An application of data fusion techniques in quantitative operational risk management

In this article we show an application of data fusion techniques to the field of quantitative risk management. Specifically, we study a synthetic dataset which represents a typical mid-level financial institution's operational risk loss as defined by the Basel Committee on Banking Supervision (BCBS) report. We compute the economic capital needed for a sample financial institution using a Loss Distribution Approach (LDA) by determining the Value at Risk (VaR) figure along with the correlation measures by using copulas. In addition, we perform computational studies to test the efficacy of using a "universal" statistical distribution function to model the losses and compute the VaR. We find that the Lognormal-Gamma (LNG) distribution is computationally robust in fusing the frequency and severity data when computing the overall VaR.