Informs Annual Meeting 2017

TB16

INFORMS Houston – 2017

TB17

4 - Evolution of Supply Networks Nikolay Osadchiy, Emory University, 1300 Clifton Rd NE, Atlanta, GA, 30322, United States, nikolay.osadchiy@emory.edu, Vishal Gaur, Maximiliano Udenio Using a large panel of firm-level buyer-supplier relationships, we study evolution of supply networks over time and implications for idiosyncratic and systematic risk.

340A Robust Optimization and Applied Probability Sponsored: Applied Probability Sponsored Session Chair: Chaithanya Bandi, Northwestern University, Evanston, IL, 60208, United States, c-bandi@kellogg.northwestern.edu 1 - Beyond Worst-case: A Probabilistic Analysis of Affine Policies in Dynamic Optimization Omar El Housni, Columbia University, 110 Morningside Drive, Apt 64, New York, NY, 10027, United States, oe2148@columbia.edu, Vineet Goyal Affine policies are widely used in dynamic optimization where computing an optimal adjustable solution is usually intractable. While the worst case performance of affine policies can be significantly bad, the empirical performance is observed to be near-optimal for a large class of problem instances. In this paper, we aim to address this stark-contrast. In particular, we consider a two-stage dynamic robust optimization problem with linear covering constraints and uncertain right hand side and show that affine policies give a good approximation on random instances generated from a large class of distributions including the Henry Lam, University of Michigan, 1205 Beal Avenue, Industrial & Operations Engineering, Ann Arbor, MI, 48109, United States, khlam@umich.edu, Jeff Hong, Zhiyuan Huang We discuss a statistical framework to integrate data into robust optimization (RO) based on prediction set learning and a simple data-splitting validation scheme that achieves finite-sample statistical guarantees on the feasibility of the underlying uncertain constraints. We demonstrate how the framework satisfies a dimension- free sample size requirement on feasibility guarantees and how it provides a platform to integrate machine learning tools into RO driven by convoluted or high-dimensional data. 3 - Distributionally Robust Inventory Control when Demand is a Martingale Linwei Xin, University of Chicago, Booth School of Business, Chicago, IL, 60637, United States, linwei.xin@chicagobooth.edu, David Goldberg Independence of random demands across different periods is typically assumed in multi-period inventory models. In this talk, we consider a distributionally robust model in which the sequence of demands must take the form of a martingale with given mean and support. We explicitly compute the optimal policy and worst-case distribution in closed form. We prove that at optimality the worst-case distribution corresponds to the setting in which inventory may become obsolete at a random time, a scenario of practical interest. We also compare to the analogous setting in which demand is independent across periods, and identify interesting differences between these two models. 4 - Tractable Distributionally Robust Optimization with Data Zhi Chen, National University of Singapore, NUS Business School, PhD B1-02 Biz2 Building, Singapore, 117592, Singapore, chenzhi@u.nus.edu, Melvyn Sim We present a unified and tractable framework for distributionally robust optimization then encompasses a variety of statistical information including, among others things, expectation constraints, uncertain probabilities of regions and Wasserstein distance. We also introduce a new region based linear decision rule that provides better approximations of adaptive distributionally robust optimization problems and this can also be applied in situations when the recourse variables are discrete.The framework can be incorporated in an algebraic model tool box to facilitate modeling of robust optimization problems. commonly used distributions such as uniform and Gaussian. 2 - Learning-based Robust Optimization: Procedures an d Statistical Guarantees

TB16

332F Operations Management Contributed Session Chair: Katsunobu Sasanuma, Stony Brook University, Stony Brook, NY, United States, katsunobu.sasanuma@stonybrook.edu 1 - New Heuristics for the Lost Sales Inventory Problem Peter L. Berling, Associate Professor, Lund University, Box 118, Lund, SE-22100, Sweden, peter.berling@iml.lth.se, Victor Martínez-de-Abénis We develop two heuristics based on the ideas of the ”Single-unit decomposition approach” that originates from Axsäter (1990) and have been further developed by Muharremoglu and Tsitsiklis (2008) among others. Both heuristics use the expected cost per demanded unit inked to the next unit to be ordered but there is a difference in how this cost is defined. The best myopic decision, i.e. the one that minimizes the expected cost per demanded unit, can then be determined and implemented in respective heuristic. The resulting policy is a simple state- dependent base-stock policy and the numerical test show an improvement of the performance compared to traditional base-stock or time-phased policies. 2 - The Optimization Model of Inventory Control of Spare Components when Inventory Fails Liying Cai, Mechanical Technology Institution, Shijiazhuang, China, cailiyingsharon@sina.cn Properly control of spare components is the core element in optimization of inventory control of spare components. This paper established an optimization model of inventory control of spare components when inventory fails. Firstly, we established a single component system with its spare components’ type being irreparable; then analyzed its system lifetime and cost; finally, a case study is carried out, which proved the effectiveness of the proposed optimal model made in this article. 3 - Stochastic Setup Cost Inventory Model with Backorders and Quasiconvex Cost Functions Yan Liang, Stony Brook University, Department of Applied Mathematics and Statistics, Stony Brook, NY, 11794, United States, yan.liang@stonybrook.edu, Eugene A. Feinberg This talk is concerned with a periodic-review setup-cost inventory model with backorders and holding/backlog costs satisfying quasiconvexity assumptions. We show that this model satisfies assumptions that imply the validity of optimality equations for discounted and average cost criteria and the family of discounted relative value functions is equicontinuous. We establish two groups of results: (i) optimality of (s,S) policies for infinite-horizon problems under discounted and average cost criteria, and (ii) convergence of optimal discounted lower thresholds s and discounted relative value functions to their average-cost counterparts as the discount factor converges to 1. 4 - Finding Repair Shop Priorities Ece Zeliha Demirci, Postdoc Researcher, Eindhoven University of Technology, Eindhoven, 5600 MB, Netherlands, e.z.demirci@tue.nl, Joachim Jacob Arts, Geert-Jan van Houtum We investigate the dynamic scheduling problem of a repair shop with a single server. The problem is to allocate the capacity of the repair shop between spare parts and decide on which part to repair first. Using the MDP formulation of the problem, we show that longest queue policy is optimal if the parts have identical characteristics and controlled with the same or different levels of base stock. 5 - Analysis of Perishable Inventory with Substitution under a Base Stock Policy Katsunobu Sasanuma, Assistant Professor, Stony Brook University, Harriman Hall, Stony Brook University, Stony Brook, NY, 11794, United States, katsunobu.sasanuma@stonybrook.edu, Mohammad Delasay, Christine Pitocco, Thomas Raymond Sexton, Alan Scheller-Wolf We study heuristic control policies for perishable inventory systems with substitution. We derive a simple state-dependent base stock policy by approximating the exact optimality condition. We demonstrate its validity by comparing with simulation results.

308