Informs Annual Meeting 2017

SB58

INFORMS Houston – 2017

3 - Shipment Scheduling in Hub Location Problems James H. Bookbinder, University of Waterloo, 421 Richview Avenue, Toronto, ON, M5P.3G7, Canada, jbookbinder@uwaterloo.ca, Mobina Masaeli, Sibel Alumur Alev We incorporate shipment scheduling decisions into hub location problems. We determine optimal hub locations, hub network structure, the number of vehicles to operate in the network, and the time period of dispatching each vehicle from a hub. We develop three MIP models for different problem versions, depending on whether holding costs are incorporated and whether the vehicles are of different types. We investigate the impact of shipment scheduling on hub network configurations, routing decisions, and total network costs. We solve the models on instances from a new dataset with real data. 4 - Heuristic Solutions for a Class of Stochastic Uncapacitated p-hub Median Problems Francisco Saldanha-da-Gama, University of Lisbon, Deio-Fcul, Bloco C6, Lisbon, 1749016, Portugal, fsgama@ciencias.ulisboa.pt, Ángel Corberán, Rafael Marti, Juanjo Peiró We extend the r-allocation p-hub median problem by including fixed allocation costs, non-stop services and uncertainty in traffic and costs. However, even for small instances of the problem, a compact optimization model for the new problem becomes too large for being tackled by a general solver. This fact motivates the development of an approximate procedure whose starting point is the determination of a feasible solution to every (deterministic) single-scenario problem. Such solutions are then embedded into an iterative process rendering feasible solutions for the overall problem. We discuss the results of the numerical experiments performed using instances build from well-known data sets. 362D Advances in Stochastic Simulation Sponsored: Simulation Sponsored Session Chair: Henry Lam, khlam@umich.edu 1 - Efficient Rare-event Simulation for Heavy-tailed Processes VIA Principle of Multiple Big Jumps Chang-Han Rhee, Centrum Wiskunde & Informatica, Jakoba Mulderplein 164, Amsterdam, 1098 XG, Netherlands, rhee@cwi.nl, Jose Blanchet, Jose Blanchet, Bohan Chen, Bert Zwart, Bert Zwart We propose a new class of strongly efficient rare event simulation estimators for heavy-tailed random walks and compound Poisson processes. The new estimators are straightforward to implement, has bounded relative errors, and are “universal” in the sense that a single importance sampling scheme applies to a very general class of rare events. In particular, our estimators can deal with rare events that are provoked by multiple big jumps (hence, beyond the usual principle of a single big jump) as well as multidimensional processes such as the buffer content process of a queueing network. We illustrate our new method with examples in mathematical finance, simulation optimization, and queueing theory. 2 - Input Model Averaging Barry L. Nelson, Northwestern University, Dept of Industrial Eng and Mgmt Sciences, 2145 Sheridan Road, Room C216, Evanston, IL, 60208-3119, United States, nelsonb@northwestern.edu, Alan T. Wan Input uncertainty is an aspect of simulation model risk that arises when the input distributions are derived or “fit” to real-world, historical data. While there has been significant progress on quantifying and protecting against input uncertainty, there have been few direct attempts to reduce it. We show that frequentist model averaging can be a provably effective way to create input models that better represent the true, unknown input distributions, thereby reducing model risk. We provide theoretical and empirical support for our approach. 3 - Ranking and Selection with Covariates Haihui Shen, City University of Hong Kong, Kowloon, Hong Kong, haihui.shen@my.cityu.edu.hk, Jeff Hong, Xiaowei Zhang We consider a new ranking and selection problem in which the performance of each alternative depends on some observable random covariates. The best alternative is thus not fixed but depends on the values of the covariates. Assuming a linear model that relates the mean performance of an alternative and the covariates, we design selection procedures producing policies that represent the best alternative as a function in the covariates. We prove that the selection procedures can provide certain statistical guarantee, which is defined via a nontrivial generalization of the concept of probability of correct selection that is widely used in the conventional ranking and selection setting. SB57

