2016 INFORMS Annual Meeting Program

TC41

INFORMS Nashville – 2016

4 - Influence Maximization In Linear Threshold And Triggering Models Po-Ling Loh, UW - Madison, Madison, WI, 53717, United States, polingloh@gmail.com We discuss upper and lower bounds for the influence of a set of nodes in certain types of contagion models. We quantify the gap between our upper and lower bounds in the case of the linear threshold model and illustrate the gains of our upper bounds for independent cascade models in relation to existing results. Furthermore, our lower bounds are monotonic and submodular, implying that a greedy algorithm for influence maximization is guaranteed to produce a maximizer within a $\left(1-\frac{1}{e}\right)$-factor of the truth. Our bounds may be evaluated efficiently, leading to an attractive, highly scalable algorithm for influence maximization with rigorous theoretical guarantees. TC41 207C-MCC Quantitative Risk Management Sponsored: Financial Services Sponsored Session Chair: Abel Cadenillas, University of Alberta, Edmonton, AB, Canada, abel@ualberta.ca 1 - Systemic Influences On Optimal Equity-credit Investment Christoph Frei, University of Alberta, cfrei@ualberta.ca, Agostino Capponi Recent events showed that the dependence structure of financial markets is more complex than what is captured by classical models. For example, the financial instability of some companies spread out to affect other companies. We analyze how such systemic influences are reflected in optimal investment decisions. To this end, we introduce a model with dependence structure between market risk and default risk of the companies. An investor can use stocks and credit default swaps (CDSs) to participate in the market. We derive an explicit expression for the optimal investment strategy in stocks and CDSs. An empirical analysis reveals the critical role of systemic risk in portfolio monitoring. 2 - Optimal Governement Debt Ceiling Abel Cadenillas, University of Alberta, abel@ualberta.ca, Ricardo Huaman-Aguilar Motivated by the debt crisis in the world, we apply methods of stochastic control to obtain an explicit formula for the optimal government debt ceiling. 3 - Optimal Cash Holdings Under Funding Risk Andreea Minca, Cornell University, acm299@cornell.edu This talk explores a one-period model for a firm that finances itsoperations through debt provided by heterogeneous creditors. Creditorsdiffer in their beliefs about the firm’s investment outcomes. We showthe existence of Stackelberg equilibria in which the firm holds cashreserves in order to provide incentives for pessimistic creditorsto invest in the firm. We find interest rates and cash holdings tobe complementary tools for increasing debt capacity. In markets witha high concentration of capital across a small interval of pessimisticcreditors or by a few large creditors, cash holdings is the preferredtool to increase the debt capacity of the firm. 4 - EM Algorithm and Stochastic Control Steven Kou, National University of Singapore, matsteve@nus.edu.sg We propose an algorithm called EM-Control (EM-C) algorithm to solve multi- period finite-time horizon stochastic control problems. Generalizing the idea of the EM algorithm, the EM-C algorithm sequentially updates the control parame- ters in each time period in a backward manner. The EM-C algorithm has monot- onicity of performance improvement in every iteration. We apply the EM-C algo- rithm to solve stochastic control problems in real business cycle and monopoly pricing of airline tickets. This is a joint work with Xianhua Peng and Xingbo Xu. TC42 207D-MCC Revenue Management with Advertising Applications Sponsored: Revenue Management & Pricing Sponsored Session Chair: John G Turner, University of California - Irvine, Room SB2 338, Irvine, CA, 92697-3125, United States, john.turner@uci.edu 1 - The Bid Adjustment Problem In Search Advertising Mustafa Sahin, University of Maryland, mustafa.sahin@rhsmith.umd.edu, Subramanian Raghavan, Abhishek Pani, Abhishek Pani We discuss the problem faced by the advertiser in search advertising in the presence of bid adjustments. Recent developments in search advertising created a

