Dissertation/Thesis Abstract

Model elicitation in nation-building simulation: Analytic network process for ranking decisions and Petri Nets for robust optimization
by Zhang, Ying, Ph.D., State University of New York at Buffalo, 2013, 131; 3554522
Abstract (Summary)

Simulation is the main tool to study nation-building problems because of their complexity. Considering the substantial work which is necessary to reconstruct a nation, there are many complex and interacting modules as well as a large number of parameters in nation-building simulation models. Thus, it is time-consuming and computationally expensive to study quantitative relationships within a nation-building system by simulation. Also, simulation cannot explicitly answer optimization-related questions which are important to nation-building decisions. For example, the optimal sequence of nation-building operations cannot be found by simulation unless all possible sequences are simulated, which is impossible in reality. Therefore, it is greatly beneficial to elicit analytic models and solutions from large-scale simulation for the nation-building interests that cannot be supported by simulation effectively and efficiently.

This dissertation defines the framework of model elicitation in nation-building simulation, and it presents systematic methodologies to elicit the Analytic Network Process (ANP) and Robust Optimization (RO) models from large-scale simulation models.

The first elicitation approach is developed to derive the ANP models from simulation models to study the entities with stochastic attributes and the input parameters in a nation-building system. Although the same purpose can be achieved by analyzing simulation results statistically, the ANP models are much more efficient to evaluate and rank model entities and input parameters by their significance without running a simulation model intensively. The numerical experiments show that the ANP models can provide good results which approximate to those by simulation with great time efficiency.

The second elicitation approach is proposed to formulate RO models from the simulation models developed by Ptolemy II software. Since it is not easy to formulate RO models from a Ptolemy II model directly, Colored Stochastic Timed Petri Nets (CSTPNs) are used to facilitate this elicitation. More specifically, a Ptolemy II model is first converted to a CSTPN, and then RO models are formulated based on the graph and network representation of the CSTPN model. The whole process is illustrated by developing two types of RO models from a Ptolemy II model which simulates a nation-building system, and the elicitation approach is validated by comparing the solutions from the RO models with simulation results.

Indexing (document details)
Advisor: Nagi, Rakesh, Sudit, Moises
Commitee: Batta, Rajan
School: State University of New York at Buffalo
Department: Industrial Engineering
School Location: United States -- New York
Source: DAI-B 74/06(E), Dissertation Abstracts International
Source Type: DISSERTATION
Subjects: Operations research
Keywords: Model elicitation, Nation-building simulation, Network process, Petri nets, Ranking decisions, Robust optimization
Publication Number: 3554522
ISBN: 9781267947024
Copyright © 2019 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy
ProQuest