Luca Attolico

KEYWORDS: Macroeconomic Nowcasting; Machine Learning; Uncertainty Quantification; Explainable AI (XAI); Dynamic Model Combination

Abstract

Effective policymaking requires timely assessment. This research investigates a generalizable GDP growth nowcasting framework, using Singapore’s highly volatile and open economy as a stress-test to validate robustness for broader contexts. The study implements a rigorous recursive pipeline combining penalized regressions, dimensionality reduction, ensemble learners, and neural networks. It targets the "black box" limitation by generating operational prediction intervals via block-bootstrap and analyzing model-aware feature importance, incorporating Explainable AI. Integrating the Model Confidence Set and dynamic aggregations, the research aims to verify that this adaptive architecture significantly outperforms a three-level econometric benchmark (Random Walk, AR(3), Dynamic Factor Model).

• EMAIL: luca.attolico@gmail.com
• LINKEDIN PROFILE
• ORCID: 0000-0002-8649-5440