Constrained Naïve Bayes with application to unbalanced data classification

Abstract The Naïve Bayes is a tractable and efficient approach for statistical classification. In general classification problems, the consequences of misclassifications may be rather different in classes, making it crucial to control misclassification rates most critical and, many realworld minority cases, possibly at expense higher less problematic classes. One traditional address this problem consists assigning costs classes applying rule, by optimizing loss function. However, fixing precise values such applications. paper we issue classifier. Instead requesting costs, threshold are used performance measures. This done adding constraints optimization underlying estimation process. Our findings show that, under reasonable computational cost, indeed, measures consideration achieve desired levels yielding user-friendly constrained procedure.