INTELLIGENT
E-LEARNING SYSTEMS
An Intelligent Approach to Flexible Learning Methodologies
Sukanya Ramabadran and Vivekanand Gopalkrishnan
School of Computer Engineering, Nanyang Technological University, Nanyang Avenue, Singapore
Keywords:
e-Learning, Learning Path/Approach, Data Mining Application.
Abstract:
The evolution of the educational industry from adoption of classroom training methods to e-learning systems
has been remarkable and has satisfied its purpose of existence. However, it has not been able to address issues
faced by students who do not want to be constricted with a set pattern of progress. Hence, an Intelligent
e-Learning Systems (IeLS) framework that facilitates flexibility and maximum learners’ satisfaction, was de-
veloped. The framework consists of a presentation component, a data mining component, a business logic
component, a content management component and a database component. The data mining component uses
techniques such as association rule discovery and conceptual clustering to generate recommendations for stu-
dents, course coordinators as well as the institute.This framework was implemented using PHP and MySQL
with various components such as registration, entry test, tutorials, guest-book and bulletin boards. This system
allows flexibility in terms of choice of learning path, change in direction of learning path and change of learn-
ing approach. In this paper we discuss the role that such an intelligent e-Learning system plays in satisfying
the diverse needs of students.
1 INTRODUCTION
Evolution of educational systems has been a dynamic
process influenced by the subject matter being taught,
the audience and the dynamics of the learning envi-
ronment. Of all the factors that have helped in mold-
ing its form, technology has been considered as the
primary factor. Prior to the boom of personal comput-
ers, instructor - led training was the primary training
method. These systems allowed students to interact
with their instructor and classmates. However, they
led to high costs and downtime in terms of travel. Ed-
ucational systems developed subsequently shifted to
ones that were powered by multimedia whose deliv-
ery was by means of CD-ROMS. The anytime, any-
where availability of CD-ROM provided time and
cost savings that the instructor-led educational sys-
tems could not, and helped to reshape the educational
industry. Despite the benefits, the courses powered
by multimedia lacked instructor interaction and dy-
namic presentation making the experience less than
satisfying and slower and less engaging for students.
Subsequent advancement in technology gave rise to
e-Learning systems that facilitated live instructor led
training via the Internet which could be combined
with real-time mentoring and improved learner ser-
vices.
The e-Learning industry by itself has also experi-
enced various stages of transitions. Early forms of
e-Learning were the result of material being trans-
formed into the electronic medium. Its benefits in
terms of flexibility, self-paced learning as well as sav-
ings in terms of cost could not be overlooked. As
e-Learning began to mature, critics pointed out is-
sues such as isolation and importance of interaction
in the context of learning. Evolution of e-Learning
into Learning Management Systems addressed some
of these criticisms (Wesley, 2002). However, apart
from being adaptive, such systems needed to com-
bat challenges such as curricula development, quality
assurance, continuing professional development, and
mutual recognition (Enemark, 2005). Moreover, we
notice that increase in sophistication of management
of the e-Learning system does not necessarily address
107
Ramabadran S. and Gopalkrishnan V. (2007).
INTELLIGENT E-LEARNING SYSTEMS - An Intelligent Approach to Flexible Learning Methodologies.
In Proceedings of the Ninth International Conference on Enterprise Information Systems - AIDSS, pages 107-112
DOI: 10.5220/0002401001070112
Copyright
c
 SciTePress

all needs of the students.
Let us imagine a scenario of a student using the
conventional e-learning systems where he enrolls for
a course say, ‘algorithms’ and opts to receive a ‘cer-
tificate’ for it. He is compelled to study all parts of al-
gorithms although his specific interest lies in dynamic
programming, which he wanted to study at an ad-
vanced level. He would have preferred to study parts
of the course using practical working examples. He
would want to know about the preferred job oppor-
tunities, interest areas as well as learning approaches
associated with the course he is pursuing, about which
he does not have information.
Conventional e-learning systems do not attempt
to satisfy these diverse needs of the students. These
needs manifest themselves in terms of flexible ways
of pursuing courses that encompass design and deliv-
ery of courses.
