Prediction of parameters of socioeconomic development of the region using standard neural networks (on the example of GRP of the Voronezh region)

Abstract

Purpose : study of possibilities of application of neural networks apparatus for prediction of parameters of regional socio-economic systems in the process of formation of strategies of administrative-territorial entities of Russia. Discussion : the use of mathematical methods to predict the socioeconomic development of regions is widely discussed in the scientific literature. In practice, however, their application is very limited due to the high complexity and labour intensity of forming models aimed at a specific task. The replication of such systems, which allowed the use of a mathematical apparatus ready for use, was difficult because of the specificity of each administrative and territorial entity. The dissemination of these methods, including neural networks, can be of practical importance when using standard computer program packages. Results : the article demonstrates the possibilities and limitations of forecasting indicators of social and economic development of the regions for the period up to 2024 on the example of GRP of the Voronezh region. It has been established that the forecast of the neural network, formed on the basis of its training for the retrospective period, can be considered adequate to the real economic processes in the region. Private forecast errors occur, but the general conclusion about the possibility of using neural networks in the tasks of predicting the dynamics of social and economic development indicators of the region is positive. For GRP, the forecast of neural networks is more "cautious" than that of experts, and neither method can be preferred a priori. In this connection, the use of a prediction method using a neural network apparatus as an auxiliary tool for generating a long-term prediction within the limits adopted for the use of most economic-statistical methods (1/3 of the period for which the original data are generated) can be considered justified.