V. S. Brezgin. Application of Neural Network Methods and Models in the Analysis of Short-Term Economic Dynamics of Regional Development

https://doi.org/10.15507/2413-1407.26343.498-518
EDN: https://elibrary.ru/jxusgz ISSN 2413-1407 
УДК / UDC 33:004.032.26:353 ISSN 2587-8549

Abstract

Introduction. In the face of structural shocks, traditional econometric methods lose their effectiveness in forecasting regional economies, especially when working with short time series. The objective of this study is to develop a methodology for analyzing and forecasting short-term regional dynamics based on neural network modeling, adapted to the specifics of regional statistics.

Materials and Methods. The empirical base consisted of annual Rosstat data for the Trans-Baikal Territory for 2003–2022, representative of resource-rich territories of the Russian Federation. Recurrent neural networks with LSTM architecture were used. The author's approach includes assessing the systemic signi­ficance of indicators using a composite index and comparing training sample generation strategies (sliding and expanding windows).

Results. The differential effectiveness of forecasting methods for different types of indicators was revealed. The expanding window method demonstrated high accuracy for indicators with stable trends (gross regional product forecast error of approximately 0.3 % in 2020–2021), while the sliding window method was more effective for volatile indicators. The developed toolkit for analyzing the systemic significance of indicators revealed a stable cluster structure of the regional economy, including production-investment and socio-demographic clusters. Verification of the models confirmed their resilience to typical fluctuations but demonstrated fundamental limitations under the unprecedented geopolitical shocks of 2022.

Conclusion. The practical significance of this study lies in the development of a comprehensive toolkit for improving the effectiveness of regional economic planning and forecasting in the face of structural change. The material may be useful in the activities of regional government and strategic planning bodies.

Keywords: regional economy, economic forecasting, neural networks, LSTM-models, short time series, structural shifts, systemic significance, Trans-Baikal Territory

Conflict of interest. The author declares no conflict of interest.

Funding. The work was carried out within the framework of the state assignment of the Institute of Natural Resources, Ecology and Cryology of the Siberian Branch of the Russian Academy of Sciences, project “Adaptation of Regional socio-ecological-economic systems of the border territories of the Russian East to geopolitical and climatic risks: Strategic approaches and mechanisms” (no. 126020216343-0).

For citation: Brezgin V.S. Application of Neural Network Methods and Models in the Analysis of Short-Term Economic Dynamics of Regional Development. Russian Journal of Regional Studies. 2026;34(3):498–518. https://doi.org/10.15507/2413-1407.26343.498-518

REFERENCES

1. Gamukin V.V. Structural Change of Gross Regional Product in the Subjects of Ural Fe­deral District. Economy of Regions. 2017;13(2):410–421. (In Russ., abstract in Eng.) https://doi.org/10.17059/2017-2-7

2. Borodin S.N. Forecasting the Economic Development of the Region Based on Regression and Neural Network Models. Russian Economic Bulletin. 2024;7(1):290–300. (In Russ., abstract in Eng.) https://www.elibrary.ru/eqfigc

3. Lim B.,  Zohren S. Time-Series Forecasting with Deep Learning: A Survey. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences. 2021;379(2194):20200209. https://doi.org/10.1098/rsta.2020.0209

4. Vidmant O.S. Forecasting Financial Time Series with LSTM Recurrent Neural Networks. Obshchestvo: politika, ekonomika, pravo. 2018;(5):63–66. (In Russ., abstract in Eng.) https://doi.org/10.24158/pep.2018.5.12

5. Siami-Namini S., Tavakoli N., Namin A.S. A Comparison of ARIMA and LSTM in Forecasting Time Series. In: 2018 17th IEEE International Conference on Machine Learning and Applications (ICMLA). 2018. P. 1394–1401. https://doi.org/10.1109/ICMLA.2018.00227

6. Richardson A., van Florenstein Mulder T., Vehbi T. Nowcasting GDP Using Machine-Learning Algorithms: A Real-Time Assessment. International Journal of Forecasting. 2021;37(2):941–948. https://doi.org/10.1016/j.ijforecast.2020.10.005

7. Hamiane S., Ghanou Y., Khalifi H., Telmem M. Comparative Analysis of LSTM, ARIMA, and Hybrid Models for Forecasting Future GDP. Ingénierie des Systèmes d'Information. 2024;29(3):853–861. (In French) https://doi.org/10.18280/isi.290306

8. Matytsyn A.V. Application of LSTM Model to Prediction of CPI and Inflation Rate on the Example of Russia. Russian Journal of Innovation Economics. 2022;12(2):1241–1252. (In Russ., abstract in Eng.) https://doi.org/10.18334/vinec.12.2.114845

9. Abbasimehr H., Shabani M., Yousefi M. An Optimized Model Using LSTM Network for Demand Forecasting. Computers and Industrial Engineering. 2020;143:106435. https://doi.org/10.1016/j.cie.2020.106435

10. Letyagina E.N., Perova V.I. Neural Network Modeling of Regional Innovation Ecosystems. Journal of New Economy. 2021;22(1):71–89. (In Russ., abstract in Eng.) https://doi.org/10.29141/2658-5081-2021-22-1-4

11. Treschevsky Yu.I., Gladkikh M.O., Opoуkova E.A., Papina O.N. Methodological Approach to Prognosis оf Target-High Parameters оf Regions Socio-Economic Development Using the Standard Neural Network Apparatus. Region: Systems, Economics, Management. 2020;(4):114–126. (In Russ., abstract in Eng.) https://doi.org/10.22394/1997-4469-2020-51-4-114-126

12. Feng Y., Zhang Y., Wang Y. Out-of-Sample Volatility Prediction: Rolling Window, Expanding Window, or Both? Journal of Forecasting. 2024;43(3):567–582. https://doi.org/10.1002/for.3046

13. Hall S.G., Tavlas G.S., Wang Y., Gefang D. Inflation Forecasting with Rolling Windows: An Appraisal. Journal of Forecasting. 2024;43(4):827–851. https://doi.org/10.1002/for.3059

14. Inoue A., Jin L., Rossi B. Rolling Window Selection for Out-of-Sample Forecas­ting with Time-Varying Parameters. Journal of Econometrics. 2017;196(1):55–67. https://doi.org/10.1016/j.jeconom.2016.03.006

About the author:

Vyacheslav S. Brezgin, Cand.Sci. (Econ.), Researcher, Laboratory of Ecological and Economic Research, Institute of Natural Resources, Ecology and Cryology, Siberian Branch of the Russian Academy of Sciences (16a Nedorezova St., Chita 672014, Russian Federation), ORCID: https://orcid.org/0000-0001-9008-4540, Researcher ID: R-5652-2016, SPIN-code: 2984-6475, monmanage@bk.ru

Availability of data and materials. The datasets used and/or analyzed during the current study are available from the author on reasonable request.

The author has read and approved the final manuscript.

Submitted 21.02.2025; revised 15.07.2025; accepted 30.07.2025.