Description
Business Intelligence in E-Learning
473
Business Intelligence in E-Learning
(Case Study of Iran University of Science and Technology)
Mohammad Hassan Falakmasir
1
, Jafar Habibi
2
, Shahrouz Moaven
1
, Hassan Abolhassani
2
Department of Computer Engineering
Sharif University of Technology
Tehran, Iran
1
{Falakmasir, Moaven}@ce.sharif.edu
1
{Jhabibi, Abolhassani}@sharif.edu
Abstract— Nowadays, e-learning platforms are widely used by
universities and other research-based and educational
institutions. Despite lots of advantages these educational
environments provide for organizations, yet there are many
unresolved problems which cause instructors and training
managers with some difficulties to get proper information about
the students’ learning behavior. On one hand, lack of tools to
measure, assess, and evaluate the performance of learners in
educational activities has led the educators to fail to guarantee
the success of learning process. On the other hand, strict
structure of learning materials prevents students to acquire
knowledge based on their learning style. Consequently,
developing tools monitor and analyze the learner’s interaction
with e-learning environment is necessary. Business intelligence
(BI) and On Line Analytical Processing (OLAP) technologies can
be used in order to monitor and analyze the learner’s behavior
and performance in e-learning environments. They can also be
used to evaluate the structure of the course content and its
effectiveness in the learning process. This article investigates the
use of business intelligence and OLAP tools in e-learning
environments and presents a case study of how to apply these
technologies in the database of an e-learning system. The study
shows that students spend little time with course courseware and
prefer to use collaborative activities, such as virtual classroom
and forums instead of just viewing the learning material.
Keywords- E-Learning; Educational Data Mining; Business
Intelligence; Data Warehouse; OLAP; Intelligent Data Analysis
I. INTRODUCTION
Today, most universities and educational institutions use e-
learning platforms to offer their educational material benefiting
from their advantages such as work “any-time, anywhere”, use
of collaborative tools, support different styles of learning, etc.
However, these e-learning platforms do not cover all teaching
aspects since they do not usually provide teachers and
instructional designers with tools which allow them to monitor
and assess all the activities performed by learners [1,2,3,4].
These environments provide the professor with access
summary information such as the date of the first and last
connection, the number of visited pages according to the
category specified by the platform (not by the professor), or the
number of messages read/sent by each learner; the total number
of page visits, the average time spent on each page but,
globally, per page not by learner. As can be deduced, this
information is not enough to analyze the behavior of learners
and their evolution. Even more, when the number of students
and the diversity of interests are high, the professor has serious
difficulties to extract useful information.
Business Intelligence techniques applied to the database of
Web-based Educational Systems could help instructors and
other educational experts to generate statistics, analytical
models, and uncover meaningful patterns from these huge
volumes of data. In this article, a framework for applying
business intelligence in e-learning environments has been
proposed, which increased both flexibility and performance of
e-learning environments. Hence, on one hand, the proposed
environment enables educational technologists to identify,
analyze and monitor relevant aspects of instruction, such as
different style, paths, and strategies of learning. On the other
hand, such parameters may be used to adapt the learning
process to each individual learner and improve the performance
of the learning process.
The reminder of this paper is organized as follows. Section
2 briefly describes business intelligence principles. A brief
history of the related work is represented in the third section. A
description of the e-learning system and the proposed
architecture are represented in section 4, moreover, three steps
to deploy the data warehouse for the e-learning environment
are demonstrated and some valuable results are presented.
Finally, the last section contains the conclusion and the future
work.
II. BACKGROUND
Nowadays, the role of business intelligence as an important
strategic tool for business management and making efficient
decisions is significant. Additionally, because of prompt
progress of business competitors, the ability to obtain
information in real-time and reduce decision-making cycle
times has become increasingly critical in recent years. Taking
advantage of business intelligence in companies can help them
to improve business performance by giving them the
opportunity to gain insight into their business and make better
decisions [5].
In short, business intelligence helps companies to gain a
comprehensive and integrated view of their business and
facilitate better and more effective decision-making and other
474
benefits such as having access to the summarized and
distributed relevant information on time. Moreover, companies
are provided with a framework capable of introducing and
measuring business key performance indicators while
analyzing its process and understanding their behavior. In order
to carry out these tasks, Business Intelligence uses a wide range
of techniques and technologies: the data warehouse as an
integrated repository of strategic information, the OLAP (On-
Line Analytical Processing) technology for the exploration of
information under different perspectives, dashboard, scorecard
and reporting tools for the analysis and visualization of
information and trends, and data mining techniques to discover
meaningful patterns and rules in large volumes of data by
automatic or semi-automatic means. Before continuing, it
should be said that some authors, such as Kimball [6] consider
that the data warehouse is the platform for business
intelligence; however other authors such as Inmon [5] consider
that the data warehouse is simply the database where the
business data are consolidated and stored. In this work, we
follow Kimball’s idea.
