Tuesday, May 29, 2018

The adoption of machine learning in credit risk modeling and management is gathering pace, with 2017 & 2018 seeing more banks implementing and/or running pilot projects with these techniques. While machine learning tools and models have been around for many years, recent increases in computing power and data storage have catalyzed the expanded opportunities for application.

The adoption of these techniques is bringing greater accuracy, efficiency in model development, and new ways to overcome data deficiencies and model biases. It concurrently brings new challenges, particularly in how data is accessed and managed, and in the human skills needed in support.

In analyzing these developments, the IIF interviewed 60 firms (58 banks and 2 mortgage insurers) between September 2017 and January 2018 on their adoption or exploration of machine learning. This study covered the specific techniques applied, areas of application (such as credit decisioning and the monitoring of deteriorating credits), benefits and challenges, budgeting and vendors, model governance, and regulation.

While distribution of the full Machine Learning in Credit Risk: detailed survey report is limited to the supervisory community and the 60 financial institutions that participated in the survey, this summary report provides an abbreviated public summary of the key themes from our report.

IIF Authors

Brad Carr

Brad
Carr
Senior Director, Digital Finance Regulation and Policy
+1-202-857-3648
bcarr@iif.com

Natalia Bailey

Natalia
Bailey
Associate Policy Advisor
+1-202-682-7440
nbailey@iif.com

Michael Kueker

Michael
Kueker
Associate Policy Advisor
+1 202 682-7454
mkueker@iif.com

Share