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.
11
,
No.
1
,
Febr
uar
y
2021
, pp.
728
~
73
5
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
11
i
1
.
pp
72
8
-
735
728
Journ
al h
om
e
page
:
http:
//
ij
ece.i
aesc
or
e.c
om
A hybrid
feature sele
ctio
n on AIRS m
eth
od for id
entifyin
g
breast c
ancer di
seases
Achm
ad
Rid
ok
1
,
N
ashi
Wid
odo
2
,
W
ay
an
Fir
da
us
Mah
mudy
3
,
Muhai
mi
n R
if
a’i
4
1
,3
Facul
t
y
of
Co
m
pute
r
Scie
n
ce
(
FILKO
M)
,
Brawij
a
y
a
Univer
sit
y
,
Indone
si
a
2
,4
Depa
rtment
of
Biol
og
y
,
Fa
cul
t
y
of
Math
ematic
s a
nd
Natu
ral Sc
ie
nc
e, Bra
wij
a
y
a
Univer
sit
y
,
Indo
nesia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
2
9
, 201
8
Re
vised
Ju
l
2
7, 2020
Accepte
d
Aug
25
, 202
0
Brea
st
c
ancer
m
a
y
c
ause
a
de
at
h
due
to
th
e
l
at
e
d
ia
gnosis.
A
che
ap
an
d
ac
cur
ate
tool
for
ea
rl
y
detec
ti
on
of
thi
s
dise
ase
i
s
essenti
al
to
pr
eve
nt
f
atal
inc
id
enc
e
.
In
gene
ra
l,
th
e
cheap
and
le
ss
inva
sive
m
et
hod
to
dia
gnos
e
the
d
isea
se
could
be
don
e
b
y
bi
ops
y
using
fine
nee
dl
e
aspira
t
es
from
bre
ast
ti
ss
ue.
How
eve
r
,
rap
id
and
ac
cur
at
e
id
entificat
io
n
of
the
c
anc
er
ce
l
l
patter
n
from
the
cell
b
i
ops
y
is
sti
ll
challe
nging
ta
sk
.
T
his
dia
gnosti
c
t
ool
ca
n
be
d
e
v
e
l
o
p
e
d
u
s
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
a
s
a
c
l
a
s
s
i
f
i
c
a
t
i
o
n
p
r
o
b
l
e
m
.
T
h
e
p
erf
orm
anc
e
of
the
class
ifi
er
depe
nds
on
the
i
nte
rre
l
ationship
bet
wee
n
sam
ple
size
s,
som
e
fea
tur
es,
and
class
ifi
er
complex
ity
.
Thus,
the
r
emoval
of
some
irre
l
eva
nt
fea
tur
es
m
a
y
in
cre
ase
c
la
ss
ifi
c
a
ti
on
ac
cur
acy
.
I
n
thi
s
stud
y
,
a
new
h
y
bri
d
fea
tur
e
sel
ec
t
ion
fast
cor
r
el
a
ti
on
base
d
fe
at
ure
(
F
CBF
)
and
infor
m
at
io
n
gai
n
(
IG
)
was
used
t
o
sele
ct
f
ea
tur
e
s
on
ide
nti
f
y
ing
bre
ast
ca
n
ce
r
using
AIRS
al
gorit
hm
.
The
result
s
of
10
tim
es
the
cro
ss
in
g
(CF)
of
our
v
al
id
at
ion
on
var
ious
AIRS
see
ds
indi
c
at
e
that
the
proposed
m
et
hod
c
an
a
chie
ve
the
b
est
per
form
anc
e
wit
h
accurac
y
=
0
.
9
797
and
AU
C=0
.
9777
a
t
k
=6 and
see
d=
50
.
Ke
yw
or
d
s
:
AI
RS
Breast
cance
r
FCB
F
Hybr
i
d feat
ure
sel
ect
ion
IG
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Ach
m
ad
Ri
dok
Faculty
of Com
pu
te
r
Scie
nc
e (FIL
KO
M)
,
Brawijaya
U
niv
ersit
y
,
8 Vete
ra
n
Roa
d
Ma
la
ng,
Indo
nesia
.
Em
a
il
:
acrido
kb@
ub.ac.i
d
1.
INTROD
U
CTION
Breast
cance
r
i
s
the
m
os
t
fam
ou
s
fem
al
e
kill
er
betwee
n
t
he
ages
of
35
an
d
54
ye
ars
ol
d
a
nd
th
e
m
os
t
com
m
on
cancer
in
wo
m
en
[
1].
The
chea
p
diag
nosti
c
too
l
fo
r
ea
rly
detect
ion
of
this
di
sease
is
essenti
al
to
pr
e
ve
nt
fatal
i
ncide
nce.
T
his
diag
no
sti
c
to
ol
can
be
de
vel
op
e
d
us
in
g
m
a
chine
le
ar
ning
to
identify
can
cer
cel
l
p
r
o
f
i
l
e
f
r
o
m
c
e
l
l
b
i
o
p
s
y
.
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
i
s
o
n
e
o
f
t
h
e
s
u
b
-
f
i
e
l
d
s
o
f
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
t
h
a
t
h
a
s
b
e
e
n
a
p
p
l
i
e
d
t
o
a
u
t
o
m
a
t
i
c
a
l
l
y
i
d
e
n
t
i
f
y
t
h
e
h
i
d
d
e
n
p
a
t
t
e
r
n
s
t
h
r
o
u
g
h
l
e
a
r
n
i
n
g
o
r
c
l
a
s
s
i
f
y
f
r
o
m
ex
p
e
r
i
e
n
c
e
o
f
p
r
e
v
i
o
u
s
data [
2].
Ther
e
a
re
se
ve
ral
cl
assifi
cat
ion
m
et
ho
ds
that
hav
e
bee
n
de
ve
lop
e
d,
s
uc
h
as
arti
fici
al
neu
ra
l
networ
ks
(
A
N
N
)
,
K
-
n
e
a
r
e
s
t
n
e
i
g
h
b
o
r
(
K
N
N
)
,
s
u
p
p
o
r
t
v
e
c
t
o
r
m
a
c
h
i
n
e
(
S
V
M
)
a
n
d
A
I
R
S
.
H
o
w
e
v
e
r
,
o
n
e
o
f
t
h
e
ad
va
ntages of
A
I
R
S
c
o
m
p
a
r
e
d
t
o
o
t
h
e
r
m
e
t
h
o
d
s
i
s
t
h
a
t
t
h
e
e
x
a
c
t
s
e
t
t
i
n
g
s
d
o
n
o
t
n
e
e
d
t
o
b
e
k
n
o
w
n
b
e
f
o
r
e
h
a
n
d
b
u
t
a
r
e
d
e
t
e
r
m
i
n
e
d
b
y
y
o
u
r
s
e
l
f
.
W
i
t
h
t
h
i
s
c
a
p
a
b
i
l
i
t
y
,
t
h
i
s
m
e
t
h
o
d
i
s
c
o
n
s
i
d
e
r
e
d
a
s
m
a
r
t
s
u
p
e
r
v
i
s
e
d
c
l
a
s
s
i
f
i
c
a
t
i
o
n
[
3
]
.
I
n
a
d
d
i
t
i
o
n
,
A
I
R
S
i
s
o
n
e
o
f
t
h
e
s
e
t
e
c
h
n
i
q
u
e
s
t
h
a
t
h
a
s
b
e
e
n
u
s
e
d
s
u
c
c
e
s
s
f
u
l
l
y
i
n
t
h
e
p
r
o
b
l
e
m
o
f
m
e
d
i
c
a
l
c
l
a
s
s
i
f
i
c
a
t
i
o
n
[
4
]
.
