A note on the choice of a weighting function to give an efficient method for estimating the probability of misclassification

摘要

The problem of estimating the performance of a given classifier on a given data set is considered. In an attempt to provide a nonparametric estimator which not only uses the data efficiently but is also essentially an unbiased estimator of the probability of misclassification, Toussaint evaluated empirically an estimator formed by weighting the resubstitution and rotation estimators. In this study theoretical consideration is given to the choice of a suitable weighting function for this estimator.