Informs Annual Meeting 2017

MD55

INFORMS Houston – 2017

MD55

2 - On the Spatial Separability of Uncapacitated Single Allocation P-hub Median Problems Taghi Khaniyev, University of Waterloo, 200 University Ave., CP.H.3669, Waterloo, ON, N2L.3G1, Canada, thanalio@uwaterloo.ca, Samir Elhedhli, F. Safa Erenay The spatial separability property of uncapacitated single assignment p-hub median problems (USApHMP) is studied. We illustrate that the optimal solutions to the well-known USApHMP instances can be partitioned into p allocation clusters, defined as the set of nodes which are allocated to the same hub, such that the convex hulls of the allocation clusters are disjoint. To exploit this property, a data driven approach to group the network nodes into regions and to obtain high quality solutions based on the solution of a smaller (low resolution) USApHMP is proposed. Experiments indicate the effectiveness of the proposed approach in generating high quality solutions within a reasonable amount of time. 3 - Solution Methods of Safe Walking Route Design Problem Yuichiro Miyamoto, Sophia University, Tokyo, Japan, miyamoto@sophia.ac.jp, Ryuhei Miyashiro, Ken-ichi Tanaka In the previous study, we proposed the safe walking route design problem (SWRDP) and analyzed exact optimal solutions using medium-sized problem instances. In this talk, we propose several solution methods of SWRDP for solving larger problem instances. We will also mention some variants of SWRDP. 4 - A Lagrangean Relaxation Algorithm for the Radius Formulation of the P-median Problem

362B Financial Econometrics Sponsored: Financial Services Sponsored Session Chair: Gustavo Schwenkler, Boston University, Boston, MA, 02215, United States, gas@bu.edu 1 - Estimating Dynamic Option Pricing Models using Large Panels Kris Jacobs, University of Houston, Houston, TX, United States, kjacob@bauer.uh.edu We estimate option valuation models on returns and large panels of option data using the particle filter. 2 - State-varying Factor Models of Large Dimension Markus Pelger, Stanford University, 312 Huang Engineering Center, 475 Via Ortega, Stanford, CA, 94305, United States, mpelger@stanford.edu, Ruoxuan Xiong This paper develops an inferential theory for state-varying factor models of large dimension. Unlike constant factor models, the loadings are general functions of some recurrent state process. Our estimator combines nonparametric methods with principal component analysis. We derive the rate of convergence and limiting normal distribution for the factors, loadings and common components. We develop a statistical test for the constancy of factor loadings in different states. In an empirical study on U.S. Treasury securities data, we find that the systematic factor structure is different in boom and recession times and in periods of high market volatility. 3 - The Discretization Filter: A Simple Way to Estimate Nonlinear State Space Models Leland E. Farmer, University of Virginia, Charlottesville, VA, United States, farmer.leland@gmail.com Existing methods for estimating nonlinear dynamic models are either too computationally complex to be of practical use, or rely on local approximations which often fail to adequately capture the nonlinear features of interest. I develop a new method, the discretization filter, for approximating the likelihood of nonlinear, non-Gaussian state space models. I apply results from the statistics literature on uniformly ergodic Markov chains to establish that the implied maximum likelihood estimator is strongly consistent, asymptotically normal, and asymptotically efficient (see paper for full abstract). 4 - Long-run Risk or Disasters? New Evidence from an Efficient Estimation Method for Multivariate Jump-diffusions Gustavo Schwenkler, Boston University, 595 Commonwealth Ave, Boston, MA, 02215, United States, gas@bu.edu, Andrea M. Buffa, Yunjeen Kim We study the dynamics of consumption growth in a series of countries over a time span of 200 years. We seek to answer whether long-run risk or disasters are features of models that yield good fit to consumption data. To accomplish this goal, we develop a new methodology for filtering and estimation of multivariate jump-diffusion processes with incomplete data. Our methodology is both statistically and computationally efficient, and enables the empirical analysis of previously intractable multidimensional models. Our estimates suggest that long-run risk is a predominant feature of consumption data, albeit of low persistence. Disasters are estimated to be small and frequent, yet necessary for a good model fit. 362C Recent Advances on Facility Location and Network Design Problems I Sponsored: Location Analysis Sponsored Session Chair: Sergio Garcia Quiles, University of Edinburgh, Edinburgh, EH9 3FD, United Kingdom, sergio.garcia-quiles@ed.ac.uk 1 - An Integer Programming Model of Route Assignment Problem for Reducing Traffic Congestion in Urban Area Naoya Uematsu, Osaka University, Suita, Japan, naoya.uematsu@ist.osaka-u.ac.jp, Shunji Umetani, Hiroshi Morita Despite of the spread of navigation services through smartphones, it has been still hard to resolve traffic congestion in urban area.We consider a large-scale combinatorial optimization problem that finds optimal routes for six million people to minimize the peak of congestion in the urban area, instead of solving shortest path problem for each user in individual.To avoid solving a large number of shortest path problems, we enumerate plausible candidate routes for each pair of origin and destination in advance and formulate this problem as a variant of the assignment problem. MD56

Sergio Garcia Quiles, University of Edinburgh, School of Mathematics, James Clerk Maxwell Bldg, King’s Building, Edinburgh, EH9 3FD, United Kingdom, sergio.garcia- quiles@ed.ac.uk, Minerva Martin del Campo Barraza

The p-median problem is a well-known problem in Discrete Facility Location, but at the same time still very challenging to solve for problems of large size. One of the proposed methods in recent years is the so-called radius formulation, which is based on using variables that rank the distances for each customer. This formulation has allowed to solve very large instances for some values of p but it fails for some others (e.g., for small values of p). In this paper we explore the usefulness of applying Lagrangian relaxation to the radius formulation.

MD57

362D Simulation Optimization Sponsored: Simulation Sponsored Session Chair: Mohammad Dehghani, Northeastern University, M.dehghani@northeastern.edu

1 - Triggering an Optimization Module within the Simulation Run: Introducing a New Iterative Optimization-based Simulation (IOS) Framework Mohammad Dehghanimohammadabadi, Northeastern University, 170 Brookline Avenue, Unit 1025, Boston, MA, 02115, United States, mdehghani@neu.edu, Thomas K. Keyser, Hossein Cheraghi A unique Iterative Optimization-based Simulation (IOS) framework is presented by integrating simulation, optimization and database managers. With this IOS model, optimization occurs frequently at the operational level in order to optimize system configuration during the simulation run. A trigger event momentarily pauses the simulation and prompts the optimization manager to solve a mathematical model in order to update the simulation model configuration. This promising IOS framework is applied in a manufacturing system and its results are compared with Simulation-Based Optimization (SBO). 2 - A Stochastic Flow Shop Maintenance Scheduling using a Simulation-optimization Approach Javad Seif, University of Tennessee-Knoxville, 411 B.H. Goethert Parkway, Tullahoma, TN, 37388, United States, jseif@utsi.edu, Mohammad Dehghanimohammadabadi, Andrew Junfang Yu In this study, preventive and corrective maintenance (PM and CM) activities are incorporated into a permutation flow shop scheduling problem. A set of usage- based and diverse PMs as well as random variables with different probability distributions that represent potential CMs are applied to the model. A hybrid simulation-optimization (SO) approach using mixed integer programming (MIP) and discrete event simulation (DES) is utilized to solve the problem. Finally, a numerical experiment is conducted to evaluate the performance of the provided methodology and its potentialities.

256