Evaluation of the sensitivity of neural network method for constructing the dynamic model of bankruptcy to identify signs of a developing process of the crisis of the Corporation
Gorbatkov Stanislav Anatol'evich
Doctor of Technical Sciences, Professor Mathematics and Computer Science
Russia, Financial University under the Government of the Russian Federation
Kasimova Liana Irikovna
graduate student of macroeconomic development and public administration
Russian, Bashkir State University
Farvaev Irik Rafitovich
graduate student of the department of electrical equipment and automation of industrial enterprises
Russia, Ufa State Oil Technical University
Pilot original neural network logistic dynamic method of building the model of bankruptcies. Investigated the convergence of the iterations of the restoration indicator bankruptcy on incomplete data, which makes the basic "core" of the proposed method and regularization model on Bayesian ensemble of neural networks. Compared with 22 known models of bankruptcies: MDA models, logit models, expert models, rating models, regulated by the method of the Government of the Russian Federation.
Dynamic neural network method, the model of an evolving bankruptcy, convergence, regularization, evaluation of the sensitivity of the method, the prediction of bankruptcy, comparison with known methods.
Full article text is available only in Russian.
Please select from the menu Russian language and continue reading.