Data Study Group Final Report: Global bank

Machine learning for predicting and mitigating operational risk

In the wake of the global financial crisis in 2007-2008, the Basel II and III global regulatory accords, amongst others, stipulate the holding of so-called regulatory capital to off-set potential losses.

The risk of potential loss needs to be estimated by statistical models. A type of risk newly emphasised in Basel II and II is operational risk - concerning losses caused by fraud, bad practices, disasters, or process failures, or, more generally, human and environment/infrastructure factors.

The challenge is to develop accurate models for predicting operational risk, with the goal of preventing avoidable risks and mitigating unavoidable ones through sufficient regulatory capital.