems with few training samples. Experimental results suggest that the performance of a PFM can be a notch above that of the simple majority voting rule (MAJ), and a PFM can work on problems where a behavior-knowledge space (BKS) might not be applicable. © 2007 Pattern Recognition Society.