The
perform
a
nce
of
a
cl
as
sifie
r
de
pe
nds
on
the
i
nterrel
at
ion
sh
i
p
be
tween
sam
ple
siz
es,
so
m
e
featur
e
s,
a
nd
c
la
ssifie
r
com
plexity
.
Cl
assifi
cat
ion
resu
lt
s
w
ou
l
d
be
bette
r
i
f
usi
ng
m
or
e
tr
ai
nin
g
dataset
s
with
few
e
r
at
tribu
te
s.
Th
us
,
the
re
m
ov
al
of
so
m
e
irreleva
nt
fe
at
ur
es
m
ay
increa
se
cl
assifi
cat
ion
accu
racy
[5
-
8].
Seve
ral
featu
r
e
sel
ect
ion
al
gorithm
s
hav
e
been
us
e
d
t
o
im
pr
ov
e
the
pe
rfor
m
ance
of
the
A
IRS
al
gor
it
h
m
,
su
c
h
as
C4.
5
[
9],
pr
i
ncipal
c
om
po
ne
nt
anal
ysi
s
(P
CA)
[
10
]
,
cor
relat
io
n
-
ba
sed
featu
re
(
CFS)
an
d
FCB
F
[11].
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
A h
y
br
id
feat
ure
selec
ti
on
on
AIR
S meth
od for
i
den
ti
fy
ing
br
e
as
t ca
ncer
disease
s
(
Ach
mad Ri
dok
)
729
Be
sides,
Util
iz
at
ion
of
feat
ur
e
wei
gh
ti
ng
al
gorithm
s
ha
s
bee
n
perf
orm
ed
by
so
m
e
resea
rch
e
rs
su
c
h
as
the
wei
gh
ti
ng
of
f
uzzy
pr
e
processes
an
d
i
nfo
rm
ation
gain
s
(
IG)
[12
-
15]
.
Be
sides
t
hat,
the
pe
rfo
rm
ance
of
the
cl
assifi
cat
ion
al
gorithm
a
lso
is
determ
i
ned
by
the
pa
ram
et
ers
of
each
al
gorithm
.
The
ef
fect
of
set
ti
ng
par
am
et
ers
on
the
perform
ance
of
the
AI
R
S
al
go
rithm
fo
r
so
m
e
data
su
ch
as
Ir
is,
I
onos
phe
re,
Dia
be
te
s
an
d
Sonar
has
bee
n
co
n
du
ct
e
d
[
2].
This
w
ork
has
r
ep
or
te
d
that
dif
fer
e
nt
seed
assi
gn
m
ents
ha
ve
a
si
gnific
ant
eff
ect
on
accu
r
acy
.
Me
anwhil
e,
the
eff
ect
of
set
ti
ng
par
am
et
ers
of
A
IRS
f
or
ide
ntifyi
ng
b
reast
cance
r
da
ta
set
s
has
not
bee
n
c
onduct
ed.
The
r
efore,
this
st
udy
pr
op
os
es
a
c
o
m
bin
ed
m
et
h
od
of
sel
ect
in
g
FCB
F
and
we
igh
ti
ng
featur
e
s IG
on
the AIRS cl
ass
ifie
r
al
gorithm
.
2.
RELATE
D
R
ESE
ARCH
In
this
sect
io
n,
so
m
e
research
es
are
e
xpla
ined
relat
ed
to
effor
ts
of
im
pr
ov
i
ng
t
he
pe
rfor
m
ance
of
AI
RS
classi
fic
at
ion
alg
ori
thm
s inclu
ding
fea
ture
sel
ect
io
n
a
nd f
eat
ur
e
wei
gh
ti
ng
.
2.1.
Ut
il
iz
at
ion
of
features
selec
t
ion in
AI
RS
Po
la
t
et
al.
intr
oduce
d
the
FS
-
AI
RS
m
et
ho
d
t
o
dia
gnos
e
br
e
ast
cancer
[
9].
In
this
r
esearc
h,
the
C
4.5
decisi
on
tree
al
gorithm
was
us
ed
to
sel
ect
featur
es
that
re
duce
from
9
featur
es
to
6
feat
ures.
Ev
al
uatio
n
with
10
-
CV
ob
ta
ine
d
acc
ur
acy
of
98.51
at
k=
1.
I
n
2007,
Po
la
t
and
Sali
h
us
e
d
PCA
-
A
IRS
for
pre
dicti
ng
he
patit
is
disease
that
ac
hieve
d
acc
ur
ac
y
94.12
at
k=
1.
Util
iz
at
ion
of
PCA
i
n
thi
s
stu
dy
has
re
du
ce
d
t
he
num
ber
of
featur
e
s f
r
om
1
9
to 5
[1
6]. The sa
m
e
m
et
ho
d has b
ee
n
us
e
d for lung
ca
ncer detec
ti
on
that
reduce
d
the nu
m
ber
of
featu
res
f
r
om
57
feat
ur
es
t
o
4
featu
res
a
nd
ac
hie
ve
d
10
0%
accur
acy
at
k
K
N
N=
1
T
[
10]
.
Katsi
s
et
al
.
ha
ve
us
e
d
(C
FS)
on
A
IRS
in
or
de
r
to
detect
ea
r
ly
br
east
ca
nc
er
.
T
he
i
nvest
igati
on
res
ult
a
t
k
values
of
1
to
7
sh
owe
d
the
bes
t
resu
lt
s
at
k=3
in
A
IR
S
a
nd
AI
RS+
CF
S
m
e
thods.
Ne
ver
t
he
le
ss,
the
best
resu
lt
s
betwee
n
them
o
c
c
u
r
o
n
A
I
R
S
w
i
t
h
o
u
t
C
F
S
.
B
e
s
i
d
e
s
,
t
h
i
s
s
t
u
d
y
h
a
s
s
h
o
w
n
t
h
a
t
t
h
e
a
c
c
u
r
a
c
y
o
f
t
h
e
A
I
R
S
a
l
g
o
r
i
t
h
m
i
s
b
e
t
t
e
r
t
h
a
n
o
t
h
e
r
c
o
m
p
a
r
a
t
i
v
e
m
e
t
h
o
d
s
,
s
u
c
h
a
s
S
V
M
a
n
d
C
4
.
5
[
1
7
]
.
R
i
d
o
k
e
t
a
l
.
,
[
1
1
]
h
a
v
e
c
o
m
b
i
n
e
d
F
C
B
F
a
s
a
f
eat
ur
es
sel
ect
ion
on AI
RS f
or
cl
assify
ing
breast
can
c
er
dataset
s.
T
hi
s m
et
ho
d
ac
hieved accu
racy
100%
at
k=
1 t
o 30.
2.2.
The w
ei
gh
tin
g o
f
t
he pre
proc
ess
All
featu
res
of
trai
ning
datas
et
are
wei
gh
te
d
us
in
g
a
s
pec
ifie
d
al
gorithm
befo
re
they
a
re
use
d
f
or
trai
ning
in
cl
assifi
cat
ion
.
P
ol
at
et
al
.
hav
e
introdu
ce
d
the
A
IRS
hybr
id
m
et
ho
d
a
nd
the
weig
hti
ng
of
pr
e
processe
d
usi
ng
fuzzy
for
diag
nosin
g
thyroi
d
disease.
T
he
accu
racy of
this
m
et
ho
d
re
ached
85 at k=
2
[
18]
.
Sh
am
sh
irba
nd
et
al
.
