| dc.description.abstract | Insurance coverages for marine shipping fleets are often created by a broker, who negotiates underwriter insurance offers on behalf of the fleet owner. The broker evaluates different combinations of offers, searching for the best insurance coverage for their customers. The coverage must have a minimal price, while earning the broker a large enough commission. This time consuming and difficult task is currently performed manually. We present a bilinear model of the broker problem, and prove that solving it is NP-hard. We propose three different solution methods: (1) solving the model exactly using the commercially available software Baron, (2) alternatingly solving linearized subproblems and (3) using a metaheuristic. Results from computational experiments show that the second solution method outperforms Baron, both with respect to running time and objective function value. The model and the alternating algorithm constitute a powerful tool for the brokers. It yields solutions with prices close to provable lower bounds in less than a second for most of the test instances. | en_US |