III. RELATED WORK
Lots of works have been performed in the literature in order
to satisfy mentioned problems in e-learning environments and
overcome their weaknesses. Student’s usage logs are the main
starting point to perform such analysis. Usage statistics can be
extracted by standard tools designed to analyze web server logs
or specific tools developed to satisfy educational needs. In this
way, some tools, like GISMO [7] were proposed which
monitors activities of students in a popular Learning
Management System (LMS) called Moodle [8]. The tool
extracts tracking data from Moodle log file and represents
results in graphical format. Moreover, it provides professors to
the information in several different ways such as reports and
graph representation.
As another tool, CourseVis [1] can be mentioned which can
track students’ data that is collected through web log files of
the LMS web server. Sinergo/ColAT [9] is another tool that
offers interpretative views of the activity developed by students
in a group learning collaborative environment and resembles
the learning process of students. Another tool presented [10] in
order to show the tutor-student interaction in a hierarchical
representation.
Finally, the most recent tool is MATEP [11] which presents
a web interface for the instructor providing a set of reports
according to his requirements. According to instructors’
opinion, this tool helps them to gain a more accurate
knowledge of what is happening in their courses since it allows
them to analyze and visualize data with different level of detail
and perspectives, discovering student behavior patterns’ and
understanding how their courses are used. That means that they
have the quantitative and qualitative information available to
take improvement actions about their courses.
IV. THE PLATFORM
The E-Learning Department of the Iran University of
Science and Technology (IUST) started its services in the
spring semester of 2004 with about 700 students and is
currently serving about 1,800 students in two Bachelors’ and
three Masters’ programs. The instructional plan in this
department is designed in a way that main the learning
materials are developed in the form of multimedia courseware
and can be accessed by students in a weekly manner. In
addition, the teacher can evaluate the process of learning by
giving the students with assignments and quizzes. Finally, the
students, having gained proper perception about the course
concepts, participate in a virtual class session and collaborate
with the teacher and other students on the problems and
learning materials. The teacher can also provide them with
complementary information about the content and get some
feedback about the level of each student’s knowledge. For
supporting the “any-time, any-where” promise of e-learning, all
the sessions are recorded and archived for the students who
cannot participate the online classes.
After three years of using a commercial e-learning system
and in response to a perceived demand for more reliable and
flexible delivery of courseware, the E-Learning Center of IUST
started to use Moodle as its LMS since March 2006. Moodle is
one of the most popular and widely used e-learning platforms
in Iran and all around the world. The transition from
commercial platforms to open source systems, such as Moodle,
is a growing trend in all around the world and the spread of
these online learning environments are under continuous
evolution. Moodle has been used as a platform for sharing
useful information, documentation, and knowledge
management in a lot of research project, yielding important
benefits to researchers [12,13]. Data mining techniques have
also been used as complementary system to Moodle, where the
results where the results are achieved through the use of
association rule mining, classification, clustering, pattern
analysis, and statistical methods [15].
Moodle keeps detailed log of all activities that students
perform [12]. It logs every click that students make for
navigational purposes and has a modest log viewing system
built into it. Log files can be filtered by course, participants,
day and activity, but they are not suitable for analytical
applications. Fortunately, Moodle does not store logs as text
files. Instead, it stores the logs in a relational database. There
are about 145 interrelated tables in Moodle database but not all
of them contain the information we need for analyzing the data
and building up the data warehouse.
V. MAIN IDEA
Although Moodle presents several reports about the
students’ activities, they are not flexible enough to satisfy the
instructors’ needs to observe their interactions with the system.
The teacher has access to a summarized report about students
such as the date of the first and last connection of students and
the number of visited pages. The information about each
learning activity is also available, but according to the
categories specified by the system, not the professor.
Consequently according to some interviews with instructional
technologists and training managers of the IUST E-learning
Center a list of data elements and analytical dimensions were
defined and corresponding information is extracted from the
Moodle database to answer their questions.
475
Figure 1: Proposed Architecture
Figure 1 illustrates the architecture of the proposed
solution. As one can observe, the data from multiple sources
such as educational system, LMS, and other legacy databases is
integrate in the form of dimension and fact tables according to
star schema and can be accessed and analyzed through the
usage of API functions.
The first step to create a BI solution is the identification of
business requirements and their associated values. As it is
mentioned before, in this step according to some interviews
with instructional experts and institutional managers, a list of
data elements and desired reports that would help to answer the
analytical questions were gathered and recorded as business
requirements document. This document consists of questions
such as: When do students connect to the system? How long do
they spend viewing learning materials? How often they use
collaborative tools? Which learning activity they prefer to
participate? Which learning resource they prefer to visit?
