Skip to Main Content
Feature selection for ensembles can often improve generalization accuracy of classifiers. In this paper we present a strategy on the feature selection for ensembles based on a hierarchical Non-dominated Sorting in Genetic Algorithms (NSGA-II) proposed by Deb. The first level of our strategy performs feature selection in order to generate a set of good classifiers, the second one deletes redundant classifiers while the third one combines classifiers left to provide a series of powerful ensembles. The proposed strategy is evaluated on data sets of UCI, using support vector machine as our classifiers. Our experiments demonstrated the effectiveness of our strategy.