Harnessing Big Data and High-Performance Computing Architecture for Loss Scenario Analysis

• Activity Number: 442
• Type: Topic Contributed
• Date/Time: Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
• Sponsor: Business and Economic Statistics Section
• Author(s): Mahesh V. Joshi*+
• Companies: SAS Institute
• Keywords: Loss Modeling ; Big Data ; High-Performance Computing ; Risk Management ; Insurance ; Banking

Abstract

Financial losses due to adverse events are unfortunate yet inescapable for many insurance and banking businesses. Quantitative modeling of losses is important for managing risk better and for estimating the risk-based capital that regulations such as Basel III and Solvency II require. Most businesses collect and record information about losses, including the frequency of losses, the severity of each loss, and the characteristics of the entity and the economic environment that incur or cause each loss. Modern data collection and storage methods have resulted in large repositories of loss data. It is better to use all the data rather than a subsample of the data, because the sampling process might ignore some important features of the data. However, you need an appropriate big data management framework and analytical tools that work within that framework. This paper proposes parallel and distributed computing algorithms and architecture to implement the three key steps of the loss modeling process: frequency modeling, severity modeling, and aggregate loss modeling for various scenarios. The first two steps are big data problems, and the last step is a big computation problem.