I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
A
I
)
V
ol
.
10
, N
o.
2
,
J
une
2021
, pp.
430
~
437
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
10
.i
2
.pp
430
-
437
430
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
C
om
p
ar
i
son
so
m
e
of
k
e
r
n
e
l
f
u
n
c
t
i
on
s
w
i
t
h
su
p
p
or
t
ve
c
t
or
m
ac
h
i
n
e
s c
l
ass
i
f
i
e
r
f
or
t
h
al
ass
e
m
i
a d
at
ase
t
I
ls
ya Wir
as
at
i,
Z
u
h
e
r
m
an
R
u
s
t
am
, Jan
e
E
va A
u
r
e
li
a, S
r
i
H
ar
t
in
i,
G
lo
r
i
S
t
e
p
h
a
n
i
S
ar
agi
h
Department of Mathemati
cs, University of
Indonesia, In
donesia
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
M
a
r
7
, 20
20
R
e
vi
s
e
d
M
a
r
11
, 20
21
A
c
c
e
pt
e
d
A
pr
1
2, 20
21
In
the
medical
field,
ac
curate
classification
of
medical
data
is
really
important
because
of
its
impact
on
disease
detection
and
patient’s
treatment.
Technology,
machine
learning,
is
needed
to
help
medical
staff
to
i
mprove
accuracy
to
classify
disease.
Thi
s
research
discussed
some
kernel
fu
nctions,
such
as
g
au
ssian
radial
basis
function
(RBF)
kernel
,
Polynomial
k
e
rn
el,
and
linear
kernel
with
support
vector
machine
(SVM)
to
classify
thala
ssemia
data.
Thalassemia
is
a
genetic
blood
disorder
which
is
also
one
of
th
e
major
public
health
problems.
In
this
paper,
there
is
an
explanation
about
thalassemia
,
SVM
,
and
some
of
the
kernel
functions
that
serve
as
a
comprehens
ive
source
for
the
next
research
about
this
topic.
Furth
ermore,
there is a
comparison
result
from
three ke
rnel
functions to
find
out wh
ich one
has
the
best
performance.
The
result
is
g
aussian
RBF
kernel
with
SV
M
is
the
best method with an average of a
ccuracy 99,63%.
K
e
y
w
o
r
d
s
:
C
la
s
s
if
ic
a
ti
on
K
e
r
ne
l
f
unc
ti
on
M
a
c
hi
ne
l
e
a
r
ni
ng
S
uppor
t
ve
c
to
r
m
a
c
hi
ne
T
ha
la
s
s
s
e
m
i
a
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
I
ls
ya
W
ir
a
s
a
ti
D
e
pa
r
tm
e
nt
of
M
a
th
e
m
a
ti
c
s
U
ni
ve
r
s
it
y of
I
ndone
s
ia
J
l.
P
r
of
. D
R
. S
udj
ono
D
. P
us
pone
gor
o, P
ondok C
in
a
, D
e
pok, J
a
w
a
B
a
r
a
t
16424, I
ndone
s
ia
E
m
a
il
:
il
s
ya
.w
ir
a
s
a
ti
@
ui
.a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
T
ha
la
s
s
e
m
ia
is
one
of
th
e
m
a
in
publ
ic
he
a
lt
h
pr
obl
e
m
s
w
it
h
hi
ghl
y
pr
e
va
le
nt
in
th
e
a
r
e
a
e
xt
e
ndi
ng
f
r
om
s
ub-
S
a
ha
r
a
n
A
f
r
ic
a
,
th
r
ough
th
e
M
e
di
te
r
r
a
ne
a
n
r
e
gi
on
a
n
d
M
id
dl
e
E
a
s
t,
to
th
e
I
ndi
a
n
s
ubc
ont
in
e
nt
a
nd
E
a
s
t
a
nd
S
out
he
a
s
t
A
s
ia
[
1
]
,
[
2
]
.
H
ow
e
ve
r
,
m
ig
r
a
ti
ons
of
pe
opl
e
c
a
us
e
d
th
a
la
s
s
e
m
ia
ge
ne
s
to
s
pr
e
a
d
th
r
oughout
th
e
w
o
r
ld
a
nd
e
xt
e
nd
to
I
ndone
s
ia
.
T
he
r
e
a
r
e
7%
of
th
e
w
or
ld
'
s
popu
la
ti
on
a
s
c
a
r
r
ie
r
s
o
f
t
ha
la
s
s
e
m
ia
w
it
h
th
e
d
e
a
th
of
a
bout
50,000
-
100,000 c
hi
ld
r
e
n
[
3]
.