2 OBJECTIVE
Considering the needs of students, approaches
adopted by current e-Learning systems and also the
challenges faced by them, we highlight the two main
objectives for our work:
• To propose a framework which structures course
contents that facilitate flexibility as well as max-
imize learner’s satisfaction. The proposed frame-
work would facilitate flexibility and be adaptive
to all prevalent scenarios. It would also use in-
puts such as expertise, interest and job preferences
entered by students during entry tests in order to
counter challenges imposed by planning the mod-
ularization of courses.
• To enable such a system with Data Mining which
employs predictive analytic techniques to gener-
ate recommended Student preferences, Institutes
partnerships and Course coordinators’ content de-
sign functions.
3 COMPARATIVE STUDY OF
EXISTING E-LEARNING
SYSTEMS
A number of e-Learning systems exist and we present
a comparative study of four of the more prominent
of these, namely VirtualU
1
, LearnLoop
2
, WBT Sys-
1
http://www.virtualsystems.com
2
http://learnloop.sourceforge.net
tems
3
and NETg
4
. This study brings to the forefront
that these applications, while maintaining high levels
of quality in provided content, have issues that remain
unsolved.
The systems being compared in Table 1, display
various inabilities on the basis of attributes of com-
parison such as flexibility, technology used and cost.
The systems also exhibited other flaws:
• The systems that act as Internet Application
providers do not provide the client with control.
The client manages hosting of e-learning systems
through back-end administration site using stan-
dard web browser. Flexibility of the systems is
also handled by the provider so may not meet de-
sired standards of the clients and would not be
easy to modify.
• For open source systems, maintenance is ques-
tionable. Since multiple people are involved with
its development, authenticity of source code is
questionable. Moreover, the systems that are built
on integration of components face issues of effi-
cient information exchange between the various
components.
These limitations lead us to propose an Intelligent e-
Learning Systems (IeLS) framework as described in
the following sections.
4 IELS FRAMEWORK
IeLS adopts a component-based approach for both
design and development. It consists of a presenta-
tion component, a data mining component, a business
logic component, a content management component
and a database component as illustrated in Figure 1.
4.1 Role of Data Mining Component
Data Mining is performed by analyzing the data re-
lated to students, course coordinators and the insti-
tute, as depicted in Figure 2. This is done to generate
recommended preferences for all the above actors of
the IeLS.
4.1.1 Students’ Perspective
Association Rules would be employed to organize
historical data collected by the IeLS with respect to
the students for:
• Predicting interest areas for a particular course
based on analysis of past preferences.
3
http://www.wbtsystems.com
4
http://www.netg.com
ICEIS 2007 - International Conference on Enterprise Information Systems
108

Table 1: Comparative study of Existing e-Learning Systems.
VirtualU Learnloop WBT Systems NETg
Technology Internet App Provider, Server side -PHP. Learning Object KnowledgeNet
use clients site to ena- Client -JavaScript Architecture using Platform integrates
ble VirtualU features. MySQL-database TopClass e-Learning learning & content
Client uses browser for stores information. Suite. management.
back-end administration.
Flexibility At discretion of VU, More flexibility, Provides modular Enabled by features
not the client. no requirement flexibility. Enables like Interact Now
for plug-ins. integration with auth- & Search Now.
entication policies.
Cost Faculty pages :$750 Freely available. Depends on Depends on
Student pages :$550 Implementation. Implementation.
Password protect :$250
Presentation Com ponent
B usiness Logic C om ponentD ata M ining C om ponent
C ontentM anagem ent
C om ponent
D atabase Com ponent
C onceptual
View
ExternalV iew
Internalview
Institute
U sers Course C oordinator
ExternalR epository D om ain Experts
C ourse
Parts
M odules
Figure 1: Framework of IeLS.
• Predicting of job opportunities for a particular
course based on analysis of data on industry
needs, quality of students who have pursued such
courses in the past as well as economic policy
prevalent in the country.
