We are in social networks:
Vkontakte Facebook Twitter

Integration of models of logit-regressions as a committee of experts to assess the creditworthiness of borrower

Svitlana Savina, Vladyslav Ben

The research is dedicated to the formation of economic-mathematical models of creditworthiness assessment of borrower-individual. To resolve this task different mathematical set of tools is applied, in particular neuronets of multilayer perceptron type and logistic regressions. In this work new methodological approach to the selection of most important factors is proposed, on the basis of which new models of creditworthiness assessment of individuals are formed. This approach is based on synthesis of probabilistic neuronet and genetic algorithm. Also, methodological approach to realization of “experts committee” due to merging of several logit-regression models of creditworthiness assessment of borrower-individual is developed. The comparative experimental analysis of effectiveness of implementation of one extended logit-regression and three logit-regression models (experts committee) is conducted, parameters of which were evaluated separately, based on data set formed according to distinguished qualitative criteria. Testing of formed economic-mathematical models confirmed the high level of their adequacy, that affirms about the ability on their basis to increase essentially the accuracy of creditworthiness assessment of potential borrower and minimize credit risk level of bank institution. As a result of implementation of modeling experiments the conclusion is received concerning the expediency of using neuronet of multilayer perceptron type in case of necessity of borrower division into two classes – reliable or unreliable; and if the research anticipates the necessity higher particularization in evaluation of creditworthiness, for instance, determination of risk level of giving credit to a particular person, then the expert committee (combination of logit-regressions) should be applied. The investigation results can be used by bank institutions and all organizations that might be interested in adequate procedure of creditworthiness assessment of borrower-individuals.

Keywords. Assessing the creditworthiness, an individual borrower, qualitative and quantitative factors of credit risk, neural network, logit-regression, the committee of experts.

Download publication