Creation of optimum spam filter on the basis of combination of statistical qualifiers

Criteria of an optimality are presented in article for classification of messages on the basis of statistical methods taking into account errors of the first and second childbirth, a share of false operations and passed spam. Features of testing and training spam filter are given. Approach to the organization of the qualifier which consists in sharing of methods of Bayes and Fischer is offered. For improvement of quality of a filtration it is offered to use the mechanism of the analysis of subsets of crossing of the sets distinguished by both methods on categories (spam \not spam, false operations and spam admission).