Fuzzy-Based Recommendation System for University Major Selection

Shaima Alghamdi, Nada Alzhrani

and Haneen Algethami

Department of Computer Science, College of Computers and Information Technology, Taif University, Taif, Saudi Arabia

Keywords: Fuzzy Expert System, Recommendation System Decision Support System, Major Selection.

Abstract: The decision of choosing a university major is one of the most important decisions in every adult life. To

make a suitable decision, a student needs an expert opinion, time, and effort. Therefore, a decision-making

system should be developed in order to help prospective students to increase their educational outcome and

productivity. In Saudi Arabia, each university requires specific criteria in order to accept students. These

criteria are made based on two factors: 1) the outcome of student’s qualification exams and 2) overall high

school grades. The student must take these calculations into consideration when selecting a major. Thus, in

this paper, a Fuzzy-Based Recommendation System (FRS) is proposed to aid students in choosing a suitable

major. This system designed using Fuzzy Expert System (FES). Additionally, a cluster-based preferences

technique is implemented to obtain the student's preferred majors, using distance measurement. The system

has been tested on fifteen prospective students to measure its accessibility. Results showed that students are

stratified by the suggested majors that fell in line with their preferences.

1 INTRODUCTION

Most high school students are uncertain of which

university major to choose after they finish their study.

The decision-making process can be overwhelming.

Hence, they usually look for guidance and support by

reaching out to teachers, relatives and colleagues.

These people are regarded as experts. However, the

problem lies in finding an expert available to help the

students. Also, some experts’ opinions tend to be

subjective to their own experience, without taking

into consideration the student preferences. Hence, the

decision support system (DSS) are more effective to

help students in making life-affecting decisions

(Turban, 1995).

In Saudi Arabia, there is a need for a DSS for the

university major selection problem. Universities

require qualification criteria in order to register

students. In addition to the student’s previous

performance, preferences should be considered

during the process to provide a more realistic result.

Finally, some majors might share courses, which

make the decision of choosing one major over the

other is a difficult task.

In this paper, a fuzzy-based recommendation

system is used to suggest a list of majors for the

student. Recommender Systems (RSs) provide

https://orcid.org/ 0000-0002-7582-4480

suggestions for items to be of use to a user (Ricci,

2011). In this paper, the knowledge-based

recommendation system (KBRS) is used, in which

the system uses inference technique to find a

relationship between the items and the user (Burke,

2000).

Fuzzy logic is used to handle uncertainty rising

from similarities between the majors. Fuzzy logic can

provide an effective means for conflict resolution of

multiple criteria and better assessment of options

(Burke, 2000).

The specific objectives of this paper are:

 To identify the features that contribute to

maximizing student satisfaction on the major

choice.

 To propose an Intelligent Decision Support

System (IDSS) to aid students in the decision-

making process

 To evaluate user satisfaction on the system

performance.

In what follows, Section II reviews related work.

Section III describes the problem definition. Section

IV proposed methodology used in this paper. Section

V presents the evaluation study and discusses the

obtained results. The paper is then concluded in

Section VI.

Alghamdi, S., Alzhrani, N. and Algethami, H.

Fuzzy-Based Recommendation System for University Major Selection.

DOI: 10.5220/0008071803170324

In Proceedings of the 11th International Joint Conference on Computational Intelligence (IJCCI 2019), pages 317-324

ISBN: 978-989-758-384-1

 2019 by SCITEPRESS – Science and Technology Publications, Lda. All rights reser ved

317

2 LITERATURE REVIEW

Recommendation systems (RSs) were found to be a

helpful tool (Yera, 2017). It can help students to find,

organize, and use resources that match their

individual goals, interests, and current knowledge

(Al-Badarenah et al, 2016). Nevertheless, recent

studies show different approaches for managing

uncertainty in recommender systems, such as

Bayesian approaches (Luis M. de Campos, 2008),

Markov models (Nachiketa Sahoo, 2012), fuzzy

approaches (Azene Zenebe, 2009), genetic

algorithms (Holland, 1992), or neural networks

(Lehr, 1990).

Recommendation systems have played an

important role in education. One of these systems is a

Markov Chain Collaborative Filtering Model for

Course Enrollment Recommendations by (Elham

S.Khorasani, 2016). Another recommendation

system is course recommender system using

association rules by (Narimel Bendakir, 2006).

