Internati
o
nal
Journal of Ele
c
trical
and Computer
Engineering
(IJE
CE)
V
o
l.
8, N
o
. 1
,
Febr
u
a
r
y
201
8,
pp
. 29
9
~
30
3
I
S
SN
: 208
8-8
7
0
8
,
D
O
I
:
10.115
91
/ij
ece.v8
i
1.p
p29
9-3
03
2
99
Jo
urn
a
l
h
o
me
pa
ge
: h
ttp
://iaesco
re.com/jo
urn
a
l
s/ind
ex.php
/IJECE
Sukuk Rating Prediction usin
g Voting Ensemble Strategy
Mira Kar
t
iw
i
1
,
Ted
d
y
Su
rya G
u
na
wan
2
, T
i
ka
Arun
dina
3
, M
o
h
d
Az
mi Omar
4
1
Departem
ent
of
Inform
ation
S
y
s
t
em
s, Int
e
rnat
ion
a
l Isl
a
m
i
c Univ
e
r
sit
y
Mal
a
y
s
ia
,
Mala
y
s
ia
2
Departem
ent
of
El
ectr
i
c
a
l
Engin
eering
,
In
tern
ati
onal Is
l
a
m
i
c Uni
v
ers
i
t
y
M
a
la
ys
ia
, M
a
l
a
y
s
ia
3
Faculty
of Econ
omy
and
Business,
Universitas
I
ndonesia, Indon
esia
4
Interna
tiona
l C
e
ntre for
Edu
cat
io
n in Isl
a
m
i
c Fin
a
nce,
Mal
a
y
s
ia
,
Mala
y
s
ia
Article Info
A
B
STRAC
T
Article histo
r
y:
Received J
u
l
3, 2017
Rev
i
sed
No
v 8, 201
7
Accepted Nov 22, 2017
Islamic fin
a
nce
development h
a
s
grown in
to a fo
cal po
int
in man
y
countries
accros th
e globe
. Sukuk, in parti
c
ular,
an Islam
i
c i
nvestm
e
nt product tha
t
has
rece
ived growi
ng atten
tion fr
om
sovereigns, m
u
ltinationa
l and nation
a
l
organizations fr
om both develo
ped a
nd
emerging economies.
I
t
s uses has
been aimed to finance
investmen
t
s in
a v
a
rieties
of econom
ic activities
an
d
development pr
ojects. Despite
the prom
ising lo
ok of Sukuk, cu
rrently
ther
e
is lack of studies had been
to examin
e and predict the rating of th
e Sukuk. As
a result, m
a
n
y
p
r
act
ition
e
rs adopted the conv
enti
onal bond hence ignore th
e
fact th
at th
ese t
w
o instrum
e
nts
are diff
erent in
nature
. In order
to fill th
e gap
in the lit
era
t
ure
,
it is the aim
of
this
research to
develop an ensemble model
that
can be us
ed
to predict Suku
k ra
ting
.
Th
e eff
ectiveness of th
e proposed
models were evaluated using data
set on Sukuk issuance for domestic from
2006 to 2016
.
The r
e
sults ind
i
cate
th
at th
e o
v
erall performance of
th
e
ensemble model is fall short behind the
inductio
n
decision tree (IDT) model.
However, the
cla
s
s
precis
i
on of the ens
e
mble model improved, p
a
r
ticularly
in
predicting
the lo
west rating of
Sukuk.
Keyword:
Ensem
b
le
Isl
a
m
i
c bo
nd
Pred
ictio
n
Ratin
g
Sukuk
Copyright ©
201
8 Institut
e
o
f
Ad
vanced
Engin
eer
ing and S
c
i
e
nce.
All rights re
se
rve
d
.
Co
rresp
ond
i
ng
Autho
r
:
Mira Kartiwi,
Depa
rt
m
e
nt
of
In
fo
rm
at
i
on Sy
st
em
s,
In
tern
ation
a
l Isla
m
i
c Un
iv
ersity Malaysia,
Jalan
G
o
m
b
ak
,
53
100
K
u
ala Lu
m
p
u
r
,
Malaysia.
Em
a
il: mira@iiu
m
.
ed
u
.
m
y
1.
INTRODUCTION
The de
vel
o
pm
ent
of t
h
e S
u
k
uk m
a
rket
has t
r
i
gge
red t
h
e i
ssue o
f
Su
k
uk
rat
i
n
g
.
As i
n
di
cat
ed i
n
t
h
e
p
r
ev
iou
s
st
u
d
i
es, rating
is essen
tial for th
e corpo
r
ation
t
h
at
i
ssue S
u
ku
k as
wel
l
as f
o
r i
n
v
e
st
ors
.
Thi
s
i
s
due
t
o
t
h
e co
nt
ri
b
u
t
i
o
n o
f
rat
i
ng i
n
pr
o
v
i
d
i
n
g a
ge
neral
pi
ct
ure
o
f
cre
d
i
t
w
o
rt
hi
ness
of a
pa
rt
i
c
ul
ar S
u
ku
k.
