2016 INFORMS Annual Meeting Program

470 / 561

INFORMS Nashville – 2016
470
WD48
210-MCC
Social Media and Marketing Analytics
Invited: Social Media Analytics
Invited Session
Chair: Amir Gandomi, Ryerson University, 350 Victoria Street,
Toronto, ON, M5B 2K3, Canada,
1 - Big Data And Marketing Analytics In Practice: Key Skills
And Competencies
Amir Gandomi, Ryerson University, 350 Victoria Street,
Toronto, ON, M5B 2K3, Canada, ,
Morteza Mashayekhi, Margaret Osborne, Michael Sparling
Several recent studies report a significant talent gap in analytics. This study aims
to address this gap by establishing a body of knowledge for marketing analytics
professionals. We perform a large-scale empirical study using the data from a
professional social network. By analyzing the unstructured textual data, we
specify the required skills and competencies in different domains of marketing.
2 - Ranking Cuisines And Customers From Geosocial
Restaurant Data
Syagnik Banerjee, University of Michigan-Flint,
We mine location based check-ins and tweets from 40 types of restaurants and
400 visitors across San Francisco, Chicago, Boston, New York, DC and Seattle to
fit a 2 parameter IRT model that estimates the cuisine conspicuity of the
restaurant and the palate diversity of the visitors.
3 - Learning Product Attribute Embedding Model From
Online Reviews
Feng Mai, Stevens Institute of Technology, Hoboken, NJ, United
States, , Xin (Shane) Wang, David J Curry,
Roger Chiang
Online reviews provide a valuable source of information to help identify the latest
attitudes, opinions, and preferences circulating among consumers. We propose a
product attribute embedding model that uses a deep learning approach to learn
computable semantic vectors from online reviews. The dimensions of the vector
space correspond to latent consumer needs, while product attributes are indexed
by their capacity to meet these needs. We demonstrate the implications of the
model in market structure analysis and conjoint analysis.
WD49
211-MCC
Teaching Analytics Using Teradata University
Network Resources
Sponsored: Education (INFORMED)
Sponsored Session
Chair: Ramesh Sharda, Oklahoma State University, Stillwater, OK,
United States,
1 - Teaching Analytics Using Teradata University Network Resources
Ramesh Sharda, Oklahoma State University,
Teradata University Network (TUN) is a group of academics supported by Teradata
to develop and share teaching and learning resources for analytics. Several
thousand faculty and students around the world are using TUN resources. These
resources include cases, assignments, software such as SAS Visual Statistics,
Teradata Aster, and others. The panelists will describe how they use various TUN
resources in their analytics courses. Sharda will introduce TUN and the Aster
platform for teaching Big Data technologies. Nestler will describe how he has used
sports analytics material. Delen will cover SAS Visual statistics. Bhaskar will
discuss how marketing analytics can be taught using TUN resources.
2 - Panelist:
Scott Nestler, University of Notre Dame,
3 - Panelist:
Dursun Delen, Oklahoma State University,
4 - Panelist:
Rahul Bhaskar, California State University-Fullerton,
WD50
212-MCC
Opt, Nonlinear Programming I
Contributed Session
Chair: Harsha Nagarajan, Postdoctoral Research Associate, Los Alamos
National Laboratory, 3000 Trinity Drive, Apt 8, Los Alamos, NM,
87544, United States,
1 - On Broyden-Conjugate Gradient Methods For Solving
Unconstrained Optimization Problems
Idowu Ademola OSINUGA, Federal University of Agriculture,
Abeokuta, Department of Mathematics, Alabata Road, Abeokuta,
234, Nigeria,
An hybrid methods known as BFGS-CG plays an importatnt role in solving large-
scaled optimization problems. Hence, we carried out computational experiments
on standard BFGS-CG methods (BFGS-FR, HS & PR). In comparison with the
newly introduced hybrid methods BFGS-BAN and BFGS-IMW, the hybrid
method BFGS-IMW shows significant improvement in the total number of
iterations and CPU time required to solve large-scaled optimization problems.
2 - Fast Solutions With Performance Guarantee For Operational
Decisions In Real Time Industrial Gas Network Problems
Pelin Cay, Lehigh University, Department of Industrial and Sys
Engineering, 200 W Packer Ave, Bethlehem, PA, 18015, United
States, , Robert H Storer, Luis Zuluaga,
Camilo Mancilla
In the gas distribution industry, meeting customer demand in real time while
meeting the physical constraints in a gas pipeline network leads to complex and
challenging optimization problems. We study the performance of different
approaches in the literature to either find global optimality or determine the
optimality gap between the best local optimum and a valid lower bound. In
industry-sized problem instances significant improvements are possible using a
better reformulation compared to solving the standard formulation of the
problem.
3 - Convex Hulls Of Graphs Of Bilinear Functions On The Unit Cube
Fabian Rigterink, University of Newcastle, University Dr,
Callaghan, 2308, Australia, , Natashia
Boland, Akshay Gupte, Thomas Kalinowski, Hamish Waterer
In his 1989 seminal paper, The boolean quadric polytope: Some characteristics,
facets and relatives, Padberg introduced five classes of valid inequalities for the
boolean quadric polytope: triangle, clique, cut, generalized cut and odd cycle
inequalities. These inequalities outer-approximate the convex hull of a bilinear
function. In this talk, we study classes of bilinear functions where some of the
Padberg inequalities characterize the convex hull. Furthermore, we study which
of the inequalities are strongest, i.e., outer-approximate the convex hull best. We
then apply the strong inequalities to (QC)QP instances from the literature to find
good lower bounds fast.
4 - A Spatiotemporal Radiotherapy Planning Approach For
Cancer Treatment
Ali Adibi, PhD Student, Wichita State University, 7413 E 18th
Street N, Wichita, KS, 67206, United States,
Ehsan Salari
Radiotherapy is one of the main modalities for cancer treatment. This research
aims at developing a spatiotemporal radiotherapy planning optimization approach
and evaluating the potential benefit of varying radiotherapy plans and thus the
spatial dose distribution over the treatment course. The proposed approach is
applied to a phantom cancer case to test its computational performance.
5 - Tightening McCormick Relaxations For Nonlinear Programs Via
Dynamic Multivariate Partitioning
Harsha Nagarajan, Postdoctoral Research Associate, Los Alamos
National Laboratory, 3000 Trinity Drive, Apt 8, Los Alamos, NM,
87544, United States, Mowen Lu,
Emre Yamangil, Russell Bent
In this work, we propose a two-stage approach to strengthen piecewise
McCormick relaxations for mixed-integer nonlinear programs with multi-linear
terms. 1st stage exploits constraint programing techniques to contract the variable
bounds. In 2nd stage, we partition the variable domains using a dynamic
multivariate partitioning scheme via sparse addition of binary variables, which is
independent of the size of variable domains. We demonstrate the performance of
the proposed algorithm on well-known MINLPLIB problems and discuss the
computational benefits of CP-based bound tightening procedures.
WD48