Informs Annual Meeting 2017

POSTER SESSION

INFORMS Houston – 2017

38 - A Regularized Inverse Optimization Method to Decide Optimal Product Characteristic Combination and Consumer Taste Segmentation CY (Chor-yiu) Sin, National Tsing Hua University, Kuang-Fu Road, Hsinchu, 30013, Taiwan, cysin@mx.nthu.edu.tw Recent advances of the random coefficient demand model uses the constrained optimization modeling technique of mathematical program with equilibrium constraints (MPEC) and connects the specially-interested problem to the continually-developing power of nonlinear programming solvers. In this work, we construct a semi-nonparametric MPEC by relaxing the distributional assumption on the random coefficient. The empirical study investigating the UK vehicle market share of 212 vehicle models in UK with a large number of potential customers from Year 2001 to Year 2015 shows the effectiveness of this approach. We predict the responses in both the business market and the consumer market. 39 - Reverse Logistics and Shipping Waste Modeling Mazen I. Hussein, Assistant Professor, University of Wisconsin- Platteville, 150 Stonebridge Road, Apartment 206, Platteville, WI, 53818, United States, mazenhussein2212@gmail.com A model based on theory and several modeling techniques is used to estimate and predict the movement and flow of different kinds of waste. Variety of data resources was used. The results can be used as an input to other projects regarding the flow of commodities, logistics, and transportation. 40 - A Dynamic Programming Approach to Determining Residential Thermostat Set Points for Demand Response Arnab Roy, Graduate Fellow, University of Louisville, 3177 South 3rd Street, Louisville, KY, 40214, United States, arnab.roy@louisville.edu, Lihui Bai We propose a dynamic programming approach to determine the optimal set point for a smart thermostat during periods surrounding a Demand Response event. The objective is to minimize the total energy consumption during the peak period (high price) period while keeping the inside temperature within user-defined tolerable range. It is applied in three stages of the event: pre-, during and post- event. The algorithm is tested on three prototypical homes with low, medium, 41 Hadi Karimi, Clemson University, 1108 Tiger Blvd., Apt. 117, Clemson, SC, 29631, United States, hkarimi@clemson.edu, Sandra D. Eksioglu, Jay Devkota In this study we are developing a machine learning framework to facilitate the industrial energy assessment process at small and medium size enterprises. The machine learning framework has two main stages: at first stage, we develop a predictive model to characterize baseline energy use by taking into account the climate conditions and production factors. At second stage, we measure the industrial energy savings by developing a multivariate regression equation and disaggregating the savings after taking the total derivative of the energy use equation. 42 - Confidence Regions for the Location of the Global Optimum of a Polynomial Model Adam Meyers, Pennsylvania State University, 1006 Golfview Avenue, Apt 3, State College, PA, 16801, United States, acm4688@gmail.com Our problem is to find a confidence region (CR) for the location of the global optimum of a polynomial model. Previous work culminated in an R package that finds CRs for quadratic polynomial regression and thin plate spline models. Our goal is to extend this package to accommodate polynomial models of higher degree and with a larger number of variables. In particular, we will discuss the solution method of the problem of finding the global minimum of a polynomial subject to polynomial inequalities. This involves solving semidefinite programming relaxations of increasing order whose solutions are guaranteed to converge monotonically to the global optimum. We will also discuss its implementation in R. 43 - Multi Attribute Trade Off in Supplier Selection Problem Somaye Ramezanpour Nargesi, University of Texas at Arlington, 807 S.Center Street, Apt 108, Arlington, TX, 76010, United States, somaye.ramezanpournargesi@mavs.uta.edu Supplier selection has been known as one of the most critical business decisions . In this research, we are going to use discrete choice setting to study how people do trade-off among price, availability and quality of the products offered by different suppliers, when they are actually choosing suppliers in an incentivized experimental setting. Aside from studying multi-attribute trade off, we will look at potential decision biases and the effect of number of choice on those biases. 41- A Machine Learning Framework for Industrial Energy Assessments at Small and Medium-sized Enterprises

