Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

TE42

2 - Support Clinical Decision Making Using Genetic Analytics Cheng Zhu, McGill University, 701-801 Sherbrooke Est, Montreal, QC, H2L 0B7, Canada Cancer has been one of the most dangerous risk factors to people’s health during last few decades. Though there have been several mainstream interventions of cancer, oncologists do NOT have a unanimous decision on the optimal treatment for an individual diagnosed with a cancer. Based on a mRNA sample from the Gene Expression Omnibus (GEO), we are identifying the key modules for PDAC metastasis, in order to provide solid support for clinical decision making. 3 - Experience of Disability Among Older Adults Manaf Zargoush, McMaster University, 1280 Main Street West- DSB 204, Hamilton, ON, L8S 4M4, Canada, Farrokh Alemi, Somayeh Ghazalbash Limited information is available regarding the complex sequence of functional change and death for older adults. Clinicians use this information to set priorities for their treatments. Policymakers can use these data to set value-based incentives for care-providers. We fill the gap by using an innovative blend of several data analytics methods based on 1.8M functional assessments of 300K residents in VA Community Living Centers. We report details on the likelihood, duration until next event, and sequence of functional change and death. Such a patient-centric benchmark is vital to ensuring optimal care for persons in nursing home settings and can be used for planning for end-of-life disabilities. n TE42 North Bldg 227A Simulation Contributed Session Chair: Xi Chen, Virginia Polytechnic Institute and State University, Blacksburg, VA, 24601, United States 1 - Four Ways to Kill a Vampire: An Agent-based Simulation of the Emergence of Dominant Technologies Christian Stummer, Prof., Bielefeld University, Universitaetsstr. 25, Bielefeld, 33615, Germany, Michelle D. Haurand Although superior technologies often establish themselves as a standard in a market, sometimes an inferior one succeeds instead. We have developed an agent-based simulation as a means for studying the corresponding formation processes and for analyzing the effectivity of some management measures (e.g., marketing campaigns). As a field of application, we chose the (fictitious) world of vampires, in which humans must adopt weapon technologies to defend themselves against their supernatural adversaries. In the model, we account for learning effects, word of mouth, and normative social influence. 2 - Gradient Based Criteria for Sequential Experiment Design Collin Erickson, Northwestern University, 2145 Sheridan Road, Room C210, Evanston, IL, 60208, United States, Bruce Ankenman, Matthew Plumlee, Susan M. Sanchez Computer simulation experiments are commonly used as an inexpensive alternative to real-world experiments to form a metamodel that approximates the input-output relationship of the real-world experiment. While a user may want to understand the entire response surface, they may also want to focus on interesting regions of the design space, such as where the gradient is large. In this paper we present an algorithm that adaptively runs a simulation experiment that focuses on finding areas of the response surface with a large gradient while also gathering an understanding of the entire surface. We consider the scenario where batches of points can be run simultaneously, such as with multicore processors. 3 - A Methodology for Using Simulation as Tools for Collecting Data Farhad Moeeni, Professor, Arkansas State University, Cit Dept, P.O. Box 130, State University, AR, 72467, United States Data collection is a challenging and time-consuming phase of simulation modeling cycle. The task becomes more challenging when the data is related to human perception or behavior. One common approach for collecting such data has been surveys and interviews. Unfortunately, good survey instruments are hard to design and interviews or open-ended questionnaires are difficult to analyze and quantify. In order to alleviate these shortcomings, we propose a method to apply simulation systems as tools for collecting data on human behavior that otherwise is hard or impossible to collect. We will demonstrate the methodology through an example. 4 - A Dual-metamodeling Approach for Robust Optimization in Simulation Wenjing Wang, Virginia Polytechnic Institute and State University, 1145 Perry Street, Durham 106, Blacksburg, VA, 24061, United States, Xi Chen In this work, we propose a dual metamodeling approach for robust optimization in simulation. We will investigate different dual-metamodeling approaches for robust optimization in simulation, including a dual Gaussian process model which models the mean and variance response surfaces separately and a variational inference-based heteroscedastic Gaussian process approach that models these two surfaces simultaneously. The performance of the proposed approaches will be demonstrated through numerical examples.

5 - An Efficient Morris Method-based Framework for Simulation Factor Screening Xi Chen, Assistant Professor, Virginia Tech, 1145 Perry Street, Blacksburg, VA, 24061, United States We study the methodological underpinnings of the Morris’s elementary effects method, a model-free factor screening technique originally proposed for deterministic simulation experiments, and develop an efficient Morris method- based framework (EMM) for simulation factor screening. Equipped with an efficient cluster sampling procedure, EMM can simultaneously screen the main and interaction (or nonlinear) effects of all factors and control the overall false discovery rate at a prescribed level. We reveal the connections between EMM and other factor screening methods and examine the resulting implications under some commonly stipulated assumptions in design of experiments. n TE43 North Bldg 227B Long-term Sustainable Energy/Power System Expansion Planning Emerging Topic: Energy and Climate Emerging Topic Session Chair: Qipeng Zheng, University of Central Florida, FL, United States 1 - A Multistage Decision-Dependent Stochastic Bi-level Programming Approach for Power Generation Investment Expansion Planning Qipeng Phil Zheng, University of Central Florida, 12800 Pegasus Dr., P.O. Box 162993, Orlando, FL, 32816, United States, Yiduo Zhan We propose a bi-level optimization model that includes an upper-level multistage stochastic nonlinear expansion planning problem and a collection of lower-level problems that solves for economic dispatch problems under different scenarios. This model seeks for the optimal sizing and siting for both thermal and wind power units to be built to maximizing the expected profit for a profit-oriented power investor. We design probability distribution functions of decision variables and input parameters based on the economies-of-scale theory in electricity systems. We develop a solution algorithm based on column generation for this program. 3 - Proximal-based Dual Decomposition Algorithms With Applications To Capacity Expansion Run Chen, Purdue University, West Lafayette, IN, United States, Andrew Lu Liu Power grids’ planning and operation exhibit extreme multi-scale, ranging from hourly operation to decades of planning. We use proximal-based predictor- corrector enabled dual decomposition to decouple the linkage between different time-scales, proposing a distributed algorithm where primal-minimization is decomposed into massive parallel sub-problems. Convergence is established, as well as the linear convergence rate. Explicit non-anticipativity constraints are used to deal with stochasticity. Applications demonstrate significance of considering ramping constraints in finer time-scale. n TE44 North Bldg 227C Renewable Energy Sponsored: Energy, Natural Res & the Environment/Electricity Sponsored Session Chair: Ming Jin, UC Berkeley, Berkeley, CA, 94720, United States Co-Chair: Somayeh Sojoudi, University of California-Berkeley, Berkeley, CA, 94703, United States 1 - Risked Investment in Renewable Electricity Generation Andy Philpott, University of Auckland, Dept of Engineering Science, Private Bag 92109, Auckland, New Zealand, Michael C. Ferris Using data from New Zealand, we discuss models of the national electricity system that will deliver 100% renewable electricity. Wind and solar energy are intermittent, and run-of-river hydroelectricity plants convert uncertain inflows into uncertain levels of energy, so the planning problem must account for these uncertainties. To compensate, hydroelectric reservoirs provide a means of storing energy, and geothermal power provides a predictable base-load energy source. We compare a socially optimal solution with one that arises from competitive equilibrium with risk-averse investors.

394