A method of determination of optimum volume of the training selection is developed. Correlation relationship between quantity of right predictions of scoring algorithm and total number of tests is used as a base of the method. Bayes's qualifier is used as an example of mathematical argumentation. The method can be applied to arbitrary scoring algorithms.
обучающая выборка, скоринг, надежность скорингового алгоритма, training selection, scoring, reliability of scoring algorithm
1. E. Meyz. Rukovodstvo po kreditnomu skoringu. Banksis, Moskva, 2008. 464 s.
2. A.A. Stroev. Raschety i operacionnaya rabota v kommercheskom banke. № 6. 28-33. (2004)
3. V.V. Karasev. Risk-menedzhment v kreditnoy organizacii, metodicheskiy zhurnal. 10. 2. 97-108. (2013).
4. J. Kruppaa, A. Schwarzb, G. Arminger, A. Ziegler Expert Systems with Applications. 40. 5125-5131. (2013).
5. Tom Fawcett. HP Laboratories (2004)
6. S.I. Pshenichnyy. // Ekonomicheskie nauki. 2010. № 63, S. 306-310.
7. K. V. Voroncov, A.S. Inyakin, A. V. Lisica. // Tr. Vseross. konf. MMRO-13. «MAKS Press», Moskva, 2007. S. 577-581.
8. E. V. Strebkov V sb. Issledovaniya po prikladnoy matematike. Vyp.21. Unipress, Kazan', 1999. S. 228.
9. Strebkov E.V. V sb. materialov mezhdunarod. nauchno-prakt. konfer. Logisticheskaya integraciya rossiyskih regionov: Institucional'nye innovacii. Izd-vo «Brig», Kazan', 2012. S. 255 -257.
