딥러닝을 활용한 실시간 GDP 예측 모형 개발
A Study on Real-time Forecasting GDP and GRDP Using Deep Learning
0
조회
0
다운로드
2
인용
0
참고문헌
키워드
초록
본 연구는 국내 경제 전망에 대한 실시간 예측 모형을 개발하는 것을 목적으로 한다. 또한 예측의 적시성과 예측의 정확성을 제고시키기 위해 비정형 데이터인 경제 뉴스를 활용하는 텍스트 마이닝과 기계학습 방법론을 활용한다. 특히, 신경망분석 중 하나인 장단기기억(LSTM: Long Short Term Memory) 알고리즘과 게이트순환유닛(GRU: Gated Recurrent Unit) 알고리즘을 활용했으며 예측의 요인별 구조를 파악하기 위해 자기회귀 혼합데이터샘플링모형(AR-MIDAS: AutoRegressive Mixed Data Sampling Model)을 분석 방법에 추가했고, 비정형 데이터인 경제 뉴스를 정량화하기 위해 TF 가중 방법인 TF-IDF 방법론을 활용했다. 분석 결과, LSTM은 예측력 측면에서 우월한 것으로 나타났지만 신경망 분석의 단점인 요인 분석이 블랙박스로 처리되는 단점이 있는 것으로 나타났다. 반면, AR-MIDAS 방법으로 추정한 결과, t기 전산업 생산지수가 상승하고 t-1기 실업률이 하락할수록 국내 경기는 양호해지는 것으로 나타났다. 또한 첫 번째 달의 경제 뉴스는 경제에 부정적인 영향을 미치는 반면 두 번째 달의 경제 뉴스는 경제에 긍정적인 영향을 미치는 것으로 나타났다. 이와 같이 본 연구는 GDP 예측의 적시성을 제고시키기 위해 GDP nowcasting 모형보다 예측 주기를 단축시켰으며 갑작스런 이벤트 발생 시 이를 GDP 예측에 즉각적으로 반영하기 위해 온라인 상의 경제 뉴스 데이터를 활용했다.
Abstract
This paper is trying to develop a real-time forecast model for the gross domestic product. And text mining and machine learning that utilize economic news, which is unstructured data are used to enhance the timeliness of predictions and the accuracy of predictions. In particular, the long short term memory(LSTM) algorithm and the gated recurrent unit(GRU) algorithm, which are one of the neural network analyses, are used and an autoregressive mixed data sampling model is added to explain the structure of each factor. And the TF-IDF methodology is used to quantify economics news which is unstructured data. As a result of the analysis, LSTM is superior in terms of predictive power, but it has a disadvantage that it is difficult to understand the factors because they are treated as a black box. By the way, as a result of estimating by the AR-MIDAS method, it is found that as the industrial production index at t increases and the unemployment rate decreases at t-1, the gross domestic product at t improves. It is also found that economic news in the first month of the quarter had a negative impact on the economy, while economic news in the second month had a positive impact on quarterly GDP. As such, this study shortened the prediction cycle compared to the GDP nowcasting model to improve the timeliness of GDP prediction, and used online economic news data to immediately reflect it in GDP prediction in the event of a sudden event.
참고문헌 (42)
- 월별지표를 이용한 GDP추정을 위한 계량모형 구축, "김덕파", 국회예산정책처, ~ (2020)
- 김치호, "GDP 성장률의 Nowcasting에 관한 연구", 한국은행 국민계정리뷰, 2, pp.1~22 (2016)
- 박강우, "GDP 예측력 제고를 위한 동태요인모형 구축", 한국은행조사통계월보, 2009(6), pp.27~60 (2009)
- 이동수, "전산업생산지수를 이용한 분기 성장률 예측", 통계연구, 18(2), pp.66~88 (2013), 통계청
- 이승윤, "기계학습과 혼합주기 모형을 활용한 단기전망시스템", 한국은행 조사통계월보, 2017(5), pp.16~41 (2017)
- 디지털 신기술을 이용한 실시간 당분기 경제전망(GDP nowcasting)시스템 개발, "이현창", BOK, ~, pp.이슈노트, 2022-7 (2022)
- 임상수, "텍스트 마이닝을 활용한 실질 국내총생산의 실시간 예측", 기업과혁신연구, 44(4), pp.91~106 (2021), 지식경영연구원[DOI]
- 최광신, "K-SuperCast: 빅데이터를 이용한 GDP 성장률 적시예측 시스템", 한국데이터정보과학회지, 30(4), pp.723~743 (2019), 한국데이터정보과학회[DOI]
- 최병재, "월별자료를 이용한 예측조합에 의한 GDP 전망", 한국은행조사통계월보, 10, pp.16~48 (2014)
- 형남원, "혼합자료 샘플링 모형을 통한 GDP 예측", 국제경제연구, 22(2), pp.83~117 (2016), 한국국제경제학회[DOI]
- 대규모 혼합주기 자료를 이용한 단기 경제전망 모형 구축, "황선웅", 산2업연구원, ~, pp.연구보고서 2012-642 (2012)
- 황종률, "월간경제지표를 이용한 실질GDP성장률 예측", 예산정책연구, 1(2), pp.103~133 (2012), 국회예산정책처[DOI]
- Nowcasting Norwegian GDP: The Role of Asset Prices Ain a Small Open Economy, "Aastveit, K.", Norges Bank, ~, pp.working paper 2007 (2007), 09
- Nowcasting New Zealand GDP using machine learning algorithms, "PAdam Richardson", oCAMA, ~, pp.Working Papers 2018-47 (2018)
- Andreou, E., "Regression Models with Mixed Sampling EFrequencies", Journal of Econometrics, 158, pp.246~261 (2010)
- Short-term Forecasting of GDP Using Large Monthly Datasets: A Pseudo Realtime Forecast Evaluation Exercise, "Barhoumi, K.", European Central Bank, ~, pp.Occasional paper series No. 84 (2008)
