fuzzy algorithm for selecting students...
TRANSCRIPT
FUZZY ALGORITHM FOR SELECTING STUDENTS FROM LOW INCOME
FAMILY
1Zamali, T.,
2Ling-Ling, U,
3Nasrah, N. &
4Tammie, S.
1,2,3,4Department of Mathematics, Faculty of Computer and Mathematical Sciences, Universiti Teknologi
MARA, Sabah Branch, Locked Bag 71, 88997 Kota Kinabalu, Sabah, Malaysia [email protected], [email protected] ,
ABSTRACT
This research introduces a new algorithm to select
students from low income family using fuzzy
approach. It focuses on the refinement and
modification of certain variables in selection process.
The technique employs the intersection of fuzzy
goals and constraints concept in judgmental process.
The initial input was directly obtained based on the
multi-person opinion and experiences. A numerical
example was fully utilized to demonstrate the
applicability of the proposed method. It shows that
the proposed approach has successful dealt with the
uncertainty of the input datasets and beneficially for
student’s selection purposes. As a result, the decision
for selection process can be derived successfully in a
simple manner and the proposed algorithm offers a
new dimension technique as well from the tradition
point of views. Finally, our new refinement and
modification of the variables can derive more precise
in terms of representing the actual situation.
KEYWORDS
Fuzzy algorithm, fuzzy goals and constraints,
mengubah destini anak bangsa (MDAB).
1.0 INTRODUCTION
Admission to higher education system in
Malaysia mainly is based on the applicants’
academic performance or score. This has caused
disadvantages to students from low income
family, as they have limited access to better
education assistance such as tuition classes, extra
courseware and internet references, compared to
their peers from better wealth of family.
Therefore, this group of students is unable to
perform as good as their friends too. Poverty is
part of the explanation. Numerous studies have
been made; found that most rural community is
unable to provide well-education assistance to
their children [1]. If this continues on, the
country will face a group of people whose socio
economy will not be improved for many years.
According to the statistic, Sabah has the
highest poverty rate compared to the other states
in Malaysia [2-5]. In fact Sabah has lower score
in big examinations such as UPSR, PMR, SPM
or STPM. Due to this problem, a greater number
of the younger generations fail to pursue their
studies to higher level.
An initiative called Mengubah Destini
Anak Bangsa (MDAB) is designed to cater the so
called disadvantage group of student, by
considering their family background as one of
the criteria in the hunt for a seat into UiTM. A
group of researchers from UiTM Sabah Branch
has studied numerous algorithms to be
implemented in selecting students into the
programme, and decided to further investigate on
fuzzy algorithm in the selection process.
Fuzzy algorithm is chosen mainly due to
its robustness in dealing with ill-defined or
complex problems. Fuzzy logic is a sub
component of Artificial Intelligence field, has
tremendously evolved over the years.
Researchers from different fields, namely from
the manufacturing, health, automotive and
management see the importance of fuzzy
algorithm in many problem solving. Fuzzy
algorithm is used in pattern clustering [6-8],
optimization of solution [9, 10] and estimation
[11, 12].
Based on the literature, existing research
very seldom explores the intersection fuzzy goal,
the constraints approach and utilizes as an
evaluation tools. Moreover, since the nature of
the criteria or attributes evaluated is uncertainty
and the lack of information, this proposed
approach is believed to be more efficient in daily
evaluation procedures. Thus, the objectives of
ISBN: 978-0-9853483-9-7 ©2013 SDIWC 8
this research are; i) to propose the intersection
fuzzy goal and the constraints concept for
MDAB students selection process, and ii) to
design and developed the user friendly fuzzy
algorithm for MDAB students selection
purposes. To do so, the structure of this paper as
follows; Section 2 provides the problem
identification based on the real situation; Section
3 briefly describe the background of the
proposed method; Section 4 focuses on the
applicability of the proposed method using
numerical example, and finally the brief
discussion and conclusions were pointed out.
