Inherently Interpretable Machine Learning in Financial Risk

Inherently Interpretable Machine Learning in Financial Risk

Speaker: Jim Coen

Major financial institutions have highlighted lack of explanation as a challenge to the deployment of ML models. This presentation describes the concerns and issues involved. It presents the two main approaches to improving interpretability of ML models. It introduces the Generalised Additive Model, which has become the standard approach for directly interpretable models. The culmination is three implementations of the Generalised Additive Model, including the Explanable Boosting Machine from Microsoft/MIT.

Jim Coen has a primary degree is in Applied Physics and Electronics. He has worked with Digital Equipment Corporation among others. He has a love of learning and have held positions as ICT Trainer in different settings. He went through a period of chronic illness. After deciding on a career change, he has been working on courses in Data Analysis from TU Dublin, edX, and Coursera since 2012. This culminated in achieving the MSc in Financial Analytics with Dublin Business School.