Show simple item record

dc.contributor.authorXu, Jie
dc.date.accessioned2016-04-05T19:11:22Z
dc.date.available2016-04-05T19:11:22Z
dc.date.issued2013
dc.identifier.isbn9781303476037
dc.identifier.other1459762873
dc.identifier.urihttp://hdl.handle.net/10477/50290
dc.description.abstractThe objective of this dissertation is to advance the study of security screening policy by considering different levels of robustness and provide insights to agencies/approvers (e.g., authority, manager, and screener). The performance of an optimal screening strategy highly depends on how accurately the applicants' behavior has been predicted. When the players are in complex games, they may be irrational or boundedly rational, non-strategic, misinformed or biased. Even if they are rational, it is likely that they are uncertain about some information, especially the other players' attributes. To address this problem, we apply robust optimization to study a wide range of possible applicants' uncertainties to provide the optimal robust screening policies for the approver. The novelty of this work includes: (a) the approver uses optimal robust screening policies with the control of the level of robustness; (b) the bad potential applicants' risk preference is studied, including risk seeking, risk neutral, or risk averse; and (c) non-strategic good potential applicants are studied. This work considers both system congestion and the strategic behavior of potential applicants, and integrates robust games and queueing theory to study the optimal nonprofiling and profiling screening policies for the approver, the price of robustness, the average waiting time, the applicants' submission probabilities and the corresponding applicants' expected payoffs. In particular, we first study that the approver, the decision maker, could adjust the desired level of robustness to represent different types of uncertainties, including (a) the potential applicants' attributes (e.g., the bad potential applicant's reward if he passed, the bad potential applicant's cost if he gets caught, the bad potential applicant's risk preference, and the good potential applicant's reward); (b) the approver's error probabilities (e.g., type I and type II error probabilities); and (c) other factors (e.g., the arrival rate of all potential applicants, the abandon rate of good potential applicants, and the proportion of strategic good potential applicants). Second, we compare nonprofiling and profiling screening policies to identify the conditions under which the difference is significant. In addition, we extend the perfect screening to imperfect screening by considering two types of screening errors, and consider behavior of abandon of some good applicants (who may withdraw their applications). Finally, we conduct simulation to validate the models. In particular, we compare the simulated optimal screening results and the simulated robust screening results under nonprofiling and profiling screening policies to identify the conditions under which the robust model works better considering the uncertainties. Nonprofiling robust screening policy works well when (a) the level of robustness is high with uncertainties on the bad applicant's reward if he passed, the bad applicant's cost if he gets caught, and the bad applicant's risk preference; and (b) the level of robustness is low with uncertainties on the approver's type I and type II error probabilities, the arrival rate of all potential applicants, and the abandon rate of good potential applicants. Profiling robust screening policy works well when (a) the level of robustness is high with uncertainties on the bad applicant's reward if he passed, the bad applicant's cost if he gets caught, and the bad applicant's risk preference; (b) the level of robustness is low with uncertainties on the approver's type II error probability, and the arrival rate of all potential applicants; and (c) the approver has uncertainty on the good applicant's reward, the abandon rate of the good potential applicants, and the proportion of strategic good potential applicants.
dc.languageEnglish
dc.sourceDissertations & Theses @ SUNY Buffalo,ProQuest Dissertations & Theses Global
dc.subjectApplied sciences
dc.subjectGame theory
dc.subjectQueueing theory
dc.subjectRobust screening
dc.subjectSecurity screening
dc.titleRobust Screening Policy at Security Queues in the Presence of Strategic Applicants with Private Information
dc.typeDissertation/Thesis


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record