[
19]
us
es
a
com
bin
at
ion
of
f
uzzy
lo
gic
an
d
A
IRS
t
o
di
agnose
T
uber
culosis.
The
fe
at
ur
e
s
wer
e
norm
al
ized
th
r
ough
a
f
uzzy
r
ule
base
d
on
a
la
belin
g
syst
e
m
.
The
l
abeled
featu
res
wer
e
cat
eg
or
i
zed
int
o
norm
al
cl
asses
an
d
tu
be
rcu
l
osi
s
us
i
ng
A
IRS
.
T
he
validat
io
n
resu
lt
s
us
in
g
a
10
-
CV
sho
wed
99.
14%
a
ccur
acy
on
t
he
le
ar
ning
rate
of
0.8
.
Se
ver
al
st
ud
ie
s
re
la
te
d
to
the
us
e
of
I
G
-
AI
RS
f
or
weig
htin
g
of
feat
ur
es
ha
ve
bee
n
cond
ucted.
Kodaz
et
al
.
[12
]
intro
duced
t
he
util
iz
at
ion
of
in
form
at
ion
gain
(IG)
on
AI
RS
al
gorithm
to
identify
a
t
h
e
r
o
s
c
l
e
r
o
s
i
s
d
i
s
e
a
s
e
w
i
t
h
s
u
c
c
e
s
s
r
a
t
e
9
9
.
0
9
%
.
W
i
t
h
t
h
e
s
a
m
m
e
t
h
o
d
,
K
o
d
a
z
e
t
a
l
.
[
1
3
]
h
a
v
e
u
s
e
d
t
h
i
s
m
e
t
h
o
d
t
o
i
d
e
n
t
i
f
y
T
h
y
r
o
i
d
d
i
s
e
a
s
e
w
i
t
h
a
c
c
u
r
a
c
y
9
5
.
9
0
.
K
a
r
a
e
t
a
l
.
,
a
l
s
o
a
p
p
l
i
e
d
t
h
i
s
m
e
t
h
o
d
t
o
t
h
e
cl
assifi
c
at
io
n
of m
ic
ro
organ
i
sm
s w
it
h
a su
c
cess rate
of 92.
35 [1
4].
2.3.
Dataset
This
stu
dy
us
e
d
W
isc
onsin
breast
cance
r
da
ta
set
s
colle
ct
ed
by
Dr
.
W.H.
W
ol
berg
f
r
om
fine
-
nee
dle
aspirates
fro
m
hu
m
an
breast
cance
r
t
issue
(
W
BC
D
)
(
https:/
/a
rc
hi
ve.
ic
s.
uci.edu/
m
l/m
achine
-
le
arn
i
ng
-
databases/
br
ea
st
-
cance
r
-
wisc
on
si
n).
The
W
BC
D
consi
sts
of
683
sam
ples
that
div
ide
d
into
two
gro
up,
i.e.,
444
f
or
be
nign,
a
nd
239
f
or
m
a
li
gn
ant.
Ea
ch
of
sam
ples
co
ns
ist
s
of
ni
ne
featur
e
s
as
f
ollow
s:
(
1)
Cl
um
p
thickne
ss
(x1
);
(
2)
U
nif
or
m
ity
of
cel
l
siz
e
(
x2);
(
3) U
ni
for
m
ity
of
cell
sha
pe
(x3
);
(
4) Marginal
ad
he
sion
(
x4);
(5)
Sin
gle
epit
helia
l
cel
l
siz
e
(x5)
;
(
6)
Ba
re
nu
cl
e
oli
(x6)
;
(
7)
Bl
an
d
c
hro
m
at
in
(x
7);
(8)
Norm
al
nu
cl
eoli
(x8)
;
and (
9) Mi
tose
s (x9).
2.4.
Metho
d
As
el
uci
dated
in
the
int
rod
uc
ti
on
that
cl
assi
ficat
ion
perf
orm
ance
can
be
enh
a
nce
d
by
s
el
ect
ion
an
d
weig
hting
of
f
eat
ur
es,
t
her
e
f
or
e
this
stu
dy
proposes
a
c
om
bin
at
ion
of
FCB
F
and
I
G
as
a
hybri
d
featur
e
sel
ect
ion
al
gor
it
h
m
to
i
m
pr
ov
e
A
IRS
cl
ass
ific
a
ti
on
al
gori
thm
'
s
per
form
ance
in
ide
ntifyi
ng
br
east
c
ancer
disease.
The
pr
opos
e
d
m
et
ho
d
can
be
il
lustra
te
d
as
i
n
Fi
g
ure
1.
I
n
t
he
first
ste
p,
the
FCB
F
al
gorithm
re
du
ce
s
the
featur
e
of
the
dataset
with
9
featu
res
in
to
a
dataset
with
6
feature
s.
I
n
the
seco
nd
s
te
p,
IG
al
gorit
hm
i
s
us
e
d
to
determ
ine
the
weig
ht
of
each
new
dataset
featur
e
wh
ic
h
will
be
us
ed
f
or
a
we
igh
ti
ng
facto
r
wh
e
n
cal
culat
ing
the
distance
bet
w
een
feat
ur
es
on
the
inter
nal
AI
RS
.
I
n
the
t
hir
d
ste
p,
t
he
new
dataset
is
div
ide
d
into
two
par
ts,
nam
e
ly
as
a
t
rainin
g
dataset
and
the
ot
her
as
a
te
st
datas
et
us
ing
10
-
f
ol
d
cr
os
s
validat
ion
as
ou
tl
ine
d
in
4.
2.
The
f
ourth
ste
p,
T
he
A
IRS
al
gorithm
ge
ner
at
e
d
cel
l
m
e
m
or
y
as
a
resu
lt
of
the
l
earn
i
ng
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.
11
, No
.
1,
Febr
uar
y
2021
:
72
8
-
73
5
730
process
of
eac
h
trai
ni
ng
data
unde
r
the c
ondi
ti
on
of the
s
pe
c
ifie
d
pa
ram
ete
rs
as d
e
picte
d i
n
Fig
ur
e
2.
T
he
fift
h
ste
p,
K
NN
det
erm
ines
the
la
bel
of
eac
h
trai
ning
data
base
d
on
k
vo
ti
ng
m
ajo
rity
.
In
th
e
la
st
ste
p,
acc
ur
acy
i
s
cal
culat
ed
bas
ed
on
the
num
ber
of
sim
il
arities
betwee
n
th
e
la
bel
res
ults
of
t
he
cl
assifi
c
at
io
n
an
d
t
he
or
i
gin
la
bel
of
each
trai
nin
g
data
set
.
The
third
ste
p
un
ti
l
the
la
st
s
te
p
is
rep
eat
ed
10
tim
es
according
to
the co
m
bin
at
io
n of dataset
pai
rs
a
nd traini
ng
data as t
he
r
esu
lt
o
f 1
0
-
fo
l
d
cr
os
s
-
validat
io
n.
Accor
ding
to
Watkin
et
al
.
[2
]
,
as
m
entioned
i
n
the
int
rod
uction,
cha
ng
e
s
in
seed
cel
ls
aff
ect
the
accu
racy
of
A
IRS
cl
assifi
cat
ion
.
T
he
refor
e
,
the
propos
ed
m
et
ho
d
w
as
evaluate
d
on
t
he
va
riat
ion
of
seed
from
10
to
10
0
to
fin
d
out
the
best
pe
rfo
r
m
ance
of
cl
ass
ific
at
ion
in
te
r
m
s
of
a
ccur
a
c
y
and
A
UC.
I
n
s
ome
pr
e
vious
stu
dies,
in
vestigat
io
ns
we
re
ra
rely
c
onduct
ed
on
K
NN
with
dif
fe
r
ent
k
val
ues.
