동적요인모형과 딥러닝 알고리즘을 결합한 당분기 경제전망(GDP nowcasting) 시스템 개발
Dynamic Factor Model and Deep Learning Algorithm for GDP Nowcasting
0
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
코로나19 팬데믹, 지정학적 리스크, 기후 변화 등으로 경제 불확실성이 높아지면서 경기상황을 적시에 판단할 수 있는 지표에 대한 관심이 높아지고 있다. 그러나, 전반적인 경기상황을 나타내는 GDP 성장률의 경우 공표 주기와 시차가 길어 현재 경기상황을 판단하는 지표로 활용하는 데 한계가 있다. 이에 본 연구는 매주 GDP 성장률 예측치를 업데이트하며 경기흐름을 파악하는 실시간 당분기 경제전망(GDP nowcasting) 시스템을 개발하였다. 동 시스템은 주요국 중앙은행의 GDP nowcasting 전망모형으로 이용되는 동적요인모형(DFM, Dynamic Factor Model)과 최근 빠르게 발전한 기계학습(ML, Machine Learning) 알고리즘 가운데 시계열 데이터에 특화된 LSTM(Long Short-Term Memory)을 상호보완적으로 적용하여 예측력을 제고하였다. 2017년 1분기부터 2021년 3분기까지 기간에 대해 실시간 GDP nowcasting 시스템의 예측력을 평가한 결과, DFM과 LSTM 전망모형 모두 실제 경기흐름을 적절히 포착하는 것으로 나타났다. 특히, 2020년 1분기와 2분기의 급격한 GDP 성장률 하락, 2020년 3분기의 GDP 성장률 반등 움직임을 조기에 포착한 것으로 나타났다. 또한, LSTM 전망모형은 코로나19 유행과 같이 경제 불확실성이 확대된 상황에서 상대적으로 높은 예측력을 보였다. 이는 인공신경망 구조에 기반한 LSTM 알고리즘이 변수 간 비선형, 상호의존적 관계를 포착하는데 효과적이기 때문인 것으로 판단된다.
Abstract
As economic uncertainty increases due to the COVID-19 pandemic and climate change, indicators that can judge the economic situation in a timely manner are becoming more important. Although GDP growth is a representative indicator of the overall economic situation, there is a limit to quickly assessing the real-time economic situation due to publication lags and the low frequency nature. In this paper, we develop real- time GDP nowcasting system that updates the forecast every week. In this system, Long Short-Term Memory(LSTM), which is specialized in time series among deep learning algorithms, and Dynamic Factor Model(DFM) were complementarily used to enhance forecasting power. Both DFM and LSTM appropriately captured the actual economic situation. In particular, the rapid decline in the GDP growth rate in the 2020:Q1~Q2 due to the COVID-19 pandemic was detected early, and signs of a rebound in GDP growth in the 2020:Q3 was also quickly captured. In addition, the LSTM showed relatively high forecasting power in a situation where economic uncertainty has increased, such as the COVID-19 pandemic.
참고문헌 (50)
- 월별지표를 이용한 GDP추정을 위한 계량모형 구축, "김덕파", 국회예산정책처, ~ (2020)
- 김치호, "GDP 성장률의 Nowcasting 에 관한 연구", 국민계정리뷰, 2016(2) (2016), 한국은행[DOI]
- 박강우, "GDP 예측력 제고를 위한 동태요인모형 구축", 조사통계월보 (2009), 한국은행[DOI]
- 이동수, "전산업생산지수를 이용한 분기 성장률 예측", 통계연구, 18(2), pp.66~88 (2013), 통계청
- 이동진, "실물ꞏ금융 정보를 종합적으로 활용한 경기상황지수(BOKCOIN) 시산", 조사통계월보 (2017), 한국은행[DOI]
- 이승윤, "기계학습과 혼합주기 모형을 활용한 단기전망시스템", 조사통계월보, pp.16~41 (2017), 한국은행
- 최동규, "인공지능으로 설명한 환율(딥러닝과 그 설명모형을 이용한 단기 환율예측)", 한국은행 행내현상논문, ~ (2020)
- 최병재, "월별자료를 이용한 예측조합에 의한 GDP 전망", 한국은행 경제 전망보고서, pp.16~48 (2014)
- 대규모 혼합주기 자료를 이용한 단기 경제전망 모형 구축, "황선웅", 산업연구원, ~, pp.연구보고서, 2012-642 (2012)
- 형남원, "혼합자료 샘플링 모형을 통한 GDP 예측", 국제경제연구, 22(2), pp.83~117 (2016), 한국국제경제학회[DOI]
- "Recurrent Neural Networks cheatsheet" (2021)[URL]
- Anderson, Roy M., "How Will Country-based Mitigation Measures Inuence the Course of the Covid-19 Epidemic?", The Lancet, 395(10228), pp.931~934 (2020)
- Anesti, Nikoleta, "Peering into the Present: the Bank's Approach to GDP Nowcasting", Bank of England Quarterly Bulletin, 2017(Q2) (2017)[DOI]
- Nowcasting, "Bańbura, M.", ECB, ~, pp.ECB Working Paper, No. 1275 (2010)
- Bańbura, M., "Handbook of economic forecasting, Vol. 2", pp.195~237 (2013), Elsevier
- Bańbura, M., "Maximum Likelihood Estimation of Factor Models on Datasets with Arbitrary Pattern of Missing Data", Journal of Applied Econometrics, 29(1), pp.133~160 (2014)
- Bell, Venetia, "Nowcasting UK GDP growth", Bank of England Quarterly Bulletin, 2014(Q1), pp.1~ (2014)[DOI]
- Nowcasting Indian GDP growth using a Dynamic Factor Model, "Bhadury, Soumya", RBI, ~, pp.RBI Working Paper Series, No. 03 (2020)
- Bok, Brandyn, "Macroeconomic Nowcasting and Forecasting with Big Data", Annual Review of Economics, 10, pp.615~643 (2018)
- Chernis, Tony, "A Dynamic Factor Model for Nowcasting Canadian GDP Growth", Empirical Economics, 53(1), pp.217~234 (2017)