I
n
I
ndone
s
ia
,
t
ha
la
s
s
e
m
ia
i
s
one
of
th
e
m
os
t
c
om
m
on c
hr
oni
c
di
s
e
a
s
e
s
[
4]
. C
ur
r
e
nt
ly
, t
ha
la
s
s
e
m
ia
r
a
nk
s
5t
h a
m
ong non
-
c
om
m
uni
c
a
bl
e
di
s
e
a
s
e
s
a
f
te
r
he
a
r
t
di
s
e
a
s
e
,
c
a
nc
e
r
,
ki
dne
y,
a
nd
s
tr
oke
w
it
h
th
e
num
be
r
of
c
a
r
r
ie
r
s
3.8%
of
th
e
e
nt
ir
e
popula
ti
on
in
I
ndone
s
ia
.
B
a
s
e
d
on
da
ta
f
r
om
th
e
I
ndone
s
ia
n
T
h
a
la
s
s
e
m
ia
F
ounda
ti
on,
th
e
r
e
ha
s
be
e
n
a
s
te
a
dy
in
c
r
e
a
s
e
in
t
ha
la
s
s
e
m
ia
c
a
s
e
s
f
r
om
2012 unti
l
2018
[
3]
.
T
ha
la
s
s
e
m
ia
is
a
ge
n
e
ti
c
di
s
e
a
s
e
be
c
a
u
s
e
of
bl
ood
di
s
or
de
r
s
in
he
r
it
e
d
f
r
om
f
a
m
il
y.
T
ha
la
s
s
e
m
ia
s
uf
f
e
r
e
r
s
'
body ma
ke
s
a
n a
bnor
m
a
l
f
o
r
m
or
a
n i
na
de
qua
te
a
m
ount
of
he
m
ogl
obi
n
[
1
]
,
[
5]
.
H
e
m
ogl
obi
n a
ll
ow
s
r
e
d
bl
ood
c
e
ll
s
to
c
a
r
r
y
oxyge
n
[
6]
.
W
he
n
th
e
r
e
is
not
e
nough
he
m
ogl
obi
n,
th
e
body’
s
r
e
d
bl
ood
c
e
ll
s
do
not
f
unc
ti
on pr
ope
r
ly
, a
nd t
he
y di
e
m
or
e
qui
c
kl
y. A
nd t
he
n, t
he
oxyge
n de
li
ve
r
e
d t
o a
ll
t
he
ot
he
r
c
e
ll
s
of
t
he
body
is
not
e
nough.
T
he
c
a
us
e
of
t
ha
la
s
s
e
m
ia
i
s
m
ut
a
ti
ons
i
n t
he
D
N
A
of
c
e
ll
s
t
ha
t
m
a
ke
he
m
ogl
obi
n [
7]
. H
e
m
ogl
obi
n i
s
m
a
de
of
two
di
f
f
e
r
e
nt
pa
r
ts
,
c
a
ll
e
d
a
lp
ha
a
nd
be
ta
.
T
he
r
e
f
or
e
,
th
e
r
e
a
r
e
two
ty
pe
s
of
th
a
la
s
s
e
m
ia
,
s
uc
h
a
s
a
lp
ha
-
th
a
la
s
s
e
m
ia
or
be
t
a
-
th
a
la
s
s
e
m
ia
.
A
c
c
or
di
ng
to
[
8]
,
th
e
ne
w
c
la
s
s
if
ic
a
ti
on
ha
s
be
e
n
s
im
pl
if
ie
d
ba
s
e
d
on
th
e
w
a
y
of
tr
e
a
tm
e
nt
na
m
e
ly
non
-
tr
a
ns
f
us
io
n
-
de
p
e
nde
nt
th
a
la
s
s
e
m
ia
(
N
T
D
T
)
a
nd
tr
a
ns
f
us
io
n
-
de
pe
nd
e
nt
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
C
om
par
is
on
s
om
e
of
k
e
r
ne
l
fu
nc
ti
ons
w
it
h s
uppor
t
v
e
c
to
r
m
a
c
h
in
e
s
c
la
s
s
if
ie
r
f
or
…
(
I
ls
y
a W
ir
as
at
i
)
431
th
a
la
s
s
e
m
ia
(
T
D
T
)
.
B
e
c
a
us
e
of
di
f
f
e
r
e
nc
e
s
in
tr
e
a
tm
e
nt
,
e
a
r
ly
de
te
c
te
d
th
a
la
s
s
e
m
ia
w
it
h
a
s
c
r
e
e
ni
ng
pr
oc
e
s
s
is
ne
c
e
s
s
a
r
y
to
he
lp
th
a
la
s
s
e
m
ia
s
uf
f
e
r
s
to
ge
t
th
e
r
ig
ht
tr
e
a
tm
e
nt
.
T
he
a
im
is
to
in
c
r
e
a
s
e
th
e
ir
li
f
e
e
xpe
c
t
a
nc
y
a
nd
r
e
duc
e
th
e
r
is
k
of
th
a
la
s
s
e
m
ia
to
th
e
ne
xt
ge
n
e
r
a
ti
on.
