Journal of
Italian
Banking
Association

year 100
 

Archive » May 2022 » Machine learning for credit risk: what role in regulatory models?

Machine learning for credit risk: what role in regulatory models?

Paolo Di Biasi, Rita Gnutti, Andrea Resti, Daniele Vergari
May 2022 - n. 5
Keywords: Machine learning, rischio di credito, banche, modelli regolamentari
Jel codes: G21, G28, G32, O33

The use of machine learning (Ml) for credit risk analysis has sparked a lively debate. Ml refers to a set of heterogeneous techniques, and one should not treat them all in the same way, since the most consolidated and stable approaches guarantee a high level of transparency and a low risk of overfitting. They can be used for regulatory capital calculation purposes, but may also serve broader purposes: reducing costs, shortening response times, strengthening credit monitoring, improving distribution channels and cementing customer relationships. Discouraging their use can put banks at a disadvantage vis à vis their most innovative competitors, making supervision less effective.

Interested in this paper?
Buy the issue