• Prediction of learning approach that could be
adopted by the students based on past data on pre-
ferred learning methods for those courses.
Association Rule Discovery
Association rule discovery finds relationships or
affinities between item sets. Each transaction consist-
ing of a set of items could be considered as an item
set. Here, the action of a student selecting a par-
ticular course is considered to be a transaction. The
Association rule is composed of two item sets called
the Antecedent and a Consequent. The rules are dis-
played with an arrow leading from the Antecedent to
the Consequent (Ye, 2003). In this case,
{Course Name} → {Job Opportunity}
{Course Name} → {Interest}
{Course Name} → {Learning Approach}
This association rule is accompanied by two statistical
terms to describe it, namely support and confidence.
These can be defined as follows:
Example 1: Let D be the database transactions of
selecting the course ’Computer Science’. Let N be the
number of transactions in D. Each transaction in D is
an item set. Let X be the event of choosing of Job
Opportunity as ’Software Architect’ by the student
choosing ’Computer Science’ as his course. Then
Support(X) is the proportion of transactions that con-
tain item set X, i.e. Support(X) = |{I|I ∈ D ∧ I ⊇
X}|/N, where I is an item set and k.k denotes the car-
dinality of a set.
The Support of an association rule is the propor-
tion of transactions that contain both antecedent and
the consequent. The Confidence of an association rule
is the proportions of transactions containing the an-
tecedent that also contain the consequent (Ye, 2003).
Example 2: For an association
{Course Name} → {Job Opportunity}, i.e., C → J
Support(C → J) = Support(C ∪ J)
Con f idence(C → J) = Support(C → J)/Support(C)
Hence the Institute would set minimum bounds on
support and confidence measures. If the support for
the job of software architect crosses the minimum
bounds then it would get listed as the one of preferred
job opportunity.
4.1.2 Course Coordinators’ Perspective
Association rules would again be employed by the
IeLS to organize historical data for Course Coordina-
tors. Here they would analyze preferences of students
with respect to difficulty levels that they opt for a par-
ticular module. Hence it can suggest to the Course
Coordinator to deliver modules at the most preferred
INTELLIGENT E-LEARNING SYSTEMS - An Intelligent Approach to Flexible Learning Methodologies
109

Database
Domain
Experts
External
Repository
Data
Pre-processing
Data
Mining
Association
rules - Students
Clustering -
Institute
Association rules
- Co-ordinator
Support &
Confidence Measure
Lift Measure
DFBnB
Data Mining
Yield
Figure 2: Data Mining Layer Functionality.
level of difficulty.
Measure of Interestingness: In this case, in or-
der to study the difficulty level preference, a finer
level of granularity is being adopted; we need to re-
duce the number of association rules identified so that
only the most interesting ones are analyzed. This is
done by employing a popular measure of interesting-
ness called ’Lift’ (Ye, 2003).
Lift: Lift is defined as the ratio of the frequency
of the consequent in the transactions that contain the
antecedent over the frequency of the consequent in the
data as a whole.
Here the association rule would be,
{Module Name} → {Level o f Di f f iculty} i.e.
M → L
Li f t(M → L) = con f idence(M → L)/support(L)
With this we would be able to determine the pref-
erence of students with respect to levels of difficulty
for particular courses.
4.1.3 Institutes’ Perspective
Conceptual Clustering would be employed to orga-
nize the historical data collected by the IeLS with re-
spect to the Institute for:
• Designing of Marketing strategies to be employed
for the launch of new courses.
• Studying performance of students belonging to
various streams as well as thee economic trends
prevalent in the country to decide the Institutes
partner both other educational institutes as well as
organizations.
Conceptual Clustering: This could also be ap-
plied in the case of decisions regarding partnerships.
The conceptual clusters would be the groups of tar-
get audiences for whom the new courses would be
launched. We would then use partitional clustering
algorithms on the conceptual clusters hence devel-
oped to partition them into mutually exclusive clus-
ters (Jain and Dubes, 1988).
Depth First Branch and Bound Searching: We
would then use the cluster generated by conceptual
clustering which is most appropriate for that particu-
lar new course to be launched.