T
her
e
fore,
the
st
ud
y
inv
est
igate
d
th
e res
ults o
f
the
KNN
cl
assi
ficat
ion
at
k=5,
k=
6
a
nd k
=
7.
Figure
1. A
IRS
classi
ficat
ion
base
d on I
G
a
nd
FCB
F
featur
e
sel
ect
io
n
Figure
2. A
IRS
classi
ficat
ion
base
d
2.5.
Clas
sific
at
i
on
perf
ormance
measure
ment
Me
asur
em
ent
of
cl
assifi
cat
io
n
accu
racy
can
be
e
valuate
d
by
cal
culat
ing
four
c
om
po
ne
nts,
s
uch
a
s
the
nu
m
ber
of
co
rr
ect
ly
rec
ognized
cl
ass
instances
(tr
ue
posit
ive)
,
the
num
ber
of
c
orrectl
y
recog
niz
e
d
instances
that
are
not
include
d
in
the
cl
ass
(
true
ne
gati
ve
),
and
the
w
r
ong
exam
ple
bu
t
assigne
d
to
the
cl
ass
(f
al
se
po
sit
ives
)
or wh
o
is n
ot r
eco
gniz
ed
as
a cla
ss instanc
e (f
al
se
neg
at
iv
es).
All of
t
hes
e com
po
ne
nts
can be
sh
ow
n
as
a
co
nfusion
m
at
rix
sh
ow
n
in
Ta
bl
e
1
for
the
ca
se
of
bin
a
ry
cl
assifi
cat
ion
.
T
he
r
o
w
of
t
he
ta
ble
r
e
p
r
e
s
e
n
t
s
p
r
e
d
i
c
t
e
d
l
a
b
e
l
,
m
e
a
n
w
h
i
l
e
,
t
h
e
a
c
t
u
a
l
l
a
b
e
l
i
s
r
e
p
r
e
s
e
n
t
e
d
b
y
a
c
o
l
u
m
n
o
f
t
h
e
t
a
b
l
e
.
F
r
o
m
t
h
e
ta
ble,
s
o
m
e
c
o
m
m
o
n
l
y
u
s
e
d
m
e
t
r
i
c
s
f
o
r
m
e
a
s
u
r
i
n
g
c
l
a
s
s
i
f
i
c
a
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
c
a
n
b
e
g
e
n
e
r
a
t
e
d
a
s
s
h
o
w
n
i
n
T
a
b
l
e
2.
The
perform
ance
of
th
e
pro
pose
d
m
et
ho
d
was
e
valuated
on
ly
us
ed
thre
e
m
e
tric
s
nam
el
y
accuracy,
e
rror
rate
a
n
d
A
U
C
.
I
n
g
e
n
e
r
a
l
,
t
h
e
a
c
c
u
r
a
c
y
m
e
t
r
i
c
m
e
a
s
u
r
e
s
t
h
e
r
a
t
i
o
o
f
c
o
r
r
e
c
t
p
r
e
d
i
c
t
i
o
n
s
o
v
e
r
t
h
e
t
o
t
a
l
n
u
m
b
e
r
o
f
i
n
s
t
a
n
c
e
s
e
v
a
l
u
a
t
e
d
.
O
t
h
e
r
w
i
s
e
,
m
i
s
c
l
a
s
s
i
f
i
c
a
t
i
o
n
e
r
r
o
r
m
e
a
s
u
r
e
s
t
h
e
r
a
t
i
o
o
f
i
n
c
o
r
r
e
c
t
p
r
e
d
i
c
t
i
o
n
s
o
v
e
r
t
h
e
total
nu
m
ber
of inst
ances ev
al
uated.
F
or
tw
o
-
cl
as
s p
r
ob
le
m
, A
U
C i
s o
ne
of
the
popula
r
ra
nk
i
ng ty
pe
m
et
rics
wh
ic
h
ref
le
ct
s the
ove
rall
r
an
king
pe
rfor
m
ance of a c
la
s
sifie
r
[
20,
21
]
.
Table
1
.
C
onf
usi
on
m
at
rix
Actu
al +label
Actu
al
lab
el
Predicted
+label
Tr
u
e +
(
T
P)
False
-
(FN
)
Predicted
-
lab
el
False +
(FP)
Tr
u
e
-
(
TN)
Tabel
2.
Met
ri
cs [20,
22]
Metr
ics
Fo
r
m
u
la
Accurac
y
(
acc)
+
+
+
+
Er
ror
r
ate
(
e
rr
)
+
+
+
+
Sen
sitiv
ity
(
Rec
all
)
+
Sp
ecif
ity
(
sp
)
+
Precisio
n
(
P)
+
AUC
1
2
(
+
+
+
)
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
A h
y
br
id
feat
ure
selec
ti
on
on
AIR
S meth
od for
i
den
ti
fy
ing
br
e
as
t ca
ncer
disease
s
(
Ach
mad Ri
dok
)
731
3.
METHO
D
3.1.
Sy
s
tem d
e
vel
opme
nt en
viro
nment
The
syst
em
was
de
velo
pe
d
us
in
g
t
he
R
uby
pro
gr
am
m
in
g
la
ngua
ge
with
JR
ub
y
1.7
.3
on
NetB
ea
ns
ID
E
8.1.
T
he
syst
e
m
ru
ns
on
a
W
in
dow
s
7
Profe
ssio
nal
32
-
bit
en
vir
onm
ent
on
a
Lapto
p
with
a
n
I
ntel
(R)
Core (T M
)
i3
-
2310M p
r
ocess
or an
d 4
GB of
RAM.
3.2.
10
-
f
old cr
os
s
-
va
li
d
at
i
on
Mod
el
validat
ion
f
or
m
achin
e
le
arn
in
g
al
gorithm
s
sh
oul
d
en
sure
that
data
we
re
tra
nsfo
rm
ed
to
the
m
od
el
pro
per
ly
an
d
the
m
od
el
rep
rese
nts
the
syst
e
m
with
an
acce
ptable
accu
rac
y
[23]
.
T
her
e
f
or
e
,
to
m
ini
m
iz
e
the
bias
ass
ociat
ed
with
the
ra
ndom
sa
m
pling
us
e
d
durin
g
t
he
trai
ni
ng
ph
ase,
the
syst
e
m
was
evaluate
d
us
in
g
n
-
fo
l
d
c
ross
-
validat
io
n.
In
this
ap
proac
h,
the
i
ns
ta
nces
are
r
an
dom
l
y
div
ide
d
into
n
eq
ual
strat
ifie
d
s
ubse
ts
.
At
eac
h
it
er
at
ion
,
n
-
1
subs
et
s
are
m
erg
ed
to
f
orm
the
trai
nin
g
set
,
a
nd
the
rem
ai
nin
g
s
et
is
us
e
d
as
t
he
te
sti
ng
set
w
hich
cl
assifi
cat
io
n
acc
ur
acy
of
the
al
go
rithm
is
m
easur
ed
on
it
.
T
his
pr
ocess
i
s
rep
eat
e
d
n
ti
m
es,
c
hoos
in
g
a
diff
e
re
nt
subse
t
as
the
te
st
set
each
ti
m
e.
Th
eref
or
e
,
al
l
da
t
a
instances
have
bee
n
us
e
d
n
-
1
ti
m
e
fo
r
trai
ni
ng
an
d
once
for
te
sti
ng.
The
fi
nal
pr
edict
ive
perfor
m
ance
is
c
om
pu
te
d
over
al
l
f
ol
ds
in
the usual m
anner.
