Learning Model for Assessing Loss Severity of Operational Risk

Abstract

Risks, deficiencies and other issues identified within the organization should be evaluated and assessed with regard to their severity and significance. Operational risk is one of the risk categories within the banking and financial services community. It is defined as the risk of loss resulting from inadequate or failed internal processes, people and systems or from external events. Scenario analyses and risk assessments based on expert opinion should be frequently validated and reassessed by comparing them to actual loss data available over time. On contrary, this paper presents a quantitative operational risk assessment using the technique of backpropagation neural network. The multiple risk causes and resulting loss form a network of interdependencies as a learning model. The risk scenarios collected from expert judgment represents training instances of causal chains and effects. The output model could be used as the substitute of expert assessments for the mature organizations where operational loss data are available.