Another recommendation system is designed by

(Desi Purwanti Kusumaningrum, 2017) entitled

Recommendation System for Major University

Determination Based on Student’s Profile and

Interest.

There are also, number of studies that have

addressed the major selection problem. One of these

studies is a Prototype Rule-based Expert System with

an Object-Oriented Database for University

Undergraduate was proposed by (Ahmar, 2012). The

study highlighted the importance of using an expert

system supported by an object-oriented database.

Also, it used Kappa-PC expert system development

environment, which supports rule-based reasoning,

object-oriented modelling, list processing, and

graphical user interface construction components.

This ES has three major components that are: 1)

Knowledgebase; 2) Inference engine; 3) User

interface.

Another study is the Decision Support System for

Major Selection Vocational High School (VHS)

using Fuzzy Logic Android-Based was proposed by

(Salaki, 2015). It is a DSS to aid the student in the

decision-making process, based on the score of

acceptance exams to specify the appropriate VHS

major for the student. The DSS consists of three main

parts: 1) Information system; 2) DSS, which has three

subsystems Database subsystem, Model subsystem,

Dialog Subsystem; 3) Fuzzy Inference System.

In the previously presented literature, fuzzy logic

was used to deal with uncertainty in relative

problems. Additionally, database systems were also

used to store data. Finally, a graphical user interface

was also used to retrieve online information. Hence,

the same components are used in developing the FRS

for the university major selection problem presented

in this paper. Even though the previous studies have

used fuzzy expert systems to solve this problem, it is

worthy to develop an efficient IDSS for tackling real-

world major selection, for students applying to Taif

University at Saudi Arabia. The intended contribution

focuses on the use of fuzzy logic to improve the

performance of knowledge-based recommender

systems. The combination harnesses its power with

the fuzzy expert system.

3 PROBLEM STATEMENT

The major selection problem aims to maximise

student satisfaction on their major choice to minimise

the number of ungraduated students.

There are two tracks in high school, science and

art. Each track has specific majors. Students from

each track can apply only to those majors. However,

the science track has more options than the art track.

For example, a student applying for mathematics

must be from the science track. On the other hand, a

student applying for linguistics can be from science

or art track. Thus, the high school track affects the

direction of the result of the system.

Universities require three qualification criteria to

accept students. third-year high school percentage

(HSA), and percentages of two tests:1) General

Ability Test (GAT) and 2) Achievement test (AT).

The proposed recommendation system suggests a

list of suitable majors based on the student’s overall

percentage and the student’s preferences. The

percentage is calculated based on the GAT, AT and

HSA values.

In Taif University (TU), there are ten colleges,

with each college having several majors to choose

from and different calculation scheme. The student's

overall percentage to be accepted in medicine and

pharmacy colleges are calculated as shown in

equation (1), where HSA and AT must be greater than

or equal 75%. The student's overall percentage to be

accepted in engineering, computers and information

technology and applied medical sciences colleges is

calculated also as shown in equation (1), however

HSA and AT must be greater than or equal 70%. The

student's overall percentage value for the science

college is calculated as shown in equation (2). The

student's overall percentage value for the art,

education, shari’a, and business administration

colleges is calculated as shown in equation (3). Note

that α =0.3, β = 0.4 and γ =0.5.

FCTA 2019 - 11th International Conference on Fuzzy Computation Theory and Applications

318









  







  







  





(1)





    

 

(2)





    

(3)

3.1 Data Collection

Prospective students find it difficult to select a

university major. If they are not satisfied with their

selection, they might change their major during their

four-year degree program. In order to understand the

factors affecting their decisions, two surveys were

conducted. The first survey targeted high school

students. The second survey targeted university

students to give their insights after spending a year in a

specific major.

The survey had 239 prospective participants and

392 university participants. In both surveys, a high

percentage of students agreed on the difficulty of

choosing a university major. In addition, more than

half of the participants recommend the need for a

system to help them in the decision-making process.

4 METHODOLOGY

4.1 Fuzzy Expert System

Implementation

The aim of this step is to define a set of available

majors for each student, i.e. the majors where the

student's overall percentage matches the major

requirements.

