2015 Informs Annual Meeting

MB30

INFORMS Philadelphia – 2015

MB28 28-Room 405, Marriott Economic Models of Auctions Cluster: Auctions Invited Session Chair: Brian Baisa, Assistant Professor, Amherst College, 100 Boltwood Ave, Amherst, MA, 01002, United States of America, bbaisa@amherst.edu 1 - A Detail-free and Efficient Auction for Budget Constrained Bidders Brian Baisa, Assistant Professor, Amherst College, 100 Boltwood Ave, Amherst, MA, 01002, United States of America, bbaisa@amherst.edu I study auctions for divisible goods where bidders have private values and private budgets. My main result shows that when bidders have full-support beliefs over their rivals’ types, a clinching auction played by proxy-bidders implements a Pareto efficient outcome. Bid behavior is derived using two rounds of iterative deletion of weakly dominated strategies. This contrasts with recent work that shows efficient auction design is incompatible with dominant strategy incentive compatibility. 2 - Weak Cartels and Collusion-proof Auctions Jinwoo Kim, Associate Professor, Seoul National University, 1 Gwanak-ro Gwanak-gu, Seoul, Korea, Republic of, jikim72@gmail.com, Yeon-koo Che, Daniele Condorelli We study collusion in auctions by cartels whose members cannot exchange side- payments (i.e., weak cartels). We provide a complete characterization of outcomes that are implementable in the presence of weak cartels, identifying the set of circumstances under which standard auctions are susceptible to them. We then solve for optimal collusion-proof auctions and show that they can be made robust to the specific details of how cartels are formed and operated. 3 - Multi-unit Auctions with a Large Bidder Recent work in IPV settings shows that the uniform-price and discriminatory auctions are approximately efficient if there are many bidders with relatively small demands bidding for a homogenous good. We study a setting where a large bidder competes against a continuum of small bidders, and show that the small bidders prefer the uniform-price over the discriminatory auction, the large bidder has the reverse ranking, and there is no clear efficiency or revenue ranking between the two formats. 4 - A Truthful-in-expectation Mechanism for the Generalized Assignment Problem Salman Fadaei, Technische Universität Mönchen, Munich, Germany, salman.fadaei@tum.de, Martin Bichler We propose a truthful-in-expectation, 1-1/e-approximation mechanism for the generalized assignment auction. In such an auction, each bidder has a knapsack valuation function. We present a novel convex optimization program for the problem which yields an MIDR allocation rule. We show how to implement the convex program in polynomial time using a fractional greedy algorithm which approximates the optimal solution within an arbitrarily small error. Justin Burkett, Wake Forest University, Box 7505, Winston-Salem, NC, 27104, United States of America, burketje@wfu.edu, Brian Baisa Sponsor: Analytics Sponsored Session Chair: Shannon Harris, Katz Graduate School of Business, 241 Mervis Hall, Pittsburgh, PA, 15213, United States of America, sharris@katz.pitt.edu 1 - Optimal Staffing of Revenue Centers in Healthcare Delivery Organizations Jerome Niyirora, SUNY Polytechnic Institute, 100 Seymour Rd, Utica, NY, 13502, United States of America, jerome.niyirora@gmail.com, Jamol Pender In the operations management literature, little attention is paid to profitability in healthcare delivery organizations. But such an important issue cannot be overlooked since an unprofitable organization is unlikely to meet the quality of service demands. To address this issue, we introduce a nonstationary queueing model and apply optimal control theory to derive a new closed form square root staffing formula to allow for optimal staffing based on the cost-to-revenue ratio. MB29 29-Room 406, Marriott Joint Session Analytics/MIF/HAS: Healthcare Analytics

2 - Managing Customer Arrivals in Service Systems with Multiple Servers

Christos Zacharias, Visiting Assistant Professor, University of Miami, School of Business Administration, Miami, FL, United States of America, czacharias@miami.edu, Michael Pinedo We analyze a discrete multi-server queueing model for scheduling customer arrivals in service systems with parallel servers. Theoretical and heuristic guidelines are provided for the effective practice of appointment overbooking to offset no-shows. The benefits of resource-pooling are demonstrated in containing operational costs and increasing customer throughput. 3 - A Hierarchical Bayes Model of No-show Rates Joseph Johnson, Associate Professor, University of Miami, Patient no-shows in US clinics can sometimes shoot up to 80%. Accurate predictions of no-shows help clinics optimally schedule appointments. We develop a Hierarchical Bayes logit model which improves prediction accuracy over the widely-used simple logit model. The accuracy gain arises from the individual patient-level coefficients provided by the Bayesian method. Comparison of model fit on 12-months of appointment data shows that the Bayesian model vastly outperforms the simple logit model. 4 - Appointment Scheduling with No-shows and Cancellations Shannon Harris, Katz Graduate School of Business, 241 Mervis Appointment no-shows and cancellations can be disruptive to clinic operations. Scheduling strategies such as overbooking or overtime slot assignments can assist with mitigating these disruptions. We propose a scheduling model that accounts for both no-show and cancellation rates, and show properties of optimal scheduling models under specific conditions. MB30 30-Room 407, Marriott Practice Presentations by INFORMS Roundtable Companies I Sponsor: INFORMS Practice Sponsored Session Chair: Stefan Karisch, Digital Aviation Optimization & Value Strategy, Boeing Commercial Aviation Services, 55 Inverness Drive East, Englewood, CO, 80112, United States of America, stefan.karisch@jeppesen.com 1 - Optimization Models for Planning and Dispatch in Large-scale Freight Operations Ted Gifford, Distinguished Member Of Technical Staff, Schneider National, Inc., P.O. Box 2545, Green Bay, WI, 54306, United States of America, GiffordT@schneider.com Schneider National operates a fleet of 13,000 tractors and 48,000 trailers in a random network and accepts 10,000 customer orders daily. In order to maximize revenue and asset utilization, Schneider current employs math programming models providing real-time decision support for load acceptance, asset re- positioning and dispatch assignment. We will describe a number of these models and the operational challenges that accompany them, as well as enhancements and new models in development. 2 - Management Science at Bank of America Merrill Lynch Russ Labe, Director, Analytics & Modeling, Bank of America, 1500 Merrill Lynch Drive, First Floor, Pennington, NJ, 08534, United States of America, russ.labe@bankofamerica.com This paper will discuss the impact of analytics at Bank of America Merrill Lynch. Russ Labe, Director of Analytics & Modeling, will provide an overview of his group and discuss a few examples of business issues, analytic approaches, and results. He will also discuss how OR/MS drives business benefits, lessons learned, and best practices. 3 - Predictive Modeling at Scale Kathy Lange, Sr. Director, Global Analytics Practice, SAS, SAS Campus Drive, Cary, NC, 27518, United States of America, Kathy.Lange@sas.com Most organizations realize that analytics can help them become more competitive, more profitable, save money, or improve lives. Now they need to expand their analytical impact. This presentation will discuss new capabilities that address how organizations can scale up the modeling process, how to generate many hundreds or thousands of predictive models simultaneously, automating the creation and management of models. 5250 University Drive, 501 Kosar Epstein Building, Coral Gables, FL, 33146, United States of America, jjohnson@bus.miami.edu, Yu Tang, Yutian Li Hall, Pittsburgh, PA, 15213, United States of America, sharris@katz.pitt.edu, Jerrold H. May, Luis Vargas

181