2.0 PROBLEM IDENTIFICATION
Mengubah Destini Anak Bangsa (MDAB) is a
programme that introduced by UiTM to give the
opportunity to Malay and Bumiputera youths
from low income family to further their studies
in pre-Diploma level (i.e., Pre-
Diploma(Commerce) and Pre-Diploma(Science))
offered by UiTM. After more than 3 years
introduced, this programme gets a lot of positive
responses but then due to the drastic phenomena
have causes the admission procedure become
more complicated and took a longer time in
order to process their application. Recent
selection process practiced is inadequate
considering the deeper elements of each criteria
selection concern. For instance, if two applicants
with the similar academic qualification apply for
MDAB programme from difference background
family income, say RM500 and RM1800,
respectively. The existing system failing to
discriminate efficiently which applicant should
be given priority. This problem occurs in other
criteria as well such as the number of credit
obtained, number of sibling, etc. Actually, this
discriminate element becoming more significant
in most cases due to UiTM Sabah has faced
financial constraints as well as limited space
available to be offered. Therefore, based on this
phenomenon and the inadequate comprehensive
procedures existing, we proposed the new
algorithm that believed can dealing with more
efficient to discriminate of above lacking.
Moreover, the proposed method is ease and more
precise to measures all criteria situation in
selection process. In addition, we not solely
proposed a new approach but in the same time
develop user friendly algorithm for UiTM
management users. As a result, the both
proposed method can improve the existing
selection process and beneficial to UiTM,
particularly for MDAB student’s selection
decision.
3.0 METHODOLOGY AND THE
PROPOSED ALGORITHM
3.1 The Background Theory and
Methodology
In this section we discuss briefly the similar
method which has been proposed by [13].
Consider a simple decision-making model
consisting of a goal described by a fuzzy set G
with membership function µG(x). A constraint
described by a fuzzy set K with membership
function µK(x) where x is an element of the crisp
set of alternatives Lalt.. Hence, the decision is a
fuzzy set D with membership function µD(x),
expressed as intersection of G and K.
D = G K = {(x, µD(x)/x [d1,d2], µD(x) [0,
h ≤ 1]}
(1)
Where [d1,d2] is the crisp set of selection from
the set of alternatives (Lalt) µD(x) is the degree to
which any x [d1,d2] belongs to the decision D
Here, the operation intersection of A and B
denoted as A B is defined by
µA B(x) = min(µA(x), µB(x)), x U;
(2)
if µA(x)= a1 < a2 = µB(x), min(a1,a2) = a1
Using the membership functions and operation
intersection –(2), formula –(1) gives
µD(x) = min(µG(x), µK(x)), x Lalt
(3)
Hence, the goal and constraint in –(1) can be
formally interchanged as follows:
D = G K = K D
(4)
To obtain [d1,d2] with the highest degree of
membership in the set D, the maximization
decision is expressed by
Xmax = {x/max µD(x)= max min(µG(x), µK(x))}
(5)
ISBN: 978-0-9853483-9-7 ©2013 SDIWC 9
Thus, formula –(1), -(3) & -(4) have been
generalized with many goals and constraints.
For goals Gi, i = 1,2,3, …,n, and constraints Kj, j
= 1,2,3, …,m, the decision is given by
D = G1 G2 G3 … Gn K1 K2 K3 ….
Km
(6)
The membership function of D is
µD(x)= min(µG1(x), …, µGm(x), µk1(x), …,
µkm(x))
and the maximization decision is given by
Xmax = {x/µD(x) is max}
(7)
Based on Eq (7), we can identify the best n-
options by descending order. For instance, if we
have n-alternatives,
XD = {x/µD1(x) > x/µD2(x), > x/µD3(x), >, …, >
x/µDn1(x) > x/µDn(x)}
(8)
Where the symbol ‘>’ means ‘is preferred or
superior to’.
