REINSURANCE ANALYTICS USING SERIAL AND PARALLEL COMPUTATION ON THE MULTIOBJECTIVE EVOLUTIONARY ALGORITHM SPEA2
This paper presents a novel and efficient application of the SPEA2 on reinsurance contract optimization considering the perspective from an insurance company. The reinsurance operation aims to transfer the risk taken by an insurance company, usually against natural catastrophes, to a bigger corporation. The process of reinsurance is similar to that one where a client wants to insure his properties upon the payment of a premium. Then, the insurance company sells his portfolio using the reinsurance market aiming to maximize the expected return and, at the same time, to maximize the risk hedged to the reinsurance company. This problem is naturally multi-objective, consequently the SPEA2 algorithm appears as an attractive approach to tackle the problem. Results show that the SPEA2 can obtain better outcomes than a sophisticated algorithm called enhanced MO-PBIL in terms of hypervolume. A parallel version based on the master-slave model showed that a speedup of 1.83 can be reached using 4 cores and 500 iterations, and 2.13 using 8 cores and 100 iterations, with little effort to parallelize the application.