Forecasting Economic Development of the Donetsk People’s Republic up to 2040 by Using Neural Net Algorithms

The article studies the approach to long-term forecasting of the economic development of the Donetsk People’s Republic (DPR) in the Russian Federation for the period up to 2040 by using neural net algorithms. The research is quite acute, which is stipulated by the necessity to build DPR and the city of Donetsk economy into unified economic space of the Russian Federation and by limited volume of statistic data of the region, by high uncertainty of the external environment and the goal of shaping well-grounded land-marks of strategic planning of social and economic development. The hybrid generative model combining Conditional Variational Autoencoder (CVAE) and generative-competitive net Wasserstein GAN (WGAN) supplemented by neural net mechanism of imputing time series gaps (TimeVAE) was put forward as a methodological basis of the research. The model was taught on multivariate time series of social and economic indicators of the Donetsk People’s Republic and regions of the Russian Federation comparable by structure and set of industries, such as Rostov, Sverdlovsk and Kemerovo regions. It can help take into account both common regularities of regional development and specific features of the territory being studied. The model was tested on selected regions and showed sustainable quality of forecasting for the period up to 2 years. On the basis of adjusted model the principle scenario of dynamics of gross regional product and key macro-economic indicators in DPR up to 2040 was developed. The obtained figures demonstrate the potential of restoration growth of regional economy with gradual transition to moderate growth rates in the longterm. The research analyzes specificities of smooth trajectory of the forecast stipulated by probabilistic nature of the generative model, character of initial data and in-built economic restrictions. Benefits and restrictions of the proposed approach are being discussed, as well as opportunities of its use in the system of strategic planning of regional development.