Legal entities creditworthiness modeling using discriminant analysis and neural networks

The aim of the article is to develop effective economic and mathematical methods and models of assessment of legal entities creditworthiness, aimed at reducing of credit risk of borrowers of banking institutions. There are determined the conceptual aspects of modeling of legal entity creditworthiness on the basis of the system of financial ratios. The article proves that it is enough to use such financial ratios as instant and overall liquidity, ratio of accounts receivable and payable, return on sales, payable turnover, autonomy and solvency ratio in order to assess the creditworthiness of domestic enterprises. The synthesis of several approaches to the assessment of creditworthiness of a borrower, based on the use of methods of discriminant analysis, classification functions and neural networks of perceptron type, as well as on radial basis functions is developed in the article. For the purpose of assessment of creditworthiness the set of legal entities is divided into 4 classes: companies with high, satisfactory, low and unsatisfactory creditworthiness. If the creditworthiness is high it’s recommended to bank to grant loan, if it is satisfactory ? to lend provided that the loan is secured by liquid collateral, if it is low or unsatisfactory ? to deny loan. Testing of the developed models proved the effectiveness of the proposed approach. So, built in research the economic and mathematical models allow to improve essentially the accuracy of assessing the creditworthiness of potential borrowers and minimize credit risk of the banking institution.