First, the high school track must be identified. the

overall percentage is calculated as explained in Section

III by using HSA, AT and GAT values. For a science

student, 



, 



, and 



values are computed, since

she/he can enroll in both science and art majors. On the

other hand, only 



value is computed for art students,

as they are only allowed to enroll in art majors. If the

user is a science student, the following step is to ask

her/him to choose their preferred track in the

university. If the student chooses the art track; the

system only deals with 



, i.e. for the art section. If the

science student chooses the science track; the system

handles three values, i.e. 



, 



, and 



The values are then passed to the FES to determine

the applicable majors for the student, with respect to

the previous criteria.

4.1.1 Fuzzy Logic Process

This process consists of a number of steps as follows:

1. Identify the linguistic variables and values, as

presented in Table 1 (Jang J. S., 1997).

Table 1: Linguistic variables.

Type

Linguistic variable

Linguistic value

Input

Science,

Engineering,

Medicine,

Art

High,

Medium,

or Low

Output

Science_Major,

Engineering _Major,

Medicine _Major,

Art _Major

High,

Medium,

or Low

2. Identify fuzzy sets and their corresponding

membership functions. The antecedent fuzzy sets

represent the overall percentage constraints imposed

by the university. For each college, the student’s

overall percentage is classified into one of the three

fuzzy sets. This definition is used further in the rule

evaluation, in order to ensure that the colleges and

majors are within the student’s range. Figures 1, 2, 3

and 4 present the antecedent fuzzy sets. In these

figures, the x-axis represents the student's overall

percentage values and the y-axis represent the

corresponding membership values. The overall

percentage values (i.e. the x-axis) are driven from Taif

University enrollment data from last year.

Table 2: Antecedent fuzzy sets and ranges.

Medicine and Pharmacy fuzzy sets and their ranges

Fuzzy set

Range

µ=1

Low

[80.00, 88.00]

80.00

Medium

[85.00, 91.00]

87.00

High

[89.00, 100.00]

100.00

Science fuzzy sets and their ranges

Fuzzy set

Range

µ=1

Low

[70.00, 78.00]

70.00

Medium

[75.00, 88.00]

81.00

High

[85.00, 100.00]

100.00

Engineering fuzzy sets and their ranges

Fuzzy set

Range

µ=1

Low

[75.00 , 80.00]

75.00

Medium

[78.00, 90.00]

84.00

High

[88.00, 100.00]

100.00

Art fuzzy sets and their ranges

Fuzzy set

Range

µ=1

Low

[70.00 , 78.00]

70.00

Medium

[75.00, 88.00]

81.00

High

[85.00, 100.00]

100.00

They depend on 1) the average of all student who were

accepted last year and 2) the number of available seats

in each major. These values are uncertain and can

change every year. The ranges provided by the

Fuzzy-Based Recommendation System for University Major Selection

319

Table 3: Consequent fuzzy sets for each applicable major.

Science_Major fuzzy sets

Applicable Majors

µ=1

Range

Fuzzy set

Biology- Zoology – Microbiology- Biotechnology- Food Science - Chemistry – Physics –

Mathematics Accounting

00.00

[00.00, 11.00]

High

Biology- Zoology – Microbiology- Biotechnology- Food Science - Chemistry – Physics –

Mathematics Accounting

15.00

[10.00, 21.00]

Medium

Biology- Zoology –Chemistry – Physics –Mathematics

30.00

[20.00, 30.00]

Low

Engineering _Major fuzzy sets

Applicable Majors

µ=1

Range

Fuzzy set

Computer Science – Information Technology –Interior Design – Industrial Engineering –

Architectural Engineering – Computer Engineering – Radiology –Nursing – Physical

Therapy – Laboratories

30.00

[30.00, 41.00]

High

Computer Science – Information Technology –Interior Design – Industrial Engineering –

Architectural Engineering – Computer Engineering – Radiology –Nursing Laboratories

45.50

[40.00, 51.00]

Medium

Computer Science – Information Technology

60.00

[50.00, 60.00]

Low

Medicine Major fuzzy sets

Applicable Majors

µ=1

Range

Fuzzy set

Medicine - Pharmacy

30.00

[60.00, 71.00]

