2016 INFORMS Annual Meeting Program

TB65

INFORMS Nashville – 2016

TB63 Cumberland 5- Omni Deterministic Network Design Sponsored: TSL, Freight Transportation & Logistics Sponsored Session Chair: Mike Hewitt, Loyola University Chicago, NA, Chicago, IL, 60611, United States, mhewitt3@luc.edu 1 - Barge Scheduled Service Network Design With Resource And Revenue Management Teodor Gabriel Crainic, Universite du Quebec a Montreal, TeodorGabriel.Crainic@cirrelt.net, Ioana Bilegan, Yunfei Wang We study the incorporation of revenue management (RM) considerations, usually tackled at the operational-planning level, into tactical planning models for consolidation-based freight transportation carriers, and the impact of this integration on the selection of customer to service and on the structure of the service network, e.g., should the carrier increase the offer of service through more departures or larger vessels in order to later be able to capture spot demand? We present a service network design model with resource and RM considerations, a meta-heuristic, and the experimental results and insights obtained in the context of intermodal barge transportation. 2 - The Value Of Flexibility In Long-haul Transportation Network Design Mike Hewitt, Loyola University Chicago, mhewitt3@luc.edu, Natashia Boland, Martin W P Savelsbergh Freight transportation carriers are facing increased demands from customers for shorter service standards. At the same time, some cusotmers are flexible in terms of when they want their shipments delivered, and will accept longer delivery times if given a discount. In this talk we present a model that will not only design a long-haul transportation network, but will do so while also determining which customers to offer a discount to in order to have more time for delivery. We present a solution approach for the model and the results of an extensive computational study. 3 - Decomposition Methods For Multi-period Network Design Problems We devise decomposition methods to solve large-scale instances of multi-period network expansion problems. For capacitated networks, we devise a custom heuristic procedure combined with arc-based Lagrange relaxation. For uncapacitated networks, we employ Bender’s decomposition, where the subproblems are decomposable per period and per commodity. We formulate the problem of generating Pareto Optimal cuts, and based on structural properties of optimal solutions we devise a heuristic approach to solve it, thereby improving the original Benders cuts. Computational results demonstrate the efficiency of this approach. 4 - New Lagrangian Relaxation For Multicommodity Capacitated Network Design Mohammad Rahim Akhavan, Universite de Montreal (DIRO), Monrteal, QC, Canada, Akhavanm@iro.umontreal.ca, Teodor Gabriel Crainic, Bernard Gendron The usual Lagrangian relaxations for multicommodity capacitated network design are the so-called shortest path and knapsack relaxations, which are obtained, respectively, by relaxing linking constraints and flow conservation equations. We present a new reformulation and Lagrangian relaxation for the problem. We show that the Lagrangian dual bound improves upon the so-called strong LP bound (known to be equal to the Lagrangian dual bounds of the shortest path and knapsack relaxations). TB64 Cumberland 6- Omni Multi-objective Optimization: Algorithms and Applications Sponsored: Multiple Criteria Decision Making Sponsored Session Chair: Lakmali Weerasena, University of Tennessee, College Dr, Chattanooga, TN, 31705, United States, lweeras@g.clemson.edu 1 - New Multicriteria Models For Robust Data Classification In Supervised Learning Alexander Engau, University of Colorado Denver, alexander.engau@ucdenver.edu Data classification is a key task for predictive analytics, data mining and supervised machine or statistical learning. In extension of its current state-of-the- art optimization approaches this presentation highlights several new multicriteria Ioannis Fragkos, Rotterdam School of Management, fragkos@rsm.nl, Jean-Francois Cordeau, Raf Jans

mixed-integer goal programming models that can further improve their performance for classification and prediction by combining a variety of different objectives including solution accuracy as well as total and minimum or maximum internal or external deviation. Computational experiments on financial and medical data sets are reported and demonstrate promising results with highly improved robustness and better classification accuracy overall. 2 - Utility Indifference Pricing Under Incomplete Preferences Via Vector Optimization Firdevs Ulus, Bilkent University, firdevs@bilkent.edu.tr Under some assumptions on an incomplete preference relation, utility maximization problem is a convex vector optimization problem. Accordingly, the utility buy and sell prices are defined as set valued functions of the claim. It has been shown that the buy and the sell prices recover the complete preference case where the utility function is univariate. Moreover, buy and sell prices satisfy some monotonicity and convexity properties as expected. It is possible to compute these set valued prices by solving convex vector optimization problems. 3 - Local Branching Algorithm For Approximating The Pareto Set Of The Multiobjective Set Covering Problem Lakmali Weerasena, University of Tennessee at Chattanooga, Chattanooga, TN, United States, lweeras@g.clemson.edu The multiobjective set covering problem (MOSCP), a challenging combinatorial optimization problem, has received limited attention in the literature. We present an algorithm to approximate the Pareto set of the MOSCP. The proposed algorithm applies a local branching approach on a tree structure and is enhanced with a node exploration strategy specially developed for the MOSCP. The key idea is to partition the search region into subregions based on the neighbors of a reference solution. Numerical experiments confirmed that the proposed algorithm performs well on the MOSCP. Results on a performance comparison with benchmark algorithms from the literature show that the new algorithm is competitive. 4 - Interactive Weight Region-based Approach For Multiobjective Optimization Problems We introduce an interactive weight region-based approach that can iteratively find the most preferred solution of a decision maker (DM) after exploring a small fraction of all nondominated solutions. To obtain preference information, the DM is given a series of questions to compare and these comparisons define constraints restricting the weight region. New solutions are obtained by using diverse weight vectors generated from the remaining weight region via a mixed integer programming formulation. We develop two finitely converging algorithms for multi-objective linear and integer programs respectively. The results show that the algorithms terminate after a reasonable number of iterations. TB65 Mockingbird 1- Omni Digital Business Models and Strategies in the Era of Analytics Sponsored: Information Systems Sponsored Session Chair: Ling Xue, Georgia State University, Georgia State University, Atlanta, GA, 30302, United States, lxue5@gsu.edu 1 - Relationships Between Online Daily Deal Promotions And Local Retailers’ Online Reputation Gang Wang, University of Delaware, gangw@udel.edu Online daily deal sites such as Groupon have recently provided an innovative marketing tool for local retailers. On the one hand, a local retailer’s online reputation is an important fact that may impact its decision whether to run a promotion. On the other hand, an online daily deal promotion may also impact the local retailer’s online reputation. In this study, we study the causal relationships between a local retailer’s promotion decision and its online reputation using data collected from Groupon and Yelp. Our current results enhance understanding of local retailers’ Groupon promotion decisions and yield important implications related to daily deal sites. Mehmet Basdere, Northwestern University, Evanston, IL, United States, mehmetbasdere2016@u.northwestern.edu, Sanjay Mehrotra, Karen Smilowitz

285