T
his st
ud
y
us
e
d
te
nfol
d
c
r
os
s
-
validat
io
n for e
valuati
on
pur
po
ses
.
3.3.
Experim
en
ta
l
fra
mew
ork
To
obta
in
the
best
perform
ance
of
the
pr
opos
e
d
m
et
ho
d
,
it
was
e
valuate
d
in
va
rio
us
c
onditi
ons
k
of
KNN
bet
ween
k=
5,
k=
6
an
d
k=7
on
eac
h
num
ber
of
see
d
as
m
entioned
i
n
2.4.
T
he
res
ul
ts
of
this
evaluati
on
will
b
e c
om
par
ed wit
h
th
ree
m
et
ho
ds as il
lu
strat
ed
in
Ta
ble 3
.
Table
3
. E
xper
i
m
ental
f
ram
e
work
seed
1
0
…
Seed
9
0
AIRS
…
FCB
F+AIRS
…
IG+A
IRS
…
FCB
F+IG+
AIRS
…
Each
m
et
ho
d
was
e
valuated
us
in
g
the
10
-
f
ol
d
cr
os
s
validat
ion
sc
hem
e.
In
this
schem
e,
each
m
et
ho
d
is
evaluated
by
te
n
rep
et
it
ions
.
In
eac
h
re
petit
ion
,
the
datase
t
was
div
i
ded
i
nto
10
sect
io
ns,
1
pa
rt
was
use
d
as
te
st
data
w
hile
the
rem
ai
nd
er
as
trai
ni
ng
da
ta
.
Dete
rm
inatio
n
of
t
he
te
st
data
sect
io
n
is
determ
ined
ba
sed
on
the
re
peat
nu
m
ber.
I
f
t
he
r
epe
at
nu
m
ber
is
1,
the
n
par
t
1
is
ta
ken
as
te
st
data
an
d
the
re
m
ai
nd
er
as
trai
ning
data
as
well
as
soon.
O
ne
lo
op
re
pr
e
sents
t
he
trai
ning
proc
ess
that
pro
du
ce
d
m
c.
At
this
sta
ge
,
t
he
tra
ining
tim
e
and
the
num
ber
of
cel
ls
of
each
m
c
w
ou
l
d
be
cal
cul
at
ed.
T
he
cl
as
sific
at
ion
proc
ess
be
pe
rfor
m
ed
by
fin
ding
the
m
a
j
ori
ty
voti
ng
l
abe
l
of
each
tr
ai
nin
g
data
f
or
k=
5,
k=
6,
an
d
k=7.
T
he
cl
assifi
cat
ion
res
ul
ts
of
each
k
we
re
c
al
culat
ed
accu
racy,
error
rate,
and
AU
C
thr
ough
the
cal
culat
ion
res
ult
fr
om
the
confu
si
on
m
at
rix.
A
t t
he e
nd of eac
h k
-
f
old
proces
s
be c
al
culat
ed
the
aver
a
ge of
eac
h
c
al
culat
io
n.
3.4.
Sett
in
g par
am
eters
As
il
lustrate
d
in
Fig
ur
e
2
,
the
AI
RS
al
gorith
m
has
so
m
e
pr
edeterm
ined
pa
ram
et
ers
as
in
Table
4
of
colum
n
1.
Cl
one
rate
is
us
ed
to
adjust
the
r
at
e
of
cl
on
i
ng,
m
utati
on
rate
to
regulat
e
m
utati
on
of
m
utatio
ns,
thres
h
sti
m
ulatio
n
is
us
e
d
a
s
t
he
a
ver
a
ge
sti
m
m
ing
lim
i
t
value
of
a
cel
l
in
the
po
ol
at
t
he
ti
m
e
of
cl
on
ing
a
nd
m
uta
ti
on
,
m
axi
m
u
m
reso
urc
e
de
no
te
s
t
he
m
axi
m
u
m
a
m
ou
nt
of
res
ource
s
duri
ng
cl
on
in
g
a
nd
m
uta
ti
on
.
Me
anwhil
e,
th
e
FCB
F
al
go
ri
thm
req
uire
s
1
pa
ram
et
er,
i
.e.
.
In
this
case
set
=
v
0.1
as
reco
m
m
end
ed
by
Yu and
Liu
[24
]
. While
on th
e
I
G
alg
or
it
hm
ther
e
is one
pa
r
a
m
et
er th
at
n
ee
ds
t
o be set
na
m
el
y st
ep
=
4
.
Tabl
e
4.
Sett
ing
par
am
et
er
s
AIRS
FCB
F
IG
clo
n
e_
rate
= 10
0
=
0,1
Step
=4
m
u
t
ate_
rate
=
0
.2
sti
m
_
th
r
esh
=
0
.5
m
a
x
_
res =
10
0
0
4.
RESU
LT
S
A
ND
DI
SCUS
S
ION
S
Th
e
e
xperim
e
nt
a
l
results
a
nd
di
s
cu
ssi
on
are
di
s
cu
sse
d
in
t
hi
s
se
ct
i
on
.
Com
par
iso
n
of
a
verage
accuracy
res
ults
betwee
n
pr
opos
e
d
m
et
hod
a
nd
a
no
t
her
m
e
tho
d
on
va
riat
ion
seeds
can
be
pointe
d
ou
t
as
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.
11
, No
.
1,
Febr
uar
y
2021
:
72
8
-
73
5
732
s
h
o
w
n
i
n
F
i
g
u
r
e
3
(
a
)
.
F
r
o
m
t
h
i
s
f
i
g
u
r
e
,
t
h
e
a
c
c
u
r
a
c
y
o
f
t
h
e
p
r
o
p
o
s
e
d
m
e
t
h
o
d
o
u
t
p
e
r
f
o
r
m
s
t
h
e
o
t
h
e
r
t
h
r
e
e
m
e
thods
on see
d=
50.
A
s a c
on
s
e
quenc
e, the p
rop
os
ed
m
e
tho
d
ha
s th
e m
os
t average
erro
r
as
sho
wn in F
i
gure
3(b
).
(a)
(b)
Figure
3
.
(a
)
A
ver
a
ge
acc
urac
y and
(
b)
A
ve
r
age
e
rror rate
The
f
urt
he
r
an
al
ysi
s
we
us
e
d
AU
C
t
o
ide
ntify
the
be
st
cl
assi
ficat
ion
m
eth
od
for
this
st
ud
y.
A
UC
is
on
e
of
the f
a
vorite
rankin
g
ty
pe
m
et
rics
wh
ic
h
ca
n
be
us
e
d
for
c
om
par
ing
learnin
g
al
go
rithm
s
[25]
.
T
he
val
ue
of
A
UC
re
flect
s
the
over
al
l
rankin
g
perfor
m
ance
of
a
cl
assifi
er.
T
he
AU
C
was
pro
ven
the
or
et
ic
al
ly
an
d
e
m
pirical
ly
bette
r
than
t
he
accura
cy
m
e
tric
[26,
27]
for
e
valuati
ng
the
cl
assifi
er
pe
rfo
rm
ance
an
d
discrim
inati
ng
an
op
ti
m
al
so
luti
on
du
rin
g
th
e
cl
assifi
cat
ion
trai
ning.
Th
e
pro
po
se
d
m
et
ho
d
sti
ll
ou
t
perf
or
m
s
com
par
e to t
he
o
the
r
t
hr
ee m
et
hods
as
s
hown in F
i
gure
4
Fig
ure
4. Com
par
is
on of
a
verage
AU
C
The
ave
ra
ge
num
ber
s
of
m
c
gen
e
rated
for
al
l
m
e
tho
ds
a
re
li
near
with
the
nu
m
ber
of
seeds
sp
eci
fi
e
d
as
sh
ow
n
in
Figure
5(
a
).