High

None

45.50

[70.00, 81.00]

Medium

None

60.00

[80.00, 90.00]

Low

Art Major fuzzy sets

Applicable Majors

µ=1

Range

Fuzzy set

All art majors

90.00

[90.00,

101.00]

High

Marketing- Management – English- Arabic- Media and communication science- Early

childhood – Sports – Graphical Design – Fabric Design and Fashion - Shari’a - alqara'at –

Art –Psychology – Economics -Systems - Sciences of Quran - Islamic Culture

105.00

[100.00,

111.00]

Medium

English

120.00

[110.00,

120.00]

Low

university represents the medium fuzzy set, where any

value above that range is considered high and any

value below is considered low.

Table 2 summaries all the antecedent fuzzy sets.

It consists of the fuzzy sets and their corresponding

ranges. The range consists of the average values

starting from 70%, i.e. the lower bound for

enrollment in TU, is 100%.

The consequent fuzzy sets will specify a range for

each value (Medicine, Engineering and Computers

and Information Technology, Science, Art). The

ranges determine the applicable majors for the

student. Generally, the ranges in the consequent part

start from 0 to 120. Figures 5, 6, 7 and 8 present the

consequent fuzzy sets. Values overlapping is

minimized to help in the elimination process to

provide the student with the available majors only.

Table 3 displays the consequent fuzzy sets. It

illustrates the applicable majors for each fuzzy set.

The system returns a value that is used to determine

the set of applicable majors for the student based on

her/his overall percentage. For example, in

Engineering_Major fuzzy sets if the value of the

consequent is 35, then the applicable majors for the

student is only computer science and information

technology. This value is used to retrieve the

available majors for the student from the database,

which stores the fuzzy sets ranges based on the

university majors. A triangular membership function

is used, as shown in equation (4), due to its suitability

of the overall percentage value. This function has four

parameters: 1) average value (AVG), 2) Lower

Bound (LB), 3) Membership Function (MP), and 4)

Upper Bound (UB) (Jang J. S., 1997).

Figure 1: Medicine fuzzy set.

Figure 2: Engineering fuzzy set.

FCTA 2019 - 11th International Conference on Fuzzy Computation Theory and Applications

320

Figure 3: Science fuzzy set.

Figure 4: Art fuzzy set.

Figure 5: Medicine_Major fuzzy sets.

Figure 6: Science_Major fuzzy sets.

Figure 7: Engineering_Major fuzzy sets.

Figure 8: Art_Major fuzzy set.



















 

 







 



 







 



(4)

3 INFERENCE MECHANISM

Table 4: System Knowledgebase.

Fuzzy Rules

IF Science IS

Low

THEN Science_Major

Low

Medium

High

Engineering

Low

THEN

Engineering_Major

Low

Medium

High

IF Medicine

Low

THEN

Medicine_Major IS

Low

Medium

High

IF Art IS

Low

THEN Art_Major IS

Low

Medium

High

Table 4 presents the knowledge base of the system as

IF-THEN rules. Rule evaluation executes the rules

based on the student’s input. Antecedent value is used

in the evaluation of the consequent part. Rule

aggregation combines all the fuzzy sets that resulted

from firing the rules in the last stage. The combined

fuzzy sets are in the same universe of discourse. This

combination is used as an input for the defuzzifier.

4 DEFUZZIFICATION

The algorithm used to defuzzify the values is the

center of gravity algorithm (COA), as shown in

equation (5). Centre-of-Area algorithm is commonly

used in the defuzzification process and there are many

studies that have used it such as (Maranate, 2014),

(X.Y. Djam, 2011), and (Enes Erkan, 2016). The

COA, as shown in equation 6, is defined for a finite

universe of discourse (Jang J.-S. R., 1997). The x is

the value in the universe of discourse, and μ(x) is the

corresponding membership value.























(5)

4.1 Cluster-based Preferences

The aim of this step is to cluster preferred majors into

groups, where majors in the same group are more

Fuzzy-Based Recommendation System for University Major Selection

321

similar than the majors in different groups. This

arrangement can help the student to 1) choose set

majors that they prefer and 2) give the student a

chance to see other majors in case the student's

average does not qualify to enter the major. The

cluster-based preferences use distance

measurements to calculate the similarity. To do so,

the following steps are applied:

 Majors are divided into regions, based on the

content similarity or the work field. Each region

has a number of keywords and associated to one

or more question, as shown in Table 5. Regions

and questions are formed and validated by the

university’s faculty members of each major.