H
a
vi
n
g
a
go
o
d
rat
i
ng t
e
n
d
s t
o
en
ha
nce t
h
e
dem
a
nd o
f
t
h
ese i
n
st
r
u
m
e
nt
s. The
rat
i
n
g
not
o
n
l
y
refl
ec
t
s
ri
sk a
n
d e
x
p
ect
ed
per
f
o
r
m
a
nce o
f
t
h
e
su
k
u
k
,
b
u
t
al
so
be
nefi
c
i
al
and as
si
st
t
h
e i
n
vest
or s
p
eci
fi
cal
l
y
bank
s w
h
i
c
h i
nve
st
i
n
t
h
a
t
p
a
rticu
l
ar security in
m
easu
r
in
g cap
ital ch
arg
e
for th
is inv
e
st
m
e
n
t
.
Howe
ver, currently there
has
bee
n
a c
h
allenge
for
th
e com
p
an
y to
reg
u
larly u
p
d
a
te and
assess the
rat
i
ng
of t
h
e
Su
ku
k t
h
r
o
ug
h
rat
i
ng age
n
ci
es. Thi
s
i
s
du
e t
o
hi
gh c
o
st
associ
at
ed wi
t
h
per
f
o
r
m
i
ng credi
t
assessm
en
ts, hen
ce im
p
e
lled
th
e n
ecessity to
acq
u
i
re a m
o
d
e
l fo
r Su
kuk
ratin
g pred
icti
o
n
. Su
ch
h
i
gh
co
st is
cont
ri
b
u
t
e
d
by
t
h
e l
a
r
g
e am
ou
nt
o
f
t
i
m
e and
hum
an res
o
urc
e
s t
o
c
o
nd
uct
c
o
m
p
rehe
nsi
v
e
anal
y
s
i
s
o
n
t
h
e
ri
s
k
statu
s
of
th
e co
m
p
an
y b
a
sed on
var
i
ou
s asp
ects need
ed
by th
e rating
agen
cies. Th
is
h
i
g
h
ligh
t
h
o
w
valu
ab
le
Su
ku
k rat
i
n
g p
r
edi
c
t
i
o
n
f
o
r
i
n
vest
m
e
nt
m
a
rket
.
Dat
a
m
i
ni
ng h
a
s bec
o
m
e
an i
n
creasi
ngl
y
i
m
po
rt
ant
c
o
m
ponent
i
n
fi
nanci
a
l
sect
ors.
The
num
ber a
n
d
vari
et
y
o
f
a
p
pl
i
cat
i
ons
has
be
en
gr
o
w
i
n
g ra
pi
dl
y
i
n
t
h
e l
a
st
deca
de, a
n
d
i
t
i
s
pre
d
i
c
t
e
d
t
o
co
nt
i
n
ue t
o
gr
o
w
.
Recently, the
use
of e
n
sem
b
le m
e
thods
has bee
n
t
h
e
hi
ghlight in the
trade
a
n
d
ac
adem
ic literature
in
m
a
nufact
uri
ng
and
heal
t
h
sect
ors
.
H
o
we
ve
r,
t
h
e spee
d o
f
i
t
s
use ha
d bee
n
l
acki
n
g o
f
fi
na
n
c
i
a
l
and eco
no
m
i
cs
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
n
t J Elec & C
o
m
p
Eng
,
Vo
l.
8
,
No
.
1
,
Feb
r
uar
y
201
8 :
2
99 –
30
3
30
0
dom
ai
n i
n
gen
e
ral
,
l
e
t
al
one I
s
l
a
m
i
c fi
nance.
There
f
o
r
e,
it is th
e ai
m
o
f
th
is p
a
p
e
r to
im
p
r
ov
e prev
iou
s
stu
d
y
on Su
k
uk rat
i
ng p
r
edi
c
t
i
o
n by
i
n
co
r
p
o
r
at
i
ng va
ri
o
u
s
S
u
ku
k
st
r
u
ct
u
r
es
t
h
r
o
u
g
h
t
h
e use of dat
a
m
i
ni
ng
t
echni
q
u
e,
p
a
rt
i
c
ul
arl
y
ensem
b
l
e
m
e
t
hod.
2.
ISSUE
S
I
N
S
U
K
U
K
RATI
NG
An
im
p
o
r
tan
t
co
nsid
eration
i
n
stru
cturing
su
kuk
is
Sh
ariah
app
licab
ility for th
e su
kuk
to
b
e
trad
ed
on
t
h
e e
x
ch
an
ges.
U
n
l
i
k
e c
o
nve
nt
i
o
nal
b
o
nds
, s
u
k
u
k
s
h
oul
d set
a
p
p
r
o
v
al
f
r
om
Shari
a
h
boa
rd
as S
h
ari
a
h
com
p
liance
m
a
rketa
b
le securi
ties before
bei
n
g issue
d
i
n
the m
a
rket. The
Sha
r
iah
board
assesses the
structure
of the tra
n
sacti
o
n and
determ
ine
on its com
p
liance
with Sh
ariah
p
r
i
o
r to
t
h
e lau
n
c
h
of su
ku
k [1
].
