Informs Annual Meeting Phoenix 2018

INFORMS Phoenix – 2018

MD71

However, there is little empirical evidence showing how the research conducted in hospitals can affect their outcomes. The structure of hospitals is a two-stage network process. Therefore, we use two-stage DEA methods for evaluating the performance in a sample of 189 Spanish hospitals over the period 1996-2009. We use the Malmquist productivity index for estimating the changes in efficiency since we use a longitudinal database. We measure the efficiency on several outcomes (i.e., the average length of stay or mortality rate. We measure the hospitals’ research activity by using a set of bibliometric indicators. 2 - Labor-cost Efficiency with Indivisible Outputs and Inputs: A Study of Indian Bank Branches Kankana Mukherjee, Babson College, 231 Forest Street, Babson Park, MA, 02457, United States, Subhash C. Ray, Abhiman Das This study uses Data Envelopment Analysis to examine the efficiency of branches of a major Indian public sector bank across four large metropolitan cities. We model branch operations following the production approach and introduce several methodological extensions to account for the product mix of branches in creating the efficient cost frontier. Overall, Chennai branches are found to be the most efficient. Across the three types of labor, attaining efficiency in the number of clerks would have the highest impact in terms of cost savings. 3 - Predicting Corporate Failure for Non-Manufacturing Firms - DEA SBM Joseph C. Paradi, Professor Emeritus, University of Toronto, 200 College Street, College St, Toronto, ON, M5S3E5, Canada Slacks-Based DEA Model is used to predict corporate failure of non- manufacturing companies. The benchmark was the Altman Z’’ model. Others used DEA models (BCC) to using Altman’s original asset-dominated Z-score model. Here, non-manufacturing firms were examined without their asset size. Data from non-manufacturing companies that filed for bankruptcy between 2000 and 2006 for up to five years before bankruptcy. Non-bankrupt companies were matched these, using SIC codes. Altman’s model classified more companies as bankrupt than DEA, whereas DEA classified more as non-bankrupt. This indicated that bankruptcy could be predicted without the use of total assets or liabilities as variables. n MD71 West Bldg 106C Decision Diagram Approaches for Optimization Sponsored: Computing Sponsored Session Chair: Merve Bodur, University of Toronto, 5 King’s College Rd., Toronto, ON, M5S 3G8, Canada 1 - Dexter: A Global Solver Based on Decision Diagrams Danial Davarnia, Carnegie Mellon University, Pittsburgh, PA, United States, Christian Tjandraatmadja, Willem-Jan Van Hoeve We develop a branch-and-cut technique to solve integer nonlinear programs, where the cuts are derived from the decision diagram representation of the model. The graphical structure of decision diagrams allows for specialized branching schemes that improve the relaxation by refining infeasible integer points, while benefiting from tighter representation due to the domain reduction. Such a modification of decision diagrams is efficient through branch-and-bound. We introduce a solver designed based on this approach. We conclude with computational results. 2 - A Binary Decision Diagram Based Algorithm for Solving a Class of Binary Two-stage Stochastic Programs Leonardo Lozano, University of Cincinnati, Cincinnati, OH, 45219, United States, Cole Smith We consider a class of two-stage stochastic integer programming problems with binary variables in both stages, for which the second-stage variables belong to the intersection of sets corresponding to first-stage binary variables that equal one. Our approach uncovers strong dual formulations to the second-stage problems by transforming them into dynamic programming (DP) problems parameterized by first-stage variables, formed using of binary decision diagrams, which then yield traditional Benders inequalities that can be strengthened based on observations on the structure of the underlying DPs. We test our approach on a set of stochastic traveling salesman problems. 3 - Network-based Approximate Linear Programming for Discrete Optimization Andre Augusto Cire, University of Toronto-Scarborough, Department of Management, UTSC, 1265 Military Trail, Toronto, ON, M1C-1A4, Canada, Selvaprabu Nadarajah We present a new hierarchy of approximate linear programming methods for a general class of discrete optimization problems. Our basis functions, in particular, are composed of network-based approximations for the problem, such as those derived from state-space relaxations and decision diagrams. Moreover, we exploit the network structure to iteratively construct a sequence of approximate linear programs with improving bounds that converge to the optimal solution value of the original problem. A numerical evaluation is discussed on challenging discrete optimization problems arising in practice.

