Solving Course Selection Problem by a Combination of Correlation Analysis and Analytic Hierarchy Process

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 7, No. 6, December 2017, pp. 3536~3551
ISSN: 2088-8708, DOI: 10.11591/ijece.v7i6.pp3536-3551  3536
Journal homepage: http://iaesjournal.com/online/index.php/IJECE
Solving Course Selection Problem by a Combination of
Correlation Analysis and Analytic Hierarchy Process
Mohammed Al-Sarem
Department of Information Science, Taibah University, Medina, Saudi Arabia
Article Info ABSTRACT
Article history:
Received Apr 4, 2017
Revised Jun 22, 2017
Accepted Jul 10, 2017
In the universities where students have a chance to select and enroll in a
particular course, they require special support to avoid the wrong
combination of courses that might lead to delay their study. Analysis shows
that the students' selection is mainly influenced by list of factors which we
categorized them into three groups of concern: course factors, social factors,
and individual factors. This paper proposed a two-phased model where the
most correlated courses are generated and prioritized based on the student
preferences. At this end, we have applied the multi-criteria analytic hierarchy
process (MC-AHP) in order to generate the optimum set of courses from the
available courses pool. To validate the model, we applied it to the data from
students of the Information System Department at Taibah University,
Kingdom of Saudi Arabia.
Keyword:
Course selection
Student preferences
Correlation analysis,
AHP method
Copyright © 2017Institute of Advanced Engineering and Science.
All rights reserved.
Corresponding Author:
Mohammed Al-Sarem,
Department of Information System, Taibah University,
PO box 344, Medina, KSA.
Email: mohsarem@gmail.com
1. INTRODUCTION
Course enrollment (CE) is one of the main administrative task that students faces each semester.
Often, the CE process starts few week before the start of the term itself and ends a week after the start of the
courses. During this period, the need to support students during selection and registration courses is increase.
At a time not so long ago, students were responsible for their own choices and the faculty advisor had
primarily become assisting students with the transition from high school to college [1]. Nowadays, situation
is extended to include guiding students to select courses, to register in each semester, and to fulfill the degree
requirement. Generally, the students aim to finish their study as soon as they can taking as many courses as
possible even if this affects negatively on their performance. From this end, colleges and universities began
to implement so-called academic advising affairs [2]. The academic advisory process is known as “process in
which advisor and advisee enter a dynamic relationship respectful of the student's concerns” [3]. Faculty
academic advising has a significant impact on a student’s academic success.
The academic advisor is responsible for: i) helping students in adaptation with specialization; ii)
following-up to the level of students each semester; iii) encouraging and drawing a good study plan that
ensures the improvement students' educational level; vi) determining which courses that may delay student’s
graduation at the specified time; finally, v) helping students to correctly register their plan of study according
to the rules of deanship of admission and registration [2]. Current work discusses the influential factors that
drive students' selection. It suggests to combine the correlation analysis with the multi-criteria hierarchy
analytic method. The proposed model aims to present a framework for the future e-academic advisory
system. The work is organized as the follow: Section 2 presents the formulation of the course selection
problem. Section 3 presents related works and the methods have been applied to solve such problem. Section
4 discusses influential factors that might drive students decision making process. Section 5 presents the work
methodology, the used methods, application domain, and the data description. Section 6 describes the

IJECE ISSN: 2088-8708 
Solving Course Selection Problem by a Combination of Correlation Analysis … (Mohammed Al-Sarem)
3537
experiential part of this work. It presents an illustrative example showing how the model should to work; and
finally, Section 7 outlines the outlines the conclusions of this research and the future work.
2. PROBLEM DESCRIPTION and FORMULATION
Let ∁ be the set of all courses to be taught during the study plan ℒ (academic curriculum plan) for
obtaining a university degree. Each course c ∈ C gives a number of credits rC ∈ ℤ+
and are might be in
prerequisite relation (might find courses without any prerequisite such university required courses). The
study plan ℒ is divided into academic years, and each academic year is divided into semesters. Each
semesters, students are faced with selecting list of courses c ∈ Ct where Ct is list of available courses at an
academic semesters in case they satisfy the courses perquisites, Ct ∈ ∁ . The prerequisites are formalized as a
directed acyclic graph D = (Vc, A), where Vc represents a course, and each arc (i, j) ∈ A represents a
precedence relation between the course i and j in case the j − th course cannot master without taught the
i − th course. Let also ℒ(Ci) = (h, ct, N) represent impact of a ith course on the study plan ℒ, where h is
hierarchical level of Ci, ct is opened course in the next semesters t + i and i = 1,2,3, . . ., n, and N is the total
opened courses in the study plan ℒ.
Let also the ith course ci is taught by different instructors T at different time. Each semesters has an
allowed academic load λ. The academic load is determined based on the student performance (the average
grade point GPA) at the semester t −1. Let λ obeys the following regulations:
- if student's GPA at the semester 𝑡 −1 is less than predefined threshold 𝜃, only the minimum course load
𝜆0 (a value of academic credits per semester required to consider a student as full time) is allowed to
register at the semester otherwise up to the maximum course load 𝜆1.
- if student is expected to graduate and still at least a quite little hours to accomplish his/her study, the
course load (extend course load 𝜆 𝑒) is extend and students are allowed to register more hours at the
semester.
However, in real educational realm, in order to avoid the second above scenario, the academic
workload per semester need to balance keeping the prerequisite conditions. In addition to that, if courses
ci , cj and ci , cj ∈ C are in prerequisite relation, then it is better if a course ci is followed as close as possible
by a course cj [4]. Based on the aforementioned formulation, the course selection problem CSP ,now, is
formulated as follows: Finding these courses per semester that are, on one hand, meet student's preferences.
On the other hand, maximize his/her graduation final grades.
Practice shows that personalizing students' study plan according to their preferences leads to
enhance their learning performance. However, with a lot of opportunities to compose the university curricula,
restrictions, prerequisites and sometimes the university's rule, students may not be able to select course set
that meet their needs and preferences. Furthermore, if they do not know in advance, which performance skills
are challenged in the particular course, they may select/enroll in courses that are not adequate, at least, at a
particular term. We defend on the idea that, providing students with suitable courses set leads to maximize
their final GPA.
The course selection is also affected by other factors: instructor's reputation who give the course[5],
the course difficulty [6], GPA value for the course [7], course time scheduling [8], market demand [9], peers'
advices, and existing friends in a particular group/section (see Section 4).
3. RELATED WORKS
During the registration period, at an academic institution, commonly students should determine
which courses will be taking or dropping within available registration system. This process provides the
teaching staff and administration with clear vision about students' preferences, required class lists, and their
number in each class. However, the situation, in reality, is on the opposite. The timetable committee
constructs the whole time tables and then asks students to choose from the available course lists. Students, in
this case, need to consult their academic advisors before access the system. In case of unavailability of the
advisor or laziness to seek advice, these may cause to delay the registration process or the students make
decisions depending on their own experience and the available information [6].
Indeed, the described above problem can be tackled several ways. Just as examples, we can
mentioned the following approach: constraint programming (CP) [10], integer linear programming (ILP)
problem [11], [12], hybrid techniques based on genetic algorithms and constraint programming [13], [14],
integer programming and hybrid local search method [4], generalized quadratic assignment problem [15],
and ant colony optimization meta-heuristic model [16].
In this work, we present the CSP as multi-criteria based decision problem (MCDP). Gunadhi et al.,
[16] proposed a decision model for course advising system on student’s need to know “what to do” and “how

