2016 INFORMS Annual Meeting Program

WB77

INFORMS Nashville – 2016

3 - How The Swine Flu Epidemic Spread: Lessons From The Data Jussi Keppo, National University of Singapore, Mochtar Riady Building, BIZ 1 8-69, 15 Kent Ridge Drive, Singapore, 119245, Singapore, keppo@nus.edu.sg, Elena Quercioli, Lones Smith Contagious diseases are passed on when contagious and susceptible individuals meet. This paper introduces and explores a new matching game, characterized by individuals meeting pairwise, possibly unwittingly passing along a disease in a contagion-like fashion. We assume that individuals can expend costly effort to avoid acquiring it. In this population game, efforts are strategic substitutes: The harder other individuals try, the more lax one can be. We solve for the unique Nash equilibrium when individuals are heterogeneous. We then estimate this structural model and argue that it improves on the explanation of the data without endogenous behavior. 4 - An Unbiased Measure Of Integrated Volatility In The Frequency Domain Fangfang Wang, University of Illinois at Chicago, 601 South Morgan Street, Chicago, IL, 60607, United States, ffwang@uic.edu This work studies the effect of market microstructure noise on volatility estimation in the frequency domain. We propose a bias-corrected periodogram- based estimator of integrated volatility. We show that the new estimator is consistent and the central limit theorem is established under a general assumption of the noise. We also provide a feasible procedure for computing the bias-corrected estimator in practice. As a byproduct, we extract a consistent frequency-domain estimator of the long-run variance of market microstructure noise from high-frequency data. 5 - A Parallel Computation Of Characteristics Of Markov Chains With “Islands” And “Ports” We present a new algorithm to calculate the invariant distribution of a large Markov chain (MC) whose state space is partitioned into “islands” and “ports”. An island is a group of states with potentially many connections inside of the island but a relatively small number of connections between islands. The states connecting different islands are called “ports”. Our algorithm is developed in the framework of the “state reduction approach”, but the special structure of our MCs allows for computations to be performed in parallel. Additional problems for such MCs, including computation of a Fundamental matrix, optimal stopping, and the case of small transitions between islands are also analyzed. Amod Basnet, University of North Carolina-Charlotte, 9201 University City Blvd, Fretwell, Charlotte, NC, 28262, United States, abasnet@uncc.edu, Isaac M Sonin

