 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
 |
|
page1
page2
page3
|
|
| LOGIC-PROBABILISTIC NUMERICAL VALUATION OF CREDIT RISK |
 |
| ||||||
| The new effective technique of logic-probabilistic numerical valuation of credit risk of consumer loan and company for commercial and state banks and appropriate software is offered. The technique permits to increase by 35 % the accuracy of distinguishing "good" or "bad" credits, essentially to reduce losses of banks, to lower the price for the credit, to increase competitiveness of the bank, to attract more customers. Western descriptions of credits in form of a set of characteristics which have gradation are used in the technique . Changes in the description of credits in view of the features and wishes of the bank are possible. At present western techniques are used for valuation of credit risk of natural and legal persons: quadratic discriminance analysis (QDA), linear discriminance analysis (LDA), the CART-method and neural network (NN). These techniques do not determine the numerical valuation risk of the risk but distinguish credits into good and bad . Share of faulty decisions is about 28 %. These techniques use mathematical apparatus inadequate to risk. Such main ideas for risk, as casual events, errors, probability, logic connections of events are not applied. These techniques are more or less successfully set-up "a black box" for classification of credit proposals. New technique of numerical valuation of credit risk as of the probability of credit unreturn is offered. The logic-probabilistic method of the risk valuation is used. Logic and probabilistic models of risk are built. The probabilistic model of risk is trained on statistical data of return of credits, that is, probabilities of nonsuccess of credits, connected with characteristics of the credit and their gradation are calculated. The optimum allowable credit risk is also calculated during training. The trained model of risk is used by bank for numerical valuation of risk of particular credit proposals and comparison with allowable risk. The technique is created with participation of German scientists and approved on the package 1000 credits. The technique has higher accuracy in classification of bad and good credits than that of the known Western techniques. Share of faulty decisions in classification of good and bad credits is 18 % instead by 28 % at existing techniques. The new technique permits: 1. To evaluate by the numerical risk of each credit, to establish how far as it is close to risks of other credits, allowable and average risk of the bank, to calculate the contributions of characteristics of the credit to risk; 2. To calculate the average probabilities and contributions of characteristics of credits of bank and to establish, which errors the bank does most often, to supervise and to operate by credit work of the bank, to nominate reasonable risk and price for the risk; 3. To increase competitiveness of the bank on base of decrease of losses from credit unreturn , reduction of the price for the credit and thus attracting a greater number of customers. Professor E.D.Solojentsev Tel.: (812) 217-31-24; Fax: (812) 217-86-14; E_mail: sol@sapr.ipme.ru ; WWW: http:/www.ipme.ru/ipme/labs/iisad |
| St.Petersburg | 199178 |
|
. |
| Visitors: 00371 |