Intelligent Information Ma nagement, 2011, 3, 43-48
doi:10.4236/iim.2011.32005 Published Online March 2011 (http://www.SciRP.org/journal/iim)
Copyright © 2011 SciRes. IIM
Data Mining Technology across Academic Disciplines
Lesley Farmer1, Alan Safe r2, Eric Chuk3
1California State University , Long Beach, US A
2California State University , Long Beach, US A
3University of California at Los Angeles, Los Angeles, USA
E-mail:{lfarmer, asafer}@csulb.edu, echuk@ucla.edu
Received December 3, 2010; revised January 7, 2011; accepted January 28, 2011
Abstract
University courses in data mining across the United States are taught primarily in departments of business,
computer science/engineering, statistics, and library/information science. Faculty in each of these depart-
ments teach data mining with a unique emphasis, although there is considerable overlap relative to course
offerings, terminology, technology, resources, and faculty publications. Content analysis research aims to
describe in detail the range of data mining technology differences and overlap across academic disciplines.
Keywords: Data Mining, Statistics, Academics
1. Introduction
Data mining is essentially the process of uncovering
meaningful new correlations, patterns and trends from
large quantities of complex data using statistical and
mathematical techniques. With the help of powerful
computers, new applications of data mining have been
developed recently and have expanded its areas of use.
Data mining is now applied in such diverse fields as
medicine, education, finance, marketing, meteorology,
and national defense, along with many applications asso-
ciated with the Internet.
Since the mid-1990s, many more university courses in
data mining are being taught across the United States.
The major departments teaching such courses are com-
puter science/engineering, business, statistics, and li-
brary/information science. In each discipline, data min-
ing is taught with a moderately different emphasis (see
for example Olson and Shi, 2006; Duda, Hart, and Stork,
2000; Hastie, Tibshirani, and Friedman, 2009). In busi-
ness, applications include: identifying credit card fraud,
insider trading patterns, and defect analyses. In the sci-
ences, applications include: Medicare fraud, astronomi-
cal variations, and disease risk. In statistics, new analytic
approaches are being developed, such as fuzzy logic
(Larose, 2005; Berry and Lindoff, 2004; Roiger and
Geatz, 2003). In library and information sciences, both
theoretical and technical approaches are used, often
bridging this field and specific professions such as law,
industry, and the health sciences.
As a result of these various applications, different
software, textbooks, and techniques are being used. To
clarify the differences and similarities in each discipline,
this study will examine the major academic variations
within the data mining field in relation to keywords, arti-
cles, books, courses offered, textbooks taught, and soft-
ware used.
2. Method
2.1. Keywords Used to Identify Data Mining
Courses across Disciplines
Data mining keywords from different disciplines were
identified in 2009 by searching a compiled list of data
mining courses for each of four academic disciplines:
business, computer science/engineering, statistics, and
library/information science. Graduate programs were
exclusively searched in this regard since these courses
are routinely taught at that level. The findings on com-
puter science and engineering were combined since there
was much overlap of courses in these disciplines. The
count for statistics cou rses was obtained only in statistics
departments. To find these courses, various accrediting
societies and associations were consulted to identify
universities offering programs in each of those disci-
plines. The university sites were then searched for
courses relating to data mining. Browsing course cata-
logs and department-specific webpages for relevant
course titles and descriptions resulted in finding key-
L. FARMER ET AL.
44
words.
The list of universities with business programs offer-
ing data mining courses was obtained from the Associa-
tion to Advance Collegiate Schools of Business, https://
www.aacsb.net/eweb/DynamicPage.aspx?Site = AACSB
&WebKey=00E50DA9-8BB0-4A32-B7F7-0A92E98DF
5C6. From that list of business schools, the keywords
used were: business intelligence, decision support, and
data mining.
The list of universities with computer science pro-
grams was obtained from the Accreditation Board for
Engineering and Technology, http://www.abet.org/school-
allcac.asp.The keywords relating to computer science
data mining courses were: machine learning, artificial
intelligence, and data mining.