Generally, the summarizing of above selection
decision process shown on Figure 1
Goal G
Constraint K
Alternative Aalt
Intersection
G K
Fuzzy decision
D
Maximizing
decision
Xmax
Figure 1: Process of decision-making by intersection
operator
Since the nature of selection process for MDAB
students involved the multiple objectives, here
we construct the membership functions for three
objectives; i) G1, parents gross monthly income
(PI) must less than RM3000, ii) G2c; obtain at
least three credit (CR) in Sijil Pelajaran Malaysia
(SPM) results including Bahasa Malaysia
subject, or G2s; obtain at least three credit (CR)
in SPM results including Bahasa Malaysia and
Mathematics subjects, plus at least passed one of
the pure science subject, and iii) G3, number of
sibling (NS), respectively, given as follows:
x
x
x
xPI 3000;
30001
3000;0
)(~
-(9)
10;1
105;10
53;5
2
)(~
x
xx
xx
xCR
(10)
6;1
63;75.0
31;5.0
)(~
x
x
x
xNS
(11)
Also the committee has a main constraint, the
candidates offered should be Bumiputera and
he/she must provide the complete necessary
documents during their application. Here, we
categorize the completeness of the applicants’
document as shown in Table 1.
Identify the best n-options by
descending order
ISBN: 978-0-9853483-9-7 ©2013 SDIWC 10
Table 1: The four difference definitions of
document status
Membership
values
Description
0
0.3
0.7
If the candidate is Bumiputera but not
provide all necessary documents
(DS1)(i.e., identification card, birth of
certificate, and pay slip/certified monthly
income statement for applicant and their
parent)
If the candidate is Bumiputera but not
provide both relevant documents (DS2)
(i.e., identification card or birth of
certificate and pay slip/certified monthly
income statement for applicant and their
parent)
If the candidate is Bumiputera but not
provide parent’s pay slip/certified
monthly income statement (DS3).
1 If the candidates is Bumiputera with
complete all necessary documents (DS4)
3.2 Our Proposed Algorithm
In this research we design the convenient
algorithm based on five steps given as follows:
i. Design the algorithm based on the
method that we have proposed in sub-
section 3.1.
ii. Illustrate and synthesis the appropriate
flow-chart thoroughly as respect to above
algorithm
iii. Design the step-by-step procedures base
on the objectives and constraints that we
have proposed earlier.
iv. Testing the proposed algorithm that we
have designed the complete procedures
using dummy variables and hypothetical
example
Thus, the next section (Section 4) we will show
the comprehensive numerical example for
application purposes.
4.0 A NUMERICAL EXAMPLE
To show our propose algorithm has applicable
and suitable for the issues concern, here we
provide an numerical example with some
modification and definition refinement from
[13].
Every semester MDAB committee of
UiTM Sabah received more than 1000
applications from low income family candidates
especially from the rural area a cross state of
Sabah. Since UiTM Sabah has faced financial
constraint and limited space available to be
offered, the committee make an initial screening
and short listed for qualified candidates (i.e., A1,
A2, A3, …,An) for Pre-Diploma (Commerce)
programmes. UiTM Sabah has three specific
objectives (goals) which the candidates have to
satisfy: i) G1, parents gross monthly income (PI)
must less than RM3000, ii) G2c; obtain at least
three credit (CR) in SPM results including
Bahasa Malaysia subject, and iii) G3, number of
sibling (NS). Also the committee has a main
constraint, the candidates offered should be
Bumiputera and he/she must provide the
complete necessary documents during their
application. Thus, the committee was
constructing the membership function as given in
Eq-(8) – (10), respectively. For the constraint,
the committee also decided to categorize the
status of documents of Bumiputera using three
difference scores (i.e., membership values)
