Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Best-subset Selection for Complex Systems using Agent-based Simulation

Wang, Yu (2012) Best-subset Selection for Complex Systems using Agent-based Simulation. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

PDF (Yu Wang's PhD Dissertation)
Updated Version

Download (3MB) | Preview


It is difficult to analyze and determine strategies to control complex systems due to their inherent complexity. The complex interactions among elements make it difficult to develop and test decision makers' intuition of how the system will behave under different policies. Computer models are often used to simulate the system and to observe both direct and indirect effects of alternative interventions. However, many decision makers are unwilling to concede complete control to a computer model because of the abstractions in the model, and the other factors that cannot be modeled, such as physical, human, social and organizational relationship constraints. This dissertation develops an agent-based simulation (ABS) model to analyze a complex system and its policy alternatives, and contributes a best-subset selection (BSS) procedure that provides a group of good performing alternatives to which decision makers can then apply their subject and context knowledge in making a final decision for implementation.

As a specific example of a complex system, a mass casualty incident (MCI) response system was simulated using an ABS model consisting of three interrelated sub-systems. The model was then validated by a series of sensitivity analysis experiments.

The model provides a good test bed to evaluate various evacuation policies. In order to find the best policy that minimizes the overall mortality, two ranking-and-selection (R&S) procedures from the literature (Rinott (1978) and Kim and Nelson (2001)) were implemented and compared. Then a new best-subset selection (BSS) procedure was developed to efficiently select a statistically guaranteed best-subset containing all alternatives that are close enough to the best one for a pre-specified probability. Extensive numerical experiments were organized to prove the effectiveness and demonstrate the performance of the BSS procedure.

The BSS procedure was then implemented in conjunction with the MCI ABS model to select the best evacuation policies. The experimental results demonstrate the feasibility and effectiveness of our agent-based optimization methodology for complex system policy evaluation and selection.


Social Networking:
Share |


Item Type: University of Pittsburgh ETD
Status: Unpublished
CreatorsEmailPitt UsernameORCID
Wang, Yuyuw9@pitt.eduYUW9
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairShuman, Larry J.shuman@pitt.eduSHUMAN
Committee CoChairLuangkesorn, Louis K.lol11@pitt.eduLOL11
Committee MemberBidanda, Bopayabidanda@pitt.eduBIDANDA
Committee MemberBalaban, Carey D.cbalaban@pitt.eduCBALABAN
Committee MemberNorman, Bryan A.banorman@engr.pitt.eduBANORMAN
Committee MemberLee, Bruce Y.byl1@pitt.eduBYL1
Date: 2 February 2012
Date Type: Publication
Defense Date: 21 November 2011
Approval Date: 2 February 2012
Submission Date: 2 December 2011
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 176
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Industrial Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Agent-based simulation, Statistical selection procedure, Ranking-and-selection, Best-subset selection, Incident response simulation, Mass casualty incident response, Complex system, Complex adaptive system, Optimization via simulation
Date Deposited: 02 Feb 2012 17:56
Last Modified: 15 Nov 2016 13:55


Monthly Views for the past 3 years

Plum Analytics

Actions (login required)

View Item View Item