On the accuracy of a bootstrap estimate of the classification error

Analytical and simulation studies show that a variance of the bootstrap estimator discussed is lower than that of the commonly used leave-one-out estimator only when sample size is extremely small or when the classification error is large. An essential feature of the bootstrap method is that observations of the training sample (TS) play the role of a general population and are used to determine the optimistic bias of the resubstitution estimate. A bootstrap training sample (BTS) is formed from the TS in a random way. A classification rule is designed using a BTS and is tested twice