Informs 2015

155 / 552

INFORMS Philadelphia – 2015
153
many have fleeting interests and only generate noise that skew analyses. Thus we
propose an automated and scalable approach to identify top hacker participants in
big online communities. Specifically, we utilize discrete-time duration modeling
with recurrent events for extracting top participating hackers among large hacker
IRC communities.
2 - Semi-supervised Learning for Structured Regression on Partially
Observed Attributed Graphs
Zoran Obradovic, Laura H. Carnell Professor Of Data Analytics;
Director, Center For Data Analytics And Biomedical Informatics;
Professor, Computer And Information Sciences Department;
Professor, Statistics Department, Temple University, 1925 N. 12th
St., 386 SERC, Philadelphia, PA, 19122, United States of America,
Jelena Stojanovic,
Djordje Gligorijevic, Milos Jovanovic
We propose a structured regression model for temporal applications where a large
fraction of observations is missing. The benefits are demonstrated on precipitation
prediction and on hundreds of evolving graphs with up to 80% labels missing
under various mechanisms. The proposed method and result were published at
SDM 2015 conference proceedings.
3 - From Spot Market to Customized Contract: Role of Reputation on
Market Choice of Crowdsourcing Venders
Xiao Han, Shanghai Jiaotong University, 1954 Huashan Rd,
Shanghai, China, Pei-yu Chen, Bin Gu,
Pengzhu Zhang
Based on a unique dataset obtained from a large online crowdsourcing platform
with different markets in China, we examine the effect of reputation on vendors’
winning chance and revenues as well as their market choices over time. Our
results show that as venders become more experienced and establish more
successful relationships with buyers, they will move from spot market to
customized contract market, at the same time, they will win more and earn more.
4 - The Heterogeneity of Online Searching and Its Impact on
Stock Returns
Qiang Ye, Harbin Institute of Technology, School of Management,
Harbin, China, Xianwei Liu
Internet search frequency on stocks has been widely used to measure investor
attention in recent studies. Researchers found that weekly searching frequency
strongly associated with weekly returns for certain stocks. But we have limited
knowledge about the heterogeneity of Internet searches under different
circumstances and its varying impact on stock market. In this study, we seek to
explore whether searches done on different circumstances guarantee equal
attention.
MA25
25-Room 402, Marriott
Examining the Social Crowd
Sponsor: Information Systems
Sponsored Session
Chair: Gordon Burtch, Assistant Professor, University of Minnesota,
321 19th Ave. S., Minneapolis, MN, 55455, United States of America,
1 - Motivating Group Donation: Evidence from a
Large Field Experiment
Tianshu Sun, University of Maryland Smith School of Business,
3330 Van Munching Hall, College Park, MD, 20740-2840,
United States of America, ,
Gordon Gao, Ginger Zhe Jin
Using a randomized field experiment involving 80,000 participants, we study how
mobile messaging can leverage recipients’ social connections to increase blood
donation. We find 1) blood banks can motivate group formation to increase
donation, but only with group reward; 2) group reward works through a sorting
mechanism; 3) participants donate more blood when donating in a group. Our
study provides insights on the optimal design of mobile messaging for charities to
leverage donors’ social network
2 - Field Experiment on the Impact of Recommender Systems on
Sales Volume and Diversity
Dokyun Lee, Carnegie Mellon University, Pittsburgh, PA,
United States of America,
Kartik Hosanagar
We investigate the impact of collaborative filtering recommender algorithms,
commonly used in e-commerce, on sales volume and diversity. We do this by
running a randomized field experiment with a top retailer in North America. We
show results for sales volume and diversity. We further investigate the genre
cross-purchase behavior to shed insight into mechanism behind diversity shift.
3 - Randomized Experiment(s) Evaluating the Drivers of User
Content Generation
Gordon Burtch, Assistant Professor, University of Minnesota, 321
19th Ave. S., Minneapolis, MN, 55455, United States of America,
, Kevin Hong, Ravi Bapna, Vladas Griskevicius
UGC under-provision is a persistent problem in many online communities. We
undertake a set of randomized experiments aimed at understanding the drivers of
increased UGC production, in terms of both quantity and quality. We consider i)
the role of peer feedback on past content (stated approval vs. download), and ii)
incentives to contribution (social vs. monetary).
MA26
26-Room 403, Marriott
2015 INFORMS Special Session for JIPE and IJOR
Cluster: Globalization and International Activities
Invited Session
Chair: Kuo-Hao Chang, National Tsing Hua University, Hsinchu,
Taiwan, China,
1 - An Activity-Driven Optimization Model for Smart Space
Kuo-Hao Chang, National Tsing Hua University, Hsinchu,
Taiwan, China,
In this paper, we propose an activity-driven optimization model, followed by the
development of a solution method, to enable the optimal and automatic control
of environmental conditions, including humidity and temperature, so as to
provide people with the most comfortable environment but with the least energy
consumption. A computational study based on instances of realistic size is
conducted at the end to show the viability of the research.
2 - Optimal Arc Restoration Schedules with Resource Constraints for
Recovering Nodes in a Network of Damaged Arcs
I-Lin Wang, Professor, Department of Industrial and Information
Management, National Cheng Kung University, Tainan,
Taiwan, China,
Consider a network containing some damaged arcs with estimated restoration
time and resource. In order to access all nodes in minimum time, we aim to find
optimal arc restoration schedules that determine the timing and resources
required for restoring some damaged arcs. We will show that special cases of this
problem can be converted to minimum spanning tree, shortest path, and NP-hard
parallel machine scheduling problems. Finally, proposed IP formulations and
solution methods will be discussed.
3 - The Coopetition Game in International Liner Shipping
Dung-Ying Lin, Associate Professor, Department of Transportation
and Communication Management Science, National Cheng Kung
University, Tainan 70101, Taiwan, China,
In maritime freight transportation, carriers build a collaborative relationship with
other carriers while competing with each other to optimize their own profits. In
this scenario, a game of coopetition is formed. We formulate a nonlinear mixed-
integer problem to determine the optimal levels of coopetition for a single
company and embed the resulting problem into a general game theoretic
framework. A diagonalization algorithm that incorporates an ascent direction
search technique is developed to effectively evaluate the game. The numerical
results show that carriers choose similar coopetition levels to maximize their
profits, and the coopetition game can reach equilibrium under general conditions.
MA27
27-Room 404, Marriott
Theory and Applications of the Analytic Network
Process
Sponsor: Multiple Criteria Decision Making
Sponsored Session
Chair: Orrin Cooper, University of Memphis, 332 Fogelman, Memphis,
TN, United States of America,
1 - Coherence within the ANP Supermatrix
Orrin Cooper, University of Memphis, 332 Fogelman, Memphis,
TN, United States of America, , Idil Yavuz
In an ANP network there is both the ability and the need to check for additional
levels of consistency and coherence. A Supermatrix with priority vectors that
were obtained from either perfect or nearly perfect consistent pairwise
comparison can generate suboptimal decisions. Simulations demonstrate the
frequency of these events in general ANP networks. A form of cross validation
within the Supermatrix called linking validation and a Linking Coherency Index
(LCI) is developed and demonstrated.
MA27