Abstract: The article deals with the problem of overdue credit debt and establishing effective management of problem debts in the financial institutions. The essence and methodological features of the scoring assessment of borrowers of commercial banks, including collection scoring, are considered.

The authors developed a mathematical model of collection scoring, based on a combination of tools of theories of fuzzy logic and artificial neural networks. The model has a hierarchical structure that takes into account the qualitative and quantitative variables which characterize the borrowers.

The feature of the model is attracting of Kohonen self-organizing maps for setting the parameters of membership functions in the process of quantitative variables fuzzyfication, as well as for the automatic construction of a knowledge base at processing of qualitative variables. For aggregation of linguistic variables at each level of the hierarchy the authors used compositional integration rule, which allows generating fuzzy knowledge base without getting expert opinions in terms of multi criterion and absence of comparison base for integral variables of top-level hierarchy.

The practical value of the constructed model of collection scoring is the ability to develop recommendations for working with each segment of the portfolio of overdue loans in accordance with the calculated level of credit risk.

Introduction to the work of financial institutions the credit risk assessment models on the basis of neuro-fuzzy techniques will have a positive impact on income from lending activities of commercial banks and contribute to the stability of the financial system as a whol