C
onver
sel
y,
the
de
te
rm
inati
on
of
the
nu
m
ber
of
seeds
in
al
l
m
et
ho
ds
is
inve
rsely
pro
portion
al
to
the
ti
m
e
fo
r
t
r
ai
nin
g
as
point
ed
ou
t
in
Fig
ure
5(
b
).
From
Figures
3
a
nd
4
it
can
be
seen
that
the
highest
pe
rf
or
m
ance
is
achieve
d
w
he
n
seed=
50.
F
ur
t
her
m
or
e,
a
de
eper
i
nv
est
i
gation
on
seed=
50
was
perform
ed
to
determ
ine
the
value
of
k
wh
ic
h
pro
duc
es
the
m
os
t
op
ti
m
al
per
fo
r
m
ance.
The
hi
gh
est
perform
ance
in
te
r
m
s
of
accu
r
acy
and
A
UC
i
s
achi
eve
d
w
he
n
k=
5
wit
h
accurac=
0.9
797
and
A
UC=
0.9
777
as
sh
ow
n
in
Fi
gur
es 6(a)
and
6(b
)
re
sp
ect
ively
.
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
A h
y
br
id
feat
ure
selec
ti
on
on
AIR
S meth
od for
i
den
ti
fy
ing
br
e
as
t ca
ncer
disease
s
(
Ach
mad Ri
dok
)
733
(a)
(b)
F
igure
5.
(a
) A
ver
a
ge n
um
ber
s of m
c and
(
b) ave
rag
e
traini
ng tim
e
(a)
(b)
Figure
6
.
(a
)
T
he best acc
ur
ac
y
and
(
b) the
be
st AU
C
5.
CONCL
US
I
O
N
This
resea
rch
has
trie
d
to
use
a
new
hy
bri
d
featu
re
sel
ect
ion
by
com
bin
in
g
FCB
F
and
IG
f
or
identify
in
g
bre
ast
cancer
disease
us
in
g
AIR
S
al
gorithm
.
Exp
e
rim
entat
i
on
on
dif
fer
e
nt
seed
with
10
-
f
ol
d
c
r
o
s
s
-
v
a
l
i
d
a
t
i
o
n
t
e
s
t
y
i
e
l
d
s
t
h
e
b
e
s
t
p
e
r
f
o
r
m
a
n
c
e
a
t
k
=
6
a
n
d
s
e
e
d
=
5
0
w
i
t
h
a
c
c
u
r
a
c
y
=
0
.
9
7
9
7
a
n
d
A
U
C
=
0
.
9
7
7
7
.
REFERE
NCE
S
[1]
M.
Ghonche
h,
Z
.
Pournam
dar
,
a
nd
H.
Sale
hiniy
a
,
“
Inc
ide
nc
e
and
Mortal
ity
and
E
pide
m
iol
og
y
of
Brea
st
Cancer
in
the
W
orld
,
”
Asia
n
Pacific
J
ournal
of
Cance
r
Prev
ent
ion
(
AP
JCP
)
,
vol. 17, pp. 43
-
46,
2016
.
[2]
A.
W
at
kins,
J.
T
imm
is,
and
L.
C
.
Boggess,
“
Arti
fic
i
al
Im
m
une
Rec
ogni
ti
on
S
y
st
e
m
(AIRS
):
AN
I
m
m
une
Inspire
d
S
u
p
e
r
v
i
s
e
d
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
A
l
g
o
r
i
t
h
m
,
”
G
e
n
e
t
i
c
P
r
o
g
r
a
m
m
i
n
g
a
n
d
E
v
o
l
v
a
b
l
e
M
a
c
h
i
n
e
s
,
v
o
l
.
5
,
n
o
.
3
,
p
p
.
2
9
1
-
3
1
7
,
2
0
0
4
.
[3]
J.
Brownlee,
“
Cle
ver
Algor
it
hm
-
Natur
e
-
Inspire
d
Program
m
ing
Rec
ip
es
,
”
F
irst Edit. Lulu
En
te
rpr
ises
,
2011
.
[4]
M.
A.
Chikh
,
M.
Saidi,
and
N.
S
et
touti,
“
Diagnosis
of
Diabete
s
Disea
ses
Us
ing
An
Artifi
c
ia
l
Im
m
une
Rec
ogni
tion
S
y
s
t
e
m
2
(
A
I
R
S
2
)
w
i
t
h
F
u
z
z
y
K
-
n
e
a
r
e
s
t
n
e
i
g
h
b
o
r
,
”
J
o
u
r
n
a
l
o
f
M
e
d
i
c
a
l
S
y
s
t
e
m
s
,
v
o
l
.
3
6
,
n
o
.
5
,
p
p
.
2
7
2
1
-
2
7
2
9
,
2
0
1
2
.
[5]
T.
M.
Moham
e
d,
“
Eff
i
ci
en
t
br
ea
st
ca
nc
er
d
et
e
ct
ion
using
seq
uent
i
al
f
ea
tur
e
s
el
e
ct
ion
techniq
ues,
”
2015
IE
E
E
S
e
v
e
n
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
I
n
t
e
l
l
i
g
e
n
t
C
o
m
p
u
t
i
n
g
a
n
d
I
n
f
o
r
m
a
t
i
o
n
S
y
s
t
e
m
s
(
I
C
I
C
I
S
)
,
C
a
i
r
o
,
p
p
.
4
5
8
-
4
6
4
,
2
0
1
5
.
[6]
J.
Ta
ng,
S.
Alelyani,
and
H.
L
iu,
“
Feat
ure
Sele
c
tion
for
Cla
ss
ifi
cation:
A
Revi
ew,”
Data
Classif.
Al
gorithms
Appl.
,
pp.
37
-
64
,
2014
.
[7]
H.
Om
ara
,
M
.
Laz
aar,
and
Y
.
Tabii,
“Ef
fect
o
f
Fe
a
ture
S
el
e
ction
on
Gene
E
x
press
ion
Datasets
Classifi
ca
ti
on
A
c
c
u
r
a
c
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
(
I
J
E
C
E
)
,
v
o
l
.
8
,
n
o
.
5
,
p
p
.
3
1
9
4
-
3
2
0
3
,
2
0
1
8
.
[8]
O.
Cigde
m
and
H.
Dem
ire
l,
“
Perform
anc
e
an
aly
sis
of
diffe
r
ent
cl
assificat
ion
a
lgori
thms
using
diffe
ren
t
featur
e
sele
c
ti
on
m
et
hod
s on
Parkinson’s di
sea
se
de
te
c
ti
o
n,
”
Journal
o
f
N
euroscie
nc
e
M
ethods
,
vol. 309, p
p.
81
-
90
,
2018
.
[9]
K.
Polat,
S.
Sah
an,
H.
Kodaz
,
a
nd
S.
Günes,
“
A
New
Cla
ss
ifi
c
ation
Method
for
Brea
st
Can
ce
r
Diagnosis:
Fea
tur
e
Sele
c
ti
on
Artific
ia
l
Im
m
une
Recogniti
on
S
y
st
em
(FS
-
AIR
S),”
Inte
rnational
Conf
ere
nce
on
Natur
al
Computati
on
,
vol.
3611
,
pp
.
83
0
-
838,
2005
.
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.
11
, No
.
1,
Febr
uar
y
2021
:
72
8
-
73
5
734
[10]
K.
P
o
l
a
t
a
n
d
S
.
