Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
3584
~3
590
IS
S
N: 20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v9
i
5
.
pp3584
-
35
90
3584
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
Perform
ance ev
alu
atio
n
of di
fferen
t classifi
ca
ti
on techni
qu
es
usin
g different d
atasets
Ab
d
ulka
dir Ö
z
demi
r, U
ğur
Yavu
z
, F
ares
Ab
d
ulha
fidh Dael
Depa
rtment
o
f M
ana
gement Inf
orm
at
ion
S
y
s
te
m
s,
Atat
urk
Unive
rsit
y
,
Turkey
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
N
ov
27
, 201
8
Re
vised
A
pr
5
,
201
9
Accepte
d
Apr
14
, 201
9
Now
aday
s
data
m
ini
ng
bec
om
e
one
of
the
techn
ologi
es
that
paly
m
aj
or
eff
ec
t
on
business
int
ellige
n
ce
.
How
ev
e
r,
to
be
ab
le
to
u
se
the
data
m
ining
outc
om
e
the
user
should
go
through
m
an
y
proc
esses
such
as
cl
assifie
d
da
ta.
Cla
ss
ifi
c
at
ion
o
f
data
is
pro
ces
sing
dat
a
and
orga
ni
ze
the
m
in
spe
ci
fi
c
ca
t
egor
ize
to
be
use
in
m
ost
eff
ec
t
ive
and
eff
ic
i
ent
use
.
In
d
at
a
m
ini
ng
on
e
te
chn
ique
is
not
appl
i
ca
bl
e
to
be
appl
i
ed
to
al
l
th
e
dat
ase
ts.
Man
y
dat
a
users
wasting
a
lot
of
ti
m
e
tr
y
ing
m
a
n
y
class
ifi
c
at
ion
te
chni
qu
es
in
orde
r
to
find
the
m
ost
an
ap
propria
t
e
t
ec
hn
i
que
to
be
used
.
Th
is
pap
er
sh
owing
the
diffe
ren
ce
resul
t
of
apply
ing
diffe
re
n
t
t
ec
hn
i
ques
on
th
e
s
ame
data.
Thi
s
pape
r
ev
aluate
s
the
per
for
m
anc
e
of
diffe
r
ent
class
ifi
catio
n
te
chni
qu
es
using
diffe
r
ent
d
at
ase
ts.
In
thi
s
st
ud
y
four
data
class
ifi
cation
te
ch
nique
s
hav
e
chose
n.
They
ar
e
as
foll
ow,
Ba
y
esNet
,
Naiv
eBay
es,
Multi
lay
e
r
per
ce
ptron
and
J48.
The
sel
ec
t
ed
data
class
ifi
cation
techniqu
es
per
form
anc
e
t
este
d
unde
r
two
par
amete
rs,
the
ti
m
e
ta
k
en
to
buil
d
the
m
odel
of
the
da
tas
et
and
th
e
per
ce
n
ta
g
e
of
a
cc
ura
c
y
to
c
la
ss
if
y
th
e
da
ta
set
i
n
the
cor
re
ct
class
ifi
cation
.
The
exp
eri
m
ent
s
are
c
arr
ie
d
ou
t
using
W
eka
3.
8
software
.
Th
e
re
sults
in
the
pape
r
demons
trate
that
the
eff
ici
ency
of
Multi
lay
er
Per
ce
p
tron
cl
assifi
er
in
over
all
the
b
est
ac
cur
acy
per
form
anc
e
to
cl
assif
y
the
insta
nce
s
,
and
Naiv
eBay
e
s
cl
assifi
ers
were
the
wors
t
outc
om
e
of
ac
cur
acy
t
o
cl
assif
y
ing
the i
nstanc
e
for ea
ch dat
ase
t.
Ke
yw
or
d
s
:
Ba
ye
sNet
Cl
assifi
cat
ion
Data m
ining
J4
8
Mult
il
ay
er p
er
ceptr
on
NaiveBay
es
Wek
a
Copyright
©
201
9
Instit
ut
e
o
f Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Fares A
bdulh
a
fid
h Dael
,
Dep
a
rtm
ent o
f M
anag
em
ent Inf
or
m
at
ion
Sys
tem
s,
Atat
urk U
niv
er
sit
y,
Ün
i
ver
sit
e Ma
halle
si, A
ta
tü
r
k Ün
i
ver
sit
esi
Kam
pü
sü,
25030, E
rzur
um
, Tu
r
key
.
Em
a
il
:
far
esal
ariqi@
gm
ai
l.com
1.
INTROD
U
CTION
Data
m
ining
is
way
of
extr
act
us
efu
l
kn
owle
dge
out
of
huge
volum
e
of
data.
T
he
disco
ver
i
ng
knowle
dge
co
m
e
through
m
any
m
ining
to
be
us
e
f
ul
kn
ow
le
dg
e
.
T
o
c
o
nve
rt
the
ra
w
data
to
knowle
dge
the
data
sho
uld
be
inte
gr
at
e
d,
cl
eaned,
cl
assifi
ed
or
cl
us
te
r
ed
an
d
so
on.
To
util
iz
e
the
process
o
f
e
xtr
act
ing
data
m
ining
,
m
any
te
chn
iq
ue
s
an
d
sta
nd
a
r
ds
s
houl
d
be
use
d.
O
ne
of
t
he
process
of
da
ta
m
ining
is
how
t
o
cl
assify
the
da
ta
and
orga
niz
e
them
in
the
correct
cat
eg
ori
es
[1
,
2].
Eve
ry
data
has
it
s
own
c
ha
racter
ist
ic
s,
so
m
e
of
them
no
m
inal
data
and
oth
e
r
are
num
erical
data
and
s
o
f
or
t
h.
Accor
ding
to
the
data
cha
rac
te
risti
cs
the
data
sho
uld
be
cl
assifi
ed
.
Howe
ver,
it
is
no
t
r
eseal
able
to
be
c
hec
k
the
data
co
ntain
li
ne
by
li
ne
to
cl
assify
the
data.
I
n
da
ta
m
ining
the
r
e
are
sever
al
te
chn
i
qu
e
s
to
be
us
ed
to
cl
as
sify
the
data.
Howe
ver
not
al
l
the
te
chn
iq
ues
ca
n
cl
assify
cor
rec
tl
y
the
data
give
the
sam
e
res
ult
or
ou
tc
om
e
of
the
data.
E
ve
ry
te
chn
i
qu
e
has
it
s
own
m
od
el
a
nd alg
ori
thm
an
d
the
w
ay
of
how
to
cl
assi
fy the
data [
3
-
6]
.
To
fin
d
ou
t
the
dif
fer
e
nces
of
the
cl
assif
ic
at
ion
te
chn
i
ques
an
d
the
r
easo
ns
of
the
diff
e
ren
ce
s,
four
cl
assifi
ca
ti
on
te
ch
nique
s
hav
e
sel
ect
ed.
T
hey
are
as
fo
ll
ow
Ba
ye
sNet,
Naiv
eB
ay
es,
Mult
i
la
ye
r
per
ce
ptr
on
a
nd
J48.