The list of engineering programs was taken from the
Accreditation Board for Engineering and Technology,
http://www.abet.org/ schoolalleac.asp. The keywords in
engineering courses relating to data mining were: pattern
recognition, artificial intelligence, and data mi ning.
The computer science and engineering programs were
obtained from the same accrediting association but dif-
ferent keywords were utilized.
The list of statistics programs obtained was on the
website of the American Statistical Association at
http://www.sci.csueastbay.edu/~mwatnik/statlist. The key-
words from statistics programs offering data mining
courses were: neural network, decision tree, and data
mining. The keyword count for statistics was obtained
from schools with graduate programs in statistics, but
math departments were exclud ed because of the extre mely
small number of data mining courses in mathematics
outside of schools offering a graduate degree in statisti cs.
The list of library and information science programs
was from the American Library Association, http://www.
ala.org/ala/education careers/education/ accreditedpro-
grams/index.cfm. The keywords associated with data
mining courses were: informatics, information retrieval,
information management, knowledge management, know-
ledge discovery in databases, competitive intelligence,
bibliometrics, biometrics, bibliomining and data mining.
2.2. Top Resources and Publications by
Discipline
Information about the most commonly used books and
software by discipline was collected from course syllabi
and instructors’ replies to email in 2009. A reply speci-
fying book(s), software, or both was counted as a re-
sponse. There is no further breakdown of the percent
who responded with each piece of information because
the counts from the two sources, syllabi and emails, were
not kept separa t e l y .
The average number of articles per year from 1990 to
September 2009 was calculated for the disciplines: busi-
ness, computer science/engineering, statistics, and li-
brary/information science. The phrase “data mining” in
the abstracts of journals, books, and conference pro-
ceedings was used to search the business database ABI
Inform Complete, the computer science/engineering da-
tabase Compendex, the statistics database Current Index
to Statistics, and the library/information science data-
bases Library Literature and Information Science and
Library, Information Science & Technology Abstracts. It
should be noted that in the Current Index to Statistics,
which is the main database for statistics, there was no
specific identifier for abstracts (as in the other two data-
bases), so title/keywords was the closest option. This is
very likely the reason for a lower number of results
found in the statistics search. If one looks at the number
of articles divided by the number of departments in a
particular discipline having data mining courses, one can
compare articles/department across disciplines to com-
pare publication productivity. The list of departments
was obtained from the same list as keywords to identify
data mining courses in the different disciplines.
3. Results
There were 75 business faculty surveyed by email and 48
responded (64%) providing information on data mining
books or related software. Of the 235 computer sci-
ence/engineering faculty surveyed who were teaching
data mining courses, 127 responded (54%). For inquiries
from statistics departments, 31 of the 44 surveyed re-
sponded (a 70% email response rate of either book or
software or both). All library/information science pro-
grams had online information. Although the degree of
response combined both texts and so ftware, the text titles
and the type of software were recorded separatel y .
3.1. Courses by Discipline
Once the data mining courses were identified by disci-
pline, the number of departments offering them was de-
termined. That 2009 data are reported in Table 1. The
courses are listed by departments because of the marked
variation in courses by department. Note that the com-
puter science/engineering departments offer the most
graduate data mining courses followed by business
school offeri ng s.
3.2. Keywords by Discipline
Keywords obtained from university catalog titles, course
listings and descriptive words relating to data mining
Copyright © 2011 SciRes. IIM
L. FARMER ET AL.
Copyright © 2011 SciRes. IIM
45
courses from each of the four major disciplines are
shown in Table 2. Keyword overlap between disciplines
is surprisingly infrequent.
3.3. Software by Discipline
In the email responses from academicians in each disci-
pline, numerous types of data mining software were re-
ported. These are presented as proportions in Figures 2,
3, 4, and 5.
3.4. Books by Discipline
Data mining books vary by discipline largely because
their focus and applications differ. Some of the leading
books as identified in 2009 are listed below by discip line.