The second step is designing a single, integrated, easy-to-
use, high performing information model that gathers the
identified business requirements and building a dimensional
schema [5]. Dimensional schema is made up of a central fact
table and its associate dimensions. It is also called star schema
because it looks like a star with the fact table in the middle and
the dimensions serving as the points on the star. According to
Star Schema, in proposed solution two fact tables have been
designed:
• CourseActivity_Fact: This fact table contains the
details about students’ learning activities in a course in
each semester such as the number of: resource view,
forum read, forum post, message read, message write,
virtual classroom participation, recorded session
review, etc.
• VirtualClassroom_Fact: Having focus on virtual
classroom sessions, this fact table contains the
information about how students participated in the
virtual classroom sessions. It contains details about
students’ activities in the virtual classroom such as the
number of: posted text message, raise hands, change
status, broadcast audio, etc.
As can be observed, all these measurements or facts are
numeric and additive, meaning they can be summed up across
all dimensions. CourseActivity_Fact contains the records of the
all activities performed by a learner in a course.
VirtualClassroom_Fact gathers the information about every
virtual classroom session conducted in the e-learning
environment. The level of detail or grain of this fact table is a
row for each completed learner session. A learner session is
defined as the time spent by a student since he or she connects
to a certain virtual class until he leaves it. There are also four
dimension tables designed to hold the attributes that describe
fact records:
• Date_Dim: It gathers each day of year with all
characteristics such as number of day, month, day of
week, week, year, and so on.
• Learners_Dim: it collects the students’ information,
such as name, gender, major, city, field of study, etc.
• Session_Dim: it stores an identifier per each session,
containing IP address and its relevant location.
• Courses_Dim: it contains the information about each
course such as name, type, semester, etc.
These tables are the foundation of dimensional modeling,
containing descriptive information relevant to analyze the fact
table attributes from different perspectives.
Once the dimensional schema is designed, the data stage
and ETL processes must be defined and programmed. The ETL
process of proposed solution was quite simple because all the
data is stored in relational database. However some
preprocessing steps have to be performed for transforming data
into suitable shape (e.g. star schema). ETL process normally is
a manual process in which the administrator has to apply a
number of general data preprocessing tasks such as, data
cleaning, user identification, session identification, path
completion, transaction identification, data transformation and
enrichment, data integration, data reduction [6]. In our case,
data preprocessing of LMS is a little more simple due to user
authentication in which logs have entries identified by users.
Although the amount of work required in data preparation
is less, it is necessary to build some new tables according to
star schema and convert the data from relational into multi
dimensional format. For this reason, we have to transform the
transactional data from several tables of Moodle database into
fact and dimension tables and integrate the most important
information for our objectives. After these steps, data is
prepared for multi dimensional analysis in the BI environment.
The only thing remaining to do is using several analytical tools
such as dashboards, scorecards and dynamic reporting tools for
investigation and analysis of the behavior of students in e-
learning environment.
There are several products to implement a BI solution. In
the past, while the BI market was strictly dominated by closed
source and commercial tools, the last few years were
characterized by the birth of open source solutions: first as
single BI tools, and later as complete BI platforms. An Open
Source BI platform provides a full spectrum of BI capabilities
within a unified system that reduces the overhead for the
development and management of each application. Oracle,
SQL Server, Sybase or DB2, to name some, are commercial
closed source Database Management Systems which support
data warehousing and OLAP technology. On the other hand,
there are some open source platform (namely JasperSoft,
Pentaho and SpagoBI) which allows educational institutes to
begin immediately to deploy the architecture without any need
to purchase commercial tools, but it requires a fair amount of
customization.
476
VI. RESULTS
The study examined the activity logs of 1,300 students, in
nearly 100 courses over the 16 week fall semester of 2008. The
key findings showed that 80% of all students accumulated less
than one hour a week with course materials in LMS and most
looked at large portions of their content for less than one
minute. Contrarily, about 65% of students participated in
virtual classroom sessions where they can directly interact with
their teacher via instant messaging and voice chat.
Around 60% of students spent less than one hour a week in
LMS within the first 10 week, but the percentage shrinks to
20% in remaining 6 weeks. There are some active students
(about 5%) that spent more than 400 hours in LMS within the
16 week semester and some passive students (about 7%) that
spent less than 50 hours.
Figure 2 demonstrates some of the reports generated using
the proposed solution. Using such information, instructors can
adopt some decisions to improve performance of students and
avoid accumulation of students’ works for the end of semesters
which avoids students to mostly focus on their exam and
review their courses. For example, suitable and uniform
distribution of assignments during the semester and forcing
students to upload their assignments exactly at the prescribed
time can be considered as an appropriate decision for achieving
the mentioned goal.