 ISSN:2088-8708
IJECE Vol. 7, No. 6, December 2017 : 3536–3551
3538
to do it”. At the core of the system lies the curriculum generator which customizes the study plan to each
individual's needs and produces a schedule for the courses chosen. Customizing the study plan is depend on
the course selection criterion. Some systems allow students to request only courses for which they have
appropriate prerequisites and co-requisites [17]. In the others, the courses are suggested based on balancing
the course load, frequency of the course offering, shortening the path length to graduation, students'
preferences and their progress in the program [18], [19].
Current academic systems provide information about available courses and professors who will
teach them, sections, number of students in each section, and schedule. However, information about students'
previous progress from current/past enrollment is usually ignored even though such information are priceless
treasure in finding interdependent courses. In this direction, the educational data mining methods have been
successfully applied. Association rules e.g., are used as a way to seek dependency among courses of a
curriculum plan [20], [21], [22]. The course characteristics similarities of former students' study were used in
optimizing curricula of current students [7], [23].
4. INFLUENTIAL FACTORS on STUDENTS SELECTION
In the universities where students have a chance to select and enroll in a particular course, selecting
the optimum set of courses from the available courses pool is a high risk decision-making situation because
the cumulative impact will effect negatively/ positively on the students' performance progress, their expected
graduate date and the final GPA as well as their career direction and future employment opportunities. As
mentioned before, course selection process is influenced several factors. Analysis the research literature and
the conducted questionnaire, we summarize these factors into three main groups of concerns: (i) course
factors, (ii) social factors, and (iii) individual factors. Indeed, these groups is decomposed into sub-groups
which influence on the whole decision-making process. Since different courses are selected with different
preferences and objectives, the decision process must take all these factors concurrently (see Figure 1 below).
Figure 1. Infuential factors on students' course selection

IJECE ISSN: 2088-8708 
3539
Next, we discuss the impact of these factors on students' decision-making process and show how
will they engage in the proposed approach. Table 1 gives a brief description of decision attributes that are
used for driving the selection process.
Table 1. Description of criteria and decision attributes used for selecting a course
Criteria Decision attributes Refers to:
Course factors
Course characteristics
Course credit hours, Distance between a course and
its prerequisites, Student competence for a given
course
Instructor characteristics
Personal instructor characteristics, Instructor
assessment approach, Instructor lecturing style
Teaching language Course teaching language
Social factors
Peer opinions Peers' feedback
Closed Friend Existing in the class a closed friend
Campus Location Location of the class room, Campus location
Individual factors
Course time scheduling Time when student attend the class
Student demands
Student's interest in a course, job opportunities, Local
labor.
Learning style
A way or an approach a student follows in the course
of learning.
4.1. Course Characteristics
- Course Characteristics
The questionnaire results show that students' choices regarding course characteristics are depend
mainly on the difficulty of the course, course weight ( course credit hour), distance between a course 𝑐𝑖 and
its prerequisites, and student competence for a given course.
Difficulty- refers to complexity level of a course taking in consideration the grades of every student
who passed that course successfully to the grades of all students who follow the same curriculum plan.
Logically, a course with high 𝑑𝑖𝑓𝑓(𝐶𝑖) is considered as difficult course, otherwise it is easy.
𝑑𝑖𝑓𝑓(𝑐𝑖) = 1 − (
∑ ∑ 𝑔𝑖,𝑘
′𝑘
𝑖=1
𝑘∈𝐶 𝑖
∑ ∑ 𝑔 𝑖,𝑘
𝑘
𝑖=1
𝑘∈𝐶 𝑖
∗
𝑚
𝑛
) (1)
where, 𝐶𝑖- is the 𝑖th course in the curriculum plan, 𝑔𝑖,𝑘
′
- is GPA of a student who passed a course 𝐶𝑖
successfully from the first attempt, 𝑔𝑖,𝑘- is GPA of the student who take the course 𝐶𝑖, 𝑚- number of student
who passed the course 𝐶𝑖 from the first attempt, and 𝑛- is number of students who follow the same
curriculum plan and take the course 𝐶𝑖.
Distance between two courses 𝐶𝑖 and 𝐶𝑗 taught by a student s is defined as the Euclidean distance of
the hierarchical level ℎ at where the courses 𝐶𝑖 and 𝐶𝑗are being taught.
𝑑𝑖𝑠 (𝐶𝑖, 𝐶𝑗) = {
√(𝐶𝑗
ℎ
− 𝐶𝑖
ℎ
)
2
+ (𝐶𝑆 𝑗
ℎ
− 𝐶𝑆 𝑖
ℎ
)
22
, 𝐶𝑖
𝑝𝑟𝑒𝑟𝑒𝑞𝑢𝑖𝑠𝑖𝑡𝑒
→ 𝐶𝑗
1 , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒
(2)
where, 𝐶𝑖
ℎ
and 𝐶𝑗
ℎ
- is the hierarchical level ℎ at where the courses 𝐶𝑖 and 𝐶𝑗
ℎ
are being taught
respectively, 𝐶𝑆 𝑖
ℎ
and 𝐶𝑆 𝑗
ℎ
- is the academic semester where courses 𝐶𝑖 and 𝐶𝑗
ℎ
are being taught.
Competence represents student's ability to study a course based on the grades he has obtained in the
prerequisites.
𝐶𝑜𝑚𝑝𝑒𝑡𝑒𝑛𝑐𝑒(𝑐𝑖
𝑠
) = {
1 , 𝐶𝑖 ℎ𝑎𝑠 𝑛𝑜𝑡 𝑝𝑟𝑒𝑟𝑒𝑞𝑢𝑖𝑠𝑖𝑡𝑒
∑ 𝑛 𝐶 𝑗
𝑆
∗ 𝑑𝑖𝑓𝑓(𝑐𝑖) ∗ 𝑑𝑖𝑠(𝐶𝑖, 𝐶𝑗)𝑘
𝑗=1,𝑗∈ℒ
𝑊𝑖 ∗ 𝑔𝑖
𝑠⁄ , 𝑜𝑡ℎ𝑒𝑟𝑤𝑖𝑠𝑒 (3)
where, 𝑔𝑖
𝑠
- is the current GPA grades of student 𝑠, 𝑑𝑖𝑓𝑓(𝑐𝑖) - is difficulty of the course 𝑐𝑖,
𝑑𝑖𝑠(𝐶𝑖, 𝐶𝑗)- is distance between a course 𝐶𝑗(prerequisite course) and course 𝐶𝑖, 𝑊𝑖- is credit hours of course
𝐶𝑖and, 𝑛𝑖
𝑆
- is number of attempts student 𝑆 was enrolled in course 𝐶𝑗.