4 - Bagging-based Estimation of Simulation Input Uncertainty Huajie Qian, University of Michigan, Ann Arbor, MI, United States, hqian@umich.edu, Henry Lam We study resampling-based methods for estimating input-induced variance in stochastic simulation as a means to construct output confidence intervals under nonparametric input uncertainty. We revisit two unbiased bootstrap estimators and investigate their efficiencies. We then investigate an infinitesimal jackknife estimator on a bagging prediction that closely connects to the input uncertainty problem. We demonstrate that the bagging-based approach is more efficient for single-run simulation experiments. We also present a subsampling technique that significantly reduces the simulation effort required by the methods. 362E Disaster/Disruption/Hazmat Logistics Sponsored: Transportation Science & Logistics Sponsored Session Chair: Elise Miller-Hooks, miller@gmu.edu 1 - Facility Location and Item Prepositioning under Demand Uncertainty and Road-Facility Vulnerabilities Melih Celik, Middle East Technical University, ODTU. Endustri Muh. Bolumu, Universiteler Mh. Dumlupinar Bl. No: 1, Ankara, 06800, Turkey, cmelih@metu.edu.tr, Ece Aslan This study proposes a multi-echelon humanitarian supply chain network design by incorporation of demand uncertainty and road-facility vulnerabilities. In the frame of the study, a two-stage stochastic programming model is formulated to make decisions on warehouse locations and item prepositioning in the first stage, and relief distribution in the second stage. As an extension, concurrent road repair decisions are also considered. A sample average approximation scheme is developed for the solution of the proposed model, which is executed under efficiency- and equity-based objectives. 2 - Network Operations Against Attacks: A Tri-Level Model in Oil & Gas Production and Distribution Mustafa Alassad, University of Arkansas, Little Rock, AR, United States, mmalassad@ualr.edu, Yupo Chan, Hamzeh Davarikia Oil and gas are distributed via pumps and pipelines that are exposed to threats. Hardening strategies against malicious attacks are needed. A tri-level leader- follower-operator game determines the optimal fortification tactics to protect the critical assets. There exist dimensions of gaming beyond outright fortification, including tactics often adapted in the “fog of war,” such as deception. This is modeled mathematically under the shared cognition concept. Our computational results show that the deception concept is much more effective than fortification, where the cost of attacker’s damages diminished significantly without substantial resource commitment on the part of the defender. 3 - A Spectral Risk Measure in Hazardous Materials Transportation Liu Su, University of South Florida, ENG 302, Tampa, FL, 33620, United States, liusu@mail.usf.edu, Longsheng Sun, Mark Henry Karwan, Changhyun Kwon Due to catastrophic consequences of accidents by hazardous materials (hazmat) transportation, risk-averse approach for routing is needed. In this paper, we consider spectral risk measures, which are coherent and more general than existing approaches such as conditional value-at-risk. We develop a mixed integer programming model in hazmat routing to minimize a special class of spectral risks and propose an efficient algorithm to solve the problem. For general spectral risk measures, we approximate the spectrum function using step functions. By applying spectral risk measures to hazmat routing, we protect the road network from undesirable route choices that may lead to severe consequences. 4 - Optimizing Disaster Resilience of the Built Infrastructure Elise Miller-Hooks, Professor, George Mason University, 208 Rosalie Cove Ct, Silver Spring, MD, 20905, United States, miller@gmu.edu, Mersedeh Tariverdi, Hossein Fotouhi, Seksun Moryadee A multi-stage stochastic, mixed integer program is presented for quantifying and maximizing disaster resilience of the built infrastructure to multiple probabilistic hazard events. Dependencies on interdependent societal lifelines, including transportation, water and power, which are themselves subject to failure, are explicitly modeled. The model is applied on a case study. SB58

55