Informs Annual Meeting 2017

WB04

INFORMS Houston – 2017

WB05

4 - Management and Effects of In-Store Promotional Displays Oguz Cetin, Kenan Flagler Business School, 213 Ivy Meadow Lane, Durham, NC, 27707, United States, oguz_cetin@kenan-flagler.unc.edu, Adam J.Mersereau, Ali Kemal Parlakturk We examine a brick-and-mortar retailer’s choice of product to include in a promotional display. The display provides a visibility advantage to both the featured product and its category, but it also has consequences on customer traffic and substitution. Retailers need to understand the promotional display problem because it is a powerful demand-shaping lever and as an input to its negotiations with manufacturers. We develop analytical insights using a problem formulation based on a nested multinomial logit model of customer choice. Our work provides guidance for how retailers can use and value promotional displays effectively. 320A Models in Sustainable Operations Management Sponsored: Manufacturing & Service Oper Mgmt, Sustainable Operations Sponsored Session Chair: Foad Iravani, University of Washington, Seattle, WA, 98195-3226, United States, firavani@uw.edu 1 - Peak Load Energy Management by Direct Load Control Contracts Ali Fattahi, ali.fattahi.1@anderson.ucla.edu, Sriram Dasu, Reza Ahmadi Energy firms use direct load control contracts (DLCCs) to curtail customers’ electricity consumption during peak load periods via remote control devices. The question is: when and which groups of customers should reduce their energy consumption, and for how long? Implementing DLCCs require solving a stochastic dynamic optimization problem each day. We develop a near-optimum heuristic procedure with two simplifications in our approximation scheme: aggregating the customer groups into a single group, and ignoring the detailed structure of the uncertainty. We show that the asymptotic relative error of our heuristic approach is zero. 2 - Optimal Allocation Rules with Waste Considerations Sara Rezaee Vessal, HEC Paris, Paris, France, sara.rezaee-vessal@hec.edu, Sam Aflaki We study capacity allocation of a scarce and perishable product among stockout- averse retailers that face stochastic demand. We focus on two commonly practiced allocation mechanisms and using a dynamic model characterize the conditions under which each allocation mechanism performs superior from a waste and profit point of view. 3 - Wind Generation Strategic Bidding in a Competitive Electricity Market Ali Shantia, HEC, Paris, France, ali.shantia@hec.edu, Owen Wu, Roman Kapuscinski We empirically study how strategic behaviour of renewable generators translates to their higher profitability. Using the day-ahead and real-time market data, provided by Midcontinent Independent System Operator (MISO), we study what differentiates the bidding behaviour of wind generators. We also analyze whether such strategic behaviours practically increase wind generation profitability. 4 - That’s Not Fair: Tariff Structures for Electricity Markets with Rooftop Solar Siddharth Prakash Singh, Carnegie Mellon University, Pittsburgh, PA, United States, sps1@andrew.cmu.edu, Alan Scheller-Wolf Increased penetration of rooftop solar has led to decreased utility company profitability and undesirable cross-subsidization among customers. Regulatory responses have been controversial: changes in Nevada induced SolarCity, the market leader in solar systems, to suspend local operations. We demonstrate that choosing the right tariff structure is crucial to fair regulation: the sort proposed in Nevada have limited ability to allocate the financial losses (or gains) from solar adoption among different parties. We then present a contract, featuring full retail price repurchasing solar customers that does offers full flexibility. We illustrate our findings numerically. WB04

320B Decision Making for Hospitals and Health Systems Sponsored: Health Applications Sponsored Session Chair: Jens Brunner, University of Augsburg, Universitÿtsstraße 16, Faculty Business Administration/Economics, Augsburg, 86159, Germany, jens.brunner@unikat.uni-augsburg.de 1 - Dynamic Capacity Allocation for Elective Surgeries Steven Shechter, University of British Columbia, Vancouver, BC, V6T.1Z2, Canada, steven.shechter@sauder.ubc.ca We model a longer-term, dynamic capacity planning problem, in which a children’s hospital semiannually apportions available operating room time to di erent surgical specialties. We consider an objective of minimizing urgency- weighted wait time, and use approximation methods to identify good solutions. We also discuss the use a probabilistic model of the queueing discipline to represent noisy patient selection patterns observed in practice. 2 - Market-Based Risk Adjustment in Capitated Healthcare Systems Zhaowei She, Georgia Institute of Technology, Atlanta, GA, United States, zshe3@gatech.edu, Turgay Ayer Risk adjustment is essential component of all capitated healthcare systems, such as Medicare Advantage and Medicaid Managed Care. However, it is hard for payers to perfectly predict their customers’ cost patterns based on pure statistical methods. To complement the existing statistical risk adjustment models, this study presents a market based risk adjustment model where perfect risk adjustment can be endogenously generated by proper market segmentation. Specifically, we explicitly characterize the competitive bidding outcomes in Medicare Advantage in a queueing framework, and show that the proposed mechanism can help CMS to achieve the targeted service levels more cost efficiently. 3 - Task-related Resident Scheduling under Uncertainty using Column Generation Jens Brunner, University of Augsburg, Universitÿtsstraße 16, Faculty Business Administration/Economics, Augsburg, 86159, Germany, jens.brunner@unikat.uni-augsburg.de, Sebastian Kraul Residents have to perform different tasks while passing several departments in a specific time period. Hospitals have to make efforts to get the best candidates for their programs but need to stay cost-efficient. We present a cyclic optimization model to generate robust training schedules determining the optimal number of residents. The model allows hospital management to derive employment policies. Solving the problem efficiently, we present a new pattern generation approach for the Dantzig-Wolfe decomposition. We present preliminary results using real- world data. 320C Machine Learning in Medicine Sponsored: Health Applications Sponsored Session Chair: Dimitris Bertsimas, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States, dbertsim@mit.edu Co-Chair: Ying Zhuo, Massachusetts Institute of Technology, 70 Lincoln Street, Unit L612, Boston, MA, 02111-2670, United States, zhuo@mit.edu 1 - Predicting Response and Adverse Events in Immunotherapy-treated Patients Ying Zhuo, Massachusetts Institute of Technology, 70 Lincoln Street, Unit L612, Boston, MA, 02111-2670, United States, zhuo@mit.edu, Dimitris Bertsimas Recent development in immunotherapies has shown remarkable efficacy in treating melanoma and lung cancer patients, yet the response and adverse events associated with use of immunotherapy are highly variable across patients. We aim to build highly individualized predictive models. We obtained 11 years of electronic health records (EHR) from a large cancer center with hundreds of features: demographics, genomics, vitals, lab tests, etc. We fitted state-of-the-art machine learning models to predict 60-, 90- and 180-day mortality and adverse events such as rash, colitis, and hepatitis. WB06

474