44 - New Algorithms for Inference in Dynamic Stochastic Block Models

Mehrnaz Amjadi, PhD Student, College of Business, University of Illinois at Chicago, University hall, 601 S.Morgan St, 24th floor, UH2401, Chicago, IL, 60607, United States, mamjad2@uic.edu, Theja Tulabandhula Although the computational and statistical trade-offs for modeling single graphs are relatively well understood, extending such results to sequences of graphs is difficult. In this work, we propose two models for such sequences that capture: (a) link persistence between nodes across time, and (b) community persistence of each node across time. In the first model, we assume that the latent community of each node does not change, and in the second model we relax this assumption. For both models, we propose computationally efficient inference algorithms, which leverage community detection methods that work on single graphs. We provide simulation results validating their performance. 45 - A Decomposition Framework for System Optimal Dynamic Traffic Assignment Problem Mehrzad Mehrabipour, Graduate Research Assistant, Washington State University, 1630 NE Valley Road, Pullman, WA, 99164, United States, mehrzad.mehrabipour@wsu.edu, Ali Hajbabaie Computational complexity of cell transmission model based Dynamic Traffic Assignment (DTA) limits its applications to large transportation networks. This study develops a scheme to decompose the network into several sub-problems. The proposed methodology is applicable to larger transportation networks and finds solutions that are within 5% gap with the optimal solution in various case study networks. 46 - Redesign of Patient Flow and Resource Allocation in a High Volume Emergency Department Melih Celik, Middle East Technical University, ODTU. Endustri Muh. Bolumu, Universiteler Mh. Dumlupinar Bl. No: 1, Ankara, 06800, Turkey, cmelih@metu.edu.tr, Sakine Batun, Nur Keskin, Alpin Ilayda Ozmen, Deniz Tanrikut, Medya Tekes, Gokcem Yigit We consider the operations of a high-volume emergency department where the waiting times are above the acceptable levels for a significant proportion of the patients. We describe the system by using a simulation model and utilize this model evaluate the impact of various improvement opportunities such as using different policies about the patient flow (e.g., triage method, patient prioritization, patient streaming) and resource allocation (e.g., allocation of staff to shifts and patient groups). 47 - Thresholding Link Weights in Complex Networks as a Percolation Process Farnaz Zamani Esfahlani, PHD Candidate, State University of New York-Binghamton, 33 Schiller Street, Apt 3N, Binghamton, NY, 13905, United States, fzamani1@binghamton.edu, Hiroki Sayama Thresholding link weights is a common technique to reduce the complexity and computational time of analyzing complex networks. However, the majority of thresholding techniques alter the properties of weighted networks and therefore influence the results of the analysis. Here we propose a percolation-based thresholding method which can preserve the macro/microscopic properties of the weighted networks more effectively. 48 - A Brain Computer Interface Approach to Examine Changes in Motion Patterns While Walking in a Virtually Infinite World Rachneet Kaur, Student, University of Illinois at Urbana Champaign, TB 21, Department of Industrial and Enterprise Systems Engineering, 104 S.Mathews Ave, Urbana, IL, 61801, United States, rk4@illinois.edu, Richard Sowers, Ma nuel Hernandez, Daan Michiels This research focuses on creating virtual reality components of a test bed for understanding responses to visual stimuli and their relation to movement disorders such as Parkinson’s disease. With electroencephalography (EEG) as brain-computer interface (BCI), we measure neurological responses of a person walking through a virtual world that elicits nervousness, compute their real time neural state and adapt the virtual world accordingly to diminish fall related anxiety. The objective is to synthesize a self-monitoring and regulated personalized VR environment based on the user’s uneasiness that allows to decrease anxiety levels in balance demanding walking conditions. 49 - Modeling Lane Changing Behavior Based on Snowdrift Coordination Game Saeed Reza Ramezanpour Nargesi, UTA, 500 S Senter St, In this research, we are going to model lane changing behavior, using game theoretical approach, to take the interdependency of drivers’ decision making into microscopic traffic analysis. The model will include mandatory and discretionary lane changing behaviors in a snowdrift type of coordination game with two players (Target and Lag vehicles), complete information and static setting. Target vehicle aims to gain speed while trying to avoid collision. Lag vehicle tries to maintain its current speed subject to safety constraint. Apt 117, Arlington, TX, 76010, United States, saeedreza.ramezanpournargesi@mavs.uta.edu

371