- Nowcasting Canadian Economic Activity in an Uncertain Environment, "Chernis, Tony", Bank of Canada, ~, pp.Bank of Canada Working Paper, No. 2018-9 (2018)
- Nowcasting Japanese GDPs, "Chikamatsu, Koysuke", Bank of Japan, ~, pp.Bank of Japan Working paper, No.18-E18 (2018)
- Nowcasting with Large Bayesian Vector Autoregressions, "Cimadomo, J.", ECB, ~, pp.ECB Working Paper, No. 20202453 (2020)
- Big data and Machine Learning in Central Banking, "Doerr, S.", BIS, ~, pp.BIS Working Papers, No. 930 (2021)
- Nowcasting GDP with a Large Factor Model Space, "Eraslan, Sercan", Deutsche Bundesbank, ~, pp.Deutsche Bundesbank Discussion Paper, No. 41 (2019), 2019
- A Survey of Econometric Methods for Mixed-frequency Data, "Foroni, Claudia", Norges Bank, ~, pp.Norges Bank Working Paper, No. 2013-06 (2013)
- Galvez-Soriano., "Nowcasting Mexico’s Quarterly GDP Using Factor Models and Bridge Equations", Estudios Economicos, 35(2), pp.213~265 (2020)
- Ghysels, E., "MIDAS Regressions : Further Results and New Directions", Econometric reviews, 26(1), pp.53~90 (2007)
- Giannone, Domenico, "Nowcasting : The Real-time Informational Content of Macroeconomic Data", Journal of Monetary Economics, 55(4), pp.665~676 (2008)
- Goulet Coulombe, P., "How Is Machine Learning Useful for Macroeconomic Forecasting?", arXiv: 2008-12447, ~ (2020)
- GDP Now: A Model for GDP 'Nowcasting', "Higgins, Patrick C.", Federal Reserve Bank, ~, pp.Federal Reserve Bank of Atlanta Working Paper, No. 2014-7 (2014)
- "H2O Documentation" (2021)[URL]
- Hochreiter, S., "Long short-term Memory", Neural computation, 9(8), pp.1735~1780 (1997)
- Economic Nowcasting with Long Short-Term Memory Artificial Neural Networks (LSTM), "Hopp, D.", UNCTAD, ~, pp.UNCTAD Research Paper, No. 62 (2021)
- Itkonen, Juha., "Nowcasting the Finnish economy with a Large Bayesian Vector Autoregressive Model", BoF Economics Review, 2017(6) (2017)
- An Algorithmic Crystal Ball:Forecasts-based on Machine Learning, "Jung, J. K.", IMF, ~ (2018)
- Kim, Hyun Hak, "Methods for Backcasting, Nowcasting and Forecasting Using Factor-MIDAS : With an Application to Korean GDP", Journal of Forecasting, 37(3), pp.281~302 (2018)
- Kose, M. A., "International Business Cycles:World, Region, and Country-specific Factors", Americal Economic Review, 93(4), pp.1216~1239 (2003)
- Kurz‐Kim, Jeong‐Ryeol, "A Note on the Predictive Power of Survey Data in Nowcasting Euro Area GDP", Journal of Forecasting, 38(6), pp.489~503 (2019)
- Liebermann, Joelle, "Real-Time Nowcasting of GDP : A Factor Model vs. Professional Forecasters", Oxford Bulletin of Economics and Statistics, 76(6), pp.783~811 (2014)
- Mariano, Roberto S., "A New Coincident Index of Business Cycles Based on Monthly and Quarterly Series", Journal of applied Econometrics, 18(4), pp.427~443 (2003)
- Mullainathan, S., "Machine Learning : an Applied Econometric Approach", Journal of Economic Perspectives, 31(2), pp.87~106 (2017)
- Ng, Andrew, "Recurrent Neural Networks", Lecture Note on Sequence Models, DeepLearning.AI and Coursera, ~ (2020)
- "Understanding LSTM Networks" (2015)[URL]
- Forecasting Canadian GDP Growth Using XGBoost, "Qureshi, S.", Carleton University, Department of Economics, ~, pp.Carleton Economic Papers, No. 20-14 (2020)
- Richardson, A., "Nowcasting GDP Using Machine-learning Algorithms : A Real-time Assessment", International Journal of Foreacsting, 37(2), pp.941~948 (2021)
- A Factor Model Approach to Nowcasting Danish GDP, "Schmith, Adrian Michael Bay", Danmarks National Bank, ~, pp.Economic memo, No. 3 (2021)
- MIDAS and Bridge Equations, "Schumacher, Christian", Deutsche Bundesbank, ~, pp.Deutsche Bundesbank Discussion paper, No. 26-2014 (2014)
- Schumacher, Christian, "A Comparison of MIDAS and Bridge Equations", International Journal of Forecasting, 32(2), pp.257~270 (2016)
- Waldrop, M. Mitchell, "News Feature: What Are The Limits of Deep Learning?", Proceedings of the National Academy of Sciences, 116(4), pp.1074~1077 (2019)