Neural Network Forecasting of Target Indicators of the Regional Socio-Economic Development

Abstract:

Testing the possibilities of using the standard apparatus of neural networks for predicting the indicators of socio-economic development of regions. The used methodology is based on the standard apparatus of neural networks included in the Statistica software products. Correlation-regression analysis was used as an alternative. The use of the apparatus of neural networks is due to the practical need to supplement the socio-economic processes used to predict the socio-economic processes in the territorial and functional subsystems of countries and macro regions. Expansion of the range of interconnections between systems of various levels required the expansion of methods for predicting the parameters of socio-economic processes. Results. Approbation of the standard apparatus of neural networks was carried out on the example of predicting the dynamics of one of the most significant indicators of socio-economic development of meso-level systems - the total fertility rate. The use of a fairly simple version of the apparatus of neural networks made it possible to establish a fundamentally different version of the dynamics of the predicted indicator, which differs from the expert and "trend" one. At the time of the calculations, the growth of the indicator values in the strategic planning documents of the model region and based on the analysis of the previous trend over a fifteen-year period was predicted. Neural network analysis made it possible to predict a decrease in the birth rate in the medium and long term as one of the possible options. In other words, a pessimistic variant of the development of the demographic situation in the region is not excluded.