Estimation of Misclassification Error Using Bayesian Classifiers

Abstract

Bayesian classifiers provide relatively good performance compared with other more complex algorithms. Misclassification ratio is very low for trained samples, but in the case of outliers the misclassification error may increase significantly. The usage of ‘summation hack’ method in Bayesian classification algorithm can reduce the misclassifications rate for untrained samples. The goal of this paper is to analyze the applicability of summation hack in Bayesian classifiers in general.