T
hu
s
,
it
is
im
por
ta
nt
to
obt
a
in
a
pr
e
c
is
e
th
a
la
s
s
e
m
ia
di
a
gnos
is
.
N
ow
a
da
ys
,
in
he
a
lt
hc
a
r
e
,
it
is
s
ig
ni
f
ic
a
nt
to
in
ve
s
t
th
e
de
ve
lo
pm
e
nt
in
c
om
put
e
r
te
c
hnol
ogy
t
o
e
nha
nc
e
pr
oc
e
s
s
in
g
th
e
m
e
di
c
a
l
da
ta
[
5]
.
M
a
c
hi
ne
le
a
r
ni
ng
te
c
hnol
ogi
e
s
,
one
of
c
om
put
e
r
te
c
hnol
ogy,
c
a
n
he
lp
us
in
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
on
la
r
ge
d
a
ta
s
e
t
s
.
I
t
ha
s
a
n
i
m
por
ta
nt
r
ol
e
be
c
a
us
e
it
c
a
n
be
a
ppl
ie
d
in
da
il
y
li
f
e
s
uc
h
a
s
bi
om
e
di
c
a
l
da
ta
.
H
ow
e
ve
r
,
th
e
r
e
a
r
e
s
e
ve
r
a
l
in
te
r
e
s
ti
ng
c
ha
ll
e
nge
s
r
e
c
e
nt
ly
s
uc
h
a
s
our
da
ta
m
a
y
c
om
e
f
r
om
m
ul
ti
pl
e
he
te
r
oge
ne
ous
s
our
c
e
s
,
our
d
a
ta
m
a
y
ha
ve
a
huge
num
be
r
of
s
a
m
pl
e
s
a
nd
r
e
qui
r
e
a
m
e
th
od
to
unde
r
s
ta
nd
th
e
c
om
pl
e
x
m
ode
l,
a
nd
our
da
ta
m
a
y
ha
ve
f
e
w
s
a
m
pl
e
s
but
li
e
in
hi
gh
di
m
e
ns
io
n
a
nd
is
s
pa
ti
ot
e
m
por
a
l.
N
e
w
de
v
e
lo
pm
e
nt
s
i
n s
ta
ti
s
ti
c
s
a
nd k
e
r
ne
l
m
e
th
ods
i
s
r
e
qui
r
e
d t
o t
he
s
e
c
ha
ll
e
nge
s
[
9]
.
Th
e
r
e
a
r
e
s
om
e
m
e
th
ods
on
pr
e
vi
ous
r
e
s
e
a
r
c
he
s
to
c
la
s
s
if
y
th
a
la
s
s
e
m
ia
,
s
uc
h
a
s
f
uz
z
y
ke
r
ne
l
r
obus
t
C
-
m
e
a
ns
,
f
uz
z
y
C
-
m
e
a
ns
,
a
nd
f
uz
z
y
k
e
r
ne
l
C
-
m
e
a
n
s
[
4]
,
ne
u
r
a
l
ne
twor
ks
a
nd
ge
n
e
ti
c
pr
ogr
a
m
m
in
g
[
10]
,
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
a
lg
or
it
hm
s
[
11]
,
a
r
ti
f
ic
ia
l
ne
ur
a
l
n
e
twor
k
[
12]
,
a
nd
na
ïv
e
ba
ye
s
[
13]
.
A
ls
o,
[
12
]
,
[
14]
us
e
d S
V
M
t
ha
t
s
ho
w
e
d good r
e
s
ul
t
w
it
h 93.2%
a
c
c
ur
a
c
y
a
nd 1
00%
A
U
C
r
e
s
pe
c
ti
ve
ly
.
T
hi
s
r
e
s
e
a
r
c
h
u
s
e
d
s
om
e
of
ke
r
ne
l
f
unc
ti
ons
w
it
h
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
to
c
la
s
s
if
y
t
ha
la
s
s
e
m
ia
.
S
V
M
c
a
n
be
m
odi
f
ie
d
w
it
h
va
r
io
us
ke
r
ne
l
f
unc
ti
ons
,
a
s
a
n
e
s
s
e
nt
ia
l
c
om
pone
nt
,
to
g
e
t
a
b
e
tt
e
r
r
e
s
ul
t.
T
he
r
e
f
or
e
,
a
c
om
pa
r
is
on
be
tw
e
e
n
th
a
t
e
s
s
e
nt
ia
l
c
om
pon
e
nt
f
or
c
la
s
s
if
yi
ng
th
a
la
s
s
e
m
ia
s
houl
d
be
done
.
I
t
w
il
l
he
lp
th
e
m
e
di
c
a
l
s
ta
f
f
to
ove
r
c
om
e
th
e
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
.