VII. CONCLUSION AND FUTURE WORKS
In this paper, in order to overcome shortcomings of
traditional e-learning environments in making appropriate
decisions, the application of Business Intelligence and Data
Warehousing tools in the field of e-learning was presented.
Moreover, the architecture proposed for these systems, tries to
solve the lack of analytical and subjective reporting tools in a
widely spread, open-source, learning management system and,
moreover, provide instructors with detailed reports about
progression of students and give them the ability to track and
assess the student performance and evaluate the design of their
virtual courses in order to take suitable managerial decisions.
The study shows that students spend little time with course
materials online and about 80% spend less than two hours a
week viewing multimedia courseware and they just have
participated in virtual classroom sessions. It is also inferred that
the students prefer to use collaborative activities, such as
instant messaging and forums, instead of just viewing the
learning material.
However, expanding the period and the number of students
involved in analytical study and applying data mining methods
and association rule mining techniques to extract other
unresolved meaningful patterns and rules among the student’s
data usage can be considered as future work. Moreover,
applying clustering and classification algorithms in order to
predict the performance of new students according to their
behavior is other future work.
Figure 2: Sample Reports Generated by the Proposed Solution
477
ACKNOWLEDGEMENT
The authors gratefully acknowledge the management and
educational technologists in the E-learning Center of the Iran
University of Science and Technology for providing the access
to the database of e-learning environment and giving valuable
suggestions and comments.
REFRENCES
[1] R. Mazza and V. Dimitrova, "CourseVis: A graphical student monitoring
tool for supporting instructors in web-based distance courses,"
International Journal of Human-Computer Studies, vol. 65, no. 2, pp.
125–139, 2007.
[2] Cristobal Romero and Sepastian Ventura, Data mining in E-Learning.
Advances in Management Information.: WIT Press, 2006, vol. 4.
[3] M. E. Zorrilla, D. Marin, and E. Alvarez, "Towards virtual course
evaluation using Web Intelligence," in EUROCAST 2007 LNCS.:
Springer, Heidelberg, 2007, pp. 392–399.
[4] O. Zaïane and J. Luo, "Web usage mining for a better web-based learning
environment," in Proceedings of conference on advanced technology for
education, Banff, Alberta, 2001, pp. 60-64.
[5] W. H. Inmon, Building the Data Warehouse. Chichester: Willey & Son,
2002.
[6] R. Kimball and J. Caserta, The Data Warehouse ETL Toolkit. Chichester:
John Wiley & Sons, 2002.
[7] (2010) GISMO. [Online].http://gismo.sourceforge.net/
[8] (2010) Moodle. [Online].http://www.moodle.org
[9] N. Avouris, V. Komis, G. Fiotakis, M. Margaritis, and G. Voyiatzaki,
"Logging of fingertip actions is not enough for analysis of learning
activities," in Proccedings of Workshop Usage Analysis in learning
systems, Amsterdam, 2005.
[10] J. Mostow et al., "An Educational Data Mining Tool to Browse Tutor-
Student Interactions: Time Will Tell!," in Workshop on educational data
mining, 2005, pp. 15–22.
[11] M. E. Zorrilla and E. Alvarez, "MATEP: Monitoring and Analysis Tool
for E-learning Platforms," in 8th IEEE International Conference on
Advanced Learning Technologies, Santander, Spain, 2008, pp. 611-613.
[12] J. Cole, Using Moodle.: O'Reilly, 2005.
[13] A. Uribe-Tirado, L. M. Melgar-Estrada, and J. A. Bornacelly-Castro,
"Moodle learning management system as a tool for information,
documentation, and knowledge management by research groups,"
Profesional de la Información, vol. 16, no. 5, pp. 468–474, 2007.
[14] Cristobal Romero, S. Ventura, and E. Garcia, "Data mining in course
management systems: Moodle case study and tutorial," Computers &
Education , vol. 51, no. 1, pp. 368–384, 2008.
[15] W. H. Rice, Moodle e-learning course development. A complete guide to
successful learning using Moodle.: Packt Publishing, 2006.
[16] J. Mundy, W. Thorthwaite, and R. Kimball, The Microsoft Data
Warehouse Toolkit: with SQL Server 2005 and the Microsoft Business
Intelligence Toolset. Chichester: Wiley Publishing Inc., 2006.
[17] C. J. White, "Corporate Performance Optimization Guide.Intelligent
Business Strategies," Oracle Corporation, January 2003.
[18] E. Thomsen, OLAP Solutions: Building Multidimensional Information
Systems, 2nd ed. Chichester: John Wiley & Sons, 2002.
