Credit Risk Assessment with Stacked Machine Learning

8 January 2026

Banca d'Italia today publishes 'Credit Risk Assessment with Stacked Machine Learning', the new issue of the series 'Markets, infrastructures, payment systems'.

Banca d'Italia's In-house Credit Assessment System (ICAS) for Italian non-financial corporations, used in the Eurosystem's collateral framework for monetary policy implementation, consists of a statistical model (S-ICAS) and of the analysts' evaluation. This paper compares the performance of S-ICAS with that of artificial intelligence, specifically of machine learning (ML) and deep learning models. The findings suggest that deep learning improves discriminative power; decision tree ensembles yield a further improvement, as does a meta-model that stacks random forests, extreme gradient boosting, and deep learning models. Applying eXplainable Artificial Intelligence (XAI) techniques to the metamodel predictions, this paper shows that XAI can support analysts in understanding the key factors behind the differences between ML and S-ICAS predictions, thus helping refine their assessment. While interpretability issues prevent ML-based models from being a full alternative to traditional models, XAI allows for their integration within the overall credit assessment process, thus increasing its effectiveness.