2016 INFORMS Annual Meeting Program

WA45

INFORMS Nashville – 2016

2 - Optimizing Adaptive Stormwater Management With Green Infrastructure: A Case Study In Wingohocking Watershed, Philadelphia Fengwei Hung, Johns Hopkins University, 3400 North Charles Street Ames Hall 313, Baltimore, MD, 21218, United States, hfengwe1@jhu.edu, Benjamin F Hobbs, Arthur E McGarity Due to heterogeneous hydrology and uncertain maintenance effectiveness, the long run performance of Green Infrastructure (GI) for managing urban stormwater and pollution is highly uncertain. Implementing GI adaptively provides opportunities to modify plans in response to learning. We apply three stochastic optimization models for adaptive GI planning that represent monitoring and active experimentation. The models recommend optimal immediate GI and learning actions. 3 - Flood Risk Management Using Artificial Avulsions In The Yellow River Delta Liang Chen, Johns Hopkins University, chenliang1468@gmail.com, Benjamin Hobbs Due to high in-channel sedimentation rates, the Yellow River Delta of China has changed course frequently in its history, with huge socioeconomic impacts. Water storage and deliberately engineered avulsions can reduce these impacts, but at a cost. Multi-objective analysis and Monte Carlo simulation is used to develop decision rules and choose sizes and locations for engineered avulsions and floodways, considering uncertain future floods and trade-offs between flood risk and management cost. 4 - The Process Of Co-producing a ClimateIndicators System Melissa A Kenney, University of Maryland, kenney@umd.edu In this talk I will discuss the development and implementation of a climate indicators system that was designed to be owned collaboratively by multiple Federal agencies and designed to support for undefined climate adaptation and mitigation decisions. The process of development involved over 200 producers and users of climate information from the Federal government, academic, and private sector/NGOs over the past 5 years. I will reflect on the implications and lessons learned for future co-production processes that similarly adopt best practices in the development of indicators. WA45 209A-MCC Learning for Simulation and Simulation Optimization Sponsored: Simulation Sponsored Session Chair: Giulia Pedrielli, National University of Singapore, TBD, Singapore, TBD, Singapore, giulia.pedrielli.85@gmail.com 1 - Simulation Analytics For Virtual Statistics Yujing Lin, Northwestern University, yujinglin2013@u.northwestern.edu “Virtual statistics” are performance measures that are conditional on the occurrence of an event; the virtual waiting time of a customer arriving to a queue at time t is one example. We describe methods for estimating virtual statistics post-simulation from the retained sample paths, examining both their small- sample properties and asymptotic consistency. 2 - The Effects Of Estimation Of Heteroscedasticity On Stochastic Kriging Xi Chen, Virginia Polytechnic Institute and State University, xchen6@vt.edu In this talk, we discuss the effects of using smoothed variance estimates in place of the sample variances on the performance of stochastic kriging (SK). Different variance estimation methods are investigated, and we show that such a replacement leads to improved predictive performance of SK. An SK-based dual metamodeling approach is further proposed to obtain more accurate prediction results given a fixed simulation budget. 3 - Extended Kernel Regression Method To Combine Analytical Methods And Simulation Andrea Matta, Shanghai Jiao Tong University, matta@sjtu.edu.cn Simulation is widely adopted to predict system performance. The main drawback is that it is slow in execution and the related computer experiments can be very expensive. On the other hand, analytical methods can rapidly provide system estimates but they are approximate. Recently the Extended Kernel Regression (EKR) has been proposed to combine simulation with analytical methods. This work has different purposes: 1) test EKR on different cases; 2) compare EKR with other state of the art techniques; 3) propose two different methods for calculation of confidence bands. Numerical results show the EKR method provides accurate predictions, particularly when DOE size is small.

4 - G-STAR A New Kriging Based Trust Region Method For Global Optimization

Giulia Pedrielli, Arizona State University, giulia.pedrielli.85@gmail.com, Szu Hui Ng

Stochastic Trust region methods iteratively generate meta-models for local optimization. We propose the Global Stochastic Trust Region Augmented Method (G-STAR): local meta-models are iteratively improved, through a new sampling criterion balancing exploration and exploitation. Specifically, a global model guides the search, while local models are fitted using sampled points in the generated trust regions. The best point is predicted at each iteration through an ensemble of the global and the local meta-models generated along the search. Preliminary numerical tests show an improved performance with respect to the previously proposed extended Two Stage Sequential Optimization algorithm. WA46 209B-MCC Revenue Management with Strategic Customers Sponsored: Revenue Management & Pricing Sponsored Session Chair: Woonghee Tim Huh, University of British Columbia, Vancouver, BC, Canada, tim.huh@sauder.ubc.ca Co-Chair: Jaelynn Oh, University of Utah, Salt Lake City, UT, United States, jaelynn.oh@business.utah.edu 1 - Dynamic Pricing In The Presence Of Strategic Consumers: Theory And Experiment Anton Ovchinnikov, Queen’s School of Business, 143 Union Street, Kingston, ON, K7L 3N6, Canada, ao37@queensu.ca, Mirko Kremer, Benny Mantin We investigate the behavior of retailers who sell a fixed inventory of products over a two period horizon (main selling season followed by a markdown period) to a mixture of myopic and strategic consumers. We present a stylized model and an experimental study. Our main result is that retailers myopically underprice when facing consumers who are strategic. We explore the drivers for such underpricing and show that it is related to a counter-intuitive model prediction that most revenue is obtained at markdowns. 2 - Choosing To Be Strategic: Implications Of The Endogenous Adoption Of Forward-looking Purchasing Behavior On Multiperiod Pricing Arian Aflaki, Duke University, 923 White Pine Drive, Durham, NC, 27705, United States, aa251@duke.edu, Pnina Feldman, Robert Swinney We consider whether strategic consumer behavior benefits consumers when they purchase from a revenue-maximizing firm that sets prices over multiple periods. We show that many consumers have lower surplus if they are strategic than if they are myopic. We then develop a model in which consumers choose to become strategic by exerting costly effort, and show that considering this choice can have a significant qualitative impact on firm and consumer decisions. In addition, we illustrate that it is possible to increase firm profit and consumer welfare simultaneously by increasing the cost of strategic behavior. Finally, we find that price commitment can encourage more strategic waiting and harm firms. 3 - Product Quality And Pricing Management Ruxian Wang, Johns Hopkins Carey Business School, Baltimore, MD, 21202, United States, ruxian.wang@jhu.edu, Shiliang Cui Product quality, price and service are arguably the most important factors consumers consider in purchasing a product. We investigate a firm’s strategy for managing multiple products under various monopolistic and oligopolistic settings. Our analytical results show that the optimal quality level of any product should always be set at the global optimum, even if the firm can change price simultaneously, faces other firms’ competition, or offers a free ancillary service. Moreover, consumer surplus may be higher.

378