T
o
te
st
the
diff
e
re
nc
es
of
the
four
sel
ect
ed
cl
assifi
cat
ion
te
ch
ni
qu
es
,
three
diff
e
ren
t
dataset
s
hav
e
colle
ct
ed,
C
on
gr
essi
onal
voti
ng
rec
ords
,
Ca
r
e
valuati
on
a
nd
C
on
t
raceptiv
e
m
et
ho
d
c
ho
i
ce.
T
o
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Perf
orma
nce e
valu
ation of
d
if
fe
rent classif
ic
ation t
ech
niqu
es
…
(
Fares
A
bdulhafi
dh
Dael
)
3585
m
easur
e
the
ef
fecti
ven
es
s
of
t
he
te
chn
i
qu
e
s
two
par
am
et
ers
hav
e
te
ste
d,
th
e
tim
e
ta
kin
g
of
each
te
chn
i
que
to
bu
il
d
the
m
odel
fo
r
the
sel
ect
ed
dataset
,
and
the
accu
racy
of
cl
assify
ing
the
data
by
each
te
chn
iq
ue
.
Wek
a
s
of
twa
re
h
as
selec
te
d
as
p
la
tf
or
m
o
f
a
pply
ing
t
he
cl
as
sific
at
ion
tec
hniqu
es
and t
he d
at
aset
s.
The
pap
e
r’s
fl
ow
is
orga
nize
d
as
fo
ll
ows.
Sect
ion
I
as
in
tro
du
ct
io
n
of
t
he
pa
pe
r.
Se
ct
ion
II
co
vers
li
te
ratur
e
re
vi
ew
of
data
m
ining
an
d
cl
as
sific
at
ion
te
c
hniq
ues.
As
w
el
l
about
the
Wek
a
s
of
twa
r
e
an
d
m
et
ho
dolo
gy.
I
n
the
sect
io
n
I
I
I
Re
su
lt
s
an
d
Discussi
on
we
re
il
lustrate
d.
I
n
the
la
st
sect
ion
IV
C
oncl
us
i
on
a
nd
su
m
m
ariz
ing
t
he
c
om
par
at
ive r
es
ults.
2.
BACKG
ROU
ND
2.1
.
Data
mi
nin
g
Data
m
ining
is
a
pr
oce
ssin
g
of
the
ra
w
da
ta
to
get
of
the
us
e
f
ul
inf
or
m
at
ion
,
or
to
disc
ove
r
the
kn
ow
le
dg
e
from
huge
da
ta
bases.
The
ou
t
pu
t
of
the
data
m
ining
is
the
patte
rn
w
hich
is
t
o
ide
ntif
y
po
te
ntial
ly
us
efu
l,
valid
,
ulti
m
at
ely
under
st
and
a
ble
a
nd
no
vel
patte
r
n
i
n
t
he
m
ining
data
.
Mostl
y
data
m
ining
app
ly
in
g
in
business
,
s
o
th
e
com
pan
ie
s
can
m
ake
eff
ec
ti
ve
m
ark
et
ing
strat
egies
by
knowin
g
w
ha
t
their
custom
ers
wa
nt
to
bu
y.
Data
m
ining
outc
om
e
dep
en
ds
on
the
way
of
c
ollec
ti
ng
data
and
how
t
he
da
ta
are
processe
d [1,
3
, 7
]
.
2.2
.
Data
w
are
ho
u
sing
Data
W
are
ho
usi
ng
is
a
cente
r
of
m
any
data
colle
ct
ion
f
rom
m
any
places
.
The
c
om
pan
ie
s
or
sect
or
s
colle
ct
their
da
ta
from
diff
ere
nt
places
a
nd
br
a
nc
hes
in
on
e
place
cal
le
d
data
wa
reho
us
i
ng.
T
he
data
in
data
war
e
hous
i
ng
a
re
integ
rated
f
ro
m
al
l
places
,
cl
eaned
from
m
issi
ng
data
and
noise
data
.
The
data
in
data
war
e
hous
i
ng ar
e o
r
ga
nized
a
nd
pr
e
par
e
d for
fu
t
ur
e
us
e
or in
d
em
and
fr
om
t
he users
. T
he d
at
a w
are
hous
e
us
e
d
to s
upport
the
decisi
on
of
m
anag
em
ent m
aki
ng pr
ocess [1].
2.3
.
Clas
sific
at
i
on
Cl
assifi
cat
ion
in
data
m
ining
are
so
m
e
te
chn
iqu
es
us
e
to
predict
,
cl
assify
and
orga
nized
the
data
in
their
s
uitable
cat
egories
[
8
]
.
Each
cl
ass
ha
s
it
s
ow
n
r
ul
es
an
d
al
go
rithm
s.
So
m
e
of
th
e
cl
assifi
c
at
ion
te
chn
iq
ues
are
f
ollow
decisi
on
tree
r
ules
s
uch
as
J48,
s
om
e
oth
er
cl
ass
es
are
f
ollo
wing
Ba
ye
sia
n
N
et
wor
k
su
c
h
as
Ba
ye
sNet
an
d
Nai
ve
Ba
ye
s,
and
ot
her
are
f
ollo
wing
A
rtific
ia
l
intel
li
gen
ce
a
nd
Ne
ur
al
Net
work.
Cl
assifi
cat
ion
te
chn
iq
ues
have
di
ff
e
ren
t
ap
pl
ic
at
ion
s
and
wh
ic
h
dataset
sh
o
ul
d
be
ap
pl
ie
d
on.
I
n
ad
di
ti
on
,
al
l
cl
assifi
cat
io
n
te
ch
niques
will
no
t
be
abl
e
to
predict
co
rr
ect
ly
the
cl
assifi
cat
ion
of
da
ta
com
par
e
to
othe
r
cl
assifi
cat
ion
t
echn
i
qu
e
s.
T
o
fin
d
ou
t
t
he
be
st
cl
assifi
cat
ion
te
ch
niq
ues
f
or
the
te
sti
ng
da
ta
,
the
data
s
houl
d
be
a com
patible
wi
th the s
el
ect
ed
classi
ficat
ion
t
echn
i
qu
e
rules,
algorit
hm
s etc [
9
].
2.4
.
J48
J4
8
cl
assifi
er
i
s
an
op
ti
m
iz
e
ver
si
on
of
C4
.
5.
T
he
J
48
is
base
d
on
Deci
sion
tr
ee.
J
48
is
on
e
of
the
data
cl
assifi
cat
ion
te
chn
i
ques
us
e
d
in
da
ta
processin
g
a
nd
data
m
ining.
The
J48
r
ules
and
al
gorithm
s
are
us
in
g
decisi
on
tree
te
ch
niques
w
hich
co
ntain
s
of
m
ai
n
le
af
and
bra
nch
e
s.
Each
of
the
br
anch
or
le
a
f
c
onta
in
a
decisi
on
that l
ead
to
di
ff
e
rent
ou
tc
om
e.