setting in which the advertiser can target specific demographics by using bid adjustments. We propose a Mixed Integer Programming formulation for the problem. However, the problem is computationally hard and cannot be solved by a generic commercial solver for any instance of reasonable size. Therefore, we offer heuristic approaches to tackle the intractability issues and present results on hard instances. 2 - Analysis Of Competitive Pricing With Multiple Overlapping Competing Bids In Revenue Management Goutam Dutta, Professor, Indian Institute of Management, House No 407, Iima Old Campus, Vastrapur, Ahmedabad, 380015, India, goutam@iima.ac.in We formulate the pricing problem from the point of view of one seller having one or multiple competitors (say n). Based on past experience, we know the distribution of bid prices of the competitors. We consider uniform and normal distribution to describe the bid price of the competitors. The prices of the competitors are mutually independent and the price ranges are either identical or different and overlapping. We maximize the expected contribution of the seller. Assuming the contribution as a linear function of price we find the conditions for maximization of the expected contribution to profit in case of n bidders. Further, we also compare the optimization results with simulation results. 3 - Markov Chain Models For Controlling The Frequency Distribution Of Online Advertising Seyed Ali Hojjat, University of New Hampshire, Durham, NH, United States, ali.hojjat@unh.edu, John G Turner Recent trends in online advertising show that explicit reach and frequency specifications are more desired over aggregate impression or budget goals. Depending on whether the frequency of ad serving to each user is measured over a fixed timespan (e.g., the number of times each user is exposed to the ad within each calendar week) or on a rolling basis (e.g., over any contiguous 24-hour period throughout the campaign’s horizon), we propose an appropriate Markov chain model for serving ads and investigate its properties in maintaining a desired frequency distribution for an online ad campaign. 4 - Planning Online Advertising Using Lorenz Curves John G Turner, University of California - Irvine, Irvine, CA, United States, john.turner@uci.edu, Miguel A Lejeune Lorenz curves are commonly-used to depict dispersion; e.g., income inequality. Motivated by online advertising campaigns that desire impressions spread over targeted audience segments and time, we formulate a problem that minimizes Gini Coefficients (area under the Lorenz curve), and develop a specialized decomposition technique to solve instances quickly. TC43 208A-MCC Decision Making in Public Policy Sponsored: Decision Analysis Sponsored Session Chair: Cameron MacKenzie, Iowa State University, Ames, IA, United States, camacken@iastate.edu 1 - Hurricane Decision Simulator Eva D Regnier, Naval Postgraduate School, eregnier@nps.edu, Cameron MacKenzie, Eric S Hodson When threatened by a hurricane, Marines in New Orleans face a classic sequential decision under uncertainty with regularly updated information, but few opportunities to learn from experience. The Hurricane Decision Simulator allows personnel to run experience key decisions in the context of many realistic simulated storms, to develop a better understanding of the interrelated decisions required, and a familiarity with the forecast products and their evolving uncertainty. This talk highlights application of both “hard” and “soft” sides of analytics in the development of the tool. This is the first hurricane training tool that allows users to explore many different decision paths. 2 - Subsidizing Cybersecurity Information Sharing: A Game Between A Government And N-Companies Ali Pala, University at Buffalo, Buffalo, NY, 14260, United States, alipala@buffalo.edu Ali Pala, Turkish Military Academy, Devlet Mahallesi, Bakanlıklar, Ankara, Turkey, alipala@buffalo.edu, Jun Zhuang More cybersecurity information sharing would lead to stronger resistance against cyber-attacks in the presence of a cooperative and trustworthy sharing network. Sharing cyber-attack information, however, could harm reputation, create disadvantages against competitors and additional costs, and cause disclosing vulnerabilities and some private information. In this research, we study what, how, and to whom government incentives should be provided in order to encourage and improve information sharing. We incorporate game theory and agent-based simulation modeling to develop a dynamic decision support tool that generates information sharing strategies in the face of strategic attackers.

316