On 11 November 2021, EBA published her ‘discussion paper on machine learning (ML) for internal ratings-based (IRB) models’. The aim of this discussion paper is to gain more insight in how market participants experience the challenges and opportunities of ML for regulatory capital models. Until now, most regulatory capital models rely on the same techniques that were used when the first regulations were published. Leveraging on the exponential increase in data availability and storing capacity coupled with the improvements in computing power, ML models are becoming more common in many areas. In terms of risk management, they are mainly used for non-regulatory applications (for example in loan approvals, early warning or as challenger models). This discussion paper is a first step in creating future regulatory guidelines regarding the use of ML for models with a regulatory application.
At Zanders we follow the developments in ML with great interest since we see the added value in many cases (if applied appropriately). See our publication for our insights on machine learning in risk management and the machine learning workshops we offer. In case you are interested in a ML workshop or to discuss what ML can offer your organization, please contact John de Kroon or Siska van Hees.