depend on the completeness of documents
provided during the application submission,
given in Table 1. For calculation example
purposes, say five candidates applied as shown
in Table 2. Here, we substitute three objectives
(i.e., G1, G2, G3) from raw datasets in Table 2
using the three memberships function
respectively. Meanwhile, for constraint (K1), we
derive the membership values based on Table 1
definition. Then, we obtain all membership
values as shown in last row (Table 3):
Table 2: The raw information for three objective attributes
and one constraint
Applicants A1 A2 A3 A4 A5
G1: Income (in
RM)
1800 1000 700 2780 550
G2: Number of
credit obtained
3 5 4 3 5
G3: Number of
sibling
3 5 4 7 7
K1: Document
status
DS4 DS1 DS2 DS3 DS4
ISBN: 978-0-9853483-9-7 ©2013 SDIWC 11
Table 3: The membership values (i.e., the scores) derived
from Table 2
Applicants A1 A2 A3 A4 A5
G1: Income (RM) 0.4 0.67 0.77 0.07 0.82
G2: Number of
credit obtained
0.2 0.5 0.4 0.2 0.5
G3: Number of
sibling
0.5 0.75 0.75 1 1
K1: Document
status
1 0 0.3 0.7 1
Based on membership values in Table 3 above,
we can write as follows:
G1 = {(A1, 0.4), (A2, 0.67), (A3, 0.77), (A4, 0.07),
(A5, 0.82)}
G2 = {(A1, 0.2), (A2, 0.5), (A3, 0.4), (A4, 0.2),
(A5, 0.5)}
G3 = {(A1, 0.5), (A2, 0.75), (A3, 0.75), (A4, 1),
(A5, 1)}
And
K1 = {(A1,1), (A2, 0), (A3, 0.3), (A4, 0.7), (A5,1)}
From Eq.-(6), we have
µD(x) = min (µG1(x), …, µGm(x), µk1(x), …,
µkm(x))
= {(A1, 0.2), (A2, 0), (A3, 0.4), (A4, 0.07),
(A5, 0.5)}
And from Eq.-(7) we can obtain as
XD = Maks{(A1, 0.2), (A2, 0), (A3, 0.4), (A4,
0.07), (A5, 0.5)} or
XD = {x/µD1(x) > x/µD2(x), > x/µD3(x), >, …, >
x/µDn1(x) > x/µDn(x)}
From above result it shows that the applicant A5
is the best or the most preferred applicants as
compared to the rest due to highest score of the
membership values. Finally, we can identify the
most five superior options using Eq-(8) as
follows:
XD = {(A5, 0.5) > (A3,04) > (A1, 0.2) > (A4, 0.07)
> (A2, 0)}
Thus we have,
A5 = 0.5 > A3 = 0.4 > A1= 0.2 > A4 = 0.07 > A2
= 0
where the symbol ‘>’ means ‘is preferred or
superior to’.
It is apparent that A5 is the best candidates,
followed by A3, A1, A4 and lastly is A2
candidates.
5.0 DISCUSSION AND CONCLUSIONS
In this research the refinement and modification
from [13] have been made for the following
variables; i) the number of credit (NC)
membership functions obtained by candidates
(see Eq. (9 – 11)) and, ii) the definition of
documents status in Table 1 from three to four
categories. It clearly seen that the new definition
that we have modified are meaningful and more
represented the actual situation. In addition,
previous research did not consider at all the
qualification criteria from Pre-Diploma (Science)
applicants which require a minimum credit in
Mathematics and at least passed one of the pure
science subject (i.e., Physic, Chemistry, Biology
or Sains Tambahan) in SPM level. Also, in this
research we rectify or improvise the second
objective into G2s; obtain at least three credits
(CR) in SPM results including Bahasa Malaysia
and Mathematics subjects plus at least passed
one pure science subjects.
Thus, in this research we have proposed
the intersection of fuzzy goals and constraints
concept in judgmental process. Since the
evaluation generally involve uncertainty, it is
important to incorporate the fuzzy approach to
derive precise results in any proposed method.
From numerical example, it can be clearly seen
that the intersection between fuzzy goal and
constraints is beneficial in terms of evaluation
perspective. Also, we provide some
straightforward procedures by constructing the
relevant membership functions to derive the
membership values in the range of [0, 1], which
is extremely significant in fuzzy environment. In
addition, we also develop the user friendly
algorithm for UiTM management users that can
utilize directly from our proposed method,
particularly for MDAB students’ selection
purposes.
