2016 INFORMS Annual Meeting Program

TC45

INFORMS Nashville – 2016

TC45 209A-MCC Multi-Objective Optimization Via Simulation Sponsored: Simulation Sponsored Session Chair: Susan R Hunter, Purdue University, West Lafayette, IN, United States, susanhunter@purdue.edu Co-Chair: Enlu Zhou, Georgia Institute of Technology, na, Atlanta, GA, na, United States, enlu.zhou@isye.gatech.edu 1 - A Partition-based Random Search For Stochastic Multi-objective Optimization Via Simulation We proposed two parallel partition-based random search methods to solve the stochastic multi-objective optimization via simulation considering Pareto optimality for constrained and unconstrained case. The idea is to explore the whole feasible region and exploit on current most promising regions in the same time. Partition methods are used to shrink current most promising regions iteratively, and simulation allocation rules are adopted to decrease the noise. Both methods are proven to converge to the global Pareto set with probability one. Numerical experiments are conducted to demonstrate the effectiveness and robustness of the proposed algorithm compared to well-known methods. 2 - An Assessment Of Model Based Methods In Multi-objective Optimization Joshua Hale, Georgia Institute of Technology, 755 Ferst Drive, NW, Atlanta, GA, Atlanta, GA, United States, jhale32@gatech.edu, Helin Zhu, Enlu Zhou We propose domination measure as a new concept to measure the quality of solutions in multi-objective optimization. The domination measure of a solution can be intuitively interpreted as the size of the portion of the decision space that dominates that solution. We reformulate the multi-objective problem to a single- objective stochastic problem and solve it using a model-based approach. The numerical experiment shows that our proposed algorithm is effective at approximating the optimal Pareto set and is competitive with some previously proposed methods. 3 - On Multi-objective Ranking And Selection Methods Susan Hunter, Purdue University, susanhunter@purdue.edu, Guy Feldman, Raghu Pasupathy Consider the context of selecting Pareto-optimal systems from a finite set of systems based on multiple stochastic objectives. We seek a characterization of the asymptotically optimal sample allocation that maximizes the rate of decay of the probability of misclassification, i.e., the probability a Pareto system is falsely estimated as non-Pareto, or a non-Pareto system is falsely estimated as Pareto. We discuss recent advances in solving this problem. 4 - Precision Irrigation System Optimization Using Subsurface Water Retention Technology For Multiple Conflicting Objectives Kalyanmoy Deb, Michigan State University, kdeb@egr.msu.edu Water is precious and recent efforts to achieve precision irrigation with minimum use of water through subsurface water retention technology (SWRT) are getting popular. In this study, we have linked a water permeation simulation process with a multi-objective optimization algorithm to obtain optimized solutions involving shape and location of subsurface impermeable membranes and simultaneously obtain optimal surface water supply. The procedure is pragmatic and is customized for specific soil and crop combination and average precipitation level. Loo Hay Lee, National University of Singapore, Singapore, Singapore, iseleelh@nus.edu.sg, Weizhi Liu, Siyang Gao

3 - Resource Allocation Decisions With Deep Uncertainty Cameron MacKenzie, Iowa State University, camacken@iastate.edu

Mathematical models to help public policy decision makers often have a great amount of uncertainty, sometimes called deep uncertainty. Decision makers may also be skeptical about solely relying on model recommendations. A solution to this deep uncertainty and a decision maker’s skepticism is for the model output to consist of ranges or intervals rather than point solutions. This presentation will offer a method for identifying intervals for resource allocation models in which every solution within the interval differs from the optimal solution by a predetermined value. 4 - Economic Contagion And The Role Of Beliefs: Findings From A Borrower-lender Game Jonathan William Welburn, University of Wisconsin - Madison, welburn@wisc.edu We present a within-period sequential-move game with multiple borrower countries and a single common lender to model cross-country contagion. We discuss the role of beliefs, modeled through Bayesian updating, and determine equilibrium solutions using nonlinear optimization. The model is calibrated to the 2010 Eurozone crisis, but sensitivity analysis is used to identify conditions under for contagion. Results demonstrate that what appears to be contagion may be the result of a crisis of confidence. Findings and their implications for decision making and policy are discussed. TC44 208B-MCC Decisions, Sensitivity and Applications Sponsored: Decision Analysis Sponsored Session Chair: Emanuele Borgonovo, Bocconi University, Via Roentgen 1, Alessandra Cillo, Assistant Professor, Bocconi University, Milan, 20146, Italy, alessandra.cillo@unibocconi.it, Enrico G De Giorgi Experimental studies have found that people reject a single lottery but accept a repeated play of the same lottery. Other studies have also found that the higher acceptance rates for the repeated play when the overall distribution is displayed depends on the type of prospect. These results have critical managerial relevance, but they are based on acceptance rates. The paper provides a theoretical framework, which allows quantifying the strength of preferences in repeated prospects. We provide an experiment to test possible editing processes in the context of repeated prospects. 2 - Tolerance Sensitivity And Maximum Regret In Linear Programming Richard E. Wendell, University of Pittsburgh, Pittsburgh, PA, 15260, United States, wendell@katz.pitt.edu, Emanuele Borgonovo Within a tolerance framework for linear programming, we present a new approach for calculating optimal coefficient sets. The approach solves an otherwise NP hard problem and, moreover, allows us to streamline the computation of regret functions. 3 - Randomized Differential Sensitivity Sumeda Siriwardena, Bocconi University, sumeda.siriwardena@phd.unibocconi.it, Emanuele Borgonovo Sensitivity analysis is an integral part of the decision analysis process. In several situations, analysts have not only the dataset of realizations of the model output but also of the corresponding partial derivatives. We introduce a new method based on the randomization of the differential importance measure. This sensitivity indicator does not require independence and possesses the additivity property, which makes the calculation of joint sensitivities seamless. We study numerical estimation and obtain the expression of the convergence rate. Managerial insights are discussed in detail. 4 - Estimating Strategic Impacts Of Foreclosed Housing Redevelopment Using Spatial Analysis Michael Johnson, University of Massachusetts Boston, MA, michael.johnson@umb.edu Community-based organizations engaged in foreclosure response wish to quantify the relative value of housing units for redevelopment. We measure the ‘strategic value’ of property acquisition candidates based on proximity to site-specific neighborhood amenities and disamenities, given the relative importance of that proximity to CDC organizational and community objectives. We show that strategic values can differ in systematic ways depending on the types of amenities and disamenities identified as relevant for acquisition decisions, the relative importance assigned to those amenities and disamenities, and the utility maximization objectives of the organization. Milano, 20833, Italy, emanuele.borgonovo@unibocconi.it 1 - Strength Of Preferences In Repeated Prospects

317