 ISSN:2088-8708
3540
- Instructor Characteristics
Although, the course characteristics have a significant impact on students' enrollment decision,
practice shows that the instructor characteristics also play important role on the future decision to enroll in
those courses taught by this instructor [5], [24] and on how useful the course can be [25]. Nowadays,
majority of universities provide online system to collect students' feedback for all offered courses at the end
of academic semester. Often, feedback takes a form of questionnaire or survey which contain a series of
items that are ranked on a five-points Likert-scale. The questionnaire/survey items address the question about
personal instructor characteristics, course value presented by the instructor, instructor assessment approach,
and instructor lecturing style. Researchers, such as, [24], [26] noted that students prefer to take courses with
teachers who are enthusiastic, well spoken, knowledgeable, caring, and helpful. Beggs et al., [27] found
that the quality of a course presented by the instructor has a large affect on whether a student chooses to
enroll in a class. Although questionnaire results show beside the quality of the course, both the instructor
assessment approach [28], and instructor lecturing style [24], [26] are critical factors in course enrollment.
- Teaching Language
Several researchers considered language as a significant factor not only in learning process but also
in their motivation to learn [29], [30]. According to Coleman [31] the use of a common language allows, on
one hand, efficient exchange of ideas, on the other hand, facilitates communication skills.
Nowadays, major of universities present course contents in English even if it is not the official/
native language. The reason behind this choice is that English has a positive impact on modernization, and on
the quality of learners' experience [31]. However, students prefer to deal with instructors who share the same
native language or with course content that is written in the native language even if they speak and
understand English.
4.2. Social Factors
It is obvious that student's preferences are influenced directly or indirectly by peers and friends
opinions. Their influences are clear in shaping and molding the course of an individual life [32]. Peer
influence is more observable in friendship [33] which is represented as succumbing to the views and opinions
of the peers, making a decision based on peer's advice, or just listening to the peer before listening to their
teacher and advisors is a form of such influence [34]. Naz et al., [32] found that peer and friends have a
positive role in selection of subjects, selection of a class and laboratory.
Analysis the feedback of students of department of Information System at Taibah University (Table
2), the majority of students (57.1%) are agree that their selection is dependent on the received advice from
their peers or friends, (55.5%) prefer to enroll in a course if some of their friends are also enrolled in the
same course, and (74.6%) indicated that their opinion about instructors are influenced by peers' and friends'
opinions. Generally speak, majority of students are agree that their selection is influenced by advice of their
peers and friends.
Table 2. Students' preferences respect peer's/friend's opinion
Question
Percentage
Strongly Disagree Disagree Neutral Agree Strongly Agree
My choice of course is mainly depend on advice of
my peers/friends
3.2 11.1% 28.6% 31.7% 25.4%
I prefer to enroll in a course if some of my friends
are enrolled in it also
7.9% 9.5% 27% 22.2% 33.3%
Enrollment in a course which is taught by an
instructor is depend on peers' /friends' opinions
about the instructor
3.2% 3.2% 19% 39.7% 34.9%
Total Impact of peer's/friend's advice on course
selection
1.3% 4.2% 19.8% 33.2% 41.5%
4.3. Individual Factor
- Course Time schedule
Although student preference respect course time schedule does not play a role in selection process
of full-time student, students have made decisions to take a course, or to not take a course, based on the fact
of whether or not it fits into their schedule [35], [8].

IJECE ISSN: 2088-8708 
3541
Table 3. Students' preferences course time schedule
Question
Percentage
Choosing the scheduling times of courses have
helped me to pass them successfully.
7.9% 12.7% 31.7% 20.6% 27%
Engagement students in scheduling courses time
enhances their motivation to study
13.8% 12.7% 20% 33.8% 19.7%
Total Impact of course time schedule on
course selection
3.1% 7.2% 23.1% 31.7% 34.9%
Table 3 illustrates that 47.6% of students found that choosing the scheduling times of courses have a
positive impact on their study and lead them to pass the courses successfully, and 53.5% of students think
that engagement them in scheduling courses time enhances their motivation to study.
- Student Demands
Several studies have considered interest in a course topic or subject as a driving force behind
students’ enrollment in classes [24], [34], [35]. The interest impact is more evident when students should to
make decision to take a course from elective courses available by the collage.
According to [26], student's interest in a course is influenced by numerous factors such as subject
matter, topics, and career goals. Enjoyment, job opportunities, and local labor trend are other factors that
influences the course selection. Students are attracted to take a course that they think that will increase their
chances to get a job.
- Learning Style
Learning style is one of the individual differences that play an important role in learning [36]. In the
literature, several definitions can be found which share the same basic idea " the term learning style refers to
a way or an approach a student follows in the course of learning”. According learning style theory, students'
interest in a course is influenced also by their preferred learning style. Adapting course content has been
applied intensively in e-learning systems where the learning styles and e-media are integrated together in the
design of their applications. Such integration showed a positive results in both learning styles detection and
e-learning application [37].
Table 4 presents how the learning style impacts on students' decision. It also presents students'
preferences regarding selecting courses. Statistical results emphasize on the fact that during making a
selection decision, beside the aforementioned factors, the learning style of a student should take in
consideration.
Table 4. Students' preferences respect to learning style
Question
Percentage
I prefer to take a course with practical nature before
those with theoretical
6.3% 12.7% 36.5% 20.6% 23.8%
I prefer to enroll with maximum allowed workload in
an academic level
15.9% 19% 38.1% 11.1% 15.9%
I prefer to postponed university required course to the
latest level
23.8% 28.6% 28.6% 9.5% 9.5%
I prefer to finish early university required course as
possible as I can
3.2% 6.3% 17.5% 36.5% 36.5%
I prefer to take the course with lowest credit hours
firstly, then the highest and so on.
20.6% 20.6% 31.7% 11.1% 15.9%
I think that allowing to take a course from any level,
in case I take its prerequisite, help me to success.
7.9% 12.7% 22.2% 31.7% 25.4%
I prefer to follow courses' order as it is in the
curriculum plan.
1.6% 3.2% 31.7% 42.9% 20.6%
5. WORK METHODOLOGY
The core of this research is to build a decision model which aim at help and support the students
during the enrollment and registration process. The model is two-phased process (Figure 2). The first phase,
is similar to those presented in [23] where the most correlated courses are generated. At the second phase, the
student preferences are taking in consideration. This preferences are prioritized using multi-criteria analytic
hierarchy process (MC-AHP). To understand the research context and the used data, in the next sections, we
present a brief explanation of the used methods, the application domain, and the gathered data.