Starting at an initial node which is associated with
a global upper bound, which is the cost of implement-
ing the marketing strategy on that space. DFBnB, se-
lects the node generated next or the deepest node to
be expanded next. Whenever a node is reached whose
cost is less than the upper bound, the latter is revised
to the cost of this new leaf.
Hence we reach an optimal target audience group,
the cost to implement a marketing strategy for who
would be minimal.
4.2 Role of Other Components
The Presentation component is accessed by the users
of the IeLS, namely students, the Institute and course
coordinators.
The Business Logic component is the Controller
component of the framework. It serves as the logical
connection between the user’s interaction through the
presentation component and the database component.
The Content Management component plays the
role of a centralized repository to handle all content
related information of courses, parts and modules.
This also archives references materials that are in-
dexed by keywords.
The Database component is fragmented into
student database, coordinator database, institute
database and a database to store course related infor-
mation.
5 IELS IMPLEMENTATION
The Intelligent e-Learning System was implemented
using PHP as the front end and MySQL to manage
the database. The most important feature of the sys-
tem is modularity. Various functionality such as Reg-
istration, Entry Test, Tutorials and Guestbook facility
have all been designed and implemented as individual
modules. Hence scalability as well as extensibility in
terms of performance of the system is very sound.
5.1 Course Design
Course Path is chosen by the student initially when he
tailors his course by giving his preference regarding
various parts and choosing various modules based on
his preferred level of difficulty. This could be altered
ICEIS 2007 - International Conference on Enterprise Information Systems
110

by him at any point of time through the duration of
the course.
Figure 3: Intelligent e-Learning System Implementation.
The screen-shot in Figure 3, depicts the Course
Map of a student who has chosen the course Algo-
rithms. It consists of various Parts, the arrows depict-
ing the direction of temporal dependency between the
various parts. Here, the parts in blue that have been
bolded indicate they have been pursued. The other
parts in blue indicate ones that could be pursued at
that point of time. The parts that have been grayed out
indicate ones that could not be pursued as the parts on
which these are dependent temporally have not been
covered yet. The one with double border indicates
the students’ current selection. The sub components
of the currently selected one are the Modules which
have been categorized as Basic, Intermediate and Ad-
vanced.
5.2 Evaluation Mechanism
The System also predicts the grades associated with
the kind of award he would obtain, which could be
a Certificate, Diploma or Degree. There is a hi-
erarchy in the awards and the grades predicted fol-
low the same hierarchy which is monotonically non-
increasing. The system also generates a progress bar
associated with very kind of degree. The progress bar
indicates the threshold in terms of completion of re-
quired parts for a particular course or completion of
required courses for the entire duration of education.
5.3 Learning Approaches
Education would be delivered to students using either
synchronous or asynchronous modes. For students
who are not connected to the system at all points in
time, teachers and students may communicate asyn-
chronously (at times of their own choosing) by ex-
changing printed or electronic media or when they use
technology such as Internet that allows them to com-
municate in real time then it would be using the syn-
chronous mode. The e-Learning system may use two
innovative approaches in order to impart education to
the students.
The Interactional approach requires the student to
form peer communities with other students pursuing
similar courses. They could then use e-bulletin boards
to post their queries which would be answered by stu-
dents themselves. Alternatively, the interactional ap-
proach could also be followed when the student com-
municates with the course coordinators for queries
and problems.
The motivational approach in turn consists of two
ways which are reward motivated and recognition mo-
tivated. Reward motivated would consist of a system
where the student would be incentivized with reward
points for the successful completion of every module
test.
For high aspirers, the system could award them
with relevant diploma or certificates for the comple-
tion of each part of that particular course. This would
be the recognition motivated approach.
6 EVALUATION OF IELS’S
FUNCTIONAL EFFICACY
Success of the implementation of any e-Learning sys-
tems would be determined by the degree to which
structural and technological issues have been over-
come (Nunes and McPherson, 2006). Here we eval-
uate the design and delivery of the IeLS learning
system by judging how it has supported an efficient
learning process. We use the famous taxonomy de-
vised by Benjamin Bloom which is considered an ex-
tremely valuable attempt at classifying the learning
process (Bloom, 1956). It lists four major categories
of learning. We then map features of IeLS that satisfy
Bloom’s categories, and evaluate the framework with
respect to some quantifiable parameters.
