Managing Operational Risk using Bayesian Networks: A practical approach for the risk manager

This paper provides a practical approach to construct and learn a Bayesian network model that will enable an operational risk manager communicate actionable operational risk information for informed decision making by senior managers. Bayesian networks and their application in operational risk management has been widely studied; however, literature and research has predominantly focused on their application in modeling and measuring operational risk for capital calculation purposes. We detail the approach to construct and learn a BN model, from an incident database, using the machine learning capabilities in the R package bnlearn. The modeling and the inference capabilities of the Bayesian Networks can be applied to business-as-usual risk management techniques such as loss analysis, scenario analysis, risk assessment, development of key risk indicators, and risk reporting