 ISSN:2088-8708
3542
Figure 2. The Research Methodology
5.1. The Used Methods
- Correlation Analysis
Observing relationship among variables is a classical data mining task. Broadly, there are four types
of relationship mining: association rule mining, correlation mining, sequential pattern mining, and causal data
mining [8]. To help student in making a decision of which course he /she should to take, it is helpful finding
positive or negative linear correlations between courses.
Often, to represent the correlation graphically, a scatter diagram is used where the pair of points/data
(x, y) is allocated on an orthogonal coordinate system.
The linear correlation coefficient measures the strength of the linear correlation between the two
variables; it reflects the consistency of the effect that a change in one variable has on the other. In educational
realm, the correlation between two courses Ci and Cj as follows:
corr(Ci , Cj) =
∑ (gi
ci−g̅ci)(gi
cj
−g̅
cj)k
i=1
√∑ (gi
ci−g̅ci)2k
i=1
√∑ (gi
cj
−g̅
cj)2k
i=1
(4)
gi
ci
- is grade points for the ith course
gi
cj
- is grade points for the jth course
g̅ci- is average grade point for all students who take the ith course
g̅cj- is average grade point for all students who take the jth course
k- is number of students who take Ci and Cj.
The linear correlation coefficient takes value between −1 and +1:
 corr(Ci , Cj) = +1 reflects a perfect positive linear correlation between both courses Ci and Cj.
 corr(Ci , Cj) = −1 reflects a perfect negative linear correlation between both courses Ci and Cj.
 corr(Ci , Cj) = 0 means that there is NO linear correlation.
if the calculated value is close to +1 or −1, we then suppose that between the two variables there is a
linear correlation.
-Multi-criteria Analytic Hierarchy Process
AHP is a well-established decision making technique for dealing with multi-dimensional and often
contradictory preferences of individuals [5]. The AHP ranks alternatives in view of criteria and sub-criteria
(factors). In AHP, we start firstly with representing the problem with a hierarchal structure which is consists
of all factors and alternatives. The hierarchal structure mainly establishes the relationships between the levels
of the hierarchy order at which we place the objective (the Goal) at the top of the hierarchy, the criteria and
sub-criteria at intermediate levels, and finally the alternatives are placed at the lowest level of the order.
In the second step, a pair-wise comparison judgments are carried out, for each criterion, using a nine
points scale (1= equivalent,..., 9= extremely preferred to).
The result of each comparison is a matrix (n × n), where the diagonal elements aii are equal to
one,i = 1,2, … , n, and if aij = x, then aij = 1
x⁄ where x ≠ 0.

IJECE ISSN: 2088-8708 
3543
A = [
a11
a12 ⋯ a1n
a21
⋮
a22 ⋯
⋮ ⋱
a2n
⋮
an1
an2 ⋯ ann
]
Next step of the AHP (scoring and weighting) is to compute eigenvectors uj=(u1, u2, … , un) by
solving AW = λmax. W, where λ- is an eigen-value and W- is eigenvector.
The final step of AHP is to perform a consistency check (consistency ratio CR) by dividing the
consistency index CI by the random index RI, where the consistency index CI is calculated as follows:CI =
(λmax − n)/(n − 1), where n is the matrix size and the random index RI which is taken according Table 5.
Table 5. Average random consistency (RI) used in Saaty
Size of matrix 1 2 3 4 5 6 7 8 9 10
Random consistency 0 0 0.58 0.90 1.12 1.24 1.34 1.41 1.45 1.49
The CR is considered acceptable only if it is less than 0.1, otherwise the pair-wise comparison
judgments should be reviewed and improved.
5.2. Application Domain
To show how the decision model supports the students during the enrollment and registration
process, the experiential part of this work was developed in the context of department of Information System
at the Taibah University, Kingdom of Saudi Arabia. Generally, study at Taibah University, as all remains
universities in Saudi Arabia, are organized in two regular academic terms by year, plus a summer term which
is opened only if there is a quite number of students who failed pass a course in regular terms. The regular
terms are spanning four months, whilst the summer term is compressed into two months. Since 2004, the
academic program is changed three times. However, number of credit hours is still the same. Each program
consists of two parts:
- the preparation period where students spent one academic year at which they took a set of courses that
prepare them to their future studies
- the regular period is consists of four academic years. The program consists of 14 credit hours of
university requirement courses, 19 credit hours of faculty requirement courses, and 46 credit hours of
department requirement courses nine of them are elective courses.
In order to pass a course, the student has to obtain at least 60 points out of 100; otherwise he will be
required to attend the course again in the next academic year or in the summer term, in case the number of
those students who failed to pass the course is quite enough (the decision is made based on the opinion of the
vice dean of the academic affairs at each faculty). The maximum number of attempts to pass a course is
depends on the student's GPA. For the student whose the GPA is less than the cut-point (2.5 out of 5) for two
sequential academic terms, he will not be able to continue his/her studies. During the enrollment period,
students should to register the selected courses including the name of the preferred time and group using the
online enrollment system or by assistance the academic advisors. The students is eligible for enrollment a
course, only if they passed the prerequisites for the said course, otherwise they are deny to take it.
5.3. Data Description
Since the academic program is changed several times, the historical records contain data from three
different curricula, each of them with 42 courses separated through eight regular academic terms and four
summer terms, for further details about the total number of students and classes, see Table 6. Due to of
modification or changes in the curricula (sometimes, only the prerequisites of a course is changed), we focus
only the curricula from 2011 to 2015 namely "new curricula".
Table 6. The number of students in each academic year and average classes to graduate students.

