Better understanding of ML models with the help of visualization tools: the visible and the invisible
The slides shown at the event can be found here.
Machine Learning models for credit risk prove to be a powerful tool to reach higher levels of discriminatory power. But the model inspection continues to be problematic. New visualization tools (available in Python) remediate some of the issues. They can be helpful for both the modeler and the validator. However, full transparency is not reached.
Marco Folpmers is an FRM expert for 20 years. He has a Ph.D. from the Free University Amsterdam and has taught FRM at the University of Tilburg from 2010 to 2019 as a part-time professor. During that time he has been involved in educational courses for TIAS Business School and in the Honors program. He has also guided multiple Ph.D. students. At Deloitte he is a partner for credit risk modeling, focusing on model development and model validation for CDM, IRB and IFRS 9. Marco is a frequent contributor to the GARP Risk Intelligence platform on credit risk.