Som
e
of
the d
at
aset
s
hav
e v
ery b
ig
tree
m
od
el
wh
ic
h
co
ntains
m
any
le
afs
le
ads
to
diff
e
ren
t
resul
t
co
m
par
in
g
to
fe
w
le
afs
of
decisi
on
tree
wh
e
n
ap
plyi
ng
J48
cl
assifi
c
at
ion
te
chn
iq
ue [
10, 11
].
2.5
.
Multila
yer
pe
rceptr
on
Mult
il
ay
er
Per
ceptr
on
cl
assif
ie
r
is
based
on
Ar
ti
fici
al
in
te
ll
igence
an
d
Ne
ur
al
Netw
ork
with
ou
t
qu
al
ific
at
io
n.
A
Mult
i
-
Lay
e
r
Perce
ptr
on
(
MLP)
ha
s
as
m
ini
m
u
m
as
th
ree
la
ye
rs.
On
e
la
ye
r
as
input,
the
seco
nd
as
hidden
la
ye
r
a
nd
the
la
st
as
t
he
ou
t
pu
t
la
ye
r
.
MLP
is
a
fee
dforwa
rd
ne
ural
netw
ork,
t
he
hidde
n
la
ye
r
can
be
on
e
la
ye
r
or
m
or
e.
In
MLP
eac
h
node
in
eac
h
la
ye
r
are
co
nn
e
ct
ed
to
al
l
la
ye
r’
s
nodes.
M
ulti
la
ye
r
per
ce
ptr
on
is
on
e
of
t
he
dat
a
cl
assifi
cat
ion
te
c
hn
i
qu
es
us
e
d
in
ne
ural
netw
ork
,
dee
p
le
ar
ning
an
d
oth
e
r
app
li
cat
io
ns
of d
at
a
processi
ng
[
12
].
2
.6
.
Bay
es
Ne
t
Ba
ye
sNet
cl
assifi
er
one
of
t
he
cl
assifi
ers
in
W
e
ka
softw
are.
The
Ba
ye
sNet
is
based
on
Ba
ye
sia
n
Netw
ork
w
hich
is
based
on
Ba
ye
s
theor
em
.
The
Ba
ye
sia
n
netw
ork
is
m
os
tly
wo
rk
i
ng
w
hen
the
re
m
igh
t
be
a p
r
ob
a
bili
ty
o
f
uncertai
nty,
or com
plexity
a
n
d (eve
n
m
or
e i
m
po
rtantl
y) ca
us
al
it
y si
tuati
on
.
Ba
ye
sia
n ne
twork
consi
st
of
two
par
ts:
A
set
of
co
ndit
ion
al
pr
oba
bili
ty
d
ist
ribu
ti
ons
an
d
a
directed
a
cy
cl
ic
gr
aph
DAG
.
Ba
ye
sia
n
netw
orks,
eac
h
node
represe
nts
a
Var
ia
ble.
A
va
riable
m
igh
t
be
discrete
or
m
igh
t
be
co
ntinuo
us.
Ba
ye
sNet
cl
assifi
er
one
of
the
data
cl
assifi
cat
ion
te
chni
qu
es
a
pp
li
ed
in
m
any
areas
of
pro
ba
bili
ty
or
un
ce
rtai
nty co
nd
it
io
ns
[
13
].
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
5
8
4
-
3
5
9
0
3586
2.7
.
Na
i
veBa
yes
NaiveBay
es
cl
assifi
ers
is
a
c
ollec
ti
on
of
al
gorithm
s
that
sh
are
com
m
on
pri
nciples
bas
ed
on
B
ay
es
The
or
em
.
In
NaïveBay
es
cl
assifi
ers
each
pair
of
featu
res
cl
assifi
ed
is
ind
epe
ndent
fr
om
oth
er
pairs
.
NaïveBay
es
cl
assifi
ers
is
one
of
the
data
cl
assifi
cat
ion
te
c
hn
i
qu
e
s
us
ed
i
n
Weka
s
of
t
w
are
or
can
be
us
e
d
in
oth
e
r
area
s
of
processi
ng d
at
a u
si
ng d
if
fe
re
nt s
of
twa
re
[1
4,
1
5
].
2.8
.
Data
mi
nin
g
p
rocess
The
data m
ining
pro
ce
ss brea
ks
in
m
any sta
ges.
T
he
fir
st st
age it’s call
ed
the integ
rati
on
sta
ge
w
hic
h
is
colle
ct
the
data
from
m
an
y
so
urces
as
r
aw
data
with
diff
e
re
nt
form
at
.
The
sec
ond
sta
ge
it
’s
cal
le
d
data
cl
eaning
i
n
thi
s
sta
ge
after
r
ecei
vin
g
t
he
da
ta
fr
om
the
first
sta
ge
so
m
e
of
the
data
are
incom
patible
or
inco
ns
ist
ency
and
oth
e
r
data
are
m
issi
ng
va
lue
and
oth
e
r
data
are
il
log
ic
al
entered
.
S
o,
it
the
data
clean
in
g
sta
ge
will
cl
ean
al
l
these
data.
Th
e
thir
d
sta
ge
of
data
m
ining
it
is
to
colle
ct
the
cl
eaning
d
at
a
in
one
pla
ce
cal
l
Data
War
e
housi
ng.
I
n
the
da
ta
war
e
housi
ng
m
os
tl
y,
the
data
rea
dy
t
o
be
us
e
d
a
nd
a
naly
zed.
H
owe
ver,
the
a
m
ou
nt
of
t
he data
in data
w
a
reho
us
in
g
is h
uge size
of
data to d
eal
wit
h
it
an
d
a
naly
ze th
e w
hole
dat
a at on
ce
,
for
this
reas
on
nex
t
sta
ge
is
presente
d.
The
f
ourth
sta
ge
is
t
he
sel
ect
io
n
st
age,
w
hich
is
t
o
sel
ect
the
rel
evan
t
data
f
ro
m
data
war
e
hous
i
ng
that
will
work
on
it
.
T
he
la
st
sta
ge
of
data
m
ining
is
a
pply
ing
the
al
gori
thm
s
an
d
te
chn
iq
ues
of
data
m
ining
to
get
the
patte
r
n,
th
at
the
us
e
r
lo
ok
i
ng
f
or
.
The
ou
tc
om
e
of
t
he
a
pp
ly
in
g
data
m
ining
tec
hn
i
ques
will
b
e
represente
d
as
gra
ph or ta
ble or
ot
her
form
at
o
f ou
t
pu
t
re
pr
ese
ntati
on
[
1
,
1
6
].
3.
RELATE
D
W
ORKS
Ther
e
a
re
va
riou
s
relat
ive
stud
ie
s
of
the
di
ff
ere
nt
cl
assif
ic
at
ion
te
chn
i
ques,
ye
t
it
has
no
t
bee
n
disco
ver
e
d
t
ha
t
on
e
sin
gle
m
et
ho
d
is
sup
erior
com
par
e
d
to
ot
her
s
.