 ISSN:2088-8708
3544
Academic Year
Enrolled Graduated Average Classes to graduate Curricula
Male Female Male Female Male Female
2010/2011 420 600 90 89 10.3 9 Old Curricula
2011/2012 600 676 95 126 10.4 9.8
New Curricula2012/2013 484 686 58 150 10.9 10.3
2013/2014 789 698 93 137 10.7 11
2014/2015 828 674 111 175 10.3 11.1 Developed
Curricula
The average classes to graduate students in Table 6 refers to the number of academic terms that
students spend to finish their study in case the fail to pass the course from the first attempt. Figure 3 shows
the increase in the required classes between both groups (male and female sections).
Figure 3. Average number of classes required to graduate students
The main goal of the current research is to give the student (who intends to register on a course) a
recommendation based on the gained grades at the previous terms. The correlation analysis is performed
based on the final grade of the students. The aim of this step is to link each course with the most correlated
courses that may be effected by the selected course. Table 7 shows the used attributes and give a brief
description for each of them, whilst Table 10 presents the data type of the attributes and a short statistical
summary for each of them. The "Period" attribute refers to the academic term in which student should take a
course. It discriminates as follows:
Period = {
x ∈ [1 − 8], 𝑥 − 𝑖𝑠 𝑎 𝑟𝑒𝑔𝑢𝑙𝑎𝑟 𝑡𝑒𝑟𝑚
x ∈ [9 − 12], 𝑥 − 𝑖𝑠 𝑎 𝑠𝑢𝑚𝑚𝑒𝑟 𝑡𝑒𝑟𝑚
Both "Registered Credit hours" and "Gained Credit hours" attributes are used to split the data set in
to training and testing set. The highest value of " Registered credit hours" denotes students has a difficulties
in finishing his study, whilst the highest value of "Gained credit hours" denotes that the student is near to
graduate.
Table 7. The used attributes
Attributes Description
Course name Identifier for each course the student is enrolled on
Course code Identifier for each course in the university system
Course credits Practical and theoretical workload for each course
Period Academic term in which student should take the course
Final grade Result obtained at the end of the term in each course
Student ID Identifier for each student
GPA Overview of the student’s performance over time
Student gender The gender of student who took a course
Registered credit hours Amount of credit hours registered in the university system
Gained credit hours Amount of credit hours already student passed

IJECE ISSN: 2088-8708 
3545
Table 8. Statistical summary of the used attributes
6. EXPERIMENTATION AND EVALUATION
Most course recommendation systems use students' personal data and social networking sites to find
out what they like or are interested in [38], e.g., proposed to use students' grades in developing a course
recommendation system. The system helps students find courses in which they can get high scores. For this
purpose, data about the courses which users learned, scores that students received, and the teachers of the
courses are collected.
Current work, as mentioned before, suggested a method with two-phased process. At the first phase,
the most correlated courses are generated. However, as we check correlation course by course, only those
courses (set of courses) that satisfy the following constrains are generated:
 Only courses with highest correlation are generated.
 The total number of credit hours of the generated courses (academic load of student 𝐴) should
not exceed the maximum course load 𝜆1.
The Pseudo-Code for generating courses based on correlation analysis is illustrated in Pseudo-Code 1.
Pseudo-Code I: Generating courses based on correlation analysis
𝐺𝑃𝐴 𝑆
𝑡−1
: GPA of a student 𝑆 at previous semester 𝑡 − 1
𝜆1 , 𝜆0: maximum and minimum course load for a student respectively
𝑐𝑡−1: course at semester 𝑡 − 1 that is passed successfully by the student, 𝑐𝑡−1 ∈ ℒ
𝑐𝑡+1: course at next semester that student can take, 𝑐𝑡+1 ∈ ℒ
𝑆: Final set of recommended courses
𝑟𝑐
𝑡+1
: Credit hours of courses at semester 𝑡 + 1
1: |𝑆| ← ∅
2: IF 𝐺𝑃𝐴 𝑆
𝑡−1
> 𝜆0 THEN
3: 𝑐𝑜𝑟𝑟(𝑐𝑡−1, 𝑐𝑡+1); # Calculate correlation
4: END IF
5: FOR ∀𝑐 ∈ 𝐶 𝑡−1
; # Loop for all courses that a student passed them successfully
6: 𝐴𝑣𝑟𝑔_𝑐𝑜𝑟𝑟[𝑖] ←
𝐶𝑜𝑚𝑝𝑒𝑡𝑒𝑛𝑐𝑒( 𝑐 𝑖
𝑡+1) ∑ 𝑐𝑜𝑟𝑟(𝑐 𝑗
𝑡−1,𝑐 𝑖
𝑡+1)
𝑛
𝑗=1
𝑛
; # Find average matrix
Attributes Data type Possible Value Statistical summary
Course name String Mathematics, Physic (1),
Programming (1), est.
42 courses
Course code String Math101, Phys101, CS102, est. 42 courses
Course credits Discrete 1,2,3,4 Course Credits Percentage
1 2%
2 12%
3 55%
4 31%
Period Discrete 1,2,..., 12. 12 terms
Final grade Continuous 〈0,100〉 Minimum 38
Maximum 100
Student ID String 31#####, 32#####, 33##### Student ID Percentage
31##### 8.4%
32#####, 43%
33#####, 48.6%
GPA Continuous 〈0.000,5.000〉 Mean 3.937297
Standard Deviation 0.552337
Sample Variance 0.305076
Minimum 2.06
Maximum 4.99
Student gender String Male, Female Student gender Percentage
Male 56%
Female 44%
Registered credit hours Discrete 〈0,250〉 Mean 173.0541
Minimum 153
Maximum 209
Gained credit hours Discrete 〈0,165〉 Mean 102.797
Minimum 14
Maximum 165

