ОЦІНЮВАННЯ КРЕДИТОСПРОМОЖНОСТІ ПОЗИЧАЛЬНИКІВ КОМЕРЦІЙНОГО БАНКУ

Abstract

In the financial and economic crisis context, which is characterized by an increase in the number of insolvent business entities of the credit market, there are an increase in the share of doubtful loans in loan portfolios of the banks and the issue of studying the assessment of the creditworthiness of borrowers of a commercial bank becomes especially relevant. The article considered various approaches to the interpretation of the essence of the creditworthiness of borrowers and various methods for assessing the creditworthiness of potential customers of banks, such as regression models, neural networks, a classification tree, genetic algorithms, scoring cards, which are the main tools for data mining. Note that the different models can be applied at various stages of assessing a bank's credit risk. The analysis of the assessment of the creditworthiness of borrowers by banks gave grounds to propose a credit-scoring model. The model is presented in the form of a scoring card, which is based on the results of evaluating the logistic regression. Scoring maps are constructed on the assumption that «the past reflects the future». Accordingly, based on data on previously opened loans and analyzing the available information, it is possible to predict the result (behavior) of future borrowers. To create a scoring card, the following business process was considered: the manager of a partner store of the bank fills in the client's personal data, after which the form is sent to several banks for consideration. Banks review the application, obtain information from external sources, and make decisions. The bank, in turn, must quickly and efficiently assess the client and immediately indicate the agreed loan. The scoring model in the form of a scoring card is based on the results of evaluating the logistic regression in the R-Studio software package. The quality of the investigated model was checked by the area under the ROC-curve and the GINI index. According to the results of the study, we can conclude that this model can be included in the automatic decision-making process on the issuance of loans, which will allow banks to 1) reduce the time for a bank to decide to issue a loan; 2) be sure that the loan will be repaid by the borrower; 3) reduce the bank's credit risks.