One of the most active areas of research in supervised learning has been the study of methods for constructing good ensembles of classi ers, that is, a set of classiers whose individual decisions are combined to increase overall accuracy of classifying new examples. In many applications classi ers are required to minimize an asymmetric loss function rather than the raw misclassi cation rate. In this paper, we present approaches to modifying existing methods for constructing ensembles to incorporate arbitrary loss functions. We compare the performance of the new algorithms with traditional ensemble learners, MetaCost-a novel method for cost-sensitive learning, and single decision tree classi ers. We evaluated the algorithms on multi-class data sets from the UC Irvine ML Repository.