 Student’s answers are recorded through an

online questionnaire to define their preferences.

This process aims to measure the student's

interest in each region on a scale from one to

five, where five is the maximum score. If the

Group has more than one question, the average

score is calculated based on the student's

answers.

 The difference between the maximum score and

the user score of a current question is

calculated, as shown in equation (7).















(6)





 

(7)

In equation (6), N is the is the number of questions

for the region. Note the minimum distance is the

closest to the preferred region.

 Record majors with the minimum distance

region score, i.e. preferred majors.

4.2 Elimination Process

The aim is to ensure that the suggested majors are

align with the student's overall percentage, with

respect to the university constraints, while satisfying

her/hid preferences. Thus, the final majors are

computed based on the stored values. Where, for

each student, the suggested majors are in the

intersection area between two sets: 1) the applicable

majors and 2) the preferred majors. i.e., Final Majors

= Applicable Majors ∩ Preferred Majors

5 SYSTEM EVALUATION

This stage is the first software development stage.

Hence, a pre-alpha version of the system was

released to test for the system accessibility in a high

school in Taif, Saudi Arabia. The sample consisted of

twelve high school students. Results showed that 66

% were strongly pleased with the system and 54%

were pleased with suggestions provided by the

system as shown in figures 9. A positive feedback

was received from the students, where each student

was led to a suitable major that fell in line with their

preferences. Accordingly, the system can help in

increasing the student's satisfaction by giving each

student the chance to succeed in the suggested major.

Figure 9: High School Students’ Feedback.

6 CONCLUSIONS

This paper focused on designing a Fuzzy

Recommendation System (FRS) that aided in

students’ decision in choosing their university major.

The more satisfied are the students about their majors,

the more productive they can be. However,

measuring student's preferences and how it can relate

specifically to the student's interests can raise the

question of system accuracy. This problem can be

solved by developing a detailed scale, with the help

experts, to measure the student's preferences.

In future work, a comparison to existing methods

must be conducted. Also, the system’s accuracy and

performance must be tested.

0.00%

20.00%

40.00%

60.00%

80.00%

100.00%

Are you happy with the system overall?

Are you satisfied with the results ?

FCTA 2019 - 11th International Conference on Fuzzy Computation Theory and Applications

322

Table 5: Preferences regions based on similarities.

The Science Path Regions

Region

Majors

Similarity

Keywords

One

Medicine

Content

Anatomy

Two

Medicine – Microbiology

Content

Immune System – Diseases

Three

Medicine –Nursing – Physical Therapy

Working field

Medical Care

Four

Chemistry – Food Science and Nutrition

Content

Food Science

Five

Pharmacy – Biotechnology – Chemistry

Content

Formulation – Drugs

Six

Radiology – Physics

Content

Radiation

Seven

Mathematics

Calculus – Numbers

Eight

Accounting

Finances – management

Nine

Interior Design – Architectural Engineering

Content

Design –Building (Interior /Exterior)

Ten

Computer Science – Information Technology – Computer

Engineering

Content / Work

filed

Programming/computers

Eleven

Laboratory – Microbiology –Chemistry

Content

Labs – substance

Twelve

Biology – Microbiology -Zoology

Content

Living Organisms

Thirteen

Physics

Natural Laws

Fourteen

Industrial engineering

---

Assembly, Numbers

The Art Path Regions

Region

Majors

Similarity

Keywords

One

English – Arabic

Languages

Two

Arabic – Al-Shari'a

Content

Arabic

Three

Al-Shari'a - Systems - Islamic Culture

Content

Religion

Four

Al-qara'at , Sciences of Quran

Content

Al-Quran

Five

Al-Shari'a – Laws

Content

Islamic Law

Six

Economic and Finance-Management-Marketing- Management

Information System

Content

Administration

Seven

Graphical Design

Computer -Design

Eight

Fabric Design and Fashion

Design – Fashion

Nine

Media and Communication Science

Media

Ten

Art

Eleven

Sports

Sport

Twelve

Psychology – Early childhood

Work field

Behaviour – psychology

REFERENCES

Ahmar, M. A. (2012). A Prototype Rule-based Expert

System with an Object-Oriented Database for

University Undergraduate Major Selection.