T
hi
s
r
e
s
e
a
r
c
h
di
s
c
u
s
s
e
d
s
om
e
of
th
e
ke
r
ne
l
f
unc
ti
ons
s
uc
h
a
s
th
e
li
ne
a
r
ke
r
ne
l,
pol
ynomi
a
l
ke
r
ne
l,
a
nd
ga
us
s
ia
n
r
a
di
a
l
ba
s
is
k
e
r
ne
l.
T
he
a
im
i
s
to
f
in
d
out
w
hi
c
h
ke
r
ne
l
f
unc
ti
on
th
a
t
gi
ve
s
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
f
or
c
la
s
s
if
yi
ng
th
a
la
s
s
e
m
ia
in
th
e
S
V
M
m
e
th
od.
2.
R
E
S
E
A
R
C
H
M
E
T
H
O
D
S
uppor
t
ve
c
to
r
m
a
c
hi
ne
s
(
S
V
M
)
is
s
upe
r
vi
s
e
d
m
a
c
hi
ne
le
a
r
ni
ng.
O
r
ig
in
a
ll
y,
S
V
M
a
lg
or
it
h
m
pr
opos
e
d
by
V
a
pni
k
a
nd
L
e
r
ne
r
[
15
]
,
[
16]
.
S
V
M
c
a
n
b
e
a
ppl
ie
d
f
or
c
la
s
s
if
ic
a
ti
on a
nd
r
e
gr
e
s
s
io
n
[
17
]
,
[
18]
.
I
t
c
la
im
e
d t
ha
t
S
V
M
i
s
a
m
e
th
od
t
ha
t
ha
s
a
hi
gh
a
c
c
ur
a
c
y f
or
c
la
s
s
if
ic
a
ti
on [
19]
. M
a
ppi
ng
f
or
m
i
nput
s
pa
c
e
t
o a
hi
ghe
r
di
m
e
ns
io
na
l
s
pa
c
e
i
s
t
he
i
de
a
of
S
V
M
. S
V
M
c
ons
tr
uc
ts
a
hype
r
pl
a
ne
t
o s
e
pa
r
a
te
da
ta
i
nt
o c
la
s
s
e
s
[
20]
.
T
he
s
e
l
e
c
te
d hype
r
pl
a
ne
s
a
r
e
t
hos
e
t
h
a
t
m
a
xi
m
iz
e
t
he
m
a
r
gi
n of
c
la
s
s
if
ic
a
ti
on
e
dge
s
[
21]
.
L
e
t
{
,
}
is
th
e
da
ta
s
e
t
w
h
e
r
e
,
is
f
e
a
tu
r
e
of
ve
c
to
r
,
is
c
la
s
s
la
be
l
f
or
,
a
nd
N
is
th
e
num
be
r
of
s
a
m
pl
e
s
. T
o f
in
d t
he
be
s
t
hype
r
pl
a
ne
, t
hi
s
i
s
m
a
in
f
o
r
m
ul
a
of
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
s
:
(
)
=
∙
+
(
1)
T
ha
t
f
or
m
ul
a
c
ont
a
in
s
w
(
w
e
ig
ht
)
a
s
th
e
or
th
ogona
l
ve
c
to
r
to
th
e
hype
r
pl
a
ne
de
te
r
m
in
in
g
it
s
or
ie
nt
a
ti
on,
b
(
bi
a
s
)
a
s
th
e
di
s
ta
n
c
e
f
r
om
th
e
or
ig
in
to
th
e
hype
r
pl
a
n,
a
nd
x
in
di
c
a
te
s
th
e
tr
a
in
in
g
s
a
m
pl
e
[
22]
.
T
he
a
im
i
s
to
m
a
xi
m
iz
e
t
he
m
a
r
gi
n.
M
or
e
ove
r
, S
V
M
goa
l
is
c
on
s
tr
uc
t
th
e
t
w
o pl
a
ne
s
, l
e
t
s
a
y H
1 a
n
d H
2, a
s
(
2)
a
nd (
3)
:
1
⟶
+
=
+
1
=
+
1
(
2)
2
⟶
+
=
−
1
=
−
1
(
3)
w
he
r
e
t
he
pl
a
ne
f
or
t
he
pos
it
iv
e
c
la
s
s
i
s
+
≥
+
1
is
a
nd t
he
pl
a
ne
f
or
t
he
ne
ga
ti
ve
c
la
s
s
i
s
+
≤
−
1
.
S
e
e
F
ig
ur
e
1
il
lu
s
tr
a
te
th
e
hype
r
pl
a
ne
i
n
S
V
M
.
T
he
pr
obl
e
m
of
S
V
M
opt
im
iz
a
ti
on c
a
n be
w
r
it
te
n a
s
:
M
in
im
iz
e
1
2
‖
‖
2
(
4)
s
.t
.
