INFORMS 2021 Program Book

INFORMS Anaheim 2021

TB24

3 - An Analysis of Operating Efficiency and Public Policy Implications in Last-Mile Transportation Following Amazon’s Vertical Integration Lina Wang, Arizona State University, Tempe, AZ, 85257, United States, Elliot Rabinovich, Harish Guda We examine how Amazon’s decision to vertically integrate its retail platform and last-mile delivery operations can lead to anticompetitive outcomes as a result of a deterioration in the operating efficiency in the routes served by a last-mile transportation firm. We also expand on public policy measures that can ameliorate these outcomes. Based on an operational analysis of the last mile transportation firm, we find that Amazon’s decision to vertically integrate increased significantly the mileage necessary to deliver parcels in the ZIP code areas where this integration occurred. Moreover, this increase was significantly amplified by the remoteness and proportion of fast deliveries in these areas. TB24 CC Room 205A In Person: Emerging Topics in Revenue Management General Session Chair: Chin-Chia Hsu, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States 1 - Dynamic Pricing and Demand Learning for a Large Network of Products Prem Talwai, MIT, Cambridge, MA, 02142, United States, N. Bora Keskin, David Simchi-Levi We consider a seller offering a large network of N products over a time horizon of T periods. The seller does not know the products’ demand model and can dynamically adjust product prices to learn the demand model based on sales observations. The seller aims to minimize its regret, i.e., the revenue loss relative to a clairvoyant who knows the underlying demand model. We consider a sparse set of demand relationships between products, and design a dynamic pricing-and- learning policy that achieves near-optimal regret performance in terms of N and T. We also show that under certain sparsity conditions, the seller’s regret can be independent of N. 2 - Hotel Demand Forecasting Using a Time Varying Arrival Rate Alexander Robinson, University of California-Irvine, Long Beach, CA, 90815-4362, United States, John G. Turner Effective pricing is important in the hotel industry. Especially for budget hotels, price is frequently the primary point of differentiation for a customer. Estimating a function price-dependent demand function presents a number of challenges, including endogeneity and censored data. In this paper, we address these issues by modeling demand as the product of a time-varying arrival probability and purchase probability. We adapt the Expectation Maximization (EM) algorithm to estimate these probabilities separately. We also present a variant of EM that can make convergence time up to 90% faster, with little to no loss of prediction accuracy. 3 - Persuasion, News Sharing, and Cascades on Social Networks Chin-Chia Hsu, PhD Candidate, Massachusetts Institute of Technology, Cambridge, MA, 02139, United States, Amir Ajorlou, Ali Jadbabaie We study a model of online news dissemination on a Twitter-like social network. Given a news item and its credibility, agents with heterogeneous priors strategically decide whether to share the news with their followers. An agent shares the news, if the news can persuade her followers to take an action (such as voting) in line with the agent’s perspectives. We describe the agent’s decision making and the conditions that lead to sharing the news with followers, and characterize the size of news spread at the equilibrium of the news-sharing game. We further identify the conditions under which the news with low credibility can spread wider than highly credible news. In particular, we show that when the network is highly-connected or the news is not a “tail event”, a sharing cascade can occur even with news that is not credible.

TB26 CC Room 206A In Person: Mixed-integer Optimization in Defense Applications General Session Chair: Robert Mark Curry, United States Naval Academy, Annapolis, MD, 21403-4616, United States 1 - Robust Minimum-Cost Flow Problems under Multiple Ripple Effect Disruptions Mehdi Ansari, Oklahoma State University, Stillwater, OK, United States, Juan Sebastian Borrero, Leonardo Lozano We study multi ripple effect disruptions over a network as a defender-attacker optimization problem. The defender acts on an uncertain objective function whose parameters are determined by the attacker who operates multi disruptions on the network. A cutting generation algorithm is presented to find the robust optimal solution of the bilevel programming. The attacker solves a mixed-integer programming on an uncertainty set to identify the worst realization of parameters. In this regard, two different cost functions are proposed and the formulation of the subproblem has been modified to enhance the performance of the method. The algorithm has been tested on generated grid networks and real- world datasets. The results help decision-makers to react immediately after severe disruptions like earthquakes in the populated urban areas. 2 - Integer Programming Models for Optimal Naval Placement in Contested Waters Robert Mark Curry, United States Naval Academy, Annapolis, MD, 21403-4616, United States Territorial claims in contested maritime settings have long been disputed. Some actors proceed by claiming previously unoccupied islands and creating massive artificial islands in order to make significant progress in expanding financial and military control over these waters. In order to halt this expansion, we assume a country’s naval forces are able to fortify islands either already occupied or currently occupied by an ally force. We explore a variety of methods for determining the valuation of islands in contested waters. We next formulate and solve an integer program to build a tree of islands that maximizes the total value of occupied islands. We then analyze the sensitivity of our solutions to determine their efficacy under varying parameters. Finally, we explore a novel interdiction problem in which the adversary reacts optimally to our naval placement decision. TB27 CC Room 206B In Person: Recent Advances in Stochastic Gradient Algorithms General Session Chair: Lam M. Nguyen, IBM Thomas J. Watson Research Center, Ossining, NY, 10562-6037, United States Co-Chair: Trang H Tran, Cornell University, NY, United States 1 - Shuffling Gradient-Based Methods Trang Tran, Cornell University, Ithaca, NY, United States, Lam M. Nguyen, Quoc Tran Dinh, Katya Scheinberg We combine two advanced ideas widely used in optimization for machine learning: shuffling strategy and momentum technique to develop a novel method with momentum for finite-sum minimization problems. We establish that our algorithm achieves a state-of-the-art convergence rate for any shuffling strategy under standard assumptions. In particular, if a random shuffling strategy is used, we can further improve our convergence rate by a fraction of the data size. When the shuffling strategy is fixed, we develop another new algorithm that is similar to existing momentum methods. We prove the same convergence rate of this algorithm under the L-smoothness and bounded gradient assumptions. We demonstrate our algorithms via numerical simulations on standard datasets and compare them with existing shuffling methods.

106