Informs Annual Meeting 2017

TC66

INFORMS Houston – 2017

TC64

2 - Ship Routing and Freight Assignment Problem for Liner Shipping Considering Northern Sea Route Dung-Ying Lin, National Cheng Kung University, 1 University Road, Transportation Management, Tainan City, Taiwan, dylin@mail.ncku.edu.tw In recent years, the Northern Sea Route (NSR) has attracted significant attention with respect to liner shipping. In this research, we investigate the NSR planning problem and construct a time-space network-based mathematical formulation to analyze the ship routing and freight assignment problem in liner shipping. To solve the resulting program, we propose a Lagrangian relaxation-based decomposition algorithm that facilitates the network features. The proposed framework is empirically applied to a case from Company C. The results show that navigation skill, bunker price, delay penalty and service commitment are the primary factors that affect the NSR’s commercial practicability. 3 - On Multi-parametric Costs Perturbation of the Integer Program with a Motivational Problem of the Dynamic Ambulance Allocation Yu-Ching Lee, National Tsing Hua University, No 101, Section 2, Kuang-Fu Road, Engineering Building I, Hsinchu, Taiwan, ylee77@uiuc.edu, Chor-Yiu Sin, Albert Y. Chen A two-stage stochastic integer programming model is developed to minimize the expected response time in every designated period by dynamically deciding the numbers of the ambulances to be allocated and re-allocated at each base. Practically, the expected response time is approximated by a finite number of draws of the scenarios whose probabilities are forecasted based on the historical counts of the ambulance requests. We evaluate the robustness of this allocation model with regard to the accuracy of the forecasted probabilities by the technique of the multi- dimensional cost sensitivity for the IP. 4 - Optimal Security Deployment over Moving Trains in Rail Transit Networks I-Lin Wang, National Cheng Kung University, Industrial Information Management, 1 University Road, Tainan, 70101, Taiwan, ilinwang@mail.ncku.edu.tw, Tai-Chia Huang, Shiue-Mei Sun, Yun-Hsuan Lu Rail transit systems usually operate with fixed routes and schedules to provide more reliable and less polluted commuting services to move more people within urban areas. Due to their predictable schedules, attacks over trains may lead to severe casualties. Yet current practices usually deploy security on static stations rather than moving trains. We investigate effective security deployment over moving trains by mathematical programming models. Maximum coverage to passengers on trains can be calculated. We also consider a variant of network interdiction problem to deploy security so that the maximum number of passengers affected by a single attack over a train segment can be minimized. 5 - Price Strategy and Capacity Allocation for Crystal Cycle of TFT-LCD Industry Kang-Ting Ma, Post-doctor Researcher, National Tsing Hua University, Hsinchu, Taiwan, ktma@ie.nthu.edu.tw, Che-Wei Chou, Chen-Fu Chien Variances of supplies and demands in TFT-LCD industry fluctuate prices namely crystal cycle. By the contrast, as supplies are shortage, manipulating prices affect the demands. A two-stage mathematical model is proposed to maximize profit of the manufacturer that determines the price strategy of product portfolio. Manufacture resources are allocated in the first stage. Rising prices for brand’s demand fulfillment are provided, and a new capacity will be reallocated in second stage. A hybrid particle swarm optimization algorithm is developed to solve the nonlinear two-stage mathematical model, and the algorithm is quite efficient by designing of initial solutions and local search procedure.

370E Complex Data Analysis and its Applications Sponsored: Data Mining Sponsored Session Chair: Neng Fan, University of Arizona, Tucson, AZ, 85721, United States, nfan@email.arizona.edu 1 - Fitting Phase-type Distribution to Healthcare Transition Data Wanlu Gu, 827 E Drachman St, Tucson, AZ, 85719, United States, wanlugu@email.arizona.edu, Neng Fan The length of stay (LOS) is a term to describe the duration of a single stay in hospital, it as well as the time interval between a discharge and the next admission are important measures of healthcare utilization. In this talk, we fit a phase type distribution to LOS data from The University of Arizona Cancer Center. In order to eliminate confounding factors, we study data of patients with one typical type of disease at a time. By selecting proper number of phases (LOS groups), we can grasp the heterogeneity of these phases and connect it with the patients’ trace after discharge. Then some actions can be taken to improve the efficiency of healthcare utilization. 2 - Distributionally-robust CVAR Formulation of SVMs using Wasserstein Metric Ammon Washburn, University of Arizona, 2525 N Los Altos There has been many variants of Support Vector Machines that use different norms or penalty functions in order to accomplish feature selection, good generalization and robustness to outliers. We propose to unify all of them under a framework that comes CVaR in financial applications. Then we will show properties of the resulting quadratic program as a function of parameters and provide an algorithm that will compute the whole solution path for two out of three of the parameters in about the time it takes to solve one using interior points methods. 3 - Use of Phase Type Distribution in ALT Data Analysis Samira Karimi, University of Arkansas, 1219 Storer Ave, Apt 39, Fayetteville, AR, 72703, United States, sakarimi@email.uark.edu, Haitao Liao Accelerated life testing (ALT) is a powerful tool for estimating a product’s reliability. In ALT, a spectrum of accelerated conditions (such as pressure, temperature, voltage, stress, etc.) is utilized to generate failures in a shorter period of time in comparison to normal conditions. To model and analyze the resulting lifetime and censoring data, one is required to relate the data to the applied stress factors over time. In this research, ALT data is modeled by a phase-type distribution, and an Expectation-Maximization algorithm is developed to estimate the model parameters. A numerical study is presented to illustrate the performance and precision of the proposed method. 4 - Graph Clustering with Uncertainty Yujia Zhang, 4901 E Sunrise Drive, Apt 1013, Tucson, AZ, 85718, United States, yujiazhang@email.arizona.edu During this talk we will review certain models and algorithms for clustering of data with uncertainty. First the methods for modeling uncertainty data will be reviewed. Then we concentrate on the graph models for clustering. In the end we will review and compare algorithms for solving these models. 370F Transportation and Logistics Research in Taiwan, a Special Session for IJOR and JIPE Sponsored: TSL, Freight Transportation & Logistics Sponsored Session Chair: I-Lin Wang, National Cheng Kung University, Industrial Information Management, 1 University Road, Tainan, 70101, Taiwan, ilinwang@mail.ncku.edu.tw 1 - Optimal Fleet Deployment for Electric Vehicle Sharing Systems Chung-Cheng Lu, National Chiao Tung University, Hsinchu City, Taiwan, jasoncclu@nctu.edu.tw This study deals with the fleet deployment problem for electric vehicle (EV) sharing systems that service tourists in scenic areas. The problem aims to determine the number of plug-in EVs deployed at each rental station such that the profit of the system operator is maximized. The operational aspects of plug-in EVs are taken into account, including electricity energy consumption, battery capacity and charging requirements. A mathematical programming model of the problem is developed. TC65 Avenue, Apt 155, Tucson, AZ, 85705, United States, wammonj@email.arizona.edu, Neng Fan, Helen Zhang

TC66

371A Best Paper and Best Dissertation session Sponsored: Transportation Science & Logistics Sponsored Session

Chair: Lavanya Marla, University of Illinois at Urbana-Champaign, 104 S Mathews Avenue, Room 216E, Urbana, IL, 61801, United States, lavanyamarla@cmu.edu 1 - Best Paper Award Session Finalists Andres Medaglia, Universidad de los Andes, Bogota, Colombia, Medaglia Finalists and winner(s) of the best paper award, announced in the TSL Business Meeting on Monday, will present their papers.

359