 ISSN:2088-8708
3546
7: 𝐶 𝑡+1
← 𝑆𝑒𝑙𝑒𝑐𝑡(𝑘, 𝑀𝑎𝑥(𝐴𝑣𝑟𝑔_𝑐𝑜𝑟𝑟(𝑐𝑡−1, 𝑐𝑡+1)); # Select 𝐾 courses with maximum correlation
that are satisfy the university regulation
8: |𝑆| ← 𝑃𝑎𝑐𝑘𝑎𝑔𝑒(𝜆1, 𝐶 𝑡+1
, 𝑟𝑐
𝑡+1
); # Generate the recommended course package
Pseudo-Code II: 𝑃𝑎𝑐𝑘𝑎𝑔𝑒(𝜆1, 𝐶 𝑡+1
, 𝑟𝑐
𝑡+1
); Generate the recommended course sets
Input: 𝜆1 : allowed maximum course load for a student
𝐶 𝑡+1
: Set of all available courses at semester 𝑡 + 1
𝑟𝑐
𝑡+1
: Credit hours of a course 𝑐 ∈ 𝐶 𝑡+1
Output: Set of the recommended courses
1: Set |𝑅| ← ∅, |𝑆| ← ∅
2: 𝑠 ← 𝐺𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝐴𝑙𝑙𝐶𝑜𝑛𝑑𝑖𝑑𝑎𝑡𝑒𝑆𝑢𝑏𝑆𝑒𝑡𝑠(𝐶 𝑡+1
);
3: Add 𝑠 to |𝑆|
4: FOR 𝑖 ← 0 to 𝑠
5: 𝑟[𝑖] ← 𝐶𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑇𝑜𝑡𝑎𝑙𝐶𝑟𝑒𝑑𝑖𝑡𝐻𝑜𝑢𝑟𝑠(𝑟𝑐
𝑡+1
);
6: Add 𝑟[𝑖] to |𝑅|
7: END FOR
8: WHILE 𝒔, 𝑟[𝑖] ≤ 𝜆1
9: 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒[𝑖] ←
∑ (𝑟 𝑐 𝑖
𝑡+1𝐶 𝑡+1
i=1,j=1,𝑐∈𝐶 𝑡+1,
)∗corr(ci,cj)
𝑟[𝑖]
10: 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒𝑆𝑒𝑡 ← 𝑆𝑒𝑙𝑒𝑐𝑡(𝑀𝑎𝑥(𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒[𝑖])); # Sort the list
12: END WHILE
13: return 𝐶𝑎𝑛𝑑𝑖𝑑𝑎𝑡𝑒𝑆𝑒𝑡
Since the recommended courses are generated based on the correlation analysis, the recorded grades
can be influenced, as mentioned above, by course content itself or experiencing a particular teacher’s style or
material of teaching [23]. Assuming that a student S passed a set of courses at semester t − 1, and the grades
in the database are recorded taking in consideration the aforementioned influential factors.
Having samples of k students S = {s1, … , sk}, who took a course Cj after a course Ci, and each
student si achieved gj
cj
GPA in course cj and gi
ci
GPA in course Cias shown in Table 9.
Table 9. Sample of students grades at a semester 𝑡
Student ID Course
ISLS101 ARAB101 MATH101 PHYS103 CS101 ENGL103
3151377 95 95 87 70 81 85
3151559 90 98 71 65 71 81
3182286 99 96 93 85 80 72
3182565 95 96 90 85 86 95
3182894 98 95 77 87 75 75
3200079 100 100 98 100 98 90
3200162 100 100 74 93 95 78
3200229 99 100 100 100 95 98
Let a student passed a set of courses as shown in Table 10, and the correlation among all curriculum
courses are already calculated.
Table 10. Student grades at current semester
Student ID Course
ISLS101 ARAB101 MATH101 PHYS103 CS101 ENGL103
3200237 95 95 87 70 60 85
Since Eq. (4) assumes that the grades of all courses are given, and because of absence the grades of
those courses of the next semester, we propose to use the regression analysis to predict the grades of each
courses based on the historical records of all its perquisite courses (see Table 11).

IJECE ISSN: 2088-8708 
3547
Table 11. Predicted grades for the courses at the next semester
Student ID Curriculum Plan 𝑃
Grades at Semester 𝑡 Predicted Grades
ISLS
101
ARAB
101
MATH
101
PHYS1
03
CS
101
ENGL
103
IS
201
CS
202
ARAB
201
MAT
H
102
IS
102
IS
221
CS
112
ENG
L 104
3200237 95 95 87 70 60 85 66 86 92 52 85 70 75 75
Table 12 presents the correlation among the already taken courses and those are offered at the next
semesters.
Table 12. Example for calculation demonstration.
Current Courses at Semester 𝑡
Next Semester 𝑡 + 1
IS
201
CS
202
ARAB
201
MATH
102
IS
102
IS
221
CS
112
ENGL
104
3151377
ISLS101 95 0.04 0.25 0.604 0.331 0.441 0.244 0.221 0.09
ARAB101 95 0.00 0.27 1.000 0.346 0.210 0.279 0.297 -0.01
MATH101 87 0.15 0.63 0.526 1.000 0.536 0.447 0.489 0.47
PHYS103 70 0.29 0.37 0.483 0.625 0.342 0.260 0.358 0.20
CS101 60 0.34 0.46 0.486 0.640 0.286 0.574 1.00 0.49
ENGL103 85 0.26 0.49 0.603 0.416 0.379 0.151 0.321 1.00
Table 13 presents average correlation matrix of each generated course. Steps 7-10 allow us to sort
and select the top k courses based on its average correlation.
Table 13. Average correlation matrix of the courses
Current Courses at Semester 𝑡
Next Semester 𝑡 + 1
IS
201
CS
202
ARAB
201
MATH
102
IS
102
IS
221
CS
112
ENGL
104
3151377
ISLS101 95 0.04 0.25 0.604 0.331 0.441 0.244 0.221 0.09
ARAB101 95 0.00 0.27 1.000 0.346 0.210 0.279 0.297 -0.01
MATH101 87 0.15 0.63 0.526 1.000 0.536 0.447 0.489 0.47
PHYS103 70 0.29 0.37 0.483 0.625 0.342 0.260 0.358 0.20
CS101 60 0.34 0.46 0.486 0.640 0.286 0.574 1.00 0.49
ENGL103 85 0.26 0.49 0.603 0.416 0.379 0.151 0.321 1.00
Average 0.18 0.41 0.617 0.56 0.366 0.33 0.45 0.37
To generate the set of the candidate courses package, first we should to follow the academic
regulation. In our case where this research is conducted, the academic regulation determines the maximum
course load λ1based on the student's GPA as follows:
λ1 = {
λ1 ∈ [17,21] , student is expected to graduate next semester
λ1 ∈ [12,17] , 5 ≤ GPA ≤ 2.8
λ1 = 12 , GPA ≥ 2.7
Let that the current GPA of a non-graduated student is 3.21 which means that the student has a right
to register courses with total credit hours up to 17. Table 14 shows the combination of all possible sets of the
eight courses. The set with the largest average correlation is suggested as a recommended package of courses
for the next phase.

