Assessment Methodology for Credit Models to Meet European Regulatory Expectations

Winston, Albert (2022) Assessment Methodology for Credit Models to Meet European Regulatory Expectations. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration manual) Download (3MB) | Preview

Abstract

This research considers the assessment of machine learning models for banking in the context of European regulatory expectations. This includes requirements beyond predictive performance, is not well addressed in existing literature and prevents the wider industry adoption. Research has shown that such models may lead to capital savings for banks, but introduce complexity and cost to risk management.

The research developed a range of models for predicting credit default risk. These were subjected to an assessment approach that considered predictive performance, explainability at a local and global level, complexity and capital impacts. Under these additional headings we see that the preferred model may change depending on the regulatory focus. In addition while models may display good predictive ability it may be difficult to promote them under other criteria. The distribution of probability may lead to unexpected capital effects due to the non-linearity of that relationship.

Further work is required to consider how to optimally weight each aspect of the assessment criteria within a model risk management framework, however the research provides an important starting point for this.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HG Finance > Banking Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	14 Mar 2023 15:26
Last Modified:	14 Mar 2023 15:26
URI:	https://norma.ncirl.ie/id/eprint/6342

Actions (login required)

View Item