n MD69 West Bldg 106A Image-Based Statistical Process Control in Advanced Manufacturing Sponsored: Quality, Statistics and Reliability Sponsored Session

Chair: Anh Bui, Northwestern University, Evanston, IL, United States 1 - Tomography Reconstruction with Sparsity Regularization for Accurate Measurement of Nanoproduct Dimensions in 3D Chen Mu, FAMU-FSU College of Engineering, Tallahassee, FL, United States, Chiwoo Park The dimensions of nanomaterials often determine their properties and functionalities. Therefore, measuring the dimensions of nanomaterial products is an essential step for quality assurance of nanomaterial production. This talk is concerned with how to measure the dimensions of nanomaterial products in 3d with a series of its low dimensional projection images from different projection angles, which is well known as a tomography reconstruction problem. We presents a novel tomography reconstruction approach that combines the computational advantage of the existing filtered back projection approach and the precision advantages of the iterative optimization approach. 2 - Image Comparison for Online Image Monitoring Peihua Qiu, University of Florida, CTRB, 5th Floor, 2004 Mowry Road, Gainesville, FL, 32610, United States Images are commonly used for quality control purposes. If the quality of the products is good, then their images should be all similar to the image of a good- quality product. So, comparison of images is a fundamental task in image-based quality control. This problem is complicated because of noise and geometrical mismatch between images. In this talk, we present an effective method for detecting difference between two images of products, and our proposed method can accommodate both noise and geometric mismatch mentioned above. Theoretical results and numerical examples show that it can work effectively in applications. This is a joint research with Dr. Long Feng. 3 - Statistical Modeling and Analyzing Methods for Dynamic Nanomaterial Video Data Yanjun Qian, Virginia Commonwealth University, 1015 Floyd Ave, Room 4134, Richmond, VA, 23220, United States A large amount of nanomaterial video data has been collected by transmission electron microscopes (TEM). To model a nanoparticle growth process from such videos, we develop dynamic nonparametric models for time-varying probability density functions (pdfs) estimation. Unlike simple statistics, a pdf contains fuller information about the nanoscale objects of interests. Characterizing the dynamic changes of the pdf as the nanoparticles are growing into different sizes, the proposed nonparametric methods are capable of analyzing an in situ TEM video to delineate growth stages in a retrospective analysis or tracking the nanoparticle growth process in a prospective analysis. 4 - Identifying Variation Patterns in Stochastic Textured Surface Data Anh Tuan Bui, Northwestern University, Evanston, IL, 60208, United States, Daniel Apley Existing methods for understanding manufacturing variation in multivariate or profile data are inapplicable for stochastic textured surface data. A primary challenge is the lack of an existing measure of the dissimilarity or distance between surface samples, due to their stochastic nature. We propose a pairwise dissimilarity measure and then use manifold learning on these dissimilarities to discover a low-dimensional parameterization of the surface variation patterns. Visualizing how the surfaces change as the manifold parameters are varied helps build an understanding of the physical nature of each variation pattern, which we demonstrate with real textile and simulation examples. n MD70 West Bldg 106B Joint Session DEA/Practice Curated: Applications in DEA Emerging Topic: Productivity, Efficiency and Data Envelopment Analysis Emerging Topic Session Chair: Kankana Mukherjee, Babson College, Wellesley, MA, 02481, United States 1 - Hospitals Efficiency Redux: The Role of Medical and Surgical Research Antonio Garcia Romero, Assistant Professor, IE Business School, Maria de Molina 31 Bis, MADRID, 28006, Spain, Josep A. Tribo, Alvaro Escribano DEA methods have been widely applied to the analysis of hospitals’ efficiency.

239