R
e
l
a
t
e
d t
o
suk
uk
rat
i
ngs
, rat
i
ng a
g
enci
es
, h
o
we
ve
r, o
n
l
y
g
i
ve an o
p
i
n
i
on
on t
h
e c
r
edi
t
a
s
pect
l
i
nke
d
with
th
e in
strumen
t
s. Sin
ce th
e ratin
g agen
ci
es argue
d t
h
a
t
Shari
a
h
-
com
p
l
i
a
nt
nat
u
re o
f
su
ku
k i
s
neut
ral
from
a cre
d
i
t
per
s
pe
ct
i
v
e, t
h
e
rat
i
n
g assi
gne
d t
o
suk
u
k
doe
s
no
t
im
pl
y
any
co
nfi
r
m
a
t
i
on o
n
Sha
r
i
a
h c
o
m
p
l
i
a
nce.
Accord
ing
to
Mo
od
y’s [2
], t
h
e
ratin
g
only
addresses
the
e
xpecte
d
l
o
ss
of
a
n
investm
e
nt associated wi
th the
pr
om
i
s
e. The k
e
y
subst
a
nce
o
f
eval
uat
i
n
g
t
h
i
s
su
ku
k i
s
t
h
e r
e
t
u
r
n
o
r
pr
ofi
t
s
, t
h
e cas
h fl
ow
pay
m
ent
as wel
l
as
th
e risk
of th
e in
stru
m
e
n
t
s.
The age
n
cies be
lieve that Shariah is an
expe
r
t
opi
ni
o
n
;
he
nc
e t
h
e rat
i
ng ag
enci
es
d
o
no
t co
mm
e
n
t on
Sh
ariah
u
n
l
ess it influen
ces th
e credit
risk
[2
]. Th
i
s
p
o
sitio
n
is co
n
s
isten
t
with
ratin
g
ag
en
cies’ lon
g
-h
el
d
po
sitio
n th
at a ratin
g d
o
e
s
n
o
t
co
nstitu
te a reco
mmen
d
a
tio
n
to
bu
y, sell o
r
h
o
l
d
a
p
a
rticu
l
ar security [3
].
Th
ere is a po
ssib
ility
th
at o
t
h
e
r Sh
ariah
scho
lars ta
k
e
d
i
fferen
t
v
i
ews of oth
e
r Sh
ariah
adv
i
sors with
reg
a
rd to
th
e
co
m
p
lian
ce o
f
th
e su
kuk
.
Nev
e
rt
h
e
less, this fact wo
u
l
d no
t b
e
affect
to
th
e
ob
lig
atio
n
’
s
en
fo
rceab
ility, sin
ce it is d
e
term
in
ed
b
y
th
e
co
mmercial
la
w in
stead
o
f
Isla
m
i
c law. Th
i
s
circu
m
stan
ce also
d
o
e
s no
t influ
e
n
ce ratin
g un
less it affects t
h
e risk of th
e
suku
k.
Nev
e
rth
e
less, th
e transactio
n
liqu
i
d
ity may b
e
affected, as investo
r
ten
t
to
rel
u
ctan
t to
inv
e
st
in
th
e tran
sactio
n
th
at
h
a
s
b
e
en
argu
ed
were n
o
t
co
m
p
lian
t
with
Sha
r
iah [1].
3
.
SU
KU
K V
E
RSU
S
CONV
EN
TION
A
L
BOND
Suk
u
k
an
d
conv
en
tion
a
l bond secu
rities h
a
ve so
m
e
si
milari
ties su
ch
as
fixed
term
m
a
tu
rity, co
up
on
s
and they a
r
e
both t
r
ade
d
i
n
t
h
e sec
o
nda
ry
market. Ta
ri
q
[
4
]
m
e
nt
i
oned t
h
at
S
u
k
u
k
has
a si
m
i
l
a
r fu
nct
i
o
n
wi
t
h
b
ond
s, wh
ich
is to
en
ab
le com
p
an
ies to
raise cap
ital, h
o
wev
e
r in
a Sh
ari
a
h
-
co
m
p
lian
t
fash
ion
,
wh
ilst at th
e
sam
e
ti
me ex
pan
d
i
n
g
th
e investo
r
b
a
se and o
f
feri
n
g
i
n
v
e
st
m
e
n
t
o
p
p
o
rtun
ities for
n
e
w g
r
ou
p
s
. Th
oug
h, to
som
e
ext
e
nt
, t
h
eo
ret
i
cal
l
y
t
h
ere sh
o
u
l
d
be
som
e
di
ffere
nc
es i
n
rat
i
ng m
e
t
h
o
dol
ogi
es
f
o
r
bo
n
d
an
d S
u
k
u
k
because these
two instrum
e
nts are diffe
rent
in nature
. Bonds a
r
e contra
ctual debt obligations where
b
y the
issu
er is con
t
ractu
a
lly o
b
lig
ed to
p
a
y to
bo
ndh
o
l
d
e
rs,
o
n
certain
sp
ecified
dates, in
terest an
d prin
ci
p
a
l.
