USING A NEURAL NETWORK TO FORECAST PROSPECTS FOR THE DEVELOPMENT OF THE AGRO-INDUSTRIAL COMPLEX OF THE RUSSIAN FEDERATION
The article raises the problem of sustainable development of the agro-industrial complex of the Russian Federation. The overall positive assessments of experts regarding the development of the domestic agro-industrial complex under the influence of the sanctions effect may in fact be less optimistic in the long term, which makes it necessary to conduct additional monitoring of the current situation, including through the use of statistical methods of in-depth analysis in order to obtain reliable estimates for the subject under consideration. Based on the developed original approach to the formation of an integral indicator for assessing the state of the agro-industrial complex, it is proposed to forecast its changes by using neural networks. Step-by-step forecasting using neural networks helped to develop a test case for predicting future system responses based on previous behavior. This study proved the success of neural network methods as a self-sufficient tool for analyzing and predicting the dynamics of changes in the state of the agro-industrial complex of the Russian Federation and studying the sustainability of its development. At the same time, based on the proposed methodology for assessing the sustainability of the agro-industrial complex, it is possible to predict the values of variables that are reference indicators in the process of making various management decisions. In addition, the developed model of forecasting using a neural network can serve as a model for applying and forecasting other economic indicators.
Keywords: agro-industrial complex, a neural network model architecture, simulation modeling, neural network, forecasting, forecast scenarios, radial basis function, sanction restrictions, integral indicator of the agro-industrial complex state assessment, the sustainability of agro-industrial complex development.
Highlights:
♦ based on the results of the program calculation, the RBF architecture 38:38-8-1:1 was recognized as the best type of model, the structural scheme of this model is given;
♦ the implementation of forecast variants of events (optimistic, realistic, pessimistic) on the basis of the obtained model confirmed the general trend of recent years: the deterioration of the agroindustrial complex state of the Russian Federation and the weakening of its stability in the near future;
♦ forecasting methods based on neural networks, along with traditional methods, can be successfully used in forecasting systems of a wide range.
Roman A. Vakhrameev, Applicant of the Department of Statistics and Econometrics at Samara State University of Economics; Mikhail N. Tolmachev, Doctor of Economics, Associate Professor, Professor of the Department of Accounting, Analysis and Audit of the Financial University under the Government of the Russian Federation, Moscow; Vladimir N. Afanasiev, Doctor of Economics, Professor, Head of the Department of Statistics and Econometrics of Orenburg State University.