Estimation of credit risks using the data mining methods
DOI:
https://doi.org/10.20535/SRIT.2308-8893.2017.1.03Keywords:
credit risk, statistical data, logit model, Bayesian networks, model quality parametersAbstract
In this research, credit risks are analyzed for financial organizations using data mining techniques applied to actual data. The two sets of actual statistical data characterizing the borrowers are employed for constructing mathematical models in the form of the nonlinear logit regression, decision trees, and Bayesian networks. The constructed models are analyzed with a set of appropriate statistical criteria, providing a basis for selecting the best alternative model. A series of computational experiments have been carried out using the two sets of actual statistical data from a Ukrainian bank. As a result of the performed computations, it was established that the best models in this application turned out to be nonlinear logit equations and Bayesian networks. In the future studies, we suppose to expand the number of model constructing techniques and to apply the idea of combining the estimates generated by the alternative models. Also, a specialized decision support system is to be constructed for the purpose of carrying research in the area of financial risks estimation and prediction.References
Matigorova I.Ju. Harakteristika osnovnyh podhodov k otsenke kreditnogo riska / I.Ju. Matigorova // Ekonomicheskaja nauka i praktika: materialy mezhdunar. nauch. konf. (g. Chita, fevral' 2012 g.). — Chita: Izd-vo "Molodoj uchenyj", 2012. — S. 68–69.
Siddiki N. Skoringovye karty dlja otsenki kreditnyh riskov / N.Siddiki. — M.: Izd-vo "Mann, Ivanov i Ferber", 2014. — 268 s.
Liu Y. The evaluation of classification models for credit scoring / Y. Liu. — Arbeitsbericht 02/2002. — Institut fur Wirtschaftsinformatik, 2002. — 19 p.
Kuzminchuk N.V. Metody otsenki kreditnogo riska v bankovskoj dejatel'nosti / N.V. Kuzminchuk, O.S. Mandryka // Biznesinform, 2009. — No. 1. — S. 113–117.
Bielecki T.R. Credit Risk: modeling, valuation and hedging / T.R. Bielecki, M. Rutkowski. — Berlin: Springer, 2002. — 500 p.
Hosmer D.W. Applied Logistic Regression / D.W. Hosmer, S. Lemeshow. — New York: John Wiley & Sons, Inc. 1989. — 400 p.
Bidjuk P.I. Analiz chasovykh rjadiv / P.I. Bidjuk , V.D. Romanenko, O.L. Tymoshchuk. — K.: Politekhnika, 2013. — 600 s.
Bidjuk P.I. Systemnyj pidkhid do prohnozuvannja na osnovi modelej chasovykh rjadiv / P.I. Bidjuk // Systemni doslidzhennja ta informatsijni tekhnolohiyi. — 2003. — No. 3. — S. 88–110.
Dovhyj S.O. SPPR na osnovi jmovirnisno-statystychnykh metodiv / S.O. Dovhyj, O.M. Trofymchuk. — K.: Lohos, 2014. — 430 s.