Bootstrap Neyman-Pearson test for knowing the value of misclassification probability

A new classification framework is developed for two-class classification problems when one needs to know the value of misclassification probability. The proposed classifier is constructed by applying bootstrap to Neyman-Pearson test (NPT). We use mixtures of latent variable models (MLVMs), which are effective in density estimation, to obtain the approximation of test statistic. The classifier, which combines bootstrap, NPT and MLVMs, can be applied to any distribution for which the maximum likelihood estimates exist. Another advantage of the classifier is that Neyman-Pearson criterion is easy satisfied with the empirical quantile of the bootstrap distribution of test statistic. Experiments on both synthetic and real world data sets are carried out. The results show the applicability and efficiency of the proposed classifier.