Ali, M. S. (n.d.). A Neuro-Fuzzy Inference System For

Student Modeling In Web-based tutoring system.

Amer Al-Badarenah, J. A. (2016). An Automated

Recommender System for Course Selection.

International Journal of Advanced Computer Science

and Applications.

Azene Zenebe, A. F. (2009). Representation, similarity

measures and aggregation methods using fuzzy sets for

content-based recommender systems. Fuzzy Sets and

Systems, pp. Volume 160, Issue 1, Pages 76-94.

Burke, R. (2000). Knowledge-based recommender systems.

Encyclopedia of Library and Information Science, 175-

186.

Desi Purwanti Kusumaningrum, N. A. ( 2017, April).

Recommendation System for Major University

Determination Based on Student’s Profile and Interest.

Journal of Applied Intelligent System (e-ISSN: 2502-

9401 | p-ISSN: 2503-0493), pp. Vol. 2 No. 1, pp. 21 –

28.

Elham S.Khorasani, Z. Z. (2016). A Markov Chain

Collaborative Filtering Model for Course Enrollment

Recommendations. IEEE International Conference on

Big Data (Big Data), pp. 3484 - 3490.

Enes Erkan, Ç. T. (2016). A Fuzzy Expert System for Risk

Self-Assessment of Chronic Diseases. IOSR Journal of

Computer Engineering (IOSR-JCE), pp. PP 29- 33.

Francesco Ricci, L. R. (2011). Introduction to

Recommender Systems. New York: Springer

Science+Business Media, LLC.

Holland, J. H. (1992). Adaptation in natural and artificial

systems: an introductory analysis with applications to

biology, control, and artificial intelligence. MIT press.

Jang, J. S. (1997). Neuro-fuzzy and soft computing-a

computational approach to learning and machine

intelligence. IEEE Transactions on automatic control,

pp. 42(10), 1482-1484.

Jang, J.-S. R. (1997). Neuro-fuzzy and soft computing: a

computational approach to learning and Machine

learning. Ann Arbor: Prentice Hall, 1997.

Fuzzy-Based Recommendation System for University Major Selection

323

Lehr, B. W. (1990, September). 30 years of adaptive neural

networks: perceptron, Madaline, and backpropagation.

in Proceedings of the IEEE, pp. vol. 78, no. 9, pp. 1415-

1442.

Luis M. de Campos, J. M.-L. (2008). A collaborative

recommender system based on probabilistic inference

from fuzzy observations. Fuzzy Sets and Systems, pp.

Volume 159, Issue 12, Pages 1554-1576.

Maranate, T. (2014). Fuzzy FMEA Application for

Variables in FactorAnalysis to Prioritize Suspected

OSA Patients for sleep study. The 3rd International

Conference on Computer Engineering and

Mathematical Sciences (ICCEMS ).

Nachiketa Sahoo, P. V. (2012, December ). A hidden

Markov model for collaborative filtering. MIS Q. 36,

pp. 1329-1356.

Narimel Bendakir, E. A. (2006). Using Association Rules

for Course Recommendation. In Proceedings of the

AAAI Workshop on Educational Data Mining., pp. pp.

31-40.

Ronald R. Yager, L. A. (1992). An Introduction to Fuzzy

Logic Applications in Intelligent Systems. Kluwer

Academic Publishers.

Salaki, C. R. (2015). Decision Support Systems Major

Selection Vocational High School in Using Fuzzy Logic

Android-Based.

Turban, E. e. (1995). Decision support and expert systems:

management support systems. University of Michigan:

Prentice Hall.

X.Y. Djam, G. M. (2011). A Fuzzy Expert System for the

Management of Malaria. International Journal of Pure

and Applied Sciences and Technology, pp. pp. 84-108.

Yera, R. &. (2017). Fuzzy tools in recommender systems: A

survey. International Journal of Computational

Intelligence Systems, 10(1), 776-803.

FCTA 2019 - 11th International Conference on Fuzzy Computation Theory and Applications

324