Only the leading books are listed. The Russell and Nor-
vig title was the most popular, used more than twice as
often as the next most cited textbook, by Duda, et al.
3.4.1. Business
Witten, I., and Frank, E. 2005. Data Mining: Prac-
tical Machine Learning Tools and Techniques.
Morgan Kaufmann, Burlington, MA.
Berry, M., and Linoff, G. 2004. Data Mining
Techniques: For Marketing, Sales, and Customer
Relati onship Managem e nt . Wil ey, New York.
Olson, D., and Shi, Y. 2005. Introduction to
Table 1. Number of U. S. university departments offer i ng data mining c ourses by discipline.
Discipline # of depts. with data mining
courses % of total # of depts. in discipline offering data mining
courses
Business 83 17.6%
Computer Science/Engineering 187 48.8%
Statistics 46 28.0%
Library/Information Science 15 30.0%
Business Computer Science/ Engineering Statistics Library/Info Science
business intelligence adaptive computation association/ link analysis automatic extracting
competitive advantage artificial intelligence clustering (K means, near-est
neighbors) bibliometrics
CRM database/ data warehouse decision trees bibliomining
database mgmt. systems intelligent agents genetic algorithms biometrics
database decision making knowledge discovery in databases machine learning business intelligence
data warehouse machine learning model validation competitive intelligence
decision support systems multidimensionality(data cubes) neural networks/ fuzzy logic content mining
intelligent enterprise neural networks/neurocom put ing
processing nonparametric learning database management
knowledge mgmt./dis-
covery mgmt. text mining pattern recognition database de cisi on making
information systems support vector machines data warehouse
market-basket analysis training/testing dataset decision support
OLAP unsupervised learning fuzzy logic
quantitative methods health informatics
informatics
information mgmt.
information retrieval
knowledge mgmt.
knowledge disc/database
quantitative methods
text mining
L. FARMER ET AL.
Copyright © 2011 SciRes. IIM
46
SAS
34%
Excel
19%
other
12%
Access
11%
SQL
7%
(none used)
7%
Oracle
5% SPSS
5%
Figure 2. Business Data Mining Software by Brand Name
(n = 57).
Figure 3. Computer Scie nce/Engineering Data Mining Soft-
ware by Brand Name (n = 118).
SAS
35%
R
30%
other
15%
Ggobi
10%
S+
10%
Figure 4. Statistics Data Mining Software by Brand Name
(n = 18).
SAS
35%
R
30%
other
15%
Ggobi
10%
S+
10%
Figure 5. Library/Info Science Data Mining Software by
Brand Name (n = 21).
Business Data Mining. McGraw-Hill, Columbus,
OH.
Marakas, G. 2002. Modern Data Warehousing,
Mining, and Visualization. Prentice Hall, Upper
Saddle River, NJ.
Shmueli, G. et. al. 2006. Data Mining for Busin ess
Intelligence. Wiley-Intersci ence, Hoboken, NJ.
3.4.2. Computer Science/Engineering
Russell S., and Norvig, P. 2009. Artificial Intelli-
gence. Prentice Hall, Upper Saddle River, NJ.
Duda, R., Hart, P., and Stork, D. 2000. Pattern
Classification Wiley- Interscience, Hoboken, NJ.
Mitchell, T. 1997. Machine Learning. McGraw-
Hill, Columbus, OH.
Luger, G. 2008. Artificial Intelligence. Addison
Wesley, Boston.
Haykin, S. 2008. Neural Networks and Machine
Learning. Prentice Hall, Upper Saddle River, NJ.
Hagan, M. et. al. 2002. Neural Network Design.
Hagan Publishing, Bosto n.
Bishop, C. 1996. Neural Networks for Pattern
Recognition. Oxford, New York.
Tan, P. et. al. 2006. Introduction to Data Mining.
Addison Wesley , Boston.
Han, J. et. al. 2005. Data Mining: Concepts and
Techniques Morgan Kaufmann, Burlington, MA.
