A0413
Title: Machine learning forecasting of industrial production in Slovakia
Authors: Adam Csapai - The Institute of Economic Research of Slovak Academy of Sciences and University of Economics in Bratislava (Slovakia) [presenting]
Abstract: An extensive evaluation applies state-of-the-art machine learning models to the task of forecasting industrial production in Slovakia, demonstrating their superior ability to capture the evolution of a small, open, industrialized economy operating within a monetary union. The purpose is to represent the first systematic assessment of machine learning forecasting performance in such an environment characterized by short time series and two crisis episodes, in contrast to prior work on larger economies with decades of data. A key finding is that regularization serves as a more effective dimension-reduction technique than principal component analysis or factor models, preserving essential information and yielding more accurate predictions. Forecast combination methods that weight models according to both error magnitude and directional accuracy are also applied, and the directional accuracy of machine learning methods is explored. The robustness of these techniques is tested across pre- and post-COVID-19 periods to evaluate model performance under heightened volatility and uncertainty. Finally, an analysis of soft indicators as standalone inputs highlights their limited predictive value, and earlier critiques are also addressed by underscoring the practical advantages of machine learning approaches in macroeconomic forecasting.
