Importance Customer solvency can be assessed with machine learning methods. To keep record of new customers, whose data may change over time significantly due to socio-economic conditions, lending terms and their own characteristics, it is very relevant to update forecasting models and their adaptation to new data, thus generating more accurate forecasts and more sound decisions on loans. Objectives The research pursues creating a method to update customer solvency forecasting models and adapt them in line with new data on customers, thus ensuring accurate and adequate forecasts. Methods We used machine learning methods, which aggregate various classifiers on the basis of the neural network, logistic regression, discriminate analysis, na?ve Bayes classifier, support vector method, etc. Illustrating logistic regression, we propose a procedure for adjusting coefficients on the basis of the adaptive pseudogradient method in line with data on new customers. Results Relying upon the proposed procedure, we created a method to update forecasting models, with its efficiency proven with data on the Russian borrowers. Conclusions and Relevance The use of a numerical technique for updating models, which is based on the pseudogradient procedure, allows for adaptation to data on new customers and helps keep on track of changes in the lending market, for purposes of the model in place. The updated model forms more accurate forecasts, thus allowing to take more sound lending decisions.
Aivazyan S.A., Bukhshtaber V.M., Enyukov I.S., Meshalkin L.D. Prikladnaya statistika. Klassifikatsiya i snizhenie razmernosti [Applied statistics. Classification and reduction of dimension]. Moscow, Finansy i statistika Publ., 1989, 607 p.
Vagin V.N., Golovina E.Yu., Zagoryanskaya A.A., Fomina M.V. Dostovernyi i pravdopodobnyi vyvod v intellektual'nykh sistemakh [Reliable and plausible inference in intellectual systems]. Moscow, Fizmatlit Publ., 2004, 704 p.
Vapnik V.N. Vosstanovlenie zavisimostei po empiricheskim dannym [Recovery of dependencies with empirical data]. Moscow, Nauka Publ., 1979, 448 p.
Vasil'ev N.P., Egorov A.A. Opyt rascheta parametrov logisticheskoi regressii metodom N'yutona-Rafsona dlya otsenki zimostoikosti rastenii [Experience in calculating logistic regression parameters with the Newton–Raphson method to evaluate cold resistance of plants]. Matematicheskaya biologiya i bioinformatika = Mathematical Biology and Bioinformatics, 2011, vol. 6, no. 2, pp. 190–199.
Draper N., Smith H. Prikladnoi regressionnyi analiz. Mnozhestvennaya regressiya [Applied Regression Analysis]. Moscow, Dialektika Publ., 2007, 912 p.
Klyachkin V.N., Dontsova Yu.S. Sravnitel'nyi analiz tochnosti nelineinykh modelei pri prognozirovanii sostoyaniya sistemy na osnove markovskoi tsepi [A comparative analysis of the accuracy of non-linear models in forecasting the system's condition on the basis of the Markov chain]. Izvestiya Samarskogo nauchnogo tsentra Rossiiskoi akademii nauk = Proceedings of Samara Scientific Centre of Russian Academy of Sciences, 2013, vol. 15, no. 4, pp. 924–927.
Sorokin A.S. [Building a scorecard using the logistic regression model]. Internet-zhurnal NAUKOVEDENIE, 2014, no. 2. (In Russ.) Available at: Link.
Wasserman P. Neirokomp'yuternaya tekhnika. Teoriya i praktika [Neural Computing: Theory and Practice]. Moscow, Mir Publ., 1992, 184 p.
Yakupov A.I. Primenenie derev'ev reshenii dlya modelirovaniya kreditosposobnosti klientov kommercheskogo banka [Applying decision trees for modeling the solvency of the commercial bank's customers]. Iskusstvennyi intellekt = Artificial Intelligence, 2008, no. 4, pp. 208–213.
Klyachkin V.N., Shunina Yu.S. Sistema otsenki kreditosposobnosti zaemshchikov i prognozirovaniya vozvrata kreditov [The system for customer solvency assessment and loan repayment forecast]. Vestnik komp'yuternykh i informatsionnykh tekhnologii = Herald of Computer and Information Technologies, 2015, no. 11, pp. 45–51.
Lukashin Yu.P. Adaptivnye metody kratkosrochnogo prognozirovaniya vremennykh ryadov [Adaptive methods for short-term forecast of time series]. Moscow, Finansy i statistika Publ., 2003, 416 p.
Box G., Jenkins G.M. Analiz vremennykh ryadov. Prognoz i upravlenie [Time Series Analysis. Forecasting and Control]. Moscow, Mir Publ., 1974, 242 p.
Valeev S.G. Regressionnoe modelirovanie pri obrabotke nablyudenii [Regression modeling as part of observation data processing]. Moscow, Nauka Publ., 1991, 272 p.
Lewis C.D. Metody prognozirovaniya ekonomicheskikh pokazatelei [Industrial and Business Forecasting Methods]. Moscow, Finansy i statistika Publ., 1986, 133 p.
Merkov A.B. Raspoznavanie obrazov: vvedenie v metody statisticheskogo obucheniya [Image recognition: an introduction to statistical learning methods]. Moscow, URSS Publ., 2010, 254 p.
Montgomery D.C. A Note on Forecasting with Adaptive Filtering. Operational Research Quarterly, 1977, vol. 28, no. 11, pp. 87–91.
Polyak B.T., Tsypkin Ya.Z. Psevdogradientnye algoritmy adaptatsii i obucheniya [Pseudogradient algorithms of adaptation and learning]. Avtomatika i telemekhanika = Automation and Remote Control, 1973, no. 3, pp. 45–68.
Vasil'ev K.K., Krasheninnikov V.R. Statisticheskii analiz izobrazhenii [Statistical analysis of images]. Ulyanovsk, Ulyanovsk State Technical University Publ., 2014, 214 p.
