Seeing in the Dark: A Machine-Learning Approach to Nowcasting in Lebanon
IMF Working Papers, March 8, 2016
Source details
- Canonical URL
- Seeing in the Dark: A Machine-Learning Approach to Nowcasting in Lebanon
Other formats
Bibliographic details
- Authors: Andrew J Tiffin
- Published: March 8, 2016
- Series: IMF Working Papers
- DOI: https://doi.org/10.5089/9781513568089.001
Summary
- Macroeconomic analysis in Lebanon faces long delays in the publication of GDP data, forcing reliance on proxy variables and creating an extended “nowcasting” challenge.
- The paper explores recent techniques from the machine learning literature to address this problem, while being mindful of pitfalls when extracting information from a large number of correlated proxies.
- Focus is placed on two popular techniques: Elastic Net regression and Random Forests.
- The paper provides an estimation procedure described as intuitively familiar and well suited to the challenging features of Lebanon’s data.
Methods and Technical Focus
- Techniques examined:
- Elastic Net regression
- Random Forests
- Related statistical and machine-learning concepts emphasized:
- LASSO
- ridge regression
- regression tree
- Ensemble methods
- Cross Validation
- Variable Selection
- Statistical Learning
Subject Areas and Keywords
- Subject: Cyclical indicators, Economic forecasting, Economic growth, Machine learning, Technology
- Keywords: coefficient estimate, Cross Validation, Cyclical indicators, Elastic Net, Ensemble, GDP, GDP data, GDP growth, GDP movement, GDP release, LASSO, Lebanon, Machine learning, machine-learning technique, Macroeconomic Forecasts, Nowcasting, Random Forests, regression tree, ridge regression, Statistical Learning, Variable Selection, WP