[19] Cristobal Romero and S. Ventura, "Educational Data Mining: A Survey
from 1995 to 2005," Expert Systems with Applications, vol. 33, pp. 125-
146, 2007.
doc_850057685.pdf
Business Intelligence in E-Learning
473
Business Intelligence in E-Learning
(Case Study of Iran University of Science and Technology)
Mohammad Hassan Falakmasir
1
, Jafar Habibi
2
, Shahrouz Moaven
1
, Hassan Abolhassani
2
Department of Computer Engineering
Sharif University of Technology
Tehran, Iran
1
{Falakmasir, Moaven}@ce.sharif.edu
1
{Jhabibi, Abolhassani}@sharif.edu
Abstract— Nowadays, e-learning platforms are widely used by
universities and other research-based and educational
institutions. Despite lots of advantages these educational
environments provide for organizations, yet there are many
unresolved problems which cause instructors and training
managers with some difficulties to get proper information about
the students’ learning behavior. On one hand, lack of tools to
measure, assess, and evaluate the performance of learners in
educational activities has led the educators to fail to guarantee
the success of learning process. On the other hand, strict
structure of learning materials prevents students to acquire
knowledge based on their learning style. Consequently,
developing tools monitor and analyze the learner’s interaction
with e-learning environment is necessary. Business intelligence
(BI) and On Line Analytical Processing (OLAP) technologies can
be used in order to monitor and analyze the learner’s behavior
and performance in e-learning environments. They can also be
used to evaluate the structure of the course content and its
effectiveness in the learning process. This article investigates the
use of business intelligence and OLAP tools in e-learning
environments and presents a case study of how to apply these
technologies in the database of an e-learning system. The study
shows that students spend little time with course courseware and
prefer to use collaborative activities, such as virtual classroom
and forums instead of just viewing the learning material.
Keywords- E-Learning; Educational Data Mining; Business
Intelligence; Data Warehouse; OLAP; Intelligent Data Analysis
I. INTRODUCTION
Today, most universities and educational institutions use e-
learning platforms to offer their educational material benefiting
from their advantages such as work “any-time, anywhere”, use
of collaborative tools, support different styles of learning, etc.
However, these e-learning platforms do not cover all teaching
aspects since they do not usually provide teachers and
instructional designers with tools which allow them to monitor
and assess all the activities performed by learners [1,2,3,4].
These environments provide the professor with access
summary information such as the date of the first and last
connection, the number of visited pages according to the
category specified by the platform (not by the professor), or the
number of messages read/sent by each learner; the total number
of page visits, the average time spent on each page but,
globally, per page not by learner. As can be deduced, this
information is not enough to analyze the behavior of learners
and their evolution. Even more, when the number of students
and the diversity of interests are high, the professor has serious
difficulties to extract useful information.
Business Intelligence techniques applied to the database of
Web-based Educational Systems could help instructors and
other educational experts to generate statistics, analytical
models, and uncover meaningful patterns from these huge
volumes of data. In this article, a framework for applying
business intelligence in e-learning environments has been
proposed, which increased both flexibility and performance of
e-learning environments. Hence, on one hand, the proposed
environment enables educational technologists to identify,
analyze and monitor relevant aspects of instruction, such as
different style, paths, and strategies of learning. On the other
hand, such parameters may be used to adapt the learning
process to each individual learner and improve the performance
of the learning process.
The reminder of this paper is organized as follows. Section
2 briefly describes business intelligence principles. A brief
history of the related work is represented in the third section. A
description of the e-learning system and the proposed
architecture are represented in section 4, moreover, three steps
to deploy the data warehouse for the e-learning environment
are demonstrated and some valuable results are presented.
Finally, the last section contains the conclusion and the future
work.
II. BACKGROUND
Nowadays, the role of business intelligence as an important
strategic tool for business management and making efficient
decisions is significant. Additionally, because of prompt
progress of business competitors, the ability to obtain
information in real-time and reduce decision-making cycle
times has become increasingly critical in recent years. Taking
advantage of business intelligence in companies can help them
to improve business performance by giving them the
opportunity to gain insight into their business and make better
decisions [5].
In short, business intelligence helps companies to gain a
comprehensive and integrated view of their business and
facilitate better and more effective decision-making and other
474
benefits such as having access to the summarized and
distributed relevant information on time. Moreover, companies
are provided with a framework capable of introducing and
measuring business key performance indicators while
analyzing its process and understanding their behavior. In order
to carry out these tasks, Business Intelligence uses a wide range
of techniques and technologies: the data warehouse as an
integrated repository of strategic information, the OLAP (On-
Line Analytical Processing) technology for the exploration of
information under different perspectives, dashboard, scorecard
and reporting tools for the analysis and visualization of
information and trends, and data mining techniques to discover
meaningful patterns and rules in large volumes of data by
automatic or semi-automatic means. Before continuing, it
should be said that some authors, such as Kimball [6] consider
that the data warehouse is the platform for business
intelligence; however other authors such as Inmon [5] consider
that the data warehouse is simply the database where the
business data are consolidated and stored. In this work, we
follow Kimball’s idea.