G
ü
n
e
ş
,
“
C
o
m
p
u
t
e
r
a
i
d
e
d
m
e
d
i
c
a
l
d
i
a
g
n
o
s
i
s
s
y
s
t
e
m
b
a
s
e
d
o
n
p
r
i
n
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
l
y
s
i
s
a
n
d
a
r
t
i
f
i
c
i
a
l
i
m
m
u
n
e
r
e
c
o
g
n
i
t
i
o
n
s
y
s
t
e
m
c
l
a
s
s
i
f
i
e
r
a
l
g
o
r
i
t
h
m
,
”
E
x
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
4
,
n
o
.
1
,
p
p
.
77
3
-
779,
2008
.
[11]
A.
Ridok,
F.
M.
W
a
y
an
,
and
M.
Rifa
i
,
“
An
Im
prove
d
Artif
ic
i
al
I
m
m
une
Rec
ogni
ti
on
S
y
s
te
m
wit
h
Fast
Corre
la
t
io
n
Based
Filt
er
(F
CBF
)
for
Feat
ure
Selecti
on
,
”
2017
Fourth
Inte
rnational
Con
fe
renc
e
on
Image
Information
Proce
ss
ing
(
ICIIP
)
,
Shim
la
,
pp
.
1
-
6,
2017
.
[12]
H.
Kodaz
,
S.
Ka
ra,
F.
Latioǧl
u
,
S.
Güneş,
F.
La
t
ioğl
u,
and
S.
Gü
neş,
“
A
new
hy
b
rid
cl
assifi
er
s
y
s
te
m
:
Inform
at
io
n
gai
n
-
base
d
ar
ti
fi
ci
a
l
imm
une
re
c
ognit
ion
s
y
st
em,”
E
xpe
rimen
tal
Techni
ques
,
vol
.
31,
no.
6,
pp.
36
-
43,
2007
.
[13]
H.
Kodaz
,
S
.
Özşen,
A.
Ars
l
an,
and
S.
Güneş,
“
Medic
al
appl
i
c
at
ion
of
informa
ti
on
ga
in
base
d
art
if
ic
i
al
imm
une
rec
ogni
ti
on
s
y
st
em
(AIRS
):
Dia
gnosis
of
th
y
roi
d
disea
se
,
”
Ex
p
e
rt
Syste
ms
wit
h
Appl
ic
a
ti
ons
,
vo
l.
36
,
no
.
2
,
par
t
2,
pp.
3086
-
3092
,
2009.
[14]
S.
Kara
,
B.
H
.
A
ksebz
eci,
H
.
Ko
daz
,
S.
Güne
ş,
E
.
Ka
y
a
,
and
H.
Özbil
ge
,
“
Medical
app
lication
of
informati
on
g
ai
n
-
base
d
art
if
ic
i
al
i
m
m
une
rec
ognit
ion
s
y
stem
(IG
-
AIRS
):
Cla
ss
ifi
ca
ti
on
of
m
ic
roorga
nism
spec
ie
s,”
Ex
pert
Syste
ms
wit
h
App
licati
on
s
,
vol. 36, no. 3
,
pa
rt
1
,
pp
.
5168
-
5172,
2009
.
[15]
S.
Chorm
unge
a
nd
S.
Jena
,
“
Eff
ic
ie
n
t
fea
tur
e
subs
et
sele
ction
al
g
orit
hm
for
high
dimensional
data,
”
Int
ernati
ona
l
Journal
of
Elec
t
rical
and
Computer
Eng
ine
ering
(
IJE
CE)
,
vol. 6,
no.
4
,
pp
.
1880
-
1888,
2016
.
[16]
K.
Polat
and
S.
Güneş,
“
A
h
y
bri
d
appr
oa
ch
to
m
edi
c
al
d
ec
ision
support
s
y
st
ems
:
combining
fe
at
u
re
sel
ec
t
ion,
f
u
z
z
y
we
i
g
h
t
e
d
p
r
e
-
p
r
o
c
e
s
s
i
n
g
a
n
d
A
I
R
S
,
”
C
o
m
p
u
t
e
r
M
e
t
h
o
d
s
a
n
d
P
r
o
g
r
a
m
s
i
n
B
i
o
m
e
d
i
c
i
n
e
,
v
o
l
.
8
8
,
n
o
.
2
,
p
p
.
1
6
4
-
17
4
,
2
0
0
7
.
[17]
C.
D.
K
at
sis,
I
.
Gkogkou,
C.
A.
Papadopoul
o
s,
Y.
Gol
et
sis,
and
P.
V
.
Bouf
ounou,
“
Us
ing
Artifi
c
ia
l
Im
m
u
n
e
Rec
ognition
S
y
s
te
m
s
in
Order
t
o
Dete
ct
E
arly
Brea
st
Cancer,”
Inte
rnational
J
ournal
of
Inte
ll
i
gent
Syste
ms
and
Appl
ic
a
ti
ons
(
IJI
SA)
,
vol. 5, no. 2
,
pp
.
34
-
40
,
201
3.
[18]
K.
Polat,
S.
Şah
an,
and
S.
Güne
ş,
“
A
novel
h
y
br
id
m
et
hod
base
d
on
artifi
ci
a
l
im
m
une
rec
ogni
ti
o
n
s
y
stem
(AIRS
)
with
fuz
z
y
we
ig
hte
d
pr
e
-
proc
essing
for
th
y
roid
d
isea
se
d
ia
gnosis,
”
E
xpe
rt
Syste
ms
wit
h
Applicatio
ns
,
vol.
32,
no.
4
,
pp.
1141
-
1147
,
2
007.
[19]
S.
Sham
shirband
et
al.,
“
Tub
erculosis
disea
se
d
i
agnosis
using
artific
i
al
imm
une
rec
ogni
ti
on
s
y
s
tem
,
”
Inte
rnat
iona
l
Journal
of
Me
d
i
cal
S
ci
en
ce
s
,
vol
.
11
,
no
.
5
,
pp
.
5
08
-
514,
2014
.
[20]
M.
Hos
sin
and
M.
N.
Sulai
m
an,
“
Revi
ew
on
Ev
al
ua
ti
on
Metr
ic
s
for
Data
Cl
a
ss
ifica
t
ion
Evaluatio
ns,”
Inte
rnat
ion
al
Journal
of
Data
Mini
ng
&
Knowl
edge
Manag
eme
nt
Proc
ess
,
vol
.
5,
no
.
2
,
pp
.
1
-
1
1,
2015
.
[21]
J.
Huang
and
C.
X.
Li
ng,
“
Us
in
g
AU
C
and
ac
c
ura
c
y
in
eva
lu
ati
ng
le
arn
ing
al
go
rit
hm
s,”
in
IEEE
Tr
ansacti
ons
on
Knowle
dge
and
Data
Engi
n
ee
rin
g
,
vol
.
17
,
no
.
3
,
pp.
299
-
310
,
20
05.
[22]
M.
Sokolova
a
nd
G.
L
apalm
e,
“
A
s
y
stema
tic
an
aly
sis
of
per
form
anc
e
m
ea
sures
for
cla
ss
ifi
ca
ti
on
t
asks,”
Information
Pro
ce
ss
ing
&
Mana
geme
nt
,
vol
.
45
,
no.
4
,
pp
.
427
-
4
37,
2009
.
[23]
C.
Cat
al,
“
Perform
anc
e
eva
lu
at
i
on
m
et
ric
s
for
s
oftwa
re
fau
l
t
pre
diction
studie
s,
”
Ac
ta
Polyt
ec
h
nic
a
Hungar
ic
a
,
vol.
9
,
no
.
