2015 Informs Annual Meeting

WD35

INFORMS Philadelphia – 2015

WD36 36-Room 413, Marriott Humanitarian Operations Management Applications Sponsor: Public Sector OR Sponsored Session Chair: Alfonso Pedraza-Martinez, Assistant Professor, Indiana University, 1309 E 10th Street, Bloomington, IN, 47405, United States of America, alpedraz@indiana.edu 1 - Fast and Frugal Disaster Response? Decisions in Typhoon Haiyan Tina Comes, University of Agder, Postboks 509, Grimstad, 4898, Norway, tina.comes@uia.no, Bartel Van De Walle In the response to sudden-onset disasters, humanitarian organisations operate under trying conditions in which response targets evolve as information on the actual impact becomes available. The pressing humanitarian needs require fast decision making, including decisions on warehouse locations and the allocation of relief items, yet with little time for more than frugal analyses. In this presentation, we study the response to Typhoon Haiyan which struck the Philippines in November 2013. 2 - Estimating and Incorporating Deprivation Costs into Humanitarian Logistic Models for Relief Response Victor Cantillo, Associated Professor, Unirversidad del Norte, This research allows mathematical formulations using discrete choice modelling to quantify externalities associated to the lack of access to critical commodities in the aftermath of a disaster. Thus the estimated deprivation cost function is explicitly incorporated into the objective function of facility location models for prepositioning supplies, which attempt to minimize the total social costs, as determined by both operational and social considerations. The models are applied to a real case. 3 - Assembling High Quality and Timely Information for Humanitarian Organizations from Social Media Eunae Yoo, Arizona State University, P.O. Box 874706, Tempe, AZ, 85287, United States of America, Eunae.Yoo@asu.edu, Mahyar Eftekhar, Elliot Rabinovich, Bin Gu To support operational decision making, humanitarian organizations require high quality and timely data. We investigate how such information can be extracted from social media using automated data mining mechanisms that rapidly process data. The effectiveness of data mining mechanisms are tested using a sample of Twitter data. Our results help shed light on what constitutes high quality information for humanitarian organizations and how it can be speedily obtained from social media. 4 - Humanitarian Funding in a Multi-donor Market with Donation Uncertainty Alfonso Pedraza-Martinez, Assistant Professor, Indiana University, 1309 E 10th Street, Bloomington, IN, 47405, United States of America, alpedraz@indiana.edu, Arian Aflaki We analyze the trade-off between earmarked funding and operational performance. If a Humanitarian Organization (HO) allows donors to earmark their donations, HO’s expected funding increases but its operational efficiency decreases. We use the Scarf’s minimax approach and the newsvendor framework, and calibrate our model using data from 15 disasters. Km 5 via Puerto Colombia, Barranquilla, Colombia, vcantill@uninorte.edu.co, Nathalie Cotes, Luis Macea, Ivan Serrano

4 - A Hybrid Computational Method based on Convex Optimization for Outlier Problems Fatma Yerlikaya Ozkurt, Middle East Technical University, Institute of Applied Mathematics, Ankara, Turkey, fatmayerlikaya@gmail.com, Aysegul Askan, Gerhard Wilhelm Weber Statistical modeling plays a central role for any prediction problem of interest. However, predictive models may give misleading results when the data contain outliers. In many applications, it is important to identify and treat the outliers without direct elimination. To handle such issues, a hybrid computational method based on conic quadratic programming is introduced and employed on earthquake ground motion data set. Results are compared against widely-used ground motion prediction models. Global Issues II Contributed Session Chair: Feifan Wang, Zhejiang University, No.38, Zheda Road, Hangzhou, China, wangfeifan@zju.edu.cn 1 - A Framework of Social Recommender System Combining Social Network and Sentiment Analysis Donghui Yang, Southeast University, Sipailou 2#, Nanjing, China, dhyang@seu.edu.cn In this contribution, we propose a new framework for a social recommender system based on both network structure analysis and social context mining. Exponential random graph modelsand sentiment similarities are used to make the social recommender system much more precise and to satisfy users’ psychological preferences.The recommendation results of diabetes accounts of Sina Weibo show that our method outperforms other social recommender systems. 2 - Decision Tree Based Method for Prediction of Preventable Readmissions in Acute Myocardial Infarction Andres Garcia-Arce, University of South Florida, 4202 E. Fowler Avenue, ENB 118, Tampa, FL, 33620, United States of America, andresg@mail.usf.edu, Jose L. Zayas-Castro, Florentino Rico, Shuai Huang Preventable readmissions are recognized as a target for quality improvement. The US government implemented economic penalties to decrease the preventable readmissions, which leads stakeholders to improve to avoid penalties. The literature show several statistical models that help hospitals understand readmissions risk in their institutions, however, these models usually fail to achieve a good discriminatory power. A random forest-based predictive model is studied, achieving an AUC=0.7494. 3 - Assess Care Coordination by Multi-criteria Ranking Wei Liu, Purdue University, Industrial Engineering, West Lafayette, IN, United States of America, liu317@purdue.edu, Ping Huang, Steven Landry Care coordination reflects the quality of care and impacts the patient outcome. It remains a challenge to quantify the interactions among providers from various services as well as the relationships among patients and providers. We use a novel method of multi-criteria ranking to assess the care coordination under consideration. It may aid decision makers to identify appropriate interventions to improve care. 4 - Performance of Different Generalized Propensity Methods in Evaluating Multi-arm Nonrandomized Study Feifan Wang, Zhejiang University, No.38, Zheda Road, Hangzhou, China, wangfeifan@zju.edu.cn, Haomiao Jin, Zhengxiao Wang Generalized propensity score (GPS) is a widely used approach to adjust the inherent bias existed in multi-arm nonrandomized study. A simulation study is conducted to assess the performance of four GPS methods: regression adjustment, matching, stratification, and inverse probability weighting. Practical implications are discussed and a case is provided. WD35 35-Room 412, Marriott

WD37 37-Room 414, Marriott Health Care Strategy and Policy II Contributed Session

Chair: Neil Desnoyers, Instructor, Saint Joseph’s University, 133 Green Valley Rd, Upper Darby, PA, 19082, United States of America, ntdesnoyers@gmail.com 1 - Revenue-based Booking Policy for Clinic Appointment with Overbooking Considering Patient No-shows Jiafu Tang, Chair Professor, Dean, Dongbei University of Finance and Economics, School of Management Science, and Engineering, Dalian, 116025, China, jftang@mail.neu.edu.cn, Pingping Cao, Xuanzhu Fan In this paper, an advanced clinic access system is designed. We formulate a Markov Decision Process model with its extension considering regular patients’ no-shows and patient choice to improve clinic revenue by overbooking same-day patients, and then to improve patient satisfaction by allowing patients to choose either a same-day or a scheduled future appointment. Numerical experiments and analysis are made finally.

464