НЕЙРОСЕТЕВАЯ СИСТЕМА ОЦЕНКИ ВЕРОЯТНОСТИ БАНКРОТСТВА БАНКОВ
Object of this research is the Russian banking system. The work purpose – creation of the comput-er program of an assessment of probability of bankruptcies of banks because of revocation of li-cense of banks and use of this system as mathematical model for detection of some regularities of the Russian bank sphere. The instrument of researches – the neural networks trained and tested on materials of financial statements of the Central Bank of the Russian Federation. After detection and removal of emissions the error of testing (generalization) of the trained and optimized of neu-ral network made 6,3% of statistical data. Researches of subject domain were carried out by carry-ing out virtual computer experiments during which calculations by means of a neural network were made at change of one of fifteen input parameters characterizing banks while other parame-ters remained the invariable. A number of regularities of studied subject domain is as a result was revealed. The conclusion was that increase of coefficient of long-term liquidity positively influ-ences for bank activity, however, there are a certain level when increase of this indicator increases probability of bankruptcy of bank. The organizational and legal form of bank, and also the status of the city in which the bank is located has essential impact on success of its functioning. Howev-er this influence is ambiguous and in each case can be shown differently, depending on a set of other parameters of bank and its activity. The example of use of the developed neuronetwork sys-tem for development of recommendations about decrease in probability of bankruptcy of one of banks is given. The created intellectual system of forecasting of probability of bankruptcy of banks can be used for an assessment of risks of the interbank credits, for carrying out internal au-dit, and also for support of the decision-making, directed on improvement of activity of banks.