I
ssu
es
li
ke
acc
ur
acy
,
t
rainin
g
tim
e,
scal
abili
ty
and
m
any
oth
ers
c
on
t
rib
ute
to
ch
oo
si
ng
the
best
te
chn
iq
ue
t
o
c
la
ssify
data
f
or
m
ining
.
T
he
s
earc
h
for
best
te
ch
ni
qu
e
for
cl
assifi
cat
ion
rem
ai
ns
a
researc
h
sub
je
ct
.
Cl
assifi
cation
is
a
data
m
ining
te
c
hn
i
que
us
ed
to
predict
group
m
e
m
ber
sh
ip
for
data
insta
nces.
The
re
ar
e
nu
m
erous
tr
aditi
on
al
cl
assi
ficat
ion
m
et
hods
li
k
e
decisi
on
tree
(
DT)
i
nductio
n,
k
-
near
est
neighb
or
cl
assifi
e
r,
Ba
ye
sia
n
ne
t
work
s
,
sup
por
t
vector
m
achines,
ru
le
-
base
d
cl
as
sific
at
ion
,
case
-
base
d
reasoni
ng,
ge
netic
al
gorithm
,
fu
zzy
l
og
ic
te
c
hn
i
qu
e
s,
r
ough
set
ap
proac
h
and
oth
er
s.
Th
e
basic
dif
fer
e
nce
betwe
e
n
t
he
al
gorithm
s
de
pends
on
w
he
ther
t
hey
are
l
azy
le
arn
e
rs
or
eage
r
le
arn
er
s [1
7
]
.
A
pre
dicti
ve
KNIME
m
od
e
l
was
de
velo
pe
d
an
d
t
hr
ee
da
ta
m
ining
al
gorithm
s;
the
N
aï
ve
Ba
ye
s,
PNN
P
red
ic
to
r
an
d
Decisi
on
T
ree
w
ere
t
raine
d
usi
ng
70%
of
the
total
sam
ples
wh
ic
h
wer
e
r
andom
ly
sel
ect
ed.
T
he
knowle
dge
ac
qu
i
red
f
ro
m
the
trai
ning
w
as
ap
plied
i
n
pr
e
dicti
ng
the
ty
pe
of
s
uppl
y
that
pro
du
ce
d
t
he
r
e
m
ai
nin
g
30%
of
the
m
oto
r
operati
onal
data
sam
ples.
The
pr
e
dicti
ve
acc
uracy
achie
ved
i
n
the
pap
e
r
is i
nd
ic
at
ive of t
he
s
uitabi
li
ty
o
f data
m
ining ap
proac
h f
or
m
oto
r
p
e
rfor
m
ance m
on
it
or
i
ng [1
8
].
In
[
8],
the
aut
hor
points
out
about
Decisi
on
Tree
(
DT
)
or
J48,
that
ad
va
ntages
of
D
T
are
easy
t
o
unde
rstan
d,
ea
sy
to
gen
e
rate
and
reduce
pro
blem
capaci
t
y.
The
lim
it
a
ti
ons
of
DT
are:
R
equ
i
r
ed
se
parat
e
te
st
set
,
trai
ning
t
i
m
e
is
so
ex
pensi
ve,
does
no
t
ha
nd
le
con
ti
nu
ous
va
riab
le
an
d
s
uffer
from
ov
er
fitt
ing
.
The
a
ppli
cat
ion
s t
hat f
it
DT a
re: Text Cat
e
gorizat
io
n
a
nd Im
age Cla
ssific
at
ion
.
In
[
1
9
]
,
MSS
QL
2005
data
base
was
util
iz
ed
t
o
gathe
r
thr
oug
h
s
urve
ys
or
I
nter
net
an
d
t
o
st
or
e
inf
or
m
at
ion
order
e
d
under
31
crit
eria
in
f
our
m
ai
n
gro
up
s
co
ntains
a
total
of
100
stu
de
nts
rec
ei
ving
vo
cat
io
nal
trai
ning
in
var
i
ous
energy
app
li
cat
ion
fiel
ds
,
who
are
al
so
i
n
the
process
of
vocat
io
nal
gu
i
dan
ce
.
This
pap
e
r
a
pp
li
ed
al
gorithm
s
us
e
d
in
m
any
cl
assifi
cat
ion
t
echn
i
qu
e
s
to
a
gro
up
of
in
div
i
du
al
s
w
ho
are
in
the
process
of
vo
c
at
ion
al
guida
nc
e
and
c
on
cl
uded
t
hat
the
m
os
t
ap
pro
pr
ia
te
al
go
rithm
to
be
us
e
d
f
or
st
udie
s
in
this
area
is
the
Naive
Ba
ye
s
al
gorithm
der
ived
fr
om
a
s
ta
ti
sti
cal
est
i
m
ation
m
od
el
that
is
cal
le
d
the
Ba
ye
s’
theo
rem
.
Since
us
in
g
m
achine
le
arn
in
g
(ML)
te
chn
iq
ues
in
cl
assifi
cat
ion
s
tud
ie
s
res
ults
in
accu
rate
ou
tc
om
e
s
with
a
sign
ific
ant
savin
g
in
te
rm
s
of
tim
e
a
nd
c
os
t,
it
is
re
com
m
end
ed
to
m
ake
us
e
of
t
ho
s
e
al
gorithm
s
us
ed
in d
at
a m
ining
and m
achine learn
i
ng tech
niques
for
t
he
s
of
t
war
e
to be
de
ve
lop
e
d
in
this
f
ie
ld.
In
[
20
]
,
the
data
set
us
ed
in
t
hi
s
researc
h
is
the
trai
ni
ng
data
set
of
the
K
DD
C
up
2009
or
a
nge
sm
all
data
set
.
T
he
data
set
c
on
ta
i
ns
190
nu
m
eri
c
featu
res
a
nd
40
no
m
inal
featur
es.
O
ut
of
these
19
0
num
eric
featur
e
s,
16
a
re
em
pty
and
132
are
s
parse
with
highe
r
than
90%
m
issi
ng
rate.
The
aut
hors
use
four
cl
assifi
cat
ion
t
echn
i
gu
e
s,
J
48
,
Nai
veBay
es,
SV
M
a
nd
KNN.
The
aut
hors
pointed
,
pro
pose
d
feat
ur
e
se
le
ct
ion
m
et
ho
d
re
so
l
ve
s
the
real
-
world
CR
M
cl
assifi
cat
ion
pro
bl
e
m
s
with
no
is
y
and
hi
gh
ly
im
balanced
data
set
.
The
var
io
us
cl
assifi
ers
are
use
d
f
or
cl
assi
ficat
ion
.
As
a
re
s
ult,
the
S
VM
ha
s
highest
acc
uracy
an
d
sen
sit
ivit
y,
Naïve Bay
es
ha
s h
i
gh
est
R
O
C an
d
S
pecific
it
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Perf
orma
nce e
valu
ation of
d
if
fe
rent classif
ic
ation t
ech
niqu
es
…
(
Fares
A
bdulhafi
dh
Dael
)
3587
4.