(
∙
+
)
≥
1
,
∀
=
1
,
…
,
(
5)
B
y s
ol
vi
ng t
he
pr
obl
e
m
a
bove
, f
or
m
ul
a
of
a
nd
c
a
n be
w
r
it
te
n a
s
:
=
∑
=
1
(
6)
=
1
∑
(
−
∑
)
(
7)
T
he
n,
de
c
i
s
io
n
f
or
m
ul
a
s
of
S
V
M
c
a
n be
w
r
it
te
n a
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
2
,
J
une
20
2
1
:
430
–
437
432
(
)
=
(
∙
+
)
(
8)
F
ig
ur
e
1. I
ll
us
tr
a
ti
on of
S
V
M
[
16]
S
V
M
ha
s
s
e
v
e
r
a
l
a
dva
nt
a
ge
s
,
s
uc
h
a
s
it
s
c
a
pa
bi
li
ty
to
pr
oc
e
s
s
d
a
ta
w
it
h
la
r
ge
a
m
ount
s
in
hi
gh
di
m
e
ns
io
ns
[
23]
.
A
ls
o,
S
V
M
im
pl
e
m
e
nt
e
d
e
a
s
il
y
us
in
g
li
ne
a
r
bounda
r
ie
s
a
s
s
how
n
in
F
ig
ur
e
1
.
H
ow
e
ve
r
,
th
e
r
e
a
r
e
c
la
s
s
ifi
c
a
ti
on
pr
obl
e
m
s
w
he
r
e
c
a
n not us
in
g a
l
in
e
a
r
b
ounda
r
y t
o s
e
pa
r
a
te
c
la
s
s
e
s
[
24]
. S
e
e
F
ig
ur
e
2
,
th
a
t
c
a
s
e
is
non
-
li
ne
a
r
s
e
pa
r
a
bl
e
d
a
ta
.
T
he
be
s
t
w
a
y
to
a
ppr
oa
c
h
a
non
-
li
ne
a
r
de
c
i
s
io
n
bounda
r
y
is
to
e
xpa
nd
th
e
or
ig
in
a
l
f
e
a
tu
r
e
s
pa
c
e
.
N
e
ve
r
th
e
le
s
s
,
it
m
a
ke
s
c
om
put
a
ti
o
ns
in
tr
a
c
ta
bl
e
be
c
a
us
e
th
e
or
ig
in
a
l
f
e
a
tu
r
e
is
e
nl
a
r
ge
d t
o hi
gh dim
e
ns
io
na
l
s
pa
c
e
. T
o t
a
c
kl
e
t
ha
t
is
s
ue
, w
e
a
p
pl
ie
d t
he
'
ke
r
ne
l
tr
ic
k'
us
in
g a
ke
r
ne
l
f
unc
ti
on.
F
ig
ur
e
2. N
on
-
li
ne
a
r
s
e
pa
r
a
bl
e
da
ta
[
25]
S
V
M
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
c
lo
s
e
ly
r
e
li
e
s
o
n
th
e
ke
r
ne
l
f
unc
ti
on
[2
6
]
.
T
he
r
e
f
or
e
,
a
ke
r
ne
l
f
unc
ti
on
is
th
e
m
o
s
t
e
s
s
e
nt
i
a
l
c
om
pone
nt
to
m
a
ke
th
e
S
V
M
m
e
th
od
ge
t
hi
ghe
r
a
c
c
ur
a
c
y
[2
7
]
.
W
he
n
a
ta
s
k
is
di
f
f
ic
ul
t
in
th
e
or
ig
in
a
l
pr
obl
e
m
s
pa
c
e
,
ke
r
ne
l
f
unc
ti
on
he
lp
s
to
tr
a
ns
f
or
m
in
put
s
pa
c
e
in
to
a
not
he
r
s
p
ace
w
he
r
e
w
e
c
a
n
w
or
k
e
a
s
ie
r
[
2
5
]
.
O
n
a
not
he
r
w
or
d,
ke
r
ne
l
f
unc
t
io
n
w
or
k
f
or
tr
a
ns
f
or
m
in
g
da
ta
in
to
a
hi
ghe
r
-
di
m
e
ns
io
na
l
s
pa
c
e
[
28
]
,
[
29]
.
I
ts
a
ppr
oa
c
h
is
m
a
ppi
ng
da
ta
in
to
ke
r
ne
l
s
pa
c
e
w
he
r
e
da
ta
b
e
c
om
e
li
ne
a
r
ly
s
e
pa
r
a
bl
e
[
26]
.
