Professional Documents
Culture Documents
The Recent State of Educational Data Mining: A Survey and Future Visions
The Recent State of Educational Data Mining: A Survey and Future Visions
DCSA, PU
Chandigarh, India
rs.karansukhija@gmail.com
PURC, PU
Muktsar, India
manishphd@refiffmail.com
UIET, PU
Chandigarh India
navagg@gmail.com
Abstract
Educational Data Mining (EDM) relates to the inter-disciplinary
research that deals with the development of various methods and
techniques to explore the data generated from different educational
sources. EDM techniques investigate the data in the pursuit of
answers to educational questions and unknown patterns which
surface after the investigations. This survey paper pictures the
evolution of EDM by bringing to light the aspects and outcomes of
various studies carried out over time (from 2001 to 2015). The
paper first introduces EDM after which the various entities
involved in the process that contains: the objectives and
components of EDM are discussed. Then it briefs the research
work carried out over a period of time providing a sequential time
line for further analysis of the scenario. Then the paper leads to
listing the various tools and techniques used in EDM. The paper
then proceeds to list the various tasks in educational environment
that have been resolved via EDM techniques. Finally the current
state of EDM is discussed leading to some of the promising future
lines and needs of EDM for better result and outcome yielding.
Keywords: Educational data mining, educational systems,
Components of EDM, EDM survey, EDM Projected
Enhancement.
1.
Introduction
c
978-1-4673-6747-9/15/$31.00 2015
IEEE
354
(b)
METHODS
AND
TECHNIQUES
DA
ATA
STAKE
HOLDERS
EDM
KNOWLEDGE
Fig. 1 Components of EDM
M
1.2 Components of EDM
(a) Stake Holders: Encompassing primary to higher
education three different stake holders can be defined on
the basis of their involvement.
Primary interest: Direct involvemennt is considered in
this group the students and faculty are
a considered the
main concern.
Indirect involvement: The individuaals leading to the
growth of institutions by indirect involvement are
included e.g. Parents and Alumni
Decision centric: This group is innvolved with the
process of decision making and the administrative
a
task
e.g. employers, planners and experts.
(b) EDM Environment
Formal: The traditional environment involving face to
face teaching and the data extraaction from the
institutionally maintained records is categorized
c
as the
formal environment.
Informal: This involves the scenarios where interaction
on an indirect basis is commuted such as
a e-learning [5].
2015 IEEE 3rd International Conference on MOOCs, Innovation and Technology in Education (MITE)
355
Zaane
and Luo,
2001[10]
356
Database
Tools
Methods
DATA FROM 2001 2005
Outcomes
Advance
d Scout
for
National
Basketba
ll
Associat
ion
Facilitation
for
knowledge
discovery
NBA data
Web Log
of
Technical
university
british
Columbia
N.A
Attribute
Focusing
Associati
on and
pattern
mining
operation
Improved
online
course
activities
Sheard et
al.
2003[11]
Behrouz,
et al,
2003[12]
Data from
students
enrolled
for distant
learning
courses
Wire
website
interactive
data set
Logs of
web based
education
system
Multist
ar
Associati
on and
classificat
ion
Better
knowledge
discovery
SPSS
data
analyze
r
Log
analysis
technique
Identificati
on of
student
learning
behavior
Grade
prediction
with
higher
accuracy
Genetic
algorithm
N.A.
2015 IEEE 3rd International Conference on MOOCs, Innovation and Technology in Education (MITE)
Freyberge
r, et al,
2004[13]
Vandam
me, et al,
2007[14]
M.
Ramaswa
mi and R.
Bhaskara
n,
2009[15]
Asma and
Nusari,
2010[23]
S.Kannan
and
R.Bhaska
ran,
2010[24]
C.
MRQU
EZ, et al,
2011 [17]
MrquezVera. et
al, 2012
3.
Data from
online
algebra
tutor
Associati
ve mining
N.A.
N.A.
Rule
ordering
for
associativ
e
classifier
Better
transfer
model for
quicker
access
[25]
Efficient
performan
ce
prediction
of students
S
Agarwal.
et al,
2012 [18]
Increase in
prediction
accuracy
due to min
number of
features
student
behavior
correspond
ing to
academic
achieveme
nt
accuracy
of
associative
classifiers
can be
improved
using
appropriate
interesting
ness
measures
instead of
supportco
nfidence
framework
Cross
validation
is shown
and
compared
predict
student
failure at
670 high
school
students
from
Zacatecas,
Mexico
Data from
Communit
y college
M.
Ramaswa
mi et al,
2012 [26]
Data base
of
secondary
students
MrquezVera. et
al, 2013
[19]
670
middleschool
students
from
Zacatecas,
Mxico
15 year
data from
college of
science
and
technology
ABED
Ahmed.
et al,
2014 [27]
M angel.
et al,
2015 [28]
Data from
a Spanish
high
school
M Wook.
et al,
2015
[29]
Survey
data of
158
students
from four
public
Institution
s of
Higher
Learning
in
Malaysia
N.A.
WEKA
N.A.
WEKA
WEKA
N.A.
algorithm
school in
secondary
education
SVM
decision
tree
operation
Matured
faith on
Data
Mining
techniques
Relation
between
socioecono
mic and
academic
factors
Better
prediction
of student
failure
Bayesian
Networks
Jrip,
NNge,
OneR,
J48, C 4.5
decision
tree
method
General
DM
technique
s
Technolo
gy
Acceptan
ce Model
Helps
improveme
nt in
academics
by
identifying
poor
students
Prediction
for early
interventio
n in
corrective
manner
Students
accept the
usefulness
of EDM
technology
for
analyzing
academic
data which
could
improve
their
performan
ce
context bigger datasets have rarely been used. Apart from size
other major factors which need to be addressed are the
versatility related issues of the data sets currently available.
The databases used in different studies such as [29] are from
single sources which may fail to reflect the versatility of the
educational system and students. Thus there is a need of a
dataset that can represent the social and geographical
panorama.
In addition to a more cognizant data base better
techniques of mining are also needed. Techniques identified in
the literature have worked predominantly in isolation from
other techniques thus there is a need of hybridized techniques
which can compensate and complement each other. The use of
hybrid techniques has been proved useful in other applications
of DM. Thus there is a need of exploring the horizons of
hybridized algorithms for EDM also.
2015 IEEE 3rd International Conference on MOOCs, Innovation and Technology in Education (MITE)
357
Future Scope
a.
b.
c.
d.
e.
5.
358
2015 IEEE 3rd International Conference on MOOCs, Innovation and Technology in Education (MITE)
2015 IEEE 3rd International Conference on MOOCs, Innovation and Technology in Education (MITE)
359