III. RELATED WORK
Lots of works have been performed in the literature in order
to satisfy mentioned problems in e-learning environments and
overcome their weaknesses. Student’s usage logs are the main
starting point to perform such analysis. Usage statistics can be
extracted by standard tools designed to analyze web server logs
or specific tools developed to satisfy educational needs. In this
way, some tools, like GISMO [7] were proposed which
monitors activities of students in a popular Learning
Management System (LMS) called Moodle [8]. The tool
extracts tracking data from Moodle log file and represents
results in graphical format. Moreover, it provides professors to
the information in several different ways such as reports and
graph representation.
As another tool, CourseVis [1] can be mentioned which can
track students’ data that is collected through web log files of
the LMS web server. Sinergo/ColAT [9] is another tool that
offers interpretative views of the activity developed by students
in a group learning collaborative environment and resembles
the learning process of students. Another tool presented [10] in
order to show the tutor-student interaction in a hierarchical
representation.
Finally, the most recent tool is MATEP [11] which presents
a web interface for the instructor providing a set of reports
according to his requirements. According to instructors’
opinion, this tool helps them to gain a more accurate
knowledge of what is happening in their courses since it allows
them to analyze and visualize data with different level of detail
and perspectives, discovering student behavior patterns’ and
understanding how their courses are used. That means that they
have the quantitative and qualitative information available to
take improvement actions about their courses.
IV. THE PLATFORM
The E-Learning Department of the Iran University of
Science and Technology (IUST) started its services in the
spring semester of 2004 with about 700 students and is
currently serving about 1,800 students in two Bachelors’ and
three Masters’ programs. The instructional plan in this
department is designed in a way that main the learning
materials are developed in the form of multimedia courseware
and can be accessed by students in a weekly manner. In
addition, the teacher can evaluate the process of learning by
giving the students with assignments and quizzes. Finally, the
students, having gained proper perception about the course
concepts, participate in a virtual class session and collaborate
with the teacher and other students on the problems and
learning materials. The teacher can also provide them with
complementary information about the content and get some
feedback about the level of each student’s knowledge. For
supporting the “any-time, any-where” promise of e-learning, all
the sessions are recorded and archived for the students who
cannot participate the online classes.
After three years of using a commercial e-learning system
and in response to a perceived demand for more reliable and
flexible delivery of courseware, the E-Learning Center of IUST
started to use Moodle as its LMS since March 2006. Moodle is
one of the most popular and widely used e-learning platforms
in Iran and all around the world. The transition from
commercial platforms to open source systems, such as Moodle,
is a growing trend in all around the world and the spread of
these online learning environments are under continuous
evolution. Moodle has been used as a platform for sharing
useful information, documentation, and knowledge
management in a lot of research project, yielding important
benefits to researchers [12,13]. Data mining techniques have
also been used as complementary system to Moodle, where the
results where the results are achieved through the use of
association rule mining, classification, clustering, pattern
analysis, and statistical methods [15].
Moodle keeps detailed log of all activities that students
perform [12]. It logs every click that students make for
navigational purposes and has a modest log viewing system
built into it. Log files can be filtered by course, participants,
day and activity, but they are not suitable for analytical
applications. Fortunately, Moodle does not store logs as text
files. Instead, it stores the logs in a relational database. There
are about 145 interrelated tables in Moodle database but not all
of them contain the information we need for analyzing the data
and building up the data warehouse.
V. MAIN IDEA
Although Moodle presents several reports about the
students’ activities, they are not flexible enough to satisfy the
instructors’ needs to observe their interactions with the system.
The teacher has access to a summarized report about students
such as the date of the first and last connection of students and
the number of visited pages. The information about each
learning activity is also available, but according to the
categories specified by the system, not the professor.
Consequently according to some interviews with instructional
technologists and training managers of the IUST E-learning
Center a list of data elements and analytical dimensions were
defined and corresponding information is extracted from the
Moodle database to answer their questions.
475
Figure 1: Proposed Architecture
Figure 1 illustrates the architecture of the proposed
solution. As one can observe, the data from multiple sources
such as educational system, LMS, and other legacy databases is
integrate in the form of dimension and fact tables according to
star schema and can be accessed and analyzed through the
usage of API functions.
The first step to create a BI solution is the identification of
business requirements and their associated values. As it is
mentioned before, in this step according to some interviews
with instructional experts and institutional managers, a list of
data elements and desired reports that would help to answer the
analytical questions were gathered and recorded as business
requirements document. This document consists of questions
such as: When do students connect to the system? How long do
they spend viewing learning materials? How often they use
collaborative tools? Which learning activity they prefer to
participate? Which learning resource they prefer to visit?