T
he
ke
r
ne
l
f
unc
ti
on
c
a
n be
w
r
it
te
n a
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
C
om
par
is
on
s
om
e
of
k
e
r
ne
l
fu
nc
ti
ons
w
it
h s
uppor
t
v
e
c
to
r
m
a
c
h
in
e
s
c
la
s
s
if
ie
r
f
or
…
(
I
ls
y
a W
ir
as
at
i
)
433
(
,
)
=
<
(
)
,
(
)
>
(
9)
E
xa
m
pl
e
,
w
e
c
ons
tr
uc
t
li
ft
in
g
m
ap
:
→
ℋ
w
it
h
:
(
1
,
2
)
→
(
1
2
+
√
2
1
2
+
2
2
)
.
T
hi
s
m
a
p
li
ft
in
g
th
e
da
ta
f
r
om
=
ℝ
2
to
ℋ
=
ℝ
3
[
30]
.
T
he
r
e
f
or
e
,
m
a
ppi
ng da
ta
f
r
om
di
m
e
ns
io
na
l
s
pa
c
e
t
o f
e
a
tu
r
e
s
p
a
c
e
.
T
he
pr
obl
e
m
of
S
V
M
opt
im
iz
a
ti
on w
il
l
be
a
s
f
ol
lo
w
s
:
M
in
im
iz
e
1
2
‖
‖
2
+
C
∑
=
1
s
.t
(
∙
(
)
+
)
−
1
+
≥
0
,
∀
=
1
,
…
,
By
s
ol
vi
ng t
he
pr
obl
e
m
a
bove
, f
or
m
ul
a
of
a
nd
w
il
l
be
a
s
(
10)
a
nd (
11)
:
∗
=
∑
(
)
=
1
(
10)
∗
=
1
∑
(
−
∑
(
)
)
(
11)
T
he
n,
de
c
i
s
io
n
f
or
m
ul
a
s
of
S
V
M
w
il
l
be
a
s
(
12)
:
(
)
=
(
∗
∙
(
)
+
∗
)
(
12)
w
he
r
e
is
s
la
c
k
va
r
ia
bl
e
or
m
e
a
s
ur
e
of
th
e
m
is
c
la
s
s
if
ic
a
ti
on
e
r
r
or
s
w
hi
c
h
s
houl
d
be
m
in
im
iz
e
.
C
is
th
e
pe
na
lt
y
or
de
te
r
m
in
e
s
th
e
tr
a
de
-
of
f
be
twe
e
n
th
e
m
in
im
iz
a
ti
on
of
e
r
r
or
a
nd
th
e
m
a
xi
m
iz
a
ti
on
of
th
e
c
la
s
s
ifi
c
a
ti
on ma
r
gi
n.
I
n t
hi
s
r
e
s
e
a
r
c
h, a
ut
hor
s
pr
opos
e
d t
hr
e
e
k
e
r
ne
ls
w
hi
c
h a
ppl
ie
d f
or
t
ha
la
s
s
e
m
ia
c
la
s
s
if
ic
a
ti
on
:
a.
G
a
us
s
ia
n r
a
di
a
l
ba
s
is
k
e
r
ne
l
(
,
)
=
e
x
p
‖
−
‖
2
2
2
(
13)
W
he
r
e
s
is
th
e
onl
y
pa
r
a
m
e
te
r
th
a
t
de
fi
ne
s
w
id
th
ke
r
ne
l.
I
ts
im
pa
c
t
to
c
lo
s
e
or
f
a
r
a
s
in
gl
e
tr
a
in
in
g
s
a
m
pl
e
r
e
a
c
h
e
s
.
A
ls
o,
σ
c
a
n
de
f
in
e
d
a
s
th
e
r
a
di
us
of
in
fl
ue
nc
e
of
s
a
m
pl
e
s
w
hi
c
h
i
s
a
f
f
e
c
te
d
by
th
e
c
la
s
s
ifi
c
a
ti
on
m
od
e
l.
F
r
om
r
e
s
e
a
r
c
h
in
[
16]
,
a
s
m
a
ll
σ
in
di
c
a
te
s
th
e
w
id
th
of
th
e
ke
r
ne
l
i
s
s
m
a
ll
s
o
th
e
m
ode
l
f
oc
us
e
s
on
a
s
m
a
ll
s
e
t
of
da
ta
a
nd
th
e
n
e
w
hype
r
s
ur
f
a
c
e
w
il
l
be
s
pi
ky.
I
t
m
a
y
le
a
ds
to
a
n
ove
r
fi
tt
in
g
pr
obl
e
m
.
T
he
oppos
it
e
,
a
hi
gh
σ
in
c
r
e
a
s
e
s
th
e
ke
r
ne
l
w
id
th
a
nd
th
e
n
m
os
t
of
th
e
da
ta
a
r
e
tr
a
ns
f
or
m
e
d
in
to
a
fl
a
t
hype
r
s
pa
c
e
w
hi
c
h l
e
a
ds
t
o t
he
unde
r
f
it
ti
ng pr
obl
e
m
.
b.