3.4.3. Statistics
Hastie, T., Tibshirani, R., and Friedman, J. 2009.
The Elements of Statistical Learning. Springer,
New York.
Larose, D. T. 2005. Discovering Knowledge in
Data. Wiley, New York .
Hand, D. et. al. 2001. Principles of Data Mining.
MIT Press, Cambridge, MA.
Tan, P. et. al. 2006. Introduction to Data Mining.
L. FARMER ET AL.47
Addison Wesley, Ne w Yo rk .
Ripley, B. 1996. Pattern Recognition and Neural
Networks. Cambridge University Press, Cambridge,
UK.
3.4.4. Library and Information Science
Han, J. et. al. 2005. Data Mining. Morgan Kauf-
mann, Burlington, MA.
Witten, I., and Frank, E. 2005. Data Mining. Mor-
gan Kaufman, Burlington, MA.
Shortliffe, E., and Cimino, J., Eds. 2006. Biomedi-
cal Informatics. Springer, New York.
3.5. Data Mining Articles by Discipline
The average annual number of published data mining
articles by discipline from 1990 through mid-2009 is
listed in Figure 6. Note that the average per year in-
crease over the decade in data mining articles in business
journals was nearly two-fold, fifteen-fold in computer
science/engineering journals, seven-fold in library/ in-
formation science articles, and there was little change in
the number of statistics journals. As mentioned previ-
ously, there was no specific identifier for abstracts in the
main database for statistics (as in the other databases), so
title/keywords were used as the closest option. Again,
this is the likely the reason for the lower number of re-
sults found within statistics.
3.6. Data Mining Articles per Department across
Disciplines
For the 2005-2009 period , when the number of pub lished
articles is divided by the number of departments having
data mining courses, the following rate pattern emerges:
5.3 articles per business department, 9.4 articles per
computer science/engineering, 0.9 articles per statistics
department, and 9.9 articles per library/information sci-
ence department. Earlier calculations were not generated
because it is not easily apparent how long data mining
courses have been offered. From this perspective, computer
science/engineering and library/information science fac-
ulty have been the most produ ctive in publishing. Again,
note that the number of articles in statistics is likely un-
derrepresented because the main database in statis tics does
not include abstracts as do databases in the other fields.
4. Discussion
Course offerings dealing with data mining reflect its im-
portance within each discipline. Business courses tended
to incorporate data mining as a way to become more
competitive financially. Computer science/engineering
courses tend to focus on the technical and logical struc-
ture of data mining. Statistics courses emphasize data
mining methodologies with an eye to applications in a
variety of settings as well as comparing methods to more
traditional parametric statistical techniques.
Library/information science reflects a broad range of
perspectives: from logical architecture of data for mining
to field-specific applications of data mining (especially
health and business). Generally, courses blend theory and
practice. Data mining is also considered a viable research
methodology in library/information science, in which
case it is were more likely to be offered at the doctorate
level than at the master’s. In no case is data mining a
required course in library/information science, although
Syracuse University and Wayne State offered specializa-
tions in data management, which included data mining as
an elective.
Beyond the term data mining, each discipline gener-
ated unique associated terms. Business terms focused on
decision-making, management, and competition. Com-
puter science/engineering used more technology-related
and intelligence-related terms. Statistics used more meth-
odological terms. Library/information science terms had
the greatest variation, from fuzzy logic to text mining,
but most terms were associated with applications (e.g.,
Figure 6. Average Annual Number of Data Mining Articles by Discipline from 1990-2009.
Copyright © 2011 SciRes. IIM
L. FARMER ET AL.
48
bibiometrics, health informatics, and information man-
agement). The greatest overlap existed between business
and library/information science due to decision-making
methodology and management issues.
