2016 INFORMS Annual Meeting Program

MB01

INFORMS Nashville – 2016

3 - Price Competition And Direct-to-consumer Advertising In Prescription Drug Markets Abhik Roy, Professor, Quinnipiac University, Department of Marketing, School of Business (SB-DNF), Hamden, CT, 06518- 1949, United States, abhik.roy@quinnipiac.edu Mary Schramm We examine the relationship between direct-to-consumer advertising (DTCA), and interdependent pricing among firms marketing competing drugs to patients within the same therapeutic area. We propose that DTCA is a coordinating mechanism, where a firm signals its willingness to be a Stackelberg price leader by spending heavily on advertising promoting the drug formulation, not just its own brand within the category. Propositions are developed about the impact of ad effectiveness, ad spending and substitutability on the occurrence of a Stackelberg system. Evidence to support these propositions is provided through empirical analysis of data from a number of prescription drug categories. 4 - A Dominant Retailer’s Strategic Response To More Efficient Weak Retailer Ehsan Bolandifar, Assistant Professor, Chinese University of Hong Kong, 9/F, Cheng Yu Tung Buliding, No., 12, Chak Cheung Street, Shatin, N.T., Hong Kong, 999077, Hong Kong, ehsan@baf.cuhk.edu.hk, Zhong Chen, Fuqiang Zhang We construct a multi-stage model to study the strategic interaction between national brand manufacturer, a dominant and weak retailers. We show that more efficiency on the weak retailer’s end makes the dominant retailer reduce its joint advertisement level for the national brand and offers lower market prices, while it also receives a discount form the national brand manufacturer. We also show that manufacturer does not always benefit from improvement in its operational efficiency of its retailers. Similarly, dominant retailer does not always benefit from cheaper store brand procurement costs. 5 - Direct Sales, Agent Selling or Reselling? Firms’ Channel Structure With Consumers’ Channel Preference Libo Sun, PhD Student, University of Science and Technology of China, 96# Jinzhai Road, Hefei, Anhui, PR China, Hefei, 230026, China, libosun@mail.ustc.edu.cn, Yugang Yu Numerous firms exert themselves to adopt multiple channels to sell products. However, even when facing identical products, consumers’ choices among these channels could be diverse due to channel preference. We use a stylized theoretical model to answer two key questions faced by a monopoly manufacturer: (1) how does consumers channel preference (CCP) affect its channel structure, namely, when should the manufacturer choose a sin-gle-channel and when should it adopt a dual-channel instead? Furthermore, should the manufacturer choose an independent reseller or an agent platform when it intends to leverage external force? We respectively derive the manufacturer’s optimal channel deci-sions when consumers show positive and negative CCP to the manufacturer’s direct channel. We find that: (1) compared with centralized case, the manufacturer prefers to adopt dual-channel in a larger area in decentralized case; (2) the manufacturer will choose a dual-channel either when consumers’ CCP to direct channel is positive or negative enough; (3) the exact forms of dual channel depends on the thresholds of the agent fee charged by the platform. MA94 5th Avenue Lobby-MCC Technology Tutorial: MathWorks/Artelys Technology Tutorial 1 - MathWorks: Data Analytics With MATLAB Mary Fenelon, MathWorks, Natick, MA, United States, mary.fenelon@mathworks.com MATLAB has evolved to become a platform for predictive and prescriptive analytics. Engineering, Finance, Data Science, and IT teams are using MATLAB to build today’s advanced analytics systems ranging from risk analysis to predictive maintenance and telematics to advanced driver assistance systems and sensor analytics. Join us to see how MATLAB can help you: • Access, explore, and analyze data stored in files, on the web, and from data warehouses • Clean, explore, visualize, and combine complex multivariate data sets • Prototype, test, and refine predictive models using machine learning methods • Build and solve prescriptive models and analyze results • Share your results with others We’ll highlight our newest features for Big Data, machine learning, deep learning, and optimization through examples such as load forecasting, Monte Carlo simulation, predictive maintenance, and embedded sensor analytics. 2 - Artelys: Solving Large Least-Squares Models With The Artelys Knitro Nonlinear Optimization Solver Richard Waltz, Artelys, 150 N Michigan Avenue, Suite 800, Chicago, IL, 60601, United States, Richard.waltz@artelys.com Artelys Knitro is the premier solver for nonlinear optimization problems. This software demonstration will highlight the latest Knitro developments, including a

new specialized API, as well as enhanced algorithms, for large-scale nonlinear least-squares models. We will demonstrate how to solve least-squares models using Knitro through a variety of interfaces such as R, MATLAB and C/C++, and also provide some benchmarking results. In addition, we will summarize some of the other recent developments in Knitro.

Monday, 10:00AM - 10:50AM

Monday Plenary

Davidson Ballroom-MCC Philip McCord Morse Lecture: Margaret L. Brandeau Plenary Session Chair: Mike Magazine, University of Cincinnati, Cincinnati, OH 45221-0130, mike.magazine@uc.edu 1 - Public Health Preparedness: Answering (Largely Unanswerable) Questions With Operations Research Margaret Brandeau, Stanford University, Stanford, CA, United States, brandeau@stanford.edu Public health security - achieved by effectively preventing, detecting, and responding to events that affect public health such as bioterrorism, disasters, and naturally occurring disease outbreaks - is a key aspect of national security. However, effective public health preparedness depends on answering largely unanswerable questions. For example: What is the chance of a bioterror attack in the United States in the next five years? What is the chance of an anthrax attack? What might be the location and magnitude of such an attack? This talk describes how OR-based analyses can provide insight into complex public health preparedness planning problems - and thus support good decisions. MB01 101A-MCC Data Mining Under Uncertainty Sponsored: Data Mining Sponsored Session Chair: Erhun Kundakcioglu, Ozyegin University, Nisantepe District Orman Street / Cekmekoy, Istanbul, 34794, Turkey, erhun.kundakcioglu@ozyegin.edu.tr 1 - Approximation Algorithms For Solving Large-scale Classification Problems Neng Fan, University of Arizona, nfan@email.arizona.edu To deal the classification of data with uncertainties, the distributionally robust optimization models are proposed for the support vector machines. First the problems are reformulated as semidefinite programs or second order cone programs. To solve these problems on large-scale data sets, we design a stochastic subgradient algorithm. The numerical experiments will be presented to show the efficiency of our algorithms. 2 - Margin Maximization Via Benders Decomposition To Solve Multiple Instance Learning Problems Emel Seyma Kucukasci, Istanbul Commerce University, Istanbul, 34840, Turkey, eskucukasci@ticaret.edu.tr Emel Seyma Kucukasci, Bogazici University, Istanbul, 34342, Turkey, eskucukasci@ticaret.edu.tr, Mustafa Gokce Baydogan Multiple instance learning (MIL) aims to solve classification problem where bags of instances form the input data. Margin maximization model of MIL classification is a MINLP problem. We develop a Benders decomposition algorithm for MINLP solution to deal with large datasets. A hybrid approach combining Benders decomposition and bagging procedure is proposed to test the generalizability of the results. Computational results on publicly-available molecular activity prediction, image annotation and text classification datasets are also provided. Monday, 11:00AM - 12:30PM

147