P
ol
ynomi
a
l
ke
r
ne
l
(
,
)
=
(
<
,
>
+
1
)
(
14)
W
he
r
e
s
d
is
d
e
gr
e
e
of
pol
ynomi
a
l
ke
r
ne
l
f
unc
ti
on.
F
r
om
r
e
s
e
a
r
c
h
in
[
16]
,
hi
gh
de
gr
e
e
w
oul
d
in
c
r
e
a
s
e
t
he
c
om
pl
e
xi
ty
of
t
he
c
la
s
s
ifi
c
a
ti
on mode
l.
I
t
c
a
n be
s
e
e
n a
s
ove
r
f
it
ti
ng pr
obl
e
m
be
c
a
us
e
t
e
s
ti
ng e
r
r
or
in
c
r
e
a
s
e
s
but
tr
a
in
in
g
e
r
r
or
de
c
r
e
a
s
e
s
.
T
he
oppos
it
e
,
w
it
h
a
s
m
a
ll
d
m
a
y
le
a
d
s
to
a
hi
gh
bi
a
s
a
nd
lo
w
va
r
ia
nc
e
or
unde
r
f
it
ti
ng pr
ob
le
m
.
c.
L
in
e
a
r
ke
r
ne
l:
(
,
)
=
(
15)
T
hi
s
k
e
r
ne
l
f
unc
ti
on
is
th
e
s
im
pl
e
s
t
k
e
r
ne
l
f
unc
ti
on
w
hi
c
h
t
he
r
e
s
ul
ts
of
le
a
r
ni
ng
a
lg
or
it
hm
s
a
r
e
of
te
n
e
qui
va
le
nt
to
S
V
M
w
it
hout
ke
r
ne
l
f
unc
ti
ons
[
16]
.
B
y
c
om
pa
r
in
g
th
e
s
e
ke
r
ne
ls
,
th
e
e
xpe
c
ta
ti
on
is
w
e
know
w
hi
c
h
ke
r
ne
l
gi
ve
s
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y.
T
o
c
a
lc
ul
a
te
th
e
a
c
c
ur
a
c
y,
a
c
onf
us
io
n
m
a
tr
ix
is
us
e
d.
T
he
f
or
m
ul
a
f
or
a
c
c
ur
a
c
y i
s
:
=
+
+
+
+
(
16)
=
+
(
17)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
2
,
J
une
20
2
1
:
430
–
437
434
=
+
(
18)
1
=
2
+
(
19)
:
N
um
be
r
of
s
a
m
pl
e
s
ha
vi
ng t
ha
la
s
s
e
m
ia
c
l
a
s
s
if
ie
d c
or
r
e
c
tl
y.
:
N
um
be
r
of
he
a
lt
hy pe
opl
e
t
ha
t
w
e
r
e
i
nc
or
r
e
c
tl
y c
la
s
s
if
ie
d t
o t
ha
la
s
s
e
m
i
a
.
:
N
um
be
r
of
s
a
m
pl
e
s
w
it
h t
ha
la
s
s
e
m
ia
t
ha
t
w
e
r
e
i
nc
or
r
e
c
tl
y c
la
s
s
if
ie
d a
s
he
a
lt
hy.
:
N
um
be
r
of
he
a
lt
hy i
ndi
vi
dua
ls
c
or
r
e
c
tl
y s
pot
te
d.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
I
n
th
is
pa
pe
r
,
th
a
la
s
s
e
m
ia
da
ta
r
e
c
e
iv
e
d
f
r
om
H
a
r
a
pa
n
K
it
a
C
hi
ld
r
e
n
a
nd
W
om
e
n'
s
H
o
s
pi
ta
l,
I
ndone
s
ia
,
a
nd
it
c
ons
is
t
of
150
s
a
m
pl
e
s
.
T
he
da
ta
s
e
t
of
th
a
la
s
s
e
m
ia
r
e
pr
e
s
e
nt
e
d
by
10
va
r
ia
bl
e
s
s
uc
h
a
s
H
e
m
ogl
obi
n
(
/
)
,
H
a
e
m
a
to
c
r
it
P
e
r
c
e
nt
(
%
)
,
L
e
ukoc
yt
e
C
ount
(
103/
)
,
B
a
s
ophi
ls
P
e
r
c
e
nt
(
%
)
,
E
os
in
ophi
ls
P
e
r
c
e
nt
(
%
)
,
R
od
N
e
ut
r
ophi
ls
P
e
r
c
e
nt
(
%
)
,
S
e
g
m
e
nt
N
e
u
tr
ophi
ls
P
e
r
c
e
nt
(
%
)
,
L
ym
phoc
yt
e
s
P
e
r
c
e
nt
(
%
)
,
M
onoc
yt
e
s
P
e
r
c
e
nt
(
%
)
,
a
nd
P
la
te
le
t
C
ount
s
(
103
/
)
.