On the ot
her
hand, accordi
n
g to AAI
OFI Standa
rd no.17 [5
], Sukuk are
certificates of equal val
u
e
t
h
at
rep
r
ese
n
t
a
n
u
n
d
i
v
i
d
ed i
n
t
e
rest
i
n
t
h
e
o
w
ners
hi
p
o
f
an
u
nde
rl
y
i
ng a
sset
,
us
uf
r
u
ct
an
d
servi
ces
or
ass
e
t
s
of
part
i
c
ul
a
r
pr
oj
ect
s o
r
s
p
eci
al
i
nvest
m
e
nt
act
i
v
i
t
y
. The
S
u
ku
k
hol
der
al
so
h
a
s a cl
ai
m
t
o
t
h
e
un
de
rl
y
i
ng
asset
s
as i
s
al
so t
h
e
case wi
t
h
no
r
m
al
convent
i
o
nal
bo
n
d
s sh
o
u
l
d
t
h
e i
ssue
r
defa
ul
t
on pay
m
ent
s
. C
onse
q
uent
l
y
,
Suk
u
k
h
o
l
d
e
rs
are en
titled
to
sh
are i
n
th
e rev
e
nu
es
g
e
n
e
rat
e
d
b
y
th
e Suku
k
assets, as well as to
sh
are
in
th
e
procee
ds of the
realization of t
h
e assets. Sukuk certificat
es a
r
e unique in t
h
e way that the inve
stor
bec
o
mes an
asset ho
ld
er,
hen
ce sh
ou
ld b
e
ar
th
e
r
i
sk
of
its un
d
e
r
l
yin
g
assets. Sukuk
certif
icate h
o
l
d
e
r
s
car
r
y
th
e bu
rden
o
f
t
h
ese uni
que
ri
sks.
Ano
t
h
e
r m
a
j
o
r d
i
fferen
ce
b
e
tween Su
kuk
an
d conv
en
tional b
ond
is in
term
s o
f
th
e inv
e
st
m
e
n
t
risk
s.
Sukuk is
ass
o
c
i
ated with a
risk term
ed as
Shariah c
o
m
p
lian
ce risk,
wh
ich
is essen
tial durin
g
t
h
e st
ru
ct
uring
stage base
d on the availabl
e Isla
m
i
c finance contract
s. Nonetheless, Sukuk ha
s so
m
e
si
mi
lariti
es to
conve
n
tional
bonds
beca
use
they are
st
ructured with physi
cal assets that
g
e
n
e
r
a
te r
e
v
e
nu
e. Th
e
und
er
l
y
in
g
reve
nue from
these assets re
presents
t
h
e
so
u
r
ce o
f
i
n
c
o
m
e
fo
r pay
m
ent
of
pr
ofi
t
s
o
n
t
h
e
Su
ku
k.
Acc
o
r
d
i
ng t
o
A
A
O
I
FI
[5
], Su
kuk
ar
e issu
ed
on
v
a
ri
ous transaction c
ontracts. The
s
e
Sukuk are
Ijara
,
Mura
baha
, Sa
la
m
,
Ist
i
s
na, M
u
da
raba a
nd M
u
shara
k
a, M
u
z
a
ra’a (s
ha
recr
op
pi
n
g
)
, M
u
q
a
sa (i
rri
gat
i
on
) an
d M
u
g
h
a
rasa
(agricu
ltural partn
e
rsh
i
p
)
.
Howev
e
r, th
e last th
ree typ
e
s
are rarely used i
n
the m
a
rket. Those struct
ures will
af
f
ect th
e coupo
n p
a
ym
en
t
meth
od
as
w
e
ll as th
e
r
i
sk
ch
aracter
i
stics. Th
e
d
i
f
f
e
r
e
n
t
n
a
t
u
re o
f
bon
d
s
and
Suk
uk
in
term
s o
f
t
h
eir resp
ectiv
e cred
it risk expo
sure
ca
uses
t
h
e n
eed fo
r diffe
re
nt
ratin
gs
a
sses
s
m
e
nt.
4.
DAT
ASET
A
N
D
METH
O
D
S
4.
1.
Suku
k Iss
uan
ce Datase
t
Th
e t
r
ain
i
ng
sam
p
le is b
a
sed
o
n
th
e rating
of th
e Su
kuk
ann
oun
ced
b
y
sev
e
ral
rating
agen
cies
fro
m
2006 to 2016 for dom
e
stic
market. T
h
e rating in this study partia
lly adopted the categor
ical orde
r gi
ven
by
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
El
ec &
C
o
m
p
En
g
ISS
N
:
2
0
8
8
-
87
08
S
u
k
u
k
Ra
ting
Pred
ictio
n u
s
ing
Vo
ting
En
semb
le S
t
ra
teg
y
(
M
ira
K
a
rtiwi
)
30
1
Moody’s [2], and
categ
orise
d
the
rating i
n
to 4 classs.
The categor
ies a
d
opted in t
h
is
study is
AAA
(as the
b
e
st rating
)
,
AA, A, and
BBB (as th
e lo
west ratin
g
)
. In
o
r
d
e
r
o
b
t
ain
reaso
n
a
b
l
e sam
p
le
size, Suk
u
k
evalu
a
tion
was col
l
ect
ed
base
d o
n
t
h
ei
r
hi
st
ori
cal
rat
i
n
g
.
I
n
ot
he
r
wo
rd
s, t
h
e sa
m
p
l
e
was t
a
ken fr
om
every
rat
i
n
g
ann
o
uncem
ent
or
rat
i
n
g a
ffi
r
m
at
i
on dat
e
.
F
o
r e
x
am
pl
e, S
u
k
u
k
X ha
s b
e
en assi
gne
d
A
A
A
rat
i
n
g i
n
t
h
e fi
rst
i
ssuance o
n
J
une 2
0
0
9
. Su
bse
que
nt
l
y
on
June 2
0
1
0
, r
a
t
i
ng
age
n
cy
ann
o
unce rat
i
n
g
af
fi
rm
at
i
on
(A
AA
).