4
,
pp
.
193
-
206,
2012
.
[24]
L.
Yu
and
H.
Li
u,
“
Feat
ure
sele
ction
for
high
-
dimensiona
l
dat
a
:
a
fast
co
rre
l
at
ion
-
b
ase
d
fil
te
r
soluti
o
n,
”
Proce
ed
ings o
f
t
he
tw
ent
i
et
h
inte
rnat
ional
conf
er
enc
e
on
mach
ine l
earning
,
W
ashi
ngton
DC,
pp
.
1
-
8,
2003
.
[25]
A.
Rakot
om
amonj
y
,
“
Optimiz
ing
Area
Under
Ro
c
Curve
wi
th
SV
Ms
,”
ROCAI
,
pp
.
71
-
80
,
2004
.
[26]
C.
X.
Li
ng,
J.
H
uang,
and
H.
Zh
ang,
“
AU
C:
A
b
et
t
er
m
ea
sure
th
an
accura
c
y
in
c
om
par
ing
le
arn
i
ng
al
gori
thms
,
”
Confe
renc
e
of
th
e
Canadian
Society f
or Computat
ional
Stud
ie
s o
f
I
nte
lligen
ce
,
vol
.
2671,
pp
.
329
-
3
41,
2003
.
[27]
C.
X.
Li
ng
,
J.
Huang,
and
H.
Zha
ng,
“
AU
C:
A
stat
isti
call
y
c
onsistent
and
m
ore
discr
iminat
i
ng
m
ea
sure
th
a
n
ac
cur
acy
,
”
in
In
t
ernati
onal
Joi
n
t
Confe
renc
e
on
A
rtif
icial
Intelli
g
e
nce
(
IJCA
I)
,
pp
.
519
-
524
,
2003
.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Achmad
Ridok
is
a
doct
or
al
st
udent
a
t
the
d
ep
art
m
ent
of
Bio
lo
gi,
Facu
lty
of
M
at
hemat
ic
s
an
d
Natur
al
Sci
ence,
Brawij
a
y
a
Uni
ver
sit
y
.
He
obt
ai
ned
his
Ba
chelor
degr
e
e
in
Mate
m
atics
from
depa
rtment
of
M
at
hemat
ic,
Facu
lty
of
Ma
the
m
at
i
cs
and
Natu
ral
Scie
n
ce,
Brawi
jaya
Univ
ersi
t
y
.
He
rec
e
ive
d
his
M.Kom
.
in
Co
m
pute
r
Scie
nc
e
from
the
Univer
sit
y
of
Indone
si
a.
His
rese
ar
ch
int
er
ests
ar
e
m
achine
le
arn
ing, a
r
t
ifi
cial
intelligen
c
e
and
biominfor
m
at
ic
s.
Widod
o,
Ph.
D
is
profe
ss
or
at
Biol
og
y
Dep
artm
ent
,
Brawij
a
y
a
Univer
sit
y
Ind
o
nesia
.
He
now
ac
t
ive
r
ese
ar
ch
on
el
uc
idation
the
m
olecula
r
m
ec
hani
sm
of
JA
MU
(Indon
esia
tr
adi
t
ional
m
edi
ci
ne)
b
y
using
Meta
bolomi
c,
Meta
g
enomic
and
Bioi
nform
at
i
c
appr
oac
h
.
Since
2019,
He
was
appoi
nte
d
b
y
Dir
ec
tor
ate
ge
ner
al
of
highe
r
educ
a
ti
on
for
co
nduct
world
cl
a
ss
rese
arc
h
to
el
uc
ida
t
e
the c
o
m
ple
xity
of
JA
MU
.
Also,
he
active
as
boar
d
m
e
m
ber
of
Inte
rna
tional
As
socia
t
ion
for
Com
par
at
ive
Medic
in
e
(IACM
).
He
got
sev
era
l
awa
rds
such
as,
y
oung
inv
esti
gat
o
r
from
Ti
ss
ue
Cult
ur
e
As
socia
ti
on
-
j
apan
(2007),
Best
o
r
igi
nal
pape
r
fro
m
nestl
e
council
rese
ar
ch
-
Japa
n
(2008),
Young
scie
nti
st
from
Kalbe
-
Rist
ek
Dikti
-
Indon
esia
(
2012)
and
Best
te
ac
h
er
from
Brawij
a
y
a
Unive
rsit
y
(2014
).
He
has
expe
ri
ence
a
s
visit
ing
Profes
sor
at
Ristum
ei
k
an
Univer
sit
y
-
Japa
n
(20017
-
20
19).
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
A h
y
br
id
feat
ure
selec
ti
on
on
AIR
S meth
od for
i
den
ti
fy
ing
br
e
as
t ca
ncer
disease
s
(
Ach
mad Ri
dok
)
735
Way
an
Firdau
s
Mahmud
y
obta
in
ed
ba
che
lo
r
degr
e
e
in
m
at
hemat
ic
s
fro
m
Univer
sita
s
Brawij
a
y
a
,
Indo
nesia
,
m
aste
r
d
e
gre
e
in
informat
i
on
te
chnol
og
y
fr
om
Instit
ut
Te
kn
ologi
Sepuluh
N
o
v
e
m
b
e
r
(
I
T
S
)
,
I
n
d
o
n
e
s
i
a
,
a
n
d
c
o
m
p
l
e
t
e
d
h
i
s
P
h
.
D
.
i
n
M
a
n
u
f
a
c
t
u
r
i
n
g
E
n
g
i
n
e
e
r
i
n
g
a
t
Uni
ver
sit
y
of
South
Aus
tra
li
a
.
He
is
a
Lectu
r
er
at
Dep
art
m
en
t
of
Com
pute
r
Scie
n
ce,
Univer
si
ta
s
Brawija
y
a.
His re
sea
r
ch
in
terests i
nc
lude opt
imiza
ti
on
of
co
m
bina
tori
a
l
prob
le
m
s a
nd
m
ac
h
in
e
l
ea
rning
.
Muhaimin
Rifa'i
,
Ph.
D
is
a
Pr
ofe
ss
or
of
Im
munol
og
y
in
Dep
art
m
ent
of
B
iolog
y
,
Brawija
y
a
Univer
sit
y
,
Indo
nesia
.
He
is
no
w
ac
t
ive
l
y
cond
uct
ing
rese
arc
h
using
her
ba
l
m
edi
ci
n
e.
He
is
a
scie
nti
st
who
is
ac
t
ive
in
conducting
r
ese
ar
ch
an
d
scie
n
ti
fi
c
publ
i
ca
t
ions
in
v
ari
ou
s
int
ern
at
ion
al
journa
ls
in
cl
udi
ng
journa
ls
th
at
have
a
high
re
puta
ti
on
.
Curre
n
tly
he
is
condu
ct
ing
r
ese
ar
ch
rel
a
te
d
to
c
anc
er
,
dia
be
te
s
m
el
li
t
us,
and
at
her
osc
l
ero
sis.
As
a
rese
arc
her
he
al
so
gi
ves
le
ct
ur
es
at
the
und
erg
rad
u
at
e
,
m
aste
rs,
a
nd
doct
or
al
d
e
gre
e.
After
hol
d
ing
the
positi
on
as
he
ad
of
the
Ph
y
s
iol
og
y
La
bora
tor
y
,
he
i
s
cur
ren
t
l
y
be
co
m
e
the
h
ea
d
o
f
Biol
og
y
Dep
artm
ent
,
Brawi
jay
a
Univer
sit
y
,
Indo
nesia
.
Evaluation Warning : The document was created with Spire.PDF for Python.