The second step is designing a single, integrated, easy-to-
use, high performing information model that gathers the
identified business requirements and building a dimensional
schema [5]. Dimensional schema is made up of a central fact
table and its associate dimensions. It is also called star schema
because it looks like a star with the fact table in the middle and
the dimensions serving as the points on the star. According to
Star Schema, in proposed solution two fact tables have been
designed:
• CourseActivity_Fact: This fact table contains the
details about students’ learning activities in a course in
each semester such as the number of: resource view,
forum read, forum post, message read, message write,
virtual classroom participation, recorded session
review, etc.
• VirtualClassroom_Fact: Having focus on virtual
classroom sessions, this fact table contains the
information about how students participated in the
virtual classroom sessions. It contains details about
students’ activities in the virtual classroom such as the
number of: posted text message, raise hands, change
status, broadcast audio, etc.
As can be observed, all these measurements or facts are
numeric and additive, meaning they can be summed up across
all dimensions. CourseActivity_Fact contains the records of the
all activities performed by a learner in a course.
VirtualClassroom_Fact gathers the information about every
virtual classroom session conducted in the e-learning
environment. The level of detail or grain of this fact table is a
row for each completed learner session. A learner session is
defined as the time spent by a student since he or she connects
to a certain virtual class until he leaves it. There are also four
dimension tables designed to hold the attributes that describe
fact records:
• Date_Dim: It gathers each day of year with all
characteristics such as number of day, month, day of
week, week, year, and so on.
• Learners_Dim: it collects the students’ information,
such as name, gender, major, city, field of study, etc.
• Session_Dim: it stores an identifier per each session,
containing IP address and its relevant location.
• Courses_Dim: it contains the information about each
course such as name, type, semester, etc.
These tables are the foundation of dimensional modeling,
containing descriptive information relevant to analyze the fact
table attributes from different perspectives.
Once the dimensional schema is designed, the data stage
and ETL processes must be defined and programmed. The ETL
process of proposed solution was quite simple because all the
data is stored in relational database. However some
preprocessing steps have to be performed for transforming data
into suitable shape (e.g. star schema). ETL process normally is
a manual process in which the administrator has to apply a
number of general data preprocessing tasks such as, data
cleaning, user identification, session identification, path
completion, transaction identification, data transformation and
enrichment, data integration, data reduction [6]. In our case,
data preprocessing of LMS is a little more simple due to user
authentication in which logs have entries identified by users.
Although the amount of work required in data preparation
is less, it is necessary to build some new tables according to
star schema and convert the data from relational into multi
dimensional format. For this reason, we have to transform the
transactional data from several tables of Moodle database into
fact and dimension tables and integrate the most important
information for our objectives. After these steps, data is
prepared for multi dimensional analysis in the BI environment.
The only thing remaining to do is using several analytical tools
such as dashboards, scorecards and dynamic reporting tools for
investigation and analysis of the behavior of students in e-
learning environment.
There are several products to implement a BI solution. In
the past, while the BI market was strictly dominated by closed
source and commercial tools, the last few years were
characterized by the birth of open source solutions: first as
single BI tools, and later as complete BI platforms. An Open
Source BI platform provides a full spectrum of BI capabilities
within a unified system that reduces the overhead for the
development and management of each application. Oracle,
SQL Server, Sybase or DB2, to name some, are commercial
closed source Database Management Systems which support
data warehousing and OLAP technology. On the other hand,
there are some open source platform (namely JasperSoft,
Pentaho and SpagoBI) which allows educational institutes to
begin immediately to deploy the architecture without any need
to purchase commercial tools, but it requires a fair amount of
customization.
476
VI. RESULTS
The study examined the activity logs of 1,300 students, in
nearly 100 courses over the 16 week fall semester of 2008. The
key findings showed that 80% of all students accumulated less
than one hour a week with course materials in LMS and most
looked at large portions of their content for less than one
minute. Contrarily, about 65% of students participated in
virtual classroom sessions where they can directly interact with
their teacher via instant messaging and voice chat.
Around 60% of students spent less than one hour a week in
LMS within the first 10 week, but the percentage shrinks to
20% in remaining 6 weeks. There are some active students
(about 5%) that spent more than 400 hours in LMS within the
16 week semester and some passive students (about 7%) that
spent less than 50 hours.
Figure 2 demonstrates some of the reports generated using
the proposed solution. Using such information, instructors can
adopt some decisions to improve performance of students and
avoid accumulation of students’ works for the end of semesters
which avoids students to mostly focus on their exam and
review their courses. For example, suitable and uniform
distribution of assignments during the semester and forcing
students to upload their assignments exactly at the prescribed
time can be considered as an appropriate decision for achieving
the mentioned goal.
