Informs 2015

241 / 552

INFORMS Philadelphia – 2015
239
MD22
22-Franklin 12, Marriott
Joint Session Prize/CPMS: 2015 Informs Prize Winner
Cluster: 2015 INFORMS Prize Presentation
Invited Session
Chair: Peter Buczkowski, Manager, Workforce Management, Disney
Parks & Resorts, P.O. Box 10000, Lake Buena Vista, FL, 32830,
United States of America,
1 - 2015 Informs Prize Presentation by Chevron
Margery Connor, Chevron, 6001 Bollinger Canyon, F-2080,
San Ramon, CA, 94583, , Bill Klimack,
Wen Chen
Chevron, 2015 INFORMS Prize Winner for excellence in analytics and operations
research, will present their long and innovative history of applying analytics and
operations research across their worldwide energy company. Highlighted projects
include: • Petro: Chevron’s refinery planning tool • Workforce forecasting to
ensure the right people on the right projects • genOpt: Optimization model to
maximize oil and gas production. Chevron will also share their journey applying
decision analysis.
MD23
23-Franklin 13, Marriott
Markov Lecture
Sponsor: Applied Probability
Sponsored Session
Chair: Tolga Tezcan, Associate Professor, London Business School,
Regent’s Park, London NW14SA, United Kingdom,
Co-Chair: Neil Walton, University of Amsterdam, Science Park 904,
Amsterdam, Netherlands,
1 - Risk Analytics
David D. Yao, Columbia University, Department of Industrial
Engineering, 500 West 120 St, New York, NY, 10027-6699,
United States of America, Jose Blanchet,
Paul Glasserman
This year’s Markov lecture and discussions will provide a survey of risk analytics
as a fundamental tool in operations research. While the focus of business analytics
is on issues of productivity and efficiency: cost savings and revenue/profit
optimization, risk analytics address the complementary issues of sustainability and
resiliency: risk-return tradeoff and related resource allocation decisions and
mitigation strategies. Some of the applications to be highlighted include: resilient
urban infrastructures, production planning with risk hedging, financial systemic
risk, and securitized insurance products.
MD24
24-Room 401, Marriott
Latent Variable Models in Biomedical Informatics
Sponsor: Artificial Intelligence
Sponsored Session
Chair: Madeleine Udell, Postdoctoral Fellow, Caltech, CMS, Mail Code
9-94, Pasadena, CA, 91125, United States of America,
1 - Computational Phenotyping from Electronic Health Records using
Tensor Factorization
Joyce Ho, University of Texas at Austin, 1 University Station
C0803, Austin, TX, 78712, United States of America,
Jimeng Sun, Joydeep Ghosh
A computational phenotype (a set of clinical features or clinical condition) can
enable cohort identification, allow decision-makers to identify patients for
interventions, and be integrated with systems for real-time clinical decision
support. We developed sparse, nonnegative tensor factorization models to obtain
phenotypes with minimal human supervision. Results on real EHRs demonstrate
the effectiveness of our models to extract medically interpretable concepts from
complex health data.
MD25
2 - Unfolding Physiological State: Mortality Modelling in
Intensive Care Units
Marzyeh Ghassemi, MIT, 32 Vassar Street,, 32-257, Cambridge,
MA, 02139, United States of America,
Accurate knowledge of a patient’s disease state and trajectory is critical in modern
clinical settings. We examined the use of latent variable models to decompose
free-text hospital notes into meaningful features, and the predictive power of
these features for patient mortality. We found that latent topic-derived features
were effective in determining patient mortality both in-hospital and post-
discharge, and a combination of structured and topic features performed best.
3 - Unsupervised Learning of Disease Progression Models
David Sontag, Assistant Professor, NYU, 715 Broadway,
12th Floor, Room 1204, New York, NY, 10003,
United States of America,
Chronic diseases such as diabetes and COPD progress slowly over many years,
causing increasing burden to patients and the healthcare system. Better
understanding progression is instrumental to early diagnosis and precision
medicine. Inferring disease progression from real-world evidence is challenging
due to the incompleteness and irregularity of observations, as well as the
heterogeneity of patient conditions. We propose a probabilistic disease progression
model that address these challenges.
MD25
25-Room 402, Marriott
Economics of IS & OM
Sponsor: Information Systems
Sponsored Session
Chair: Lin Hao, University of Notre Dame, 351 Mendoza College of
Business, Notre Dame, IN, United States of America,
1 - Exploring a New Marketing Platform of Credit Card Companies
Soohyun Cho, University of Florida, 355F STZ, Gainesville, FL,
United States of America, ,
Subhajyoti Bandyopadhyay, Liangfei Qiu
Some credit card companies (CCs) and partner merchants have launched an
exclusive marketing platform for their cardholders. The platform provides either
public promotion through Social Network Services (SNS) or targeted promotion
through their websites. We examine which promotion is more profitable to CCs
and to competitive partner merchants.
2 - Bundling of Digital Products in Music Industry: An Empirical Study
Kyungsun Rhee, PhD Student, University of Washington,
University of Washington, Seattle, WA, 98105, United States of
America, , Yong Tan, Jianping Peng
It is becoming increasingly competitive for music websites nowadays. Due to
highly heterogeneous demand, offering music bundles is a popular strategy to
attract consumers. In this work, we examine the effectiveness of various bundling
strategies using a unique dataset from a music mobile application which contains
variables such as music downloads, ringtone purchase logs and user behavior in
monthly subscription.
3 - E-book Platform Competition in the Presence of Two-sided
Network Externalities
Yabing Jiang, Florida Gulf Coast University, 10501 FGCU Blvd,
Fort Myers, FL, United States of America,
The success of the Kindle e-book platform and the increased popularity of e-books
among readers have attracted extensive competition in the e-book market. We
model the direct competition in the e-book platform market through a two-sided
network externality model and show that publishers can influence consumers’ e-
book platform adoption decisions and the total e-book sales by strategically
deciding the size of contents available on each platform.
4 - The Effect of Online “Following” on Contributions to
Open Source Communities
Mohammadmahdi Moqri, University of Florida,
299 Diamond Blvd, Apt. 5, Gainesville, United States of America,
Liangfei Qiu,
Subhajyoti Bandyopadhyay, Ira Horowitz
Although numerous studies have examined members’ motivation to contribute to
online communities, the positive effect of social factors has not been unanimously
confirmed in different settings. In this study, we estimate the effect of social
factors on members’ contributions in an open source software (OSS) community,
using a large scale dataset of 4 million online members. The results have
implications for online community designers and OSS scholars.