Stress-testing applications of Machine Learning Models

Jorge A Chan-Lau

This chapter will discuss stress-testing applications of machine-learning methods against the background of forecasting accuracy, interpretability of results, and the ability to capture the adaptive behaviour of firms and households facing structural breaks in the economic and business environment in which they operate. Adequate forecasting accuracy in stress tests can be difficult to accomplish given our imperfect knowledge about macro-financial linkages and their impact on financial firms’ profitability, liquidity and soundness. The chapter will therefore argue that machine learning offers a viable option for improving forecast accuracy due to the models’ ability to capture nonlinear effects between the scenario variables and the risk factors driving the soundness of a financial firm. Stress-testing applications are reviewed, highlighting the advantages of machine-learning models over more standard econometric-based stress-test models.

The traditional methods used for stress testing originated in statistics. In the words of Larry Wasserman of Carnegie Mellon: “Statistics emphasises formal statistical inference (confidence intervals, hypothesis tests, optimal estimators) in

To continue reading...

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an individual account here: