We introduce and advocate a new paradigm in simulation experiment design and analysis, called ``green simulation,'' for the setting in which experiments are performed repeatedly with the same simulation model but different input parameters. In this dissertation three classes of green simulation estimators are proposed: the likelihood-ratio-based estimators, the metamodeling-based estimators, and the green Database Monte Carlo estimators. These estimators reuse old simulation outputs in different ways and thus have different requirements, features, and merits. We identify conditions under which these methods are most effective, establish convergence properties for some of the methods, and conduct numerical experiments on practical applications such as catastrophe bond pricing and credit risk evaluation. We find that green simulation can greatly improve computational efficiency in different applications.

Last modified

• 02/26/2018

Creator

• Mingbin Feng

DOI

• https://doi.org/10.21985/N2HQ25

Subject

• Industrial Engineering and Management Sciences

Keyword

• simulation
• database monte carlo
• metamodeling
• nested simulation
• likelihood ratio method
• stochastic kriging

Date created

• 2016-01-01

Resource type

• Dissertation

Rights statement

• In Copyright