BACKG
ROU
ND
4.1
.
WEKA
The
We
ka
s
oft
war
e
is
a
m
achine
-
le
ar
ni
ng
platf
orm
fo
r
ap
plyi
ng
m
achine
le
ar
ni
ng
.
Wek
a
is
abbre
viati
on
of
W
ai
kat
o
En
vi
ronm
ent
fo
r
Kno
wled
ge
A
naly
sis
(
W
E
K
A)
.
T
he
W
e
ka
’s
nam
e
al
so
r
efers
to
nam
e
of
bir
d
in
Ne
w
Zeal
an
d.
Wek
a
is
m
a
chine
le
ar
ning
wh
ic
h
it
s
colle
ct
ion
of
m
achine
le
ar
ning
al
gorithm
s
and
sta
ndar
ds
for
processi
ng
data
m
ining
.
I
n
Wek
a
the
al
gorithm
s
and
t
echn
i
qu
e
s
ca
n
be
a
pp
li
ed
f
rom
inp
ut
file
su
ch
as
E
xc
el
file
s,
Java
form
at
and
ot
he
rs
or
can
be
a
pp
li
ed
directl
y
fr
om
the
so
ft
war
e
it
sel
f.
A
s
sh
ows
in
Fig
ure
1.
T
he
Wek
a
E
xpl
or
e
r
wi
ndow
div
ide
d
i
n
to
6
ta
bs,
each
ta
b
ha
s
di
ff
e
re
nt
ta
sk
s.
Tab
1
cal
ls
Pr
e
process:
Pe
r
proce
ss’s
f
unct
ion
is
to
l
oad
a
dataset
f
ro
m
diff
e
re
nt
sour
ces
an
d
m
anipu
la
te
the
data
i
nto
a
desire
f
orm
.
T
ab
2
cal
ls
Cl
assify
:
is
to
sel
ec
t
a
cl
assifi
er
to
process
the
da
ta
set
that
has
sel
ect
ed
in
pr
e
proce
s
s
sta
ge.
Tab
3
c
al
ls
Cl
us
te
r:
is
to
sel
ect
a
cl
ust
er
to
proce
ss
the
dataset
th
at
has
sel
ect
ed
in
pr
e
proces
s
sta
ge.
Tab 4 call
s A
s
so
ci
at
e:
is to r
un ass
ociat
ion
a
lgorit
hm
s r
ules to
extract i
nsi
gh
ts
from
d
at
aset
. Tab
5
cal
ls Sel
ect
Attrib
utes:
it
t
o
r
un
at
trib
ute
sel
ect
ion
al
gor
it
h
m
s
on
datas
et
to
sel
ect
tho
se
at
tribu
te
s
th
at
are
relevan
t
to
the
desire
featu
re t
o pr
e
dict. Ta
b 6 call
s
Visu
al
i
ze:
is to
visu
al
i
ze the
relat
ion
s
hip
betwee
n
at
t
rib
utes [
21
].
Figure
1.
W
e
ka
interf
a
ce e
xplorer
4.2
.
Datasets
inf
or
mat
i
on
Fo
r
this
pa
per,
th
ree
data
se
t
hav
e
sel
ect
e
d.
Eve
ry
data
set
ha
s
it
s
own
cha
racteri
s
ti
cs
and
the
par
am
et
ers
tha
t
dif
fer
e
ntiat
e
it
from
oth
er
two
data
set
s.
The
T
able
1
il
lustrate
t
he
differences
betwe
en
eac
h
of
t
he
dataset
.
The
dataset
s
hav
e
te
ste
d
one
by
on
e
with
sam
e
setting
s
in
W
e
ka
s
of
t
war
e
for
al
l
da
ta
set
s.
Each
on
e
of
da
ta
set
s
has
te
ste
d
agai
ns
t
the
four
cl
assi
fiers
.
The
outp
ut
de
te
rm
ines
by
the
pa
ram
et
ers
of
t
he
tim
e
ta
ken
to
bu
il
d
t
he
m
od
el
of
the
datas
et
and
the
pe
r
centage
of
acc
ur
acy
to
cl
assi
fy
the
ta
rg
et
da
ta
set
s.
Each
dataset
ha
s test
ed fo
ur ti
m
es f
or the
four classi
fiers
.
Table
1.
Illustr
at
e the d
i
ff
e
rence
s b
et
ween ea
ch of
the
datas
et
Na
m
e
of
datas
ets/
Para
m
eters
Co
n
g
ressio
n
al Vot
in
g
Reco
rds
Car E
v
alu
atio
n
Co
n
trace
p
tiv
e M
et
h
o
d
Ch
o
ice
Data Set
Ch
arac
te
r
istics
Multiv
ariate
Multiv
ariate
Multiv
ariate
Attribu
te Ch
arac
te
ristics
Categ
o
rical
Categ
o
rical
Categ
o
rical
Ass
o
ciated Task
s
Clas
sif
icatio
n
Clas
sif
icatio
n
Clas
sif
icatio
n
Nu
m
b
e
r
o
f
I
n
stan
c
es
435
1728
1473
Nu
m
b
e
r
o
f
Attr
ib
u
tes
16
6
9
Missin
g
Valu
es
Yes
No
No
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
5
8
4
-
3
5
9
0
3588
5.
RESU
LT
S
AND A
N
ALYSIS
A
com
par
iso
n
of
eval
uatio
n
perform
ance
of
cl
assifi
ers
for
diff
e
re
nt
dataset
s
based
on
the
accu
rac
y
of
each
cl
assifi
er
a
nd
ti
m
e
ta
k
en
t
o
bu
il
d
the
m
od
el
.
Acc
ur
a
cy
is
de
fine
d
a
s
the num
ber
of
insta
nces
cl
assifi
ed
correct
ly
.
The
T
able
2
s
ummari
ze
the
ou
t
put
o
f
the
cl
assifi
cat
ion
data
te
chn
i
qu
e
s
for
the
three
dataset
s
ba
s
e
d
on
t
he
ti
m
e
tak
en
t
o
buil
d
t
he
m
od
el
.
It
is
ob
s
er
ved
f
or
t
he
fir
st
dataset
of
Ca
r
Eval
ua
ti
on
the
J
48
cl
assifi
er
giv
e
the
best
r
esult
of
the
ti
m
e
ta
ken
to
bu
il
d
the
m
od
el
.
In
the
sec
ond
and
t
he
thir
d
dataset
s
Con
tr
acepti
ve
Me
thod
Cho
ic
e
and
Co
ngres
sion
al
V
otin
g
Re
cords
res
pe
ct
ively
sh
ow
s
NaiveBay
es
cl
assifi
ers
gi
ve
the
best
ou
tc
om
e.
Ho
wev
e
r,
the
M
ulti
la
ye
r
Percep
tr
on
cl
assifi
e
r
is
the
lon
ge
s
t
tim
e
ta
ken
to
bu
il
d
the
m
od
el
for
eac
h datase
t.