H
o
w
e
v
e
r, it is
p
o
s
sib
l
e du
e to sev
e
r
a
l f
actor
s, th
e r
a
tin
g
is th
en
d
o
wn
gr
aded
in
Jun
2
011 to
A
A
.
A
s
su
ch
, th
is
study c
o
llect each Sukuk rating announc
e
d as
diffe
rent
sa
m
p
les. In
this study,
320 sam
p
les had be
e
n
col
l
ect
ed t
o
be
use
d
as
dat
a
set
f
o
r t
h
e
pre
d
i
c
t
i
on m
odel
.
4.
2.
Vari
abl
e
Sel
e
c
t
i
o
n
In
t
h
i
s
st
udy
,
t
h
e
vari
a
b
l
e
sel
ect
i
on
was
bas
e
d
on
pre
v
i
o
us
st
u
d
y
co
n
duct
e
d
by
Al
tm
an [
6
]
.
T
h
e m
o
st
com
m
on vari
abl
e
s use
d
i
n
pr
evi
o
us st
u
d
i
e
s whi
c
h im
por
t
a
nt
and rel
e
vant
t
o
Suk
uk rat
i
ng
were i
n
cl
u
d
e
d i
n
th
e bu
ild
i
n
g the pred
iction
m
o
d
e
l. Accord
i
n
g
to
p
r
ev
iou
s
research, liqu
i
dity, p
r
o
f
itab
ility an
d lev
e
rag
e
ratio
s
are also consta
ntly use
d
and
considere
d
t
o
be s
o
m
e
i
m
p
o
r
tan
t
ind
i
cators fo
r
bo
nd
ratin
g pred
iction
.
So
m
e
stu
d
i
es con
s
ider qu
alitativ
e v
a
riab
le as add
itio
n
a
l in
fo
rmatio
n
o
f
th
e co
m
p
an
y, such
as sub
o
rd
in
atio
n,
gua
ra
nt
ee st
at
u
s
, o
r
t
a
x
bu
r
d
e
n
.
Mark
et v
a
riab
l
e
s su
ch
as credit sp
read, sto
c
k p
r
ice vo
latilit
y, o
r
GDP are
rarely u
s
ed
b
y
b
ond
rating
pre
v
i
o
us st
u
d
i
e
s. H
o
we
ve
r,
N
i
kl
i
s
, Do
um
po
s, an
d Z
o
p
o
uni
di
s [
7
]
and
Ha
j
e
k an
d M
i
chal
ak [
8
]
bel
i
e
ve
t
h
at
a
mark
et v
a
riab
le is an
i
m
p
o
r
tan
t
in
d
i
cator to
cap
t
u
re th
e situ
atio
n
of th
e co
m
p
an
y o
r
particular securi
ty. As
su
ch
, t
h
e m
a
rk
et v
a
riab
le is inclu
d
e
d
in th
e pred
iction
m
o
d
e
l.
4.
3.
E
n
sembl
e
Mo
del
De
vel
o
p
m
ent
The
use
o
f
e
n
s
e
m
b
l
e
m
e
t
hods
has
bee
n
i
n
cre
a
si
ng
i
n
t
h
e
pa
st
few
y
ears
[
9
]
,
[1
0]
. T
h
e m
a
i
n
o
b
j
ect
i
v
e
of an e
n
sem
b
le
m
e
thod is to
train
m
u
ltiple
base learne
rs
to solve the sa
m
e
problem
[11]. This is different t
o
t
h
e o
r
di
nary
m
odel
devel
o
p
m
ent
app
r
oa
ch
whi
c
h ai
m
e
d t
o
b
u
i
l
d
one
l
earne
r f
r
o
m
t
h
e t
r
ai
ni
n
g
dat
a
set
.
I
n
en
sem
b
le
m
e
th
o
d
, th
e aim
is t
o
con
s
tru
c
t a set o
f
learn
e
r an
d
co
m
b
in
e th
e
m
, h
e
n
ce called
m
u
ltip
le cla
ssifier
system
s [11].
One
of the m
a
in adva
ntage
of usi
n
g en
sem
b
le m
e
thod is
to addre
ss the
weaknesse
s of each
classifier alg
o
rith
m
,
sin
ce th
e g
e
n
e
ralizatio
n
ab
ility o
f
an
en
sem
b
le is o
f
ten
m
u
ch
stron
g
e
r th
an
of b
a
se
lear
n
e
rs [14
]
.
In
t
h
is stud
y, an
ensem
b
le strateg
y
is ad
op
t
e
d
to
con
s
tru
c
t a m
o
d
e
l th
at b
e
st pred
ict the ratin
g
of a
Suk
u
k
.
In
ev
al
u
a
tin
g
th
e m
o
d
e
ls, in
th
is research
th
e au
th
ors do
no
t li
mit
th
e ter
m
‘b
est’ is to
th
e h
i
gh
est
accuracy. Instead, we foc
u
s
e
d on
c
o
nstructing
a
m
odel
th
at could
be
st explain and pre
d
ict the
fe
ature
of
‘BBB’
rating
,
as
th
e
l
o
west ratin
g
with
g
r
eate
r risk
, as co
m
p
ared
t
o
o
t
h
e
r
h
i
gh
er rating
s
.