3 - Optimal Selection Of Parsimonious Arima Forecasting Models George G Polak, Professor, Wright State University, 227 Rike Hall, Raj Soin College of Business, Dayton, OH, 45435, United States, george.polak@wright.edu, Bogdan C Bichescu In a parsimonious ARIMA Forecasting model the total number of terms chosen is constrained, and treatment for nonstationarity is limited to differencing of a pre- specified order. Accordingly, a term for non-stationary variation is included along with autoregressive, moving average and white noise terms in a decomposition of the time series. Mixed integer quadratic optimization problems are then formulated for prescribing model parameters. Walsh functions and Haar wavelets are employed to represent white noise and non-stationary variation, respectively. 4 - A Comparative Analysis Of MILP Formulations For The Multi-family Capacitated Lot-sizing Problem (MFCISP) Andrea L. Arias, Graduate Student, Texas Tech University, 4425 82nd Street, Apt 2258, Lubbock, TX, 79424, United States, andrea.arias@ttu.edu, Andrea L. Arias, Graduate Student, P. Universidad Católica de Valparaíso, Valparaíso, Chile, andrea.arias@ttu.edu, Ricardo A. Gatica The MFCLSP is a variant of the CLSP with setup times in which items are organized into families. We present three MILP models for the MFCLSP (TRAD, LI and RC), and develop a comparative analysis in order to evaluate their performance using Cplex. Since solving large-scale problems has been shown to consume a great amount of computational time (very unviable for application purposes), we examine the performance of the formulations under solving time constraints. The results show a major advantage for both LI and RC regarding to linear relaxation bounds, and that RC performs much better in terms of computational time to obtain the first feasible solution, yielding at the same time, very good optimality gaps. Chair: Clara Novoa, Associate Professor, Texas State University, 601 University Dr, San Marcos, TX, 78666, United States, cn17@txstate.edu 1 - Information Supply Chain Optimization With Bandwidth Limitations Michael Hirsch, ISEA TEK, LLC, 3283 Hickory Lane, Longwood, FL, 32779, United States, mhirsch@iseatek.com, Hector Juan Ortiz-Pena Workflow management systems allow for visibility, control, and automation of many business processes. Recently, non-business domains have taken an interest in the management of workflows, and the optimal assignment and scheduling of workflow tasks to users across a network. This research aims at developing a rigorous mathematical programming formulation of the workflow optimization problem. The resulting formulation is non-linear, but a linearized version is produced. Three heuristics are developed to find solutions efficiently. Computational experiments are presented and analyzed, comparing solutions to the original linearized formulation with the three heuristics. 2 - An Intelligent Production Control System For Automotive Parts Manufacturing With A CPS Approach Jun Kim, Sungkyunkwan University, Suwon, Korea, Republic of, tomatoes10@skku.edu, Hyun Jung Kim This paper examines an intelligent production control system for automotive parts manufacturing based on a cyber-physical systems (CPS) approach. We propose a CPS-based operations monitoring platform that helps decision-making processes in production management such as predicting undesirable events and reacting to such problems in advance. Various issues in designing and implementing a CPS platform for engine piston manufacturing processes are addressed. 3 - An Efficient Heuristic For The Capacitated Lot Sizing Problem With Setup Carryover And Splitting Srimathy Mohan, Associate Professor, Arizona State University, Department of Supply Chain Management, W.P. Carey School Of We present a new formulation for the Capacitated Lot Sizing Problem with Setup Carryover and Setup Splitting. Setup carryover transfers a completed machine setup from one period to the next. Setup splitting completes a setup across period boundaries. We present a Fix-and-Optimize heuristic that is easy to adapt for model extensions. Our extensive experimentation shows that the heuristic produces results within 6% and 8% of optimality for problems without and with demand backlogging, respectively. Management, Phoenix, AZ, 85069-7100, United States, srimathy@asu.edu, Cheng-Lung Chen, Muhong Zhang WB78 Legends F- Omni Opt, Metaheuristics II Contributed Session

WB77 Legends E- Omni Opt, Integer Programing VI Contributed Session

Chair: Andrea L. Arias, Graduate Student, Texas Tech University, 4425 82nd Street, Apt 2258, Lubbock, TX, 79424, United States, andrea.arias@ttu.edu 1 - A Branch And Price Approach For Deployment Of Multi-tier Software Services In Clouds

Bjorn Nygreen, Retired, Dept. of Ind. Econ. & Techn. Management, NTNU, Trondheim, NO 7491, Norway, bjorn.nygreen@iot.ntnu.no, Anders Nordby Gullhav

This talk considers a problem of a cloud service provider (SP) that offers a set of applications to its end-users. In short, the SP seeks to find the minimum cost deployment of its services on an infrastructure consisting of a private and public cloud, while maintaining a certain quality of service. We propose to solve the problem using branch and price, where the sub problem is solved both by a MIP solver and a heuristic label-setting algorithm. Our results show the benefits of using the heuristic in addition to the exact MIP solver. 2 - Project Scheduling And Multi-skill Workforce Assignment For Property Management Applications Anthony Vatterott, Assistant, University of Missouri Saint Louis, 1 University Boulevard, Saint Louis, MO, 63121, United States, anthonygvatterott@gmail.com, Haitao Li, Norman Keith Womer We present a new optimization framework to schedule, assign and route maintenance tasks with a multi-skilled workforce. Mixed integer programming methods are applied to obtain effective and efficient solutions to advance the state-of-the-art of property management. Practical emphasis on job priority, skill development and knowledge transfer are considered to form the temporal, spatial and resource constraints while minimizing the total cost of project delivery.

422