Therefore, based on the algorithm and the
new refinement that we have proposed, it fulfil
the following significant advantages; i) the
algorithm is successfully dealing with the
uncertainty of the initial information/datasets
which was rarely explored previously,
particularly in students selection process, ii) the
ISBN: 978-0-9853483-9-7 ©2013 SDIWC 12
modification of the memberships function of
both parent gross monthly income and number of
credits obtained will representing the actual
situation. As a result, it gives an alternative
judgment and allowed the committee to judge
beyond the traditional method using the existing
system available, and iii) it offers more
convenient and confident decision process by
equipping the convenient step-by-step
procedures so that the users can utilize and
applies directly the concept from our proposed
algorithm.
6.0 ACKNOWLEDGEMENTS
The authors acknowledge University Teknologi
MARA (Sabah Branch) and Research
Management Institute (RMI), UiTM for
supporting this research by Research Excellent
Grant (Grant No. 500-
UiTMKSH(PJI/UPP.5/1)(1/2013).
7.0 REFERENCES
1. Rashid, S.M.R.A., Peluang dan cabaran pendidikan
dalam kalangan masyarakat luar bandar: satu kajian
kes isi rumah melayu miskin di jajahan Bachok,
Kelantan, in Pusat Pengajian Ilmu Kemanusiaan2008,
Universiti Sains Malaysia.
2. Khoo, B.T., Policy Regimes and the Political Economy
of Poverty Reduction in Malaysia2012: Palgrave
Macmillan.
3. Siwar, C., et al., A review of the linkages between
climate change, agricultural sustainability and poverty
in Malaysia. Inter. Rev. Bus. Res. Papers, 2009. 5(6):
p. 309-321.
4. Shaffril, H.A.M., et al., Agriculture project as an
economic development tool to boost socio-economic
level of the poor community: The case of Agropolitan
project in Malaysia. Afr. J. Bus. Manage, 2010. 4(11):
p. 2354-2361.
5. Woo, W.T., Understanding the Middle-Income Trap in
Economic Development: The Case of Malaysia. World
Economy Lecture delivered at the University of
Nottingham, Globalization and Economic Policy, 2011.
6. Xie, X.L. and G. Beni, A validity measure for fuzzy
clustering. Pattern Analysis and Machine Intelligence,
IEEE Transactions on, 1991. 13(8): p. 841-847.
7. Bezdek, J.C., Fuzzy mathematics in pattern
classification. 1973.
8. Wu, H., Q. Chen, and M. Yachida, Face detection from
color images using a fuzzy pattern matching method.
Pattern Analysis and Machine Intelligence, IEEE
Transactions on, 1999. 21(6): p. 557-563.
9. Kahraman, C., T. Ertay, and G. Büyüközkan, A fuzzy
optimization model for QFD planning process using
analytic network approach. European Journal of
Operational Research, 2006. 171(2): p. 390-411.
10. Sasikumar, K. and P. Mujumdar, Fuzzy optimization
model for water quality management of a river system.
Journal of water resources planning and management,
1998. 124(2): p. 79-88.
11. Ahmed, M.N., et al., A modified fuzzy c-means
algorithm for bias field estimation and segmentation of
MRI data. Medical Imaging, IEEE Transactions on,
2002. 21(3): p. 193-199.
12. Chiu, S.L., Fuzzy model identification based on cluster
estimation. Journal of intelligent and Fuzzy systems,
1994. 2(3): p. 267-278.
13. Zamali, T., Nasrah, N., Tammie, S. & Ling-Ling Fuzzy
based-approach for selecting MDAB students.
Proceeding in Seminar Kebangsaan Mengubah Destini
Anak Bangsa (SKMDAB’2012), 2012.
ISBN: 978-0-9853483-9-7 ©2013 SDIWC 13