Template-type: ReDIF-Article 1.0
Author-Name: Oleg Kryzhanovskiy
Author-Email: kryzhanovskiyoa@mail.cbr.ru
Author-Workplace-Name: Bank of Russia; Tyumen State University
Author-Name: Anastasia Mogilat
Author-Email: mogilatan@cbr.ru
Author-Workplace-Name: Bank of Russia
Author-Name: Zhanna Shuvalova
Author-Email: shuvalovazhd@cbr.ru
Author-Workplace-Name: Bank of Russia
Author-Name: Dmitry Gvozdev
Author-Email: dgvozdev@nes.ru
Author-Workplace-Name: HSE University
Title: Using LSTM Neural Networks for Nowcasting and Forecasting GVA of Industrial Sectors
Abstract: This paper evaluates the potential application of long short-term memory (LSTM) neural networks for economic forecasting. We compare the accuracy of short-term forecasts of the gross value added of industrial sectors obtained using an LSTM model against several benchmarks, such as a random walk model, an autoregressive integrated moving average model, and an approximate dynamic factor model. Compared to the other models, the LSTM model demonstrates a lower mean absolute forecast error in 16 out of 18 cases and a lower root mean square error in 13 out of 18 cases.
Classification-JEL: C45, C53, C82, E17, L60
Keywords: GDP, GVA, neural networks, long short-term memory network, nowcasting, forecasting
Journal: Russian Journal of Money and Finance
Pages: 93-104
Volume: 84
Issue: 1
Year: 2025
Month: March
DOI:
File-URL: https://rjmf.econs.online/upload/iblock/089/4rgxi7zeruwecj0e00eykrvkxcz0glol/Using-LSTM-Neural-Networks-Forecasting-GVA-Industrial-Sectors.pdf
Handle: RePEc:bkr:journl:v:84:y:2025:i:1:p:93-104