Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TC04

3 - Robust Multi-product Newsvendor Model with Substitution under Cardinality-constrained Uncertainty Jie Zhang, Virginia Tech, 820 Newport Terrace, Blacksburg, VA, 24060, United States, Weijun Xie This paper studies robust multi-product newsvendor model with product substitutions (RMNMPS). The objective of RMNMPS is to determine the optimal order quantities, which maximize the worst-case total profits against budget uncertainty set of the demand. Although RMNMPS is in general nonconvex and NP-hard, we are able to identify several special cases, where the optimal order quantities can be completely characterized, and interesting managerial insights are drawn. For a general RMNMPS, we develop an efficient cutting plane based solution approach by exploring the submodularity of inner minimization problem. The numerical study demonstrates the effectiveness of the proposed algorithm. 4 - A Finite E-convergence Algorithm for Two-stage Convex 0-1 Mixed-integer Nonlinear Stochastic Programs with Mixed-integer First and Second Stage Variables Can Li, Carnegie Mellon University, Pittsburgh, PA, United States, Ignacio E. Grossmann We propose a generalized Benders decomposition-based branch and bound algorithm, GBDBAB, to solve two-stage convex 0-1 mixed-integer nonlinear stochastic programs with mixed-integer variables in both first and second stage decisions. We construct the convex hull of each subproblem by applying basic steps to convert each subproblem from conjunctive normal form (CNF) to disjunctive normal form (DNF). We prove the algorithm has finite ?-convergence if we branch on the continuous first stage variables. Since constructing the convex hull can be expensive, we propose a sequential convexification scheme that progressively applies basic steps to the CNF. 5 - Computational Evaluation of New Dual Bounding Techniques for Sparse PCA Guanyi Wang, Georgia Institute of Technology, Atlanta, GA, United States, Santanu Subhas Dey, Rahul Mazumder Principal component analysis (PCA) is one of the most widely used dimensionality reduction method in statistics. For additional interpretability, it is desirable to require cardinality constraint, known as the sparse principal component analysis (SPCA). However, the SPCA problem and its $\ell_1$ relaxation are hard to compute. We give a framework (convex integer program, IP) that certificates the optimality of solutions of SPCA problem, via dual bounds. We show that, in theoretical, the dual bound obtained from convex IP problem is affinely upper bounded by the optimal value of the SPCA problem, and in practical, plausible dual bounds are obtained via the convex IP method in acceptable time. n TC03 North Bldg 121C Product and Process Innovation in Healthcare and Energy Industries Sponsored: Technology, Innovation Management & Entrepreneurship Sponsored Session Chair: Zhili Tian, Florida International University, Weston, FL, 33327- 2444, United States 1 - Capacity Planning with Probabilistic Outcome Ambiguity Heejung Kim, University of California, Berkeley, Albany, CA, 94706, United States, Philip Kaminsky We consider a capacity planning model where demand in each period is determined by the outcome of series of binary events, such as clinical trials in pharmaceutical industries. The outcome of these events is often uncertain, and estimating exact probability of outcome is difficult. We develop approaches to solve capacity expansion planning models that are robust to ambiguities in probability of success for different objectives - minimizing expected cost, value at risk and conditional value at risk. We formulate these models as (stochastic) robust integer programs with scenario trees. We develop and test a variety of heuristics for this setting. 2 - Operating Room Scheduling with Sequential Tasks and Defined Wait Intervals James W. Hamister, Wright State University, Raj Soin College of Business, Dayton, OH, 45435, United States, Michael Magazine, George Polak We develop an optimization model for scheduling paired tasks subject to intervening procedures in an operating room environment. The durations of the intervening procedures are bounded from below and above. The objective is to minimize overall makespan. The model is tested with empirical data from a tissue bank service.

3 - Optimal Investment in Product Development and Marketing Communication in Pharmaceuticals Zhili Tian, Coastal Carolina University, 1305 Harvestor Cir, Myrtle Beach, SC, 33327-2444, United States Marketing spending on drugs is often close to that for developing a new drug. Only around 20 new drugs are approved every year for the pharmaceutical industry. The low rate of new drug development stems from the financing constraints at many firms because such high investments from both marketing and new drug development. Because of high risk in new drug development, firms have to use internal funding to support their drug development. We develop models to address the investment in the current drugs and development of new drugs where the funding comes from the internal source. Our methods optimally allocate the fund between marketing and new drug development while considering financing constraints. 4 - Evaluating Emerging Energy Technology Lifecycles using Mixed Integer Programming Joshua Pearson, Colorado School of Mines, Golden, CO, 80401, United States We develop methodologies to evaluate the feasibility of market penetration for new technologies based economic viability. We use as a case study a concentrated solar power device. With an existing piece of engineering software, the System Advisory Model, we examine the financial implications of various industry- standard designs. For each design, we employ a mixed-integer program to determine a cost-minimizing operational strategy for the device under certain market conditions. Finally, we assess long-term economic viability of the device, and draw conclusions about the types of methodologies that can predict the success of investment strategies for emerging technologies. n TC04 North Bldg 122A Integrated Methods and Decomposition Approaches Sponsored: Optimization/Integer and Discrete Optimization Sponsored Session Chair: Joris Kinable, PhD, Eindhoven University of Technology, P.O. Box 513, Eindhoven, 5600 MB, Netherlands 1 - Home Healthcare Integrated Staffing and Scheduling Louis-Martin Rousseau, Ecole Polythechnique de Montreal, Cp 6079 Succ Centre-Ville, Montreal, QC, H3C 3A7, Canada, Maria Isabel Restrepo Ruiz Workforce planning for home healthcare represents an important and challenging task involving complex factors associated with labour regulations, caregivers’ preferences, and demand uncertainties. Motivated by these challenges, we present a two-stage stochastic programming model for employee staffing and scheduling which relies on Context-free Grammars hyper-graphs. The proposed model is tested on real-world instances, where we evaluate the impact in costs, caregiver utilization, and service level, by using different scheduling policies and recourse actions. 2 - Consistency for Mixed Integer Programming John Hooker, Carnegie Mellon University, Tepper School of Business, Pittsburgh, PA, 15213, United States, Danial Davarnia Concepts of consistency have long played a key role in constraint programming but have not arisen in mathematical programming. Consistency is fundamental, because problems that satisfy consistency properties can be solved with less backtracking. We show how a basic type of consistency can be adapted to mixed integer programming (MIP) by taking linear relaxations into account. This novel perspective on MIP helps explain why certain cuts are effective and suggests new techniques for accelerating search. 3 - Optimizing Trip Sharing in Community-based Urban Commuting Pascal Van Hentenryck, University of Michigan, 1813 IOE Building, 1205 Beal Avenue, Ann Arbor, MI, 48108-2117, United States, Mohd Hafiz Hasan We describe optimization techniques for optimizing trip sharing in urban commuting. The starting point is the identification of important properties that must be satisfied by any successful trip-sharing platform, which give rise to complex optimization problems. The presentation describes potential solution techniques and evaluates them on a real case study.

307