The Scheme of Sliding Examination for Optimal Set Features Determination in Discriminant Analysis Problem

摘要

It is analytically shown that the traditional scheme of sliding examination allows one to solve the discriminant analysis problem stated in a broad sense on final samples of observations. The existence of optimal set of features is shown which corresponds to a maximum of mathematical expectation of offered generalized distance between observations from two parent populations. It is analytically shown that the parameters of parent population and sample size influence complexity of optimal discriminator. The conditions are found at which the optimal discriminator becomes reduced.