- Camacho, Maximo, "Short-term Forecasting for Empirical Economists: A Survey of the Recently Proposed Algorithms", Foundations and Trends(R) in Econometrics, 6(2), pp.101~161 (2013)
- Machine learning at central banks, "Chakraborty, Chiranjit", Bank of England, ~, pp.Bank of England working papers, 674 (2017)
- Chernis, Tony, "A dynamic factor model for nowcasting Canadian iGDP growth", Empirical Economics, 53(1), pp.217~234 (2017)
- Chris Bloor, "The use of statistical forecasting models at the Reserve Bank of New Zealand", Reserve Bank of New Zealand Bulletin, 72, pp.21~26 (2009), Reserve Bank of New Zealand
- Clements, M. P., "Macroeconomic forecasting with mixed-frequency data: Forecasting output growth in the United States", Journal of Business & Economic Statistics, 26(4), pp.546~554 (2008)
- Big data econometrics: Nowcasting and early estimates, "D. Buono", Universita Bocconi, ~, pp.Technical Report, 82 (2018)
- Diron, M., "Short-term forecasts of Euro area real GDP growth: An assessment of real-time performance based on vintage data", Journal of Forecasting, 27, pp.371~390 (2008)
- Eickmeier, Sandra, "Forecasting national activity using lots of international predictors: An application to New Zealand", International Journal of Forecasting, 27(2), pp.496~511 (2011), Elsevier
- Elena Andreou, "Should Macroeconomic Forecasters Use Daily Financial Data and How?", Journal of Business &Economic Statistics, 31(2), pp.240~251 (2013)
- Frank Schorfheide, "Real-Time Forecasting With a Mixed-Frequency VAR", Journal of Business & Economic Statistics, 33(3), pp.366~380 (2015)
- Foroni C, "A comparison of mixed frequency approaches for nowcasting Euro area macroeconomic aggregates", International Journal of Forecasting, 30, pp.554~568 (2014)
- Ghysels, E., "MIDAS regressions: Further results and new directions", Econometric reviews, 26(1), pp.53~90 (2007)
- Giannone, D., "A new coincident index of business cycles based on monthly and quarterly series", Journal of Applied Econometrics, 18, pp.427~443 (2017)
- Jansen, W. Jos, "Forecasting and nowcasting real GDP: Comparing statistical models and subjective forecasts", International Journal of Forecasting, 32(2), pp.411~436 (2016)
- GDPNow: A model for GDP nowcasting, "Higgins P.", Federal Reserve Bank of Atlanta, ~, pp.Working Paper, 2014-7 (2014)
- Hrishikesh Deka, "A Machine Learning Approach for Text and Document Mining", International Journal of Computer Science Engineering, 6(5), pp.115~123 (2017)
- Koop, Gary, "Bayesian Multivariate Time Series Methods for Empirical Macroeconomics", Foundations and Trends(R) in Econometrics, 3(4), pp.267~358 (2010)
- Kuzin, V., "Pooling versus Model selection for nowcasting GDP with many predictors: empirical evidence for six industrialized countries", Journal of Applied Econometrics, 28(3), pp.392~411 (2013)
- When are Google data useful to nowcast GDP? An approach via pre-selection and shrinkage, "Laurent Ferra", Center for Research in Economics and Statistics, ~, pp.Working Papers 2019-04 (2019)
- Paolo Fornaro, "Nowcasting Finnish Real Economic Activity: a Machine Learning Approach", Empirical Economics, 58(1), pp.55~71 (2020)
- Proietti, T., "Temporal disaggregation by state space methods :Dynamicregression methods revisited", Econometrics Journa, 9, pp.357~372 (2006)
- Golinelli, R., "Tracking world trade and GDP in real time", International Journal of Forecasting, 30(4), pp.847~862 (2014)
- Scott, S., "Bayesian variable selection for nowcasting economic time series" (2015), Economic Analysis of the Digital Economy
- Stock, R. S., "Dynamic factor models, Oxford handbook on economic forecasting" (2011), Oxford University Press
- Nowcasting Canadian Economic Activity in an Uncertain Environment, "Tony Chernis", Bank of Canada, ~, pp.Discussion Papers 18-9 (2018)
- Troy D. Matheson, "Factor Model Forecasts for New Zealand", International Journal of Central Banking, 2(2), pp.169~237 (2006)