High performance computing for reinsurance analytics

This project explores the design, implementation, and evaluation of next-generation Catastrophe Risk Modeling systems and Reinsurance Analytics that exploit advances in High Performance Computing (HPC).Natural catastrophe reinsurers provide insurance coverage to primary insurance companies against the massive losses that can occur in the event of natural catastrophes such as earthquakes, hurricanes, and floods. In order for there to be a market for catastrophic risk coverage, and thereby the possibility for individuals to buy insurance that covers these events, reinsurers need simulation and analysis systems that quantify and manage portfolios of catastrophic risk. This project explores two intertwined research themes, namely, the design of 1) scenario-based catastrophe models and 2) analytical systems for the management and analysis of portfolios of catastrophic risk. In the case of catastrophe scenario modeling, the impact of thousands of events must be analyzed against detailed descriptions of millions of buildings to determine the intensity of the physical hazard, the building's vulnerability, and the implied financial loss, in order to product a single Event Loss Table (ELT). In the case of reinsurance analytics, simulations involving thousands of Event Loss Tables and potentially millions of simulation trials must be run to generate a portfolio view of risk. High Performance Computing (HPC), in particular Cluster and Cloud computing techniques, offer a potential way to address these highly computational and data intensive problems. This project looks to accelerate catastrophe scenario modeling and reinsurance analytics, while addressing the elastic demand for computational cycles that these applications create.