Fi
gu
re
1 s
h
ows
t
h
e
pr
ocess
o
f
ensem
b
l
e
m
o
d
e
l
devel
opm
ent
u
nde
rt
ake
n
i
n
t
h
i
s
resea
r
ch
.
DA
TAS
E
T
Mo
de
l
A
Mo
de
l
B
Mo
de
l
C
VOT
I
N
G
Pr
e
d
ic
t
i
o
n
En
s
e
m
b
l
e
Mo
de
l
Dev
e
l
o
p
m
en
t
Fi
gu
re
1.
Pr
oce
ss o
f
e
n
sem
b
l
e
m
odel
devel
o
p
m
ent
5
.
R
E
SU
LTS AN
D ANA
LY
SIS
This study adopte
d
three main base learner al
go
ri
t
h
m
s
, nam
e
ly
Ind
u
c
t
i
on Deci
si
on
Tree (
I
DT
),
N
a
ïv
e Bayes,
an
d Ru
le inductio
n
.
Pr
ior
t
o
ad
op
tion
o
f
t
h
e en
sem
b
le str
a
teg
y
, th
e
p
e
rfo
r
m
an
ce of
each
b
a
se
learner algorithm
s
was assesse
d.
If the
r
e are
n
total ratings t
o
be
predicted, t
h
en the ac
cura
cy is stated as:
%
100
#
n
errors
n
Accuracy
(
1
)
Table 1 s
h
ows
the summ
ary
of the
performance accuracy of all base
learners al
gorithms, includi
ng
the ensem
b
le m
e
thod.
As c
a
n be se
en i
n
Table 1,
the
highest pe
rform
a
nce
accura
cy is calculated for
Induction Deci
sion Tree
(IDT
) algo
rith
m
with
9
0
.38
%
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
SN
:
2
088
-87
08
I
n
t J Elec & C
o
m
p
Eng
,
Vo
l.
8
,
No
.
1
,
Feb
r
uar
y
201
8 :
2
99 –
30
3
30
2
Table
1.
Performance Acc
u
ra
cy of Selected
Models
M
ode
l A
c
c
u
r
a
cy
(%)
E
n
sem
b
le 83.
78
I
nduction Decision
T
r
ee
90.
38
Naïve Bayes
68.
06
Rule induction
72.
83
To furt
her exa
m
ine the perform
a
nce of the m
ode
ls, a detail assess
m
e
nt
on the class precision was
u
n
d
e
rtak
en
in
th
is stu
d
y
. Th
i
s
is to
ensure t
h
at th
e m
a
in
pu
rpo
s
e of th
is
stu
d
y
, i.e. a mo
d
e
l t
h
at cou
l
d b
e
st
ex
p
l
ain and
pred
ict th
e
feature of ‘BBB’ rati
n
g
, is ach
i
ev
ed.
Fi
gu
re
2
sh
o
w
s t
h
e
com
p
ari
s
on
of
cl
ass
p
r
eci
si
on
s am
ong
t
h
e
m
odel
s
de
vel
o
pe
d i
n
t
h
i
s
st
udy
.
Despite a l
o
wer pe
rform
a
nce accura
cy of
ensem
b
le
m
o
de
l (see Ta
ble
1) a
s
com
p
are
d
to
IDT m
o
del, this
m
o
d
e
l h
a
s prov
en
t
o
h
a
v
e
t
h
e b
e
st ab
ility to
pred
ict th
e ratin
g
wit
h
th
e
h
i
gh
est risk
,
wh
ich
is BBB in
th
is
st
udy
.
Figure
2. Comparis
on of cl
as
s precision for
each m
odel
In
add
itio
n
,
it
is in
terestin
g
t
o
no
te on
th
e
d
eci
si
o
n
t
r
ee g
e
nerat
e
d f
r
om
IDT al
go
ri
t
h
m
i
n
cl
ude
d i
n
th
is
en
sem
b
le
m
o
d
e
l.
On
e o
f
th
e b
e
n
e
fits
of ad
op
ting
d
ecisio
n
tree in
kn
owledg
e
d
i
scovery is to v
i
su
al
ize th
e
pat
h
t
o
ea
ch
de
ci
si
on,
or
o
f
t
e
n
cal
l
e
d as rul
e
i
n
d
u
ct
i
o
n. F
o
r e
v
ery
pat
h
f
r
o
m
t
h
e r
oot
no
de
of t
h
e t
r
ee t
o
o
n
e o
f
its leaves ca
n
be tra
n
slated into
a ru
le
b
y
co
nn
ecting
t
h
e t
e
sts alon
g th
e
p
a
th to
fo
rm
th
e an
teced
e
n
t
part, and
selecting the
le
af’s
class
pre
d
i
c
tion as
the cla
ss val
u
e.