T
able
2.
C
om
par
iso
n of t
i
m
e t
aken f
or v
a
rio
us cl
assifi
ers
Na
m
e
of
datas
ets/
class
if
icatio
n
techn
iq
u
es
Co
n
g
ressio
n
al
Vo
tin
g
Reco
rds
Car
Evalu
atio
n
Co
n
trace
p
tiv
e
Metho
d
Ch
o
ice
BayesNet
0
Sec
0
.03
Sec
0
.03
Sec
Naiv
eBayes
0
Sec
0
.03
Sec
0
Sec
Multilaye
r
Pe
rcept
ron
0
.69
Sec
4
.49
Sec
3
.56
Sec
J4
8
0
.03
Sec
0
.01
Sec
0
.09
Sec
In
t
he
T
a
ble
3
sh
ow
t
he
Com
par
is
on
of
Acc
ur
acy
of
cl
assifi
ers
for
diff
e
r
ent
dataset
s.
From
F
igu
re
2
and
T
a
ble
2
It
is
obser
ve
d
t
hat,
for
the
fir
st
an
d
the
sec
ond
dataset
of
Ca
r
E
valuati
on
a
nd
Co
ntrac
eptive
Me
thod
Ch
oice
resp
ect
ively
,
the
Mult
il
ayer
Perce
ptr
on
cl
assifi
er
give
the
best
res
ult
of
the
Ac
cur
ac
y
com
par
e
to
othe
r
cl
assifi
ers.
I
n
the
third
dataset
s
C
on
gressi
on
al
V
otin
g
Re
cords
sho
ws
J
48
cl
assifi
er
giv
e
the
best
outc
om
e
of
acc
ur
a
cy
to
cl
assify
ing
th
e
instance.
H
oweve
r,
t
he
Nai
veBay
es
cl
assifi
ers
wer
e
t
he
worst
ou
tc
om
e o
f
acc
ur
acy
t
o
cl
assif
yi
ng
the
insta
nc
e f
or
eac
h dat
aset
.
Table
3.
C
om
par
iso
n of t
i
m
e t
aken f
or v
a
rio
us cl
assifi
ers
Na
m
e
of
datas
ets/
class
if
icatio
n
techn
iq
u
es
Co
n
g
ressio
n
al
Vo
tin
g
Reco
rds
Car
Evalu
atio
n
Co
n
trace
p
tiv
e
Metho
d
Ch
o
ice
BayesNet
9
0
.11
4
9
%
8
5
.70
6
%
5
1
.05
2
3
%
Naiv
eBayes
9
0
.11
4
9
%
8
5
.53
2
4
%
5
0
.78
0
7
%
Multilaye
r
Pe
rcept
ron
9
4
.71
2
6
%
9
9
.53
7
%
5
2
.34
2
2
%
J4
8
9
6
.32
1
8
%
9
2
.36
1
1
%
5
2
.13
8
5
%
Figure
2. G
raphical
v
ie
w
of a
ccur
acy
for dif
fer
e
nt classi
fie
rs on di
ff
e
ren
t
dataset
s
6.
CONCL
US
I
O
N
This
pa
per
s
ho
wed
t
he
pe
rform
ance
evaluati
on
of
dif
fer
e
nt
data
cl
assifi
er
s
te
chn
iq
ues
on
dif
fer
e
nt
dataset
s.
It
f
ou
nd
t
hat
the
out
com
e
of
the
da
ta
te
ste
d
are
di
ff
e
ren
t
f
r
om
dataset
to
ano
t
he
r.
T
her
e
a
re
r
easo
ns
for
the d
if
fer
e
nt
o
utput becaus
e the d
at
aset
s ch
rem
at
ist
ic
s
are
diff
er
ent f
r
om
each
ano
the
r
d
at
aset
.
Fact
ors that
m
ay
aff
ect
the
cl
ass
ifie
r’
s
perform
ance
as
fo
ll
ow
1.
D
at
a
set
,
2.
N
um
ber
of
insta
nce
an
d
at
trib
utes,
3.
Com
patibil
i
ty
of
t
he
data
with
the
cl
assifi
er
,
4.
Ty
pe
of
at
tribu
te
s,
5.
Mi
ssing
data
a
nd
data
in
structi
ons,
6.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N: 20
88
-
8708
Perf
orma
nce e
valu
ation of
d
if
fe
rent classif
ic
ation t
ech
niqu
es
…
(
Fares
A
bdulhafi
dh
Dael
)
3589
Syst
e
m
con
fig
ur
at
io
n.
Mult
il
ay
er
Percep
t
ron
cl
assifi
er
shows
in
ov
e
rall
the
best
accuracy
perform
a
nce
t
o
cl
assify
the
i
nst
ances,
an
d
NaiveBay
es
cl
assifi
ers
were
the
w
orst
ou
tc
om
e
of
acc
ur
a
cy
to
cl
assify
i
ng
the
instance
f
or
e
ach
dataset
.
F
uture
wor
k
m
ay
fo
cus
on
s
pecific
dataset
s
that
wo
r
king
in
har
m
on
y
with
cl
assif
ie
rs
sho
uld
be
sel
ect
ed
.
The
f
ut
ur
e
w
ork
m
ay
fo
cus
on
im
pr
ovin
g
the
perf
or
m
ance
of
eac
h
cl
as
sifie
rs
by an
al
yz
in
g
t
heir
al
gorithm
s
and the
rules.
ACKN
OWLE
DGE
MENTS
We
w
ould
li
ke
to
express
our
gr
at
it
ude
to
Wek
a
s
oft
ware
dev
el
op
e
rs
a
nd
m
ake
it
as
a
n
ope
n
source
so
ft
war
e
without
it
ou
r
w
ork
would
be
m
uc
h
dif
ficult
to
achieve
the
ou
t
com
e
of
this
pap
er
.
W
e
al
so
woul
d
li
ke
to
t
hank
t
he
creato
r
of
“C
on
t
raceptive
M
et
hod
C
ho
ic
e”
dataset
s
Mr.
T
je
n
-
Sien
Lim
(li
m
t@st
at
.w
isc
.ed
u)
and
the
D
onor:
Tj
e
n
-
Sien
Li
m
(li
m
t@st
at
.
wisc.e
du
)
with
res
pect
to
the
or
i
gin
al
s
ource
of
the
1987
N
at
ion
al
Ind
on
esi
a
Co
nt
raceptive
P
re
valence
S
urve
y
fo
r
s
har
in
g
the
dataset
s
and
m
akes
it
a
vaila
ble
for
ev
eryo
ne.
We
al
so
woul
d
li
ke
to
tha
nk
the
cr
eat
or
of
“C
a
r
Eval
ua
ti
on
Data
base
s”
dataset
s
Mr
.
Ma
rko
Bo
ha
nec
an
d
the
D
onors:
Ma
rko
Bo
ha
ne
c
(m
ark
o.b
ohanec@i
j
s.si)
,
and
Bl
az
Zu
pa
n
(
blaz.z
upa
n@
i
j
s.si)
f
or
sh
ari
ng
the d
at
aset
s
an
d
m
akes it avai
la
ble for
e
ve
ryon
e
.
