Class-switching neural network ensembles

Abstract

This article investigates the properties of class-switching ensembles composed of neural networks and compares them to class-switching ensembles of decision trees and to standard ensemble learning methods, such as bagging and boosting. In a class-switching ensemble, each learner is constructed using a modi ed version of the training data. This modi cation consists in switching the class labels of a fraction of training examples that are selected at random from the original training set. Experiments on 20 benchmark classi cation problems, including real-world and synthetic data, show that class-switching ensembles composed of neural networks can obtain signi cant improvements in the generalization accuracy over single neural networks and bagging and boosting ensembles. Furthermore, it is possible to build mediumsized ensembles ( 200 networks) whose classi cation performance is comparable to larger class-switching ensembles ( 1000 learners) of unpruned decision trees.