6.1 Cost
The costs are estimated with a fair degree of accuracy
and are done in the following categories:
• Implementation costs: These are calculated by
considering the cost involved with IeLS with ex-
isting educational systems in the institution. Since
IeLS is very modular and exhibits low cohesion it
INTELLIGENT E-LEARNING SYSTEMS - An Intelligent Approach to Flexible Learning Methodologies
111

Table 2: Evaluation of Functional Effectiveness of IeLS based on Bloom’s Taxonomy (Bloom, 1956).
Category Bloom’s context Mapped Feature of IeLS
Knowledge Ability to recall facts, Periodic Evaluation tests
rules and concepts. for both Modules and Parts.
Comprehension Ability to understand facts Adopting Learning Approach as per the
and concepts. requirement of the Course content or
need of student enables understanding.
Application Ability to use facts and Case studies, lab sessions, bulletin boards
concepts to solve problems. enable putting the gained knowledge to use.
Synthesis Ability to integrate Ability to choose components of diverse
components into whole. backgrounds, mapping to student’s interest.
is very easy implementable and these costs are es-
timated to be negligible.
• Start up costs: These are estimated to be zero as
there is no requirement of purchase of external
components either in the form of software plug-
ins or hardware devices for the implementation of
IeLS.
• Operating costs: These are estimated in terms of
the maintenance costs of a workable system, and
are negligible, since IeLS is developed using open
source components such as PHP and MySQL.
• Expenses incurred by Course coordinator: Staff
involved with development of the course content
would not put in any time that are not budgeted
for in the instructional costs (Knapper, 1980) as
they would simply be providing the URLs of their
course material as values to the relevant column
of the course content database.
6.2 Time
• Learning Time: This would vary with the speed
adopted by the student and the nature of education
he wishes to pursue.
• Accessibility Time: This is considered to be zero
as the entire system is Web based and hence the
students could reach both instructors as well as
Course content at any time they wish to.
• Evaluation Time: Periodic Tests that are con-
ducted to evaluate the students could be done so
at the initiation of the student, and he could also
choose to end the test as per his convenience after
which he would be evaluated based on the Parts
covered.
6.3 Acceptance Criteria
The IeLS Framework was constructed based on the
study of student preferences regarding the most de-
sired features of an ideal e-Learning system and hence
it conforms well to their requirements. So acceptance
of such a system can be estimated to be over 95%.
7 CONCLUSION
The framework of IeLS was developed that structures
course contents to facilitate flexibility as well as max-
imum learners’ satisfaction. This framework is adap-
tive to technological advancements. It is enabled with
data mining functionality that uses predictive analytic
techniques to generate recommended student prefer-
ences, institute partnerships and course contents.
REFERENCES
Bloom, B. S., editor (1956). Taxonomy of educational
objectives: The classification of educational goals.
Handbook I, cognitive domain. Longmans, Green,
New York.
Enemark, S. (2005). Global trends in surveying education -
and the role of FIG. Azimuth, 43(3):19–21.
Jain, A. K. and Dubes, R. C. (1988). Algorithms for clus-
tering data. Prentice-Hall Inc., NJ, USA.
Knapper, C. K. (1980). Evaluating Instructional Technol-
ogy. Croom Helm, London.
Nunes, M. B. and McPherson, M. (2006). Organisational
issues for e-learning: Critical success factors as iden-
tified by HE practitioners. International Journal of
Educational Management, 20(7):542–558.
Wesley, D. (2002). A critical analysis on the evolution
of e-learning. International Journal on E-Learning,
1(4):41–48.
Ye, N., editor (2003). The Handbook of Data Mining.
Lawrence Erlbaum Associates.
ICEIS 2007 - International Conference on Enterprise Information Systems
112