Majid Mojirsheibani - This research was supported in part by a grant from NSERC Canada. Scholl of Mathematics & Statistics, Carleton University, Ottawa, Ontario, KIS ۵B۶ Canada.

Data-based procedures are proposed for combining a number of individual classifiers in order to construct more effective classification rules. The resulting combined classifiers turn out to be almost surely superior to each individual classifier, under appropriate regularity conditions. Here, superiority means lower asymptotic misclassification error rate. Both the mechanics and the asymptotic validity of the proposed procedures are discussed.

Classification, Bayes classifier, consistency.