Data mining software varied by discipline. SAS was
the dominant software used in the business and statistics
departments. Statistics had the most stable set of soft-
ware brands. Matlab and C++ were the most frequently
cited software in computer science/engineering courses
for data mining. Computer programming languages, in
general, were used by a majority of those courses. SPSS,
SQL, and Excel were the dominant software used in li-
brary/information science courses. It appears that the
choice of tools depended on the status of the databases to
be utilized. One might assume that courses where com-
puter programming software was used would address
both database creation as well as data mining. Software
also reflected the type of data needed, such as SPSS vs.
RefEVAL or TextQuest. In addition, the choice of soft-
ware might also reflect the technical sophistication
within the academic community, with business using the
least complicated software and computer science/en-
gineering and statistics using the most complex products.
A good deal of overlap exists in textbook choices
across disciplines--and in some cases within disciplines,
especially for library/information science. Tan, et al.’s
Introduction to Data Mining was used in computer sci-
ence/engineering and in statistics, Han and Kamber’s
Data Mining was used in computer science/engineering
and library/information science, and Witten and Frank’s
Data Mining was used in business and library/informa-
tion science. Russell and Norvig’s Artificial Intelligence
was by far the most popular computer science/engineering
textbook. Han and Kamber was the favorite title in li-
brary/information science, although Shortliffe and Ci-
mino’s Biomedical Informatics was the standard text-
book for health informatics within library/information
science. The picture that emerges shows little agreement
on standard textbooks except in computer science/en-
gineering. In specialized subsets of the field, such as
biometrics, few titles may be available from which to
choose. Instead, it appears that textbook choice depends
on the specific course objectives and content focus, the
academic “lens” determining the title to be used. It
would be useful to survey faculty as to the basis for their
textbook choice.
The number of articles over time varies by discipline.
Business published the greatest number before the year
2000, but the rate leve led in the 21st century. By contr ast,
the library/information science article publication rate
has shown a continuing rise, increasing a little over
threefold from the late 1990s to the early 200 0s and then
a bit over twice as many in the past five years. Computer
science articles rose dramatically (over tenfold) from the
late 1990s to the early 2000s, and continued to rise by
nearly 50% in the past five years.
A potential limitation in organizing data mining arti-
cles by discipline is that database aggregators may not
have captured all relevant publications. It should be
noted that another interpretation involving data mining
articles by discipline is that the database aggreg ators may
vary. In addition, deeper investigation into the quality of
the articles would also shed light on the extent of schol-
arly contributions.
5. Conclusions
Data mining courses in the U. S. are available in various
academic disciplines, and the overall field is rapidly ex-
panding. Evidence presented in the figures and tables
makes this abundantly clear. Detailed information con-
cerning overlapping emphases in data mining disciplines
has not been reported heretofore and deserves attention.
Certain other academic areas include data mining courses
and have associated texts and software. Nonetheless, the
four disciplinary fields described in this review cover the
major academic areas at this time. The emerging picture
reveals a blend of theory and practice that reflects each
academic discipline rather than a unified system. Hope-
fully, a productive merging of data mining approaches
through increased cross-disciplinary research can de-
velop and advance all these related fields. The rate of
change in the data mining field is so rapid that the infor-
mation is likely to be measurably different in the next ten
to twenty years.
6. References
[1] M. Berry and G. Linoff, “Data Mining Techniques for
Marking, Sales and Customer Support,” 2nd Edition,
Wiley, New York, 2004.
[2] R. Duda, P. Hart and D. Stork, “Pattern Classification,”
2nd Edition. Wiley-Interscience, New York, 2000.
[3] T. Hastie, R. Tibshirani and J. Friedman, “The Elements
of Statistical Learning,” 2nd Edition. Springer, New
York, 2009. doi:10.1007/978-0-387-84858-7
[4] D. Larose, “Discovering Knowledge in Data,” Wiley-In-
terscience, Hoboken, 2005.
[5] D. Olson and Y. Shi, “Introduction to Business Data Min-
ing,” McGraw-Hill, Columbus, OH, 2006.
[6] R. Roiger and M. Geatz, “Data Mining,” Addison-
Wesley, Boston, 2003.
Copyright © 2011 SciRes. IIM