VII. CONCLUSION AND FUTURE WORKS
In this paper, in order to overcome shortcomings of
traditional e-learning environments in making appropriate
decisions, the application of Business Intelligence and Data
Warehousing tools in the field of e-learning was presented.
Moreover, the architecture proposed for these systems, tries to
solve the lack of analytical and subjective reporting tools in a
widely spread, open-source, learning management system and,
moreover, provide instructors with detailed reports about
progression of students and give them the ability to track and
assess the student performance and evaluate the design of their
virtual courses in order to take suitable managerial decisions.
The study shows that students spend little time with course
materials online and about 80% spend less than two hours a
week viewing multimedia courseware and they just have
participated in virtual classroom sessions. It is also inferred that
the students prefer to use collaborative activities, such as
instant messaging and forums, instead of just viewing the
learning material.
However, expanding the period and the number of students
involved in analytical study and applying data mining methods
and association rule mining techniques to extract other
unresolved meaningful patterns and rules among the student’s
data usage can be considered as future work. Moreover,
applying clustering and classification algorithms in order to
predict the performance of new students according to their
behavior is other future work.
Figure 2: Sample Reports Generated by the Proposed Solution
477
ACKNOWLEDGEMENT
The authors gratefully acknowledge the management and
educational technologists in the E-learning Center of the Iran
University of Science and Technology for providing the access
to the database of e-learning environment and giving valuable
suggestions and comments.
REFRENCES
[1] R. Mazza and V. Dimitrova, "CourseVis: A graphical student monitoring
tool for supporting instructors in web-based distance courses,"
International Journal of Human-Computer Studies, vol. 65, no. 2, pp.
125–139, 2007.
[2] Cristobal Romero and Sepastian Ventura, Data mining in E-Learning.
Advances in Management Information.: WIT Press, 2006, vol. 4.
[3] M. E. Zorrilla, D. Marin, and E. Alvarez, "Towards virtual course
evaluation using Web Intelligence," in EUROCAST 2007 LNCS.:
Springer, Heidelberg, 2007, pp. 392–399.
[4] O. Zaïane and J. Luo, "Web usage mining for a better web-based learning
environment," in Proceedings of conference on advanced technology for
education, Banff, Alberta, 2001, pp. 60-64.
[5] W. H. Inmon, Building the Data Warehouse. Chichester: Willey & Son,
2002.
[6] R. Kimball and J. Caserta, The Data Warehouse ETL Toolkit. Chichester:
John Wiley & Sons, 2002.
[7] (2010) GISMO. [Online].http://gismo.sourceforge.net/
[8] (2010) Moodle. [Online].http://www.moodle.org
[9] N. Avouris, V. Komis, G. Fiotakis, M. Margaritis, and G. Voyiatzaki,
"Logging of fingertip actions is not enough for analysis of learning
activities," in Proccedings of Workshop Usage Analysis in learning
systems, Amsterdam, 2005.
[10] J. Mostow et al., "An Educational Data Mining Tool to Browse Tutor-
Student Interactions: Time Will Tell!," in Workshop on educational data
mining, 2005, pp. 15–22.
[11] M. E. Zorrilla and E. Alvarez, "MATEP: Monitoring and Analysis Tool
for E-learning Platforms," in 8th IEEE International Conference on
Advanced Learning Technologies, Santander, Spain, 2008, pp. 611-613.
[12] J. Cole, Using Moodle.: O'Reilly, 2005.
[13] A. Uribe-Tirado, L. M. Melgar-Estrada, and J. A. Bornacelly-Castro,
"Moodle learning management system as a tool for information,
documentation, and knowledge management by research groups,"
Profesional de la Información, vol. 16, no. 5, pp. 468–474, 2007.
[14] Cristobal Romero, S. Ventura, and E. Garcia, "Data mining in course
management systems: Moodle case study and tutorial," Computers &
Education , vol. 51, no. 1, pp. 368–384, 2008.
[15] W. H. Rice, Moodle e-learning course development. A complete guide to
successful learning using Moodle.: Packt Publishing, 2006.
[16] J. Mundy, W. Thorthwaite, and R. Kimball, The Microsoft Data
Warehouse Toolkit: with SQL Server 2005 and the Microsoft Business
Intelligence Toolset. Chichester: Wiley Publishing Inc., 2006.
[17] C. J. White, "Corporate Performance Optimization Guide.Intelligent
Business Strategies," Oracle Corporation, January 2003.
[18] E. Thomsen, OLAP Solutions: Building Multidimensional Information
Systems, 2nd ed. Chichester: John Wiley & Sons, 2002.
[19] Cristobal Romero and S. Ventura, "Educational Data Mining: A Survey
from 1995 to 2005," Expert Systems with Applications, vol. 33, pp. 125-
146, 2007.
doc_850057685.pdf