Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

SC13

4 - Supply Chain Network Structure and Environmental Information Disclosure

n SC13 North Bldg 126B Healthcare Management: Meet the Editors Sponsored: Manufacturing & Service Oper Mgmt/Healthcare Operations Sponsored Session Chair: Nicos Savva, London Business School, London, NW1 4SA, United Kingdom Co-Chair: Susan F. Lu, Purdue University, Purdue University, West Lafayette, IN, 47907, United States 1 - Healthcare Operations Management at POM Sergei Savin, The Wharton School, 570 Jon M. Huntsman Hall, 3730 Walnut Street, Philadelphia, PA, 19104, United States The POM Journal has maintained a strong tradition of publishing high-quality research in healthcare operations. The talk will review the mission of the Healthcare Operations Management Department at POM and the types of papers published in recent years. 2 - Healthcare Operations Management at OR Ozlem Ergun, Northeastern University, Mechanical and Industrial Engineering, Boston, MA, 02115, United States The Policy Modeling and Public Sector OR area seeks papers that define important problems and use innovative mathematical models and analytics to solve them for improving outcomes. Especially welcome are manuscripts that utilize data to develop original models linking the design and operations of public programs and systems to recognizable policy outcomes and recommendations. Priority is given to papers that present novel and convincing data-driven model-based analyses of issues likely to generate widespread public interest, awareness, and impact. 3 - Introduction of the New Healthcare Management Department of Management Science Stefan Scholtes, University of Cambridge Management Science has recently introduced a new department of Healthcare Management. I will introduce the new department and its editorial statement. 4 - MSOM Submissions Perspective on Health Care Management Research Morris A. Cohen, Professor, University of Pennsylvania, 546 Huntsman Hall, 3730 Walnut Street, Philadelphia, PA, 19104, United States This presentation will provide an overview of the historical record of submissions to the Service Operations Department of MSOM that fall in to the area of Healthcare Management. This will include a summary of the volume and outcome of papers submitted to the department during the period 2016 to the present. We will also review the major topic areas, source (authors affiliation), reasons for acceptance or rejection, and the nature of the research (i.e. analytical, empirical, policy, other). The presentation will conclude with observations on trends in healthcare management research as well as challenges, and opportunities for future research directions as suggested by the MSOM experience. 5 - Healthcare Management at the INFORMS Journal on Computing Paul Brooks, Virginia Commonwealth University, Dept of Stat Sci and OR, P.O. Box 843083, Richmond, VA, 23284, United States The Applications in Biology, Medicine, & Healthcare area of the INFORMS Journal on Computing publishes articles and welcomes manuscripts in the area of healthcare management. We seek manuscripts that include OR, computing, and a relevant application, and provide a significant contribution in at least one of the these or in a combination.

Marcus A. Bellamy, Rafik B. Hariri Building, 595 Commonwealth Avenue, Boston, MA, 02215, United States, Suvrat Dhanorkar, Ravi Subramanian Recognizing that supply network structure has implications for a focal firm’s ability to access environmental information embedded in its supply network, this paper draws on structural, environmental, and financial data from Bloomberg to test the relationship between a focal firm’s supply network structure and its extent of environmental information disclosure. n SC12 North Bldg 126A Joint Session Frontiers/Practice Curated: Statistical Learning and Optimization Emerging Topic: OR Frontiers Emerging Topic Session Chair: Emma Frejinger, Universite de Montreal, Montreal, QC, H3C 3J7, Canada 1 - Learning MILP Resolution Outcomes before Reaching Time-limit Andrea Lodi, École Polytechnique de Montréal, Montréal, QC, Canada, Martina Fischetti, Giulia Zarpellon The solution of some MILPs still presents challenges for solvers and may require hours of computations, so that a time-limit is often provided by the user. Nevertheless, it could be useful to get a sense of the optimization trends after only a fraction of the time-limit, and ideally be able to tailor the use of the remaining solution time in a more strategic way. Looking at the evolution of a partial branch-and-bound tree for a MILP, up to a certain fraction of the time-limit, we aim to predict whether the problem will be solved to proven optimality before timing out. We exploit Machine Learning tools, and summarize the progress of a MILP solution process to cast a prediction within a classification framework. 2 - Optimizing Decision Diagrams Size and Bound via Reinforcement Learning Louis-Martin Rousseau, École Polythechnique de Montréal, Cp 6079 Succ Centre-Ville, Montréal, QC, H3C 3A7, Canada, Quentin Cappart, David Bergman Decision Diagrams are a recent technology enhancing optimization methods. They can be used in Integer Programming for tightening relaxation bounds. Their performances are highly dependent on the variable ordering chosen. Finding an optimal ordering is NP-complete. Recent research in Machine Learning has also shown that reinforcement learning can be used for solving NP-hard problems. Following this trend, we propose to use a similar approach in order to reduce the size of Decision Diagrams. 3 - Learning Heuristics for the TSP by Policy Gradient Yossiri Adulyasak, HEC Montréal, 3000 Cote-Sainte-Catherine, Montreal, QC, H3T 2A7, Canada, Michel Deudon, Pierre Cournut, Alexandre Lacoste, Louis-Martin Rousseau We extend the neural combinatorial optimization framework to solve the traveling salesman problem (TSP). The neural network is trained using reinforcement learning to predict a distribution over city permutations. We designed our own critic to compute a baseline for the tour length which results in more efficient learning. We further enhance the solution approach with the well- known local search heuristic and the approach could outperform a high performance heuristic (OR-Tools). Our approach based on machine learning techniques could learn good heuristics which, once being enhanced with a simple local search, yield promising results. 4 - A Machine Learning Algorithm for Fast Prediction of Solution Descriptions to an ILP Emma Frejinger, Université de Montréal, FAS, Pavillon Andre- Aisenstadt, Montreal, QC, H3C 3J7, Canada, Eric Larsen, Sébastien Lachapelle, Yoshua Bengio, Simon Lacoste-Julien, Andrea Lodi We propose a methodology to predict descriptions of solutions to discrete stochastic optimization problems in short computing time. We approximate the solutions based on supervised learning and the training dataset consists of a large number of deterministic problems that have been solved independently (offline). Uncertainty regarding a subset of the inputs is addressed through sampling and aggregation methods. Our application concerns booking decisions of containers on double-stack trains. The results show that deep learning algorithms make predictions high accuracy in milliseconds or less.

65