Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

WB41

availability (including soil capacity) and local nutritional needs to evaluate the potential impact of urban food systems in the growing urban Southwest. 2 - Long-term Production Planning Under Demand Uncertainty Joost T. de Kruijff, Eindhoven University of Technology, Den Dolech 2, Eindhoven, 5600 MB, Netherlands, Nico P. Dellaert, Cor A. Hurkens, Ton de Kok We focus on long-term production planning. Where mid-term production planning coordinates the release of materials and capacity aiming to minimize the costs of inventory and backlog given the demand, long-term planning focusses on decisions regarding investments in resources and materials to support long-term sales. Taking into account the uncertainty in the long-term sales, we make a tradeoff between revenue creation and investments in capacity and/or inventory. These investments do not take effect immediately; we explicitly take into account the possibility to change plans when we learn more about the future. We base our Yuji Nakagawa, Professor, Kansai University, Department of Informatics, 2-1-1 Ryouzenji-Cho, Takatsuki-City, 569-1095, Japan, Yoshiko Hanada, Youichi Takenaka, Youichi Takenaka, Chanaka Edirisinghe In high-dimensional regression, where the number of explanatory variables is much larger than the sample size, a statistical problem known as the æcurse of dimensionality’ arises. The strategy of performing all possible regressions is computationally-impractical. We model using non-convex discrete optimization to minimize MSE. We describe an application involving a large number of SNPs in genomic studies for cancer detection. 4 - Using Json and Nosql Databases to Deliver Data for Deploying Optimization Models on the Cloud Bjarni Kristjansson, Maximal Software Inc., 2111 Wilson Boulevard, Suite 700, Arlington, VA, 22201, United States When preparing optimization models for deployment, it is often not the formulation of the model or the solving that is the most difficult part, but rather how to organize and deliver the data efficiently to the model. Using modeling languages, this problem can be alleviated by importing data directly from SQL databases and spreadsheets, but this is still not easy enough. In this presentation, we will demonstrate how using the JSON data format can be particularly well suited for organizing optimization data. Furthermore, using NoSQL databases such as MongoDB and CouchDB, which are very efficient at storing dynamic data, give the model developer new opportunities for deploying optimization on the cloud. Chair: Gilberto Montibeller, Loughborough University, School of Business & Economics, Loughborough, LE11 3TU, United Kingdom 1 - Technology Adoption in a Declining Market Maria Lavrutich, Norwegian University of Science and Technology, Trondheim, Norway Rapid technological developments are inducing the shift in consumer demand from existing products towards new alternatives. When operating in a declining market, the profitability of incumbent firms islargely dependent on the ability to correctly time the introduction of product innovations. This paper considers the optimal innovation I nvestment in the context of the declining market. We study the problem of a firm that has an option to undertake the innovation investment and thereby either to add a new product to its portfolio or to replace the established product by the new one. We are able to quantify the value of the option to adopt a new technology, as well as the optimal timing to exercise it. 2 - Integrated Scenario based Robust Planning Approach for Foresight in Blockchain Technology Leili Soltanisehat, The University of Oklahoma, 815 Russell Circle, E. Brooks St, Norman, OK, 73071, United States, Reza Alizadeh Blockchain faces major future challenges related to technological, economic, social, and probably political aspects. Here we present a scenario-building framework based on the Global Business Network method to help blockchain to develop more resilient conservation policies when it faces unpredictable and external uncertainties. The approach combines several foresight methods such as Delphi, Political, Economic, Social, and Technological analysis, and Cross-Impact Analysis. modelling on practices in the high-tech low-volume industry. 3 - Nonconvex Discrete Optimization Approach for Highdimensional Regression n WB41 North Bldg 226C DA Arcade (Technology) Sponsored: Decision Analysis Sponsored Session

n WB39 North Bldg 226A Recent Advances in Simulation Analysis Methodology Sponsored: Applied Probability Sponsored Session Chair: Eunhye Song, Penn State University, University Park, PA, 16802, United States 1 - Mo-score for Multi-objective Ranking and Selection Eric Applegate, Purdue University, West Lafayette, IN, 47907, United States, Guy Feldman, Susan R. Hunter, Raghu Pasupathy Consider the context of selecting Pareto-optimal systems from a finite set of systems based on multiple stochastic objectives. We characterize the asymptotically optimal sample allocation that maximizes the rate of decay of the probability of misclassification, and we provide a multi-objective SCORE (MO- SCORE) allocation for use when the number of non-Pareto systems is large relative to the number of Pareto systems. The MO-SCORE allocation exploits phantom Pareto systems in high dimensions, which we find using a dimension- sweep algorithm. 2 - Effort Allocation and Statistical Inference for Multistart Stochastic Gradient Descent Saul Toscano, Cornell University, Peter Frazier Multistart stochastic gradient descent methods are widely used for gradient-based stochastic global optimization. However these methods seem to waste computational resources: when several starts are run to convergence at the same local optimum, all but one fail to produce useful information; when a start converges to a local optimum worse than an incumbent solution, it fails to produce useful information. We propose a rule for allocating computational effort across starts, which allocates more resources to the most promising starts. This allocation rule is based on a new statistical model, which agrees with known convergence rates for SGD. Numerical results show the effectiveness of our rule. 3 - Online Quantification of Input Uncertainty for Parametric Models Enlu Zhou, Georgia Institute of Technology, 755 Ferst Drive NW, Atlanta, GA, 30332-0205, United States, Tianyi Liu It has become increasingly important to assimilate “online data that arrive sequentially in time for real-time decision. Input uncertainty quantification in stochastic simulation has been developed extensively for batch data that are available all at once, but little has been studied for online data. We propose a computationally efficient method to incorporate online data in real time for input uncertainty quantification of parametric models. We show finite-sample bounds and asymptotic convergence for the proposed method, and demonstrate its performance on a simple numerical example. 4 - Efficient Simulation Design for Risk Management of Large Variable Annuity Portfolios Ben Feng, University of Waterloo, 217 Holbeach Cres, Waterloo, ON, N2J 4Y3, Canada Variable Annuities (VAs) are popular insurance products in practice. Monte Carlo simulation is usually required to valuate VA contracts due to their complexities. However, computations required for valuation of every contracts in a large VA portfolio via standard MC could be prohibitively expensive. We develop and examine experiment designs that can significantly improve the efficiency of simulation experiments for large VA portfolios. We also identify pitfalls in some of the existing methods and propose corresponding improvements. We show that the proposed simulation procedure has both higher accuracy and lower computational requirement than the state-of-art procedures. n WB40 North Bldg 226B Practice – Modeling and Optimization for Decision Making I Contributed Session Chair: Bjarni Kristjansson, Maximal Software Inc., Arlington, VA, 22201, United States 1 - Optimize Urban Agriculture to Address Food Deserts in Semi-arid Areas Daoqin Tong, Associate Professor, Arizona State University, School of Geographical Sciences and Urban Planning, 975 S. Myrtle Ave., Tempe, AZ, 85287, United States, Qing Zhong, Courtney Crosson Accessing to healthy, affordable food remains a great challenge in many neighborhoods, leading to the existence of food deserts. Recently, urban agriculture has been identified as an important strategy to help address urban food deserts. However, producing food in the urban setting can be challenging due to limited land availability and high water cost. This paper develops a spatial optimization model to integrate rainwater harvesting, urban municipal land

455