8
dimensional, unknown domains. Also, while AI is commonly
associated with another buzzword, “Big Data”, we wish to prove
that AI can be useful for dealing with problems for which we
possess little or no data. Here, expert knowledge modeling
is critical, and we describe how even a minimal amount of
expertise can serve as a basis for sound reasoning aided by AI.
Sunday, November 13, 11:45am–12:30pm
Technology Tutorial:
AArtificial Intelligence in Marketing
Science: Marketing Mix Modeling and Optimization with Bayesian
Networks & BayesiaLab , “Half the money I spend on advertising is
wasted; the trouble is I don’t know which half.” Various versions of
this quote have been attributed to John Wanamaker, Henry Ford,
and Henry Procter, yet 100 years after these marketing pioneers, in
this day and age of big data and advanced analytics, the quote still
rings true. The current practice is often “more art than science.”
The lack of a well-established marketing mix methodology
has little to do with the domain itself. Rather, it reflects the fact
that marketing is yet another domain that typically has to rely
on nonexperimental data for decision support. Marketing mix
modeling is a causal problem, which means we are not looking for
a prediction of an outcome based on the observation of marketing
variables, but attempting to manipulate the marketing variables to
optimize the outcome. Thus, we must simulate interventions, not
observations, and switch from observational to causal inference.
This brings us to deriving causal inference from observational data.
We introduce the fundamental concepts of graphical models and
how they can help us perform causal identification, i.e., determine
whether it is possible to estimate causal effects from observational
data, which requires causal assumptions about the domain plus
a decision criterion, e.g., the Adjustment Criterion. However,
the complexity of the marketing domain limits the practical
application of this criterion. We introduce the Disjunctive Cause
Criterion, which reduces the number of assumptions required for
causal identification and, thus, confounder selection. Proceeding
from causal identification to estimation requires an “inference
engine.” In the simplest case, we could use a regression, but with
dozens of interacting variables, that is not practical. Instead we
use Artificial Intelligence by employing BayesiaLab’s machine-
learning algorithms, which builds a high-dimensional Bayesian
network model that represents the joint probability distribution of
all variables. This causal inference engine plus BayesiaLab’s Target
Optimization algorithm enable us to search efficiently for the ideal
marketing mix.
Cambridge University Press
34
www.cambridge.org/academicCambridge University Press’
publishing in books and
journals combines state-of-the-
art content with the highest standards of scholarship, writing
and production. Visit our stand to browse new titles, available at
20% discount, and to pick up sample copies of our journals. Visit
our website to find out more about what we do.
Clemson University/COIN-OR Foundation
23
www.coin-or.orgThe Computational Infrastructure for
Operations Research publishes high quality,
free, open-source tools for OR professionals
and students, suitable for commercial,
educational, and personal use. COIN-OR is
the place to go when you need a “white box”
for algorithm research and development. COIN-OR is a strategic
partner of the INFORMS Computing Society.
Cornell Tech
7
http://tech.cornell.eduAt Cornell Tech, students across programs learn
and work side-by-side, spending one-third of their
experience together working on a studio-based core
curriculum. They collaborate with the tech industry and
postdoc-level researchers to build start-up companies
and new products. By bringing these talents together
at the start, there is enormous potential for better,
more impactful and ultimately more successful
companies and products. Programs offered: Master
of Engineering in Computer Science, Master of
Engineering in Electrical and Computer Engineering,
Master of Engineering in Operations Research and
Information Engineering, Technion-Cornell Dual Degree in
Connective Media, Technion-Cornell Dual Degree in Health
Tech, Johnson Cornell Tech MBA, and Master of Laws in Law,
Technology and Entrepreneurship.
Darden Business Publishing
16
http://store.darden.virginia.eduDarden Business Publishing markets
case-based educational materials written
by the renowned faculty at the University
of Virginia Darden School of Business.
Darden maintains a catalogue of student-centered learning
materials that energize classrooms around the world with dynamic
interactive simulations and thought-provoking paper cases.
Dynamic Ideas, LLC
49
www.dynamic-ideas.comDynamic Ideas, LLC is a publisher of scientific
books that have quality and originality in the
areas of Operations Research and Applied
Mathematics. The key objective of our titles is
to “educate the next generation.” Many of our
books are currently being used as the main textbook in academic
courses in some of the finest universities and research institutions
in the world.
Elsevier
26
www.elsevier.com/decisionsciencesElsevier publishes leading journals in OR/MS
and Decision Sciences, including
European
Journal of Operational Research
,
Computers
& Operations Research
, and
Omega-
International Journal of Management Science
.
Elsevier journals occupy 7 of the Top 10 Impact
Factor positions in the Thomson Reuters
‘Operations Research & Management Science’ category. Come
to the booth to find out more, including how to use Elsevier’s
researcher centric tools to develop your research.
FDA/Center for Drug Evaluation and Research
38
www.fda.gov/drugs
The Center for Drug Evaluation
and Research (CDER) performs
an essential public health
task by making sure that safe and effective drugs are available to
improve the health of people in the United States.
Exhibit Listings & Technology Tutorials
All Technology Tutorials will take place in the Music City Center, 5
th
Avenue Lobby.