 ISSN:2088-8708
3548
Table 14. Combination of all possible sets
ARAB
201
MATH
102
CS
112
CS
202
ENGL
104
IS
102
IS
221
IS
201
𝜆1
Overall Priority
Average
Rank
Credit
Hours
3 4 4 4 3 3 4 3
3200237
ARAB-201, MATH-102, ENGL-104, IS-221, IS-201 17 0.219992771 1
ARAB-201, MATH-102, ENGL-104, CS-112, IS-201 17 0.197847972 4
ARAB-201, MATH-102, ENGL-104, CS-202, IS-201 17 0.213518998 2
⋯ ⋮ ⋮
ARAB-201, MATH-102, CS-112, CS-202 15 0.191914578 5
ARAB-201, MATH-102, CS-112, ENGL-104 14 0.207102686 3
⋯ ⋮ ⋮
MATH-102, CS-112, CS-202 12 0.161333772 6
MATH-102, CS-112, IS-102 12 0.16062525 7
⋯ ⋮
Let the timetable of these courses are already predefined and each course (in this set) is linked with
one/ many instructors in different time (see Figure 4). The next step, as shown in Figure 2, is to prioritize
time, instructors, and sections of the courses based on the student preferences. As mentioned previously, the
preferences are prioritized using multi-criteria analytic hierarchy process (MC-AHP) (See Section 5.a).
Figure 4. Snapshot of IS-102 schedule
Following the procedure of MC-AHP, there are a total of six pair-wise comparison matrix tables:
1. The pair-wise comparison matrix of the criteria relating to the goal. This is illustrated in Table 15.
2. The pair-wise comparison matrices for the five options (set of the recommended courses) regarding all
the “criteria concerned”, where the criteria in all levels are connected to the options.
The consistency ratios of all comparisons were less than 0.1, which indicates that the weights used
are consistent.
Table 15. Pair-wise comparison matrix of the key criteria with regards to the goal
Criteria Course factors Social factors Individual factors
Global Priority
Vector
Course factors 1 9 3 0.67
Social factors 1/9 1 1/5 0.06
Individual factors 1/3 5 1 0.27
𝐶𝐼 = 0.014606 𝐶𝑅 = 0.025182 ≤ 0.1
Tables 16 illustrates the pair-wise comparisons of the alternatives for the first offered sections of IS102
course in terms of aforementioned criteria.

IJECE ISSN: 2088-8708 
3549
Table 16. Pair-wise comparison matrix with regards to the sub-criteria
Criteria
Course
characteristics
Instructor
characteristics
Teaching
language
Local Priority
Vector
Priority respect
Global Vector
Course characteristics 1 5 3 0.63 0.4221
Instructor characteristics 1/5 1 1/3 0.11 0.0739
Teaching language 1/3 3 1 0.26 0.1742
𝐶𝐼 = 0.019357 𝐶𝑅 = 0.033375 ≤ 0.1
Criteria
Peer opinions Friendship Campus
Location
Priority Vector
Priority respect
Global Vector
Peer opinions 1 3 1/3 0.29 0.0174
Friendship 1/3 1 1/3 0.14 0.00084
Campus Location 3 3 1 0.57 0.0342
𝐶𝐼 = 0.076 𝐶𝑅 = 0.1
Criteria Course time
scheduling
Student demands
Learning
style
Priority Vector Priority respect
Global Vector
Course time scheduling 1 7 1 0.51 0.1372
Student demands 1/7 1 1/3 0.10 0.027
Learning style 1 3 1 0.39 0.1053
𝐂𝐈 = 𝟎. 𝟎𝟒𝟎𝟒𝟕𝟒 𝐂𝐑 = 𝟎. 𝟎𝟔𝟗𝟕𝟖𝟒
The output of the MC-AHP algorithm is summarized in the overall priority matrix as shown in Table 17.
Table 17. Overall priority matrix
Course Name IS-102 (Foundation of Information Systems)
Course factors Social factors Individual factors Overall Priority Rank
IA_4 0.6702 0.05244 0.2695 0.992 I
IB_4 0.4763 0.3661 0.109 0.9514 II
The procedure is continuing for all courses that are generated at the first phase. The aim of this step
is to prioritize the offered sections in the timetable. Thus, students are provided by a list of courses and its
sections taking in consideration their preferences.
7. CONCLUSION and FUTURE WORK
Course enrollment (CE) as administrative task is a repetitive process which faces students each
semester. Students, during the enrollment period, often, need to support. At a time not so long ago, students
were responsible for their own choices. Since the students aim to finish their study as soon as they can taking
as many courses as possible, their choices might affect negatively on their performance. From this end,
colleges and universities began to implement so-called academic advising affairs. Although, the faculty
academic advising has a significant impact on a student’s academic success, several issues may lead to limit
this success specially when the ratio of the academic advisors to the students is high.
In this paper, we have presented the course enrollment task as a function of maximization of GPA.
For this purpose, we have proposed a two-phased process. The first phase, is similar to those presented in
[23] where the most correlated courses are generated. At the second phase, the courses are prioritized based
on the student preferences. The students selection is influenced several factors which have been categorized
into three main groups of concerns: (i) course factors, (ii) social factors, and (iii) individual factors.
Through this work, we have evaluated our decision model in the context of department of
Information System at the Taibah University, Kingdom of Saudi Arabia. Since the collected data were from
different curriculum plans, our concern was focused only on one curriculum plan because the change in
perquisite courses will cause different recommended courses. In the future, we intend to integrate with
timetable system of the admission and registration deanship, and build an unified model that can deal with
different curriculum plans. Further analysis should cover more factors that influence the course selection
itself. It will be more appropriate to shift the research towards collaborative recommended systems and
cluster students with same preferences and analyze their behaviors. The date mining approach would also be
very interesting.

