2015 Informs Annual Meeting

SB02

INFORMS Philadelphia – 2015

SB03 03-Room 303, Marriott Improving Efficiency and Effectiveness of Supply Chains Cluster: Scheduling and Project Management Invited Session

4 - Discovering an Unknown Network: An Optimization Based Approach

Piya Pal, University of Maryland, Electrical and Computer Engineering, College Park, United States of America, ppal@umd.edu

A central challenge in network tomography is to discover the structure of the network from partial observations. Depending on the type of the network, these measurements can provide us with different kinds of information. A key question in this regard is: how many measurements (or sensors) are needed to find out the topology of the network, and how should they be placed? We describe a discovery algorithm that iteratively maps the graph, by using entropy as a criterion for sensor placement.

Chair: Chelliah Sriskandarajah, Hugh Roy Cullen Chair In Business Administration, Texas A&M University, 320Q Wehner, 4217 TAMU, College Station, TX, 77843, United States of America, chelliah@mays.tamu.edu 1 - Outpatient Appointment Scheduling under Patient Heterogeneity and Patient No-shows Seung Jun Lee, PhD Student, Texas A&M University, 320N Wehner Building, College Station, TX, 77845, United States of America, sjlee@mays.tamu.edu, Chelliah Sriskandarajah, Gregory Heim, Yunxia Zhu We study an outpatient appointment scheduling system under conditions of patient heterogeneity in service times and patient no-shows. We contribute by using more sophisticated sequential block scheduling policies, leading to effective appointment schedules when scheduling two patient types. We extend our algorithm to incorporate patient no-shows. Next, our block scheduling algorithm is adapted where outpatient clinics use an open-access policy. 2 - A Framework for Analyzing the U.S. Coin Supply Chain Yiwei Huang, Mays Business School, Texas A&M University, 320M Wehner Building - 4217 TAMU, College Station, TX, 77843-4217, United States of America, yhuang@mays.tamu.edu, Subodha Kumar, Bala Shetty, Chelliah Sriskandarajah We present a framework of analyzing the supply side problem for increasing cost- effectiveness of the U.S. Coin Supply Chain (CSC). We investigate the U.S. CSC from following perspectives: new coin production in the U.S. Mint, circulating coin distribution for the Federal Reserve System (FRS), and coin inventory management and coin demand forecasting at coin vaults (CV). We provide an optimal operating policy for the FRS using a minimum cost flow (MCF) network model for multi-products. 3 - Scheduling Operating Rooms with Elective and Emergent Surgeries Kyung Sung Jung, University of Florida, P.O. Box 117169, Gainesville, Fl, 32611-7169, United States of America, kyungsung.jung@warrington.ufl.edu, Chelliah Sriskandarajah, Vikram Tiwari Operating rooms (ORs) generate the greatest revenue source for hospitals while they are the largest cost centers. Scheduling ORs are challenging tasks due to the significant uncertainty in the arrival of emergent patients. To increase the responsiveness and efficiency for OR scheduling, we develop an optimization model which deals with block schedules and determines the sequence of elective patients so that the emergency patients who arrive randomly can be accommodated without incurring delays. 4 - Operations in Currency Supply Chains – A Review Yunxia Zhu, Rider University, Sweigart Hall 358, Lawrenceville, KS, United States of America, yuzhu@rider.edu, Chelliah Sriskandarajah, Neil Geismar This paper provides an overview of studies of various currency supply chains across the world. The structure of a general banknote supply chain is given before the discussion of the problems from three different perspectives: the supply side, the demand side, and the secure third-party logistics providers. We also provide a framework for analyzing the U.S. coin supply chain and descriptions of the coin supply chains in other countries. Future research directions are also proposed.

SB02 02-Room 302, Marriott INFORMS 2015 Data Mining Best Student Paper Award

Sponsor: Data Mining Sponsored Session Chair: Kamran Paynabar, Georgia Institute of Technology, 755 Ferst Drive, Atlanta, GA, 30332, United States of America, kamran.paynabar@isye.gatech.edu 1 - Falling Rule Lists Fulton Wang, MIT, 5 Cambridge Center #792, Cambridge, MA, 02142, United States of America, fultonwang@gmail.com Falling rule lists are classification models consisting of an ordered list of if-then rules, where (i) the order of rules determines which example should be classified by each rule, and (ii) the estimated probability of success decreases monotonically down the list. These kinds of rule lists are inspired by healthcare applications where patients would be stratified into risk sets and the highest at-risk patients should be considered first. 2 - Statistical Models for Characterizing the Heterogeneous Wake Effects in Multi-turbine Wind Farms Mingdi You, PhD Candidate, University of Michigan, 1205 Beal Avenue, IOE 1773, Ann Arbor, MI, 48109, United States of America, mingdyou@umich.edu, Eunshin Byon, Giwhyun Lee Wind turbines in a wind farm exhibit heterogeneous power generations due to wake effects. Because upstream turbines absorb kinetic energy in wind, downstream turbines produce less power. Moreover, the power deficit at downstream turbines shows heterogeneous patterns, depending on weather conditions. This study introduces a new approach for characterizing heterogeneous wake effects. A case study demonstrates the proposed approach’s superior performance over commonly used alternative methods. 3 - Sparse Precision Matrix Selection for Fitting Gaussian Random Field Models to Large Data Sets Sam Davanloo Tajbakhsh, Visiting Assistant Professor, Virginia Tech, 412 Hutcheson, Blacksburg, VA, 24060, United States of America, sdt144@vt.edu, Serhat Aybat, Enrique Del Castillo Fitting Gaussian random field models and finding the Maximum Likelihood Estimate (MLE) of the parameters requires a nonconvex optimization. The problem is aggravated in big data settings since the per iteration computational complexity of MLE is O(n^3) where n is the number of distinct spatial locations. We propose a theoretically provable two-stage algorithm which solves a semidefinite program in the first stage and a least square problem in the second stage. 4 - Sensor Driven Condition Based Generation Maintenance and Operations Scheduling Murat Yildirim, PhD Student, Georgia Institute of Technology, 755 Ferst Drive, Atlanta, GA, 30332, United States of America, murat@gatech.edu, Nagi Gebraeel, Andy Sun We propose an integrated framework, which combines (1) predictive analytics methodology that uses real-time sensor data to predict future degradation and remaining lifetime of generators, with (2) novel optimization models that transforms these predictions into cost-optimal maintenance and operational decisions. We present extensive computational experiment results to show proposed models achieve significant improvements in cost and reliability.

66