EBA Consults on Use of Machine Learning for IRB Credit Risk Models

This discussion paper is a first step to engage the industry and the supervisory community to investigate the possible use of machine learning in IRB models and to build a common understanding of the general aspects of machine learning and the related challenges in complying with the regulatory requirements. The discussion paper provides a general definition of machine learning models, discusses the main learning paradigms used to train machine learning models, and discusses the current limited use of machine learning models in the context of IRB models. It also analyzes the challenges and the benefits institutions may face in using machine learning to develop compliant IRB models. The paper provides a set of principle-based recommendations that aim to ensure that machine learning models adhere to the regulatory requirements set out in the CRR, should they be used in the context of the IRB framework.

EBA recommends for institutions to ensure that the staff working in the model development unit, credit risk control unit, and the validation unit is sufficiently skilled to develop and validate machine learning models. EBA recommends that institutions should ensure that the management body and senior management are in a position to have good understanding of the model, by providing them with appropriate high-level documentation. It is also recommended for institutions to avoid any unnecessary complexity in the modeling approach, unless it is justified by a significant improvement in the predictive capacities. Institutions should avoid including an excessive number of explanatory drivers or drivers with no significant predictive information and using unstructured data if more conventional data is available that provides similar predictive capacities. Institutions should also avoid overly complex modeling choices if simpler approaches yielding similar results are available. In addition, to ensuring that the model is correctly interpreted and understood, institutions are recommended to:

• analyze, in a statistical manner, the relationship of each single risk driver with the output variable and the overall weight of each risk driver in determining the output variable.
• assess the economic relationship of each risk driver with the output variable to ensure that the model estimates are plausible and intuitive.
• provide a summary document in which the model is explained in an easy manner based on the outcomes of the analysis.
• ensure that potential biases in the model (for example, overfitting to the training sample) are detected.

Related Links

• Press Release
• Discussion Paper (PDF)

Comment Due Date: February 11, 2022

Keywords: Europe, EU, Banking, CRR, Basel, Machine Learning, Regulatory Capital, Credit Risk, Regtech, IRB Approach, EBA