An efficient method for estimating the probability of misclassification applied to a problem in medical diagnosis

摘要: Abstract The problem of estimating the performance a given classifier on data set is discussed for case when no knowledge available concerning underlying distributions. A new method probability misclassification proposed which yields essentially unbiased results similar to Lachenbruch's U -method with far less computation involved. While theoretical work presented, practical rule thumb choosing parameters estimator. are based experiments performed six diseases related epigastric pain, and underline importance reporting both testing training data. Whereas previous papers have continually reported correct classification as high 74.3 per cent raw 92.0 “processed” data, in this paper it shown that much more significant estimate 51.0 cent.