REFERE
NCE
S
[1]
Jiawe
i
Han
,
Mi
che
l
ine
Kam
ber
and
Jian
Pe
i
,
Data
Mini
ng:
Conce
pts
and
Techni
ques,
3rd
edi
t
ion,
Morga
n
Kaufm
ann,
2011
.
[2]
Saouabi
Moham
ed
,
Abdel
la
h
E
z
za
t
i
,
"
A
dat
a
m
ini
ng
proc
ess
u
sing
cl
assificat
i
on
te
chni
qu
es
for
emplo
y
ab
il
i
t
y
pre
diction
,"
Ind
onesian
Journal
of
E
le
c
tric
al
Engi
ne
ering
an
d
Computer
Scienc
e
(
IJE
E
C
S
),
v
ol.
14
,
n
o
.
2
,
pp.
1025
-
1029
,
2019.
[3]
Akibu
Mahm
ou
d
Abdulla
hi,
M
okhai
ri
Makht
ar
,
Suhail
an
Saf
ie
,
"
The
pat
t
ern
s
of
ac
c
essing
le
ar
ning
m
ana
geme
nt
s
y
stem
among
student
s
,"
Indon
e
sian
J
ournal
of
El
e
ct
rica
l
Eng
in
ee
ring
and
Computer
Sc
ie
nc
e
(
IJE
ECS)
,
v
ol
.
13
,
no.
1
,
pp
.
15
-
21
,
2019.
[4]
Longa
dge
R
.
,
D
ongre
S.
,
La
t
esh
Mali
k
,
"
Cl
ass i
m
bal
an
ce
prob
le
m
in
data
m
ini
ng
:
rev
ie
w
,
"
In
te
rna
ti
onal Journal
o
f
Computer
Scien
ce
and
Ne
twork (
IJCSN
)
,
vol.
2,
no.
1
,
Feb
2013
.
[5]
Alm
oham
m
edi
Akram
A.,
Nor
K.
Noordin
and
Sabri
Saee
d
,
"
E
val
ua
ti
ng
th
e
Im
pac
t
of
Tr
ansm
is
sion
Range
on
t
h
e
Perform
anc
e
of
VA
NET
,"
Inte
rnational
Journal
of
El
e
ct
rica
l
and
Computer
Engi
nee
ring
(
IJE
C
E)
,
vol.
6
,
no
.
2
,
pp.
800
-
809
,
20
16
.
[6]
J
abba
r
W
ahe
b
A.,
e
t
al
.
"D
esi
gn
and
Im
ple
m
ent
at
ion
o
f
IoT
-
B
ase
d
Autom
at
io
n
S
y
stem
for
Sm
art
Hom
e
,
"
2018
Inte
rnational
Sy
mpos
ium
on
Networ
ks,
Compute
rs
and
Comm
unic
ati
ons (
ISNCC)
,
IE
EE
,
2018
.
[7]
Alm
oham
m
edi
Akram
A.,
et
al
.
,
"
An
ac
cur
ate
p
erf
orm
anc
e
ana
l
y
sis
of
h
y
brid
ef
fic
i
ent
and
r
elia
ble
MA
C
protocol
in
VA
NET
und
er
non
-
satura
t
ed
condi
ti
ons
,"
Inte
rnational
Jou
rnal
of
El
ec
tri
c
al
and
Comput
er
Engi
nee
rin
g
(
IJE
CE)
,
vol.
7
,
no.
2
,
pp
.
999
-
1
011
,
2017
.
[8]
Mus
ta
kim
Mus
ta
kim,
Novia
Kum
al
a
Sari,
Jasril
Ja
sril,
Ism
u
Kus
um
ant
o,
Nurul
G
a
y
a
tri
Inda
h
R
eza
,
"
Ei
g
envalue
of
Anal
y
tic
Hier
ar
c
h
y
Proce
ss
as
The
Dete
rm
ina
nt
for
Cla
ss
Ta
rge
t
on
Cla
ss
ifi
cation
Algorit
hm
,"
Indone
sian
Journal
of
E
le
c
tric
al
En
gine
ering
and
C
omputer
Scienc
e
(
IJE
ECS)
,
vol
.
12,
n
o
.
3
,
pp
.
12
57
-
1264,
2018
.
[9]
Deulka
r
M.
D.
S.
and
Deshmukh
R.
R.
,
"
Da
ta
Mining
Cl
assific
a
ti
on
,"
Imp
erial
Journal
of
Inte
rdisciplinar
y
Re
search
,
vol
.
2
,
no.
4,
2016.
[10]
Kaur
G.,
Chh
abr
a
A.
,
"
Im
prove
d
J48
cl
assifi
ca
t
io
n
al
gori
thm
for
t
he
pre
di
ction
of
dia
be
te
s
,"
Int
ern
ati
onal
Journal
of
Computer
Ap
pli
cations,
vo
l.
98
,
no
.
22
,
2014
.
[11]
Koturwar
Praful, Shee
tal
Gira
se
a
nd
Deba
j
y
ot
i
Mukhopadh
y
a
y
,
"
A surve
y
of cl
assific
a
ti
on
t
ec
hn
i
ques
in
the a
r
ea
of
big
da
ta
,"
IJ
AF
R
C,
v
ol
.
1
,
no
.
11
,
Nov 2014
.
[12]
Gardne
r
M.
W
.
and
Dorling
S
.
R.
,
"
Artif
icial
ne
ura
l
n
et
works
(t
he
m
ult
ila
y
er
pe
rce
ptron)
a
r
evie
w
of
appl
i
cation
s
in
th
e at
m
ospheric
sc
ie
nc
es
,"
At
m
ospheric
en
vi
ron
ment
,
vo
l.
32
,
no
.
14
-
15
,
pp
.
262
7
-
2636,
1998
.
[13]
Vait
hi
y
a
nat
h
an
V.,
Rajesw
ari
K.
,
Ta
j
ane
K.
and
Pita
le
R
.
,
"
Com
par
ison
of
diffe
r
ent
class
ifi
c
at
ion
te
chni
qu
es
using
diffe
ren
t
d
at
ase
t
s
,"
Inte
rnatio
nal
Journal
o
f Adv
a
nce
s in
Eng
ine
er
ing
&
Technol
ogy
,
vo
l.
6
,
no.
2,
p
p.
764
,
2013
.
[14]
Sharm
a
A.
and
Kaur
B.
,
"
A
rese
arc
h
r
evi
ew
on
compara
ti
v
e
ana
l
y
sis
of
data
m
ini
ng
tool
s,
te
chni
qu
es
and
par
amete
rs
,"
Int
ernat
ional J
ourn
al
of
Adv
an
ce
d
Re
search in Co
mputer
Scienc
e
,
vol.
8
,
no
.
7
,
201
7.
