Aggregating Operational Risks for Risk Reporting – Machine Learning Bayesian Networks

Operational risk is managed through internal/external loss data, key risk indicators, risk and control selfassessments, scenario analysis, and requires to be measured at various organisational levels. Educated risk management decisions can be taken better when the framework allows for the integration of risk and control information, with aggregation happening across the bank. The use of a Bayesian Network (BN) provides such a model that allows for the integration of risk and control information to be delivered in a structured process. This article provides a practical framework for the aggregation of operational risk data for risk reporting through the learning of Bayesian networks. The parameters of the BN is initially learnt from an incident database and subsequently updated with expert opinion. This framework can be adopted and adapted by a risk manager enhancing their ability to deliver dynamic and timely risk intelligence for effective management of operational risk on a day-to-day basis.