As
ca
n
be see
n
i
n
Fi
gure
3, the e
x
t
r
acted
rules
for the
BBB ratin
g
s
in
th
is stud
y can
b
e
articu
l
ated
in
to
th
e
ru
le:
“If th
e co
m
p
an
y p
r
od
u
ces i
n
d
u
s
t
r
ial p
r
o
d
u
c
t, an
d
th
e bo
ok
v
a
l
u
e p
e
r sh
are is less th
an
eq
u
a
l t
o
0.367
, and
t
h
e to
tal asset v
a
lu
e is less than
equ
a
l to
578
.6
27
,
th
en
t
h
e rating
o
f
t
h
e su
ku
k is BBB”. Su
ch
set
o
f
ru
les can
t
h
en
b
e
sim
p
li
fi
ed
t
o
im
p
r
ov
e its
com
p
rehe
nsibi
lity to a hum
an use
r
, a
n
d
possibly its accu
rac
y
[12],
[13].
In
d
u
s
t
ri
a
l
Se
c
t
o
r
Bo
o
k
va
l
u
e
pe
r
sh
a
r
e
=
I
n
f
r
a
s
tr
u
c
tu
r
e
an
d
u
t
ilit
i
e
s
To
t
a
l
As
s
e
t
≤
0.3
6
7
BBB
≤
57
8
.
6
2
7
Pr
o
f
i
t
Ma
r
g
i
n
...
To
t
a
l
As
s
e
t
..
.
BB
B
≤
34
48
6
.
8
9
...
≤
1.5
3
7
=
In
d
u
s
t
r
i
al
Pr
o
d
u
c
t
Fi
gu
re
3.
Ext
r
a
c
t
e
d r
u
l
e
s
fr
om
deci
si
o
n
t
r
e
e
m
odel
i
n
cl
ude
d i
n
v
o
t
i
n
g e
n
s
e
m
b
l
e
st
rat
e
gy
Evaluation Warning : The document was created with Spire.PDF for Python.
Int
J
El
ec &
C
o
m
p
En
g
ISS
N
:
2
0
8
8
-
87
08
S
u
k
u
k
Ra
ting
Pred
ictio
n u
s
ing
Vo
ting
En
semb
le S
t
ra
teg
y
(
M
ira
K
a
rtiwi
)
30
3
Si
m
ilarly, an
o
t
h
e
r ru
le th
at can
b
e
ob
serv
ed
in
Fig
u
re 2
is th
e ru
le for th
e co
m
p
an
y in
in
frastru
c
tu
re
an
d u
tilities secto
r
s. Th
e ru
le
can
b
e
articu
l
ated
in
to
: “If t
h
e co
m
p
an
y is in
infrast
ru
ct
u
r
e and
u
tilities secto
r
s,
an
d
t
h
e
p
r
o
f
it
marg
in
is less t
h
an
equ
a
l to
1
.
5
3
7
,
and
th
e t
o
tal asset v
a
lu
e
is less th
an
equ
a
l to
3
448
4.89
, then
th
e rating
of the su
kuk
is BBB”. Hence, th
ese ru
les
h
a
v
e
prov
id
e so
m
e
in
sig
h
t
s for th
e i
n
v
e
sto
r
s and
ratin
g
ag
en
cies in
l
o
ok
ing
in
t
o
the
p
r
o
cess
of
d
e
termin
in
g
th
e Su
ku
k rating
.
6.
CO
NCL
USI
O
N
Thi
s
st
udy
a
d
o
p
t
e
d t
h
e e
n
se
m
b
l
e
st
rat
e
gy
to
pre
d
i
c
t
co
rp
orat
e
Su
k
uk
ra
t
i
ngs.
Fr
om
t
h
e pe
rf
orm
ace
accuracy, the
result show
s that IDT
has
bett
er perform
a
nce as com
p
are
d
t
o
othe
r ba
se le
arne
r algorithm
s
and
en
sem
b
le
m
o
del. Ho
wev
e
r, it is in
terestin
g to
n
o
t
e th
at
when
it
c
o
m
e
s
to the
class precision,
e
n
sem
b
le
m
odel
h
a
s b
e
tter
p
r
edictin
g
ab
ility t
o
id
en
tify th
e ratin
g
with
th
e h
i
gh
est risk, wh
ich
in
th
is stu
d
y
is BBB. Th
ese
find
ing
s
are ex
p
ected
to
enrich
th
e literatu
re and
h
a
v
e
practical i
m
p
lica
tio
n
s
. Th
is m
o
d
e
l is ex
p
ected to
b
e
usef
ul
f
o
r t
h
e
rat
i
n
g
age
n
ci
es t
o
p
e
r
f
o
r
m
a sha
d
o
w
rat
i
ng,
f
o
r i
s
s
u
i
n
g c
o
m
p
ani
e
s and
f
u
n
d
m
a
nagers
t
o
co
ndu
ct th
eir
own
cred
it an
alysis fo
r
risk
m
a
n
a
g
e
m
e
n
t
an
d trad
ing
purpo
ses. In
add
itio
n, th
ese m
o
d
e
ls can
b
e
o
f
u
s
e to b
a
nk
s t
h
at rely
on
ratin
g system
s in
o
r
d
e
r to
im
p
r
ov
e t
h
e risk-assessm
e
n
t
techn
i
qu
es, pricing
st
rat
e
gi
es, a
n
d
pr
o
v
i
s
i
oni
ng
l
e
vel
s
as
re
qui
re
i
n
B
a
sel
I
I
I
.