[15]
Nur’A
in
Maula
t
Sam
sudin,
Cik
Fere
sa
bint
i
M
ohd
Fooz
y
,
Na
bil
ah
Ali
as,
Pal
ani
app
an
Sham
al
a,
Nur
Fadzila
h
Othm
an,
W
an
Isni
Sofiah
W
an
Din
,
"
Youtube
spam
det
ec
t
ion
fra
m
ework
using
naï
ve
ba
y
es
and
logi
st
ic
reg
ression
,"
Indone
sian
Journal
of
El
ec
tri
cal
Engi
ne
ering
and
Computer
Sci
enc
e
(
IJE
ECS)
,
vol.
14,
n
o.
3,
pp.
1508
-
1517
,
2019
.
[16]
Alal
wan
S.
A.
D.
,
"
Diab
et
i
c
an
aly
tics:
Propos
e
d
conc
ep
tual
da
t
a
m
ini
ng
appr
o
a
che
s
in
t
y
pe
2
d
ia
be
te
s
da
ta
set
,"
Indone
sian J
our
nal
of
Elec
tric
al
Engi
ne
ering
and
Computer
Sc
ie
n
ce
(
IJEECS)
.
v
ol
.
14
,
no
.
1
,
p
p
.
92
-
99,
2019
.
[17]
K
.
Sara
n
y
a
,
"
Su
rve
y
on
Cl
assifi
ca
t
ion
T
ec
hniqu
es
Us
ed
in
Da
ta
Mining
and
th
e
ir
Re
ce
nt
Advan
ce
m
ent
s
,
I
JSE
T
R
,
vol.
3
,
pp
.
2380
-
2385,
2014
.
[18]
Adekit
an
A
.
I.
,
Adewal
e
Ade
y
i
nka
and
Al
ashir
i
Olaitan
,
"
Determ
ini
ng
the
op
e
rat
ion
al
sta
tus
o
f
a
Thr
ee
Phase
Induc
ti
on
Motor
using
a
pre
di
ctive
da
ta
m
ini
ng
m
odel
,
Inte
rna
ti
onal
Journal
of
Powe
r
Elec
tronic
s
and
Dr
ive
Syste
ms
(
IJP
ED
S)
,
v
ol.
10
,
no
.
1
,
pp
.
93
-
103
,
20
19.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
20
19
:
3
5
8
4
-
3
5
9
0
3590
[19]
Bulbul
Hal
il
Ibr
ahi
m
and
Özka
n
Uns
al
,
"
Com
par
ison
of
c
la
ss
if
ic
a
ti
on
t
ec
hn
iqu
es
used
in
m
achine
l
ea
rn
ing
as
appl
i
ed
on
voca
t
iona
l
guida
n
ce
d
at
a
,"
2011
10th
Inte
rnational
Co
nfe
renc
e
on
Mac
hine
Learning
and
Appl
ic
at
ion
s
and
Workshops
,
IEE
E
,
2011
, v
o
l. 2.
[20]
Natc
hi
ar
S.
Um
m
ugult
hum
and
S.
Baul
kan
i
,
"
C
ustom
er
rel
ation
ship
m
ana
gement
class
ifi
cation
using
dat
a
m
ini
n
g
te
chn
ique
s
,"
20
14
Inte
rnationa
l
Confe
ren
ce
o
n
Sci
en
ce
Enginee
ring
and
M
anageme
nt
R
ese
arch
(
ICSEMR
)
,
IEE
E
,
2014
.
[21]
Jagta
p
S.
B.
,
"
C
ensus
Data
m
ining
and
d
at
a
anal
y
sis
using
W
EK
A
,"
Inte
rnationa
l
Confe
ren
ce
in
Eme
rging
Tr
ends
in
Sc
ie
nc
e, Tec
h
nology
and
Man
ageme
nt
-
2013
,
S
inga
pore
,
2013
.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
As
soc.
Prof.
Dr
Abdulkadi
r
ÖZDEMIR,
Le
ct
ur
e
r
at
Atat
urk
Univer
sit
y
,
Facu
lty
of
Ec
onom
ic
s
and
Adm
ini
strat
ive
Scie
nc
es,
Dep
ar
tment
of
Man
a
gement
Inform
a
ti
on
S
y
s
te
m
s,
Turkey
.
Area
of
rese
arc
h
intere
st
s
Com
pute
r
Hardware
,
Vir
tua
l
izati
on
and
Cloud
Com
puti
ng,
Co
m
pute
r
Networks
and
Mana
ge
m
ent
,
Dat
aba
se
Mana
gement
Sy
st
ems
,
M
ana
gement
Inf
orm
at
ion
S
y
ste
m
s.
htt
ps://
ww
w.ata
uni.
edu
.
tr/abdul
kadi
r
-
ozd
emir
,
a
bdulka
dir@atau
ni.
edu
.
tr
,
/
+90
4
42
231
1961
.
Prof.
Dr
Uğur
Y
AV
UZ,
Lectu
r
er
at
Atat
urk
Univ
ersity
,
Facult
y
o
f
Ec
onom
ic
s
and
Adm
ini
strat
iv
e
Scie
nc
es,
Depa
r
tment
of
Mana
gement
Inform
at
i
on
Sy
st
ems
,
Turkey
.
Area
of
r
ese
arc
h
in
te
r
ests
Inform
at
ion
Te
c
hnologi
es,
Mana
gement
Scie
nc
e,
W
eb
Design,
M
ana
gement
Infor
m
at
ion
S
y
stems
,
Graphi
c
D
esign
and
Anim
at
ion
,
Artifi
c
ia
l
Int
el
l
i
genc
e
.
h
ttp
://www.ataun
i.edu
.tr
/#
p
erso
n
el=u
g
u
r
-
y
av
u
z
,
u
g
u
r@
atau
n
i.edu
.tr
/
+9
0
44
2
23
1
5752
.
Fare
s
Abdulhaf
idh
Dae
l
,
PhD
ca
ndid
at
e
at
Atat
urk
Univer
sit
y
,
Facu
lty
of
Ec
onom
ic
s
an
d
Adm
ini
strat
ive
Scie
nc
es,
Depa
r
tment
of
Mana
gement
Inform
at
i
on
Sy
st
ems
,
m
aste
r’s
degr
ee
in
comm
unic
at
ion
and
computer
E
ngine
er
ing
from
UK
M
Univer
sity
,
Mal
a
y
s
ia
.
Ba
c
hel
or’s
degr
ee
i
n
comm
unic
at
ion
engi
n
ee
ring
f
rom
IIUM
un
ive
rsit
y
Ma
lay
si
a.
Ar
ea
of
re
sea
rch
int
e
rests
Mana
gement
In
form
at
ion
S
y
st
ems
,
Artifi
cial
int
ellige
n
ce,
Radi
o
comm
unic
a
ti
ons
W
ire
less
comm
unic
at
ion,
Data
Ne
twork
a
nd
Com
pute
r. fa
resa
la
r
iqi
@gm
ail.
com
Evaluation Warning : The document was created with Spire.PDF for Python.