 ISSN:2088-8708
3550
REFERENCES
[1] Winston RB, Ender SC, Miller TK. Academic advising as student development. Developmental approaches to
academic advising. New Directions for Student Services. 1982 (17).
[2] Al-Sarem M. Predictive and statistical analyses for academic advisory support. arXiv preprint arXiv:1601.04244.
2015 Nov 7.
[3] O'Banion T. An academic advising model. NaCADA Journal. 1994 Sep; 14(2): 10-6.
[4] Chiarandini M, Di Gaspero L, Gualandi S, Schaerf A. The balanced academic curriculum problem revisited.
Journal of Heuristics. 2012 Feb 1; 18(1): 119-48.
[5] Ognjanovic I, Gasevic D, Dawson S. Using institutional data to predict student course selections in higher
education. The Internet and Higher Education. 2016 Apr 30; 29: 49-62.
[6] Vialardi C, Chue J, Peche JP, Alvarado G, Vinatea B, Estrella J, Ortigosa Á. A data mining approach to guide
students through the enrollment process based on academic performance. User modeling and user-adapted
interaction. 2011 Apr 1; 21(1-2): 217-48.
[7] Takada K, Miyazawa Y, Yamamoto Y, Imada Y, Tsuruta S, Knauf R. Curriculum Optimization by Correlation
Analysis and Its Validation. InHuman-Computer Interaction and Knowledge Discovery in Complex, Unstructured,
Big Data 2013 (pp. 311-318). Springer Berlin Heidelberg.
[8] Baker RS. Data mining for education. International encyclopedia of education. 2010 May; 7(3): 112-8.
[9] Hsia TC, Shie AJ, Chen LC. Course planning of extension education to meet market demand by using data mining
techniques–an example of Chinkuo technology university in Taiwan. Expert Systems with Applications. 2008 Jan
31; 34(1): 596-602.
[10] Monette JN, Schaus P, Zampelli S, Deville Y, Dupont P. A CP approach to the balanced academic curriculum
problem. InSeventh International Workshop on Symmetry and Constraint Satisfaction Problems 2007 Sep 23 (Vol.
7).
[11] Hnich B, Kiziltan Z, Walsh T. Modelling a balanced academic curriculum problem. In Proceedings of CP-AI-OR-
2002 2002 Mar.
[12] Castro C, Crawford B, Monfroy E. A quantitative approach for the design of academic curricula. InSymposium on
Human Interface and the Management of Information 2007 Jul 22 (pp. 279-288). Springer Berlin Heidelberg.
[13] Lambert T, Castro C, Monfroy E, Riff MC, Saubion F. Hybridization of genetic algorithms and constraint
propagation for the BACP. InInternational Conference on Logic Programming 2005 Oct 2 (pp. 421-423). Springer
Berlin Heidelberg.
[14] Lambert T, Castro C, Monfroy E, Saubion F. Solving the balanced academic curriculum problem with an
hybridization of genetic algorithm and constraint propagation. Artificial Intelligence and Soft Computing–ICAISC
2006. 2006: 410-9.
[15] Ünal YZ, Uysal Ö. A new mixed integer programming model for curriculum balancing: Application to a Turkish
university. European Journal of Operational Research. 2014 Oct 1; 238(1): 339-47.
[16] Rubio JM, Palma W, Rodriguez N, Soto R, Crawford B, Paredes F, Cabrera G. Solving the balanced academic
curriculum problem using the ACO metaheuristic. Mathematical Problems in Engineering. 2013 Dec 17; 2013.
[17] Laghari MS, Khuwaja GA. Electrical engineering department advising for course planning. InGlobal Engineering
Education Conference (EDUCON), 2012 IEEE 2012 Apr 17 (pp. 1-6). IEEE.
[18] Hashemi RR, Blondin J. SASSY: A petri net based student-driven advising support system. InInformation
Technology: New Generations (ITNG), 2010 Seventh International Conference on 2010 Apr 12 (pp. 150-155).
IEEE.
[19] Nambiar AN, Dutta AK. Expert system for student advising using JESS. InEducational and Information
Technology (ICEIT), 2010 International Conference on 2010 Sep 17 (Vol. 1, pp. V1-312). IEEE.
[20] Mashiloane L. Using association rule mining to find the effect of course selection on academic performance in
computer science i. InMining Intelligence and Knowledge Exploration 2014 (pp. 323-332). Springer International
Publishing.
[21] Hsieh TC, Wang TI. A mining-based approach on discovering courses pattern for constructing suitable learning
path. Expert systems with applications. 2010 Jun 30; 37(6): 4156-67.
[22] Aher SB. EM&AA: An Algorithm for Predicting the Course Selection by Student in e-Learning Using Data Mining
Techniques. Journal of the Institution of Engineers (India): Series B. 2014 Jan 1; 95(1): 43-54.
[23] Tsuruta S, Knauf R, Dohi S, Kawabe T, Sakurai Y. An intelligent system for modeling and supporting academic
educational processes. InIntelligent and Adaptive Educational-Learning Systems 2013 (pp. 469-496). Springer
Berlin Heidelberg.
[24] Curran JM, Rosen DE. Student attitudes toward college courses: An examination of influences and intentions.
Journal of Marketing Education. 2006 Aug; 28(2): 135-48.
[25] Wilhelm WB. The relative influence of published teaching evaluations and other instructor attributes on course
choice. Journal of Marketing Education. 2004 Apr; 26(1): 17-30.
[26] Babad E, Tayeb A. Experimental analysis of students' course selection. British Journal of Educational Psychology.
2003 Sep 1; 73(3): 373-93.
[27] Beggs JM, Bantham JH, Taylor S. Distinguishing the Factors Influencing College Students'choice of Major.
College Student Journal. 2008 Jun 1; 42(2): 381.
[28] Ferrer-Caja E, Weiss MR. Cross-validation of a model of intrinsic motivation with students enrolled in high school
elective courses. The Journal of Experimental Education. 2002 Jan 1; 71(1): 41-65.

IJECE ISSN: 2088-8708 
3551
[29] Barak M, Watted A, Haick H. Motivation to learn in massive open online courses: Examining aspects of language
and social engagement. Computers & Education. 2016 Mar 31; 94: 49-60.
[30] Lee A. Virtually Vygotsky: Using Technology to Scaffold Student Learning: By. Technology in Pedagogy. 2014;
20: 1-9.
[31] Coleman JA. English-medium teaching in European higher education. Language teaching. 2006; 39(01): 1-4.
[32] Naz, A., Saeed, G., Khan, W., Khan, N., Sheikh, I., & Khan, N. (2014). Peer and Friends and Career Decision
Making: A Critical Analysis. Middle-East Journal of Scientific Research, 22(8), 1193-1197.
[33] Webber DJ, Walton F. Gender‐specific peer groups and choice at 16. Research in Post-Compulsory Education.
2006 Mar 1; 11(1): 65-84.
[34] Malgwi CA, Howe MA, Burnaby PA. Influences on students' choice of college major. Journal of Education for
Business. 2005 May 1; 80(5): 275-82.
[35] Anderson N, Lankshear C, Timms C, Courtney L. ‘Because it’s boring, irrelevant and I don’t like computers’: Why
high school girls avoid professionally-oriented ICT subjects. Computers & Education. 2008 May 31; 50(4): 1304-
18.
[36] M. Al-Sarem, M.M. Bellafkih and Ramdani. Adaptation Patterns with respect to Learning Styles. Proceeding of 9th
International Conference on Intelligent Systems: Theories and applications (SITA’14), Rabat, Morocco.
[37] Truong HM. Integrating learning styles and adaptive e-learning system: Current developments, problems and
opportunities. Computers in Human Behavior. 2016 Feb 29; 55: 1185-93.
[38] Chang PC, Lin CH, Chen MH. A Hybrid Course Recommendation System by Integrating Collaborative Filtering
and Artificial Immune Systems. Algorithms. 2016 Jul 22; 9(3): 47.
BIOGRAPHY OF AUTHOR
Mohammed Al-Sarem is an assistant professor of information science at the Taibah University,
Al Madinah Al Monawarah, KSA. He received the PhD in Informatics from Hassan II
University, Mohammadia, Morocco in 2014. His research interests center on E-learning,
educational data mining, Arabic text mining, and intelligent and adaptive systems. He published
several research papers and participated in several local/international conferences

Solving Course Selection Problem by a Combination of Correlation Analysis and Analytic Hierarchy Process

More Related Content

What's hot (18)

Similar to Solving Course Selection Problem by a Combination of Correlation Analysis and Analytic Hierarchy Process (20)

More from IJECEIAES (20)

Recently uploaded (20)

Solving Course Selection Problem by a Combination of Correlation Analysis and Analytic Hierarchy Process