Em
pirical results are also expected to c
ont
ribute
a wealth
of kn
owledg
e to
th
e d
e
velo
p
m
en
t o
f
Islamic financ
e while enc
o
uragi
ng a
n
alysts and aca
dem
i
c researc
h
ers t
o
de
velop ot
her potential re
search
r
e
lated
to
th
is to
p
i
c.
Fu
t
u
r
e
r
e
sear
ch
is need
ed
to
co
m
p
ar
e th
e asset-b
ack
ed
Suku
k
and
the asset b
a
sed
Sukuk.
In add
itio
n, it i
s
no
ted th
at t
h
e issu
e of
gu
aran
tee statu
s
h
a
s no
t
b
e
en
ex
ten
s
iv
ely ex
p
l
o
r
ed
i
n
th
is study. Y
e
t
,
there are
vari
ous types of guarantee status
or
binding
agree
m
ents in accorda
n
ce with
t
h
e structure
of Sukuk.
Hence
,
f
u
ture
researc
h
is nee
d
ed to
fu
rthe
r study
the r
o
le o
f
v
a
riou
s types o
f
th
is
guarantee status so as to
g
i
v
e
a b
e
tter p
i
ctu
r
e
with
regard to
Suku
k cred
it risk pro
f
ile.
ACKNOWLE
DGE
M
ENTS
Th
e research
ers in
th
is stud
y wou
l
d
lik
e to
ack
nowled
g
e
t
h
e In
ternatio
n
a
l Islam
i
c Un
i
v
ersity
Malaysia (IIUM) for the
financial funding of t
h
is re
sea
r
ch through t
h
e
Research
Initiatives Gra
n
t S
c
hem
e
(
R
I
G
S) RIG
S
15
-07
0
-
007
0.
REFERE
NC
ES
[1]
Fitch R
a
tings, "
R
ating
Sukuk Cr
oss-S
ector Criter
ia,"
London Aug
u
st 2011.
[2]
Mood
y
’
s Investo
r
Service, "Shari ah, Sukuk an
d
Credit Risk: A
Mood
y
'
s Primer," London 2010
.
[3]
E. R. Ahmed, M. A. Isla
m, and T. T. Y. Alabdu
llah
,
"Islamic
Sukuk: Pricing Mechanism and Rating,"
Journal of
Asian Scient
ifi
c Res
e
ar
ch
, vol. 4
,
pp. 640, 2014.
[4]
A. A. Tariq, "Managing Finan
c
ial
Risks of Suk
uk Structures," Loughborough
University
, Leicestershire, United
Kingdom, 2004.
[5]
AAOIFI, "Accounting
and Auditing Organization
of
Islamic Finan
c
ial Institutions," 2005.
[6]
E. I. Altm
an, "Financ
i
al ra
tios,
discrim
i
nant an
aly
s
is and the pr
ediction of corpo
r
ate bankrup
tcy
,
"
The journal of
finance
, vol. 23,
pp. 589-609
, 19
68.
[7]
D. Niklis, M. D
oumpos, and C. Zopouni
dis, "C
ombining market and accountin
g
-
based models for credit scorin
g
using a classification scheme b
a
sed on support vector mach
ines,"
Applied Ma
thematics and Computation
, vo
l. 23
4,
pp. 69-81
, 2014
.
[8]
P
.
Hajek and K
.
M
i
cha
l
ak
, "F
eature s
e
lec
tion i
n
corporat
e cr
ed
it rat
i
ng predi
c
t
i
on,"
Knowledge-Based Systems
,
vol. 51
, pp
. 72-8
4
, 2013
.
[9]
A. Rojarath
, W. Songpan, and
C. Pong-inwong,
"Improved ens
e
mble learning f
o
r classification
techniques base
d
on majority voting,
"
in Software Engine
ering and Service
Sc
ienc
e (ICSESS)
, 2016 7th IEEE Internat
ion
a
l
Conference on
,
pp. 107-110
, 20
16.
[10]
Y.
Zhu, C. Xie,
G.
-J. Wa
ng,
and X.-G. Yan
,
"Comparison of indivi
du
al
, ens
e
m
b
le and
int
e
grated
ens
e
m
b
le
machine
learn
i
n
g
methods to pr
edict Chin
a’
s SME cred
it r
i
sk in supply
chain
finance,"
N
e
ural Computing and
Applica
tions
, pp. 1-10, 2016.
[11]
B. Clarke, E. Fokoue, and H. H. Zha
ng, Principles and theor
y
for data
mining and machine learning
, Springer
Science & Busin
e
ss Media, 2009
.
[12]
J. R. Quin
lan,
"Sim
plif
y
i
ng
dec
i
sion trees
,"
International Journ
a
l of Human-Co
mputer Studies
,
vol. 51
, pp. 497
-
510, 1999
.
[13]
D.
Delen,
C.
Kuzey
,
and A.
U
y
ar,
"M
easuring firm performance
us
ing financial r
a
tios: A decisio
n
tree
approach,"
Expert S
y
stems with
Application
s
, vol. 40, pp. 39
70-3983, 2013
.
[14]
Thalor
, M. A., & Patil, S. (2
016). “Incremental Learn
i
ng o
n
Non-stationar
y
Data
Stream using Ensemble
Approach”.
International Journal
of Electrical an
d Co
mputer Eng
i
neering
, 6(4)
, 1
811.
Evaluation Warning : The document was created with Spire.PDF for Python.