B
y
de
f
a
ul
t
a
ut
hor
s
ut
il
iz
e
d
th
e
S
ha
pi
r
o
-
W
il
k a
lg
or
it
hm
t
o a
s
s
e
s
s
t
he
nor
m
a
li
ty
of
t
he
di
s
tr
ib
ut
io
n of
i
ns
ta
nc
e
s
w
it
h r
e
s
pe
c
t
to
t
he
f
e
a
tu
r
e
. A
ba
r
pl
ot
a
s
s
how
n
in
F
ig
ur
e
3
,
is
th
e
n
dr
a
w
n
s
how
in
g
th
e
r
e
la
ti
ve
r
a
nks
of
e
a
c
h
f
e
a
tu
r
e
.
P
la
te
t
C
ount
s
ha
s
th
e
hi
ghe
s
t
r
a
nki
ng.
F
ig
ur
e
3. R
a
nki
ng of
t
ha
la
s
s
e
m
ia
da
ta
f
e
a
tu
r
e
s
w
it
h s
a
phi
r
o a
lg
or
it
hm
T
hi
s
r
e
s
e
a
r
c
h
us
e
d
tr
a
in
in
g
da
ta
di
ve
r
s
e
f
r
om
10%
to
90%
a
nd
u
s
e
d
=
0
.
1
f
or
G
a
us
s
ia
n
R
B
F
ke
r
ne
l
a
nd
d=
3
f
or
pol
ynomi
a
l
ke
r
ne
l.
T
h
e
r
e
a
s
on
is
,
f
r
om
th
e
num
be
r
of
th
e
e
xp
e
r
im
e
nt
th
a
t
is
obt
a
in
e
d,
=
0
.
1
a
nd d=
3 ha
s
t
he
b
e
s
t
pe
r
f
or
m
a
nc
e
.
T
hi
s
c
ho
s
e
n
=
0
.
1
is
a
ls
o s
uppor
te
d by
[
16]
.
I
t
is
s
how
n i
n T
a
bl
e
1, t
he
S
V
M
m
ode
l
w
it
h a
g
a
us
s
ia
n r
a
di
a
l
b
a
s
is
f
unc
ti
on ke
r
ne
l
pr
oduc
e
s
t
he
be
s
t
a
c
c
ur
a
c
y
f
or
c
la
s
s
if
yi
ng
t
ha
la
s
s
e
m
ia
da
ta
w
it
h
a
n
a
ve
r
a
ge
of
a
c
c
ur
a
c
y
99.63%
.
T
he
s
e
c
ond
-
be
s
t
is
a
li
ne
a
r
ke
r
ne
l
w
it
h
98.23%
a
c
c
ur
a
c
y.
T
he
l
a
s
t
one
i
s
a
po
ly
nom
ia
l
ke
r
ne
l
w
it
h
97.9%
a
c
c
ur
a
c
y.
L
in
e
a
r
ke
r
ne
l
S
V
M
ha
s
th
e
be
s
t
a
c
c
ur
a
c
y
of
100%
w
it
h
10%
a
nd
30%
tr
a
in
in
g
da
ta
.
O
n
th
e
ot
he
r
s
id
e
,
th
e
pol
ynomi
a
l
k
e
r
ne
l
ha
s
th
e
be
s
t
a
c
c
ur
a
c
y
of
100%
if
th
e
m
ode
l
us
e
s
10
-
30%
a
nd
50%
tr
a
in
in
g
da
ta
.
A
nd
f
or
g
a
us
s
ia
n
r
a
d
ia
l
ba
s
is
f
unc
ti
on
gi
ve
s
th
e
be
s
t
a
c
c
ur
a
c
y
w
it
h
10
-
50%
,
70
%
,
a
nd
80
%
tr
a
in
in
g
da
ta
.
F
or
F
1
S
c
or
e
,
ga
u
s
s
ia
n
r
a
di
a
l
ba
s
i
s
s
ti
ll
t
he
be
s
t
one
. I
n T
a
bl
e
2,
th
e
ga
u
s
s
ia
n r
a
di
a
l
ba
s
is
ke
r
ne
l
gi
ve
s
t
he
be
s
t
pe
r
f
or
m
a
nc
e
w
it
h a
n a
ve
r
a
g
e
pr
e
c
is
io
n of
99.56%
a
nd a
n
a
ve
r
a
ge
r
e
c
a
ll
of
99.78%
.
H
ow
e
ve
r
, t
he
r
e
i
s
a
di
f
f
e
r
e
nc
e
i
n s
e
c
ond pla
c
e
be
twe
e
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
C
om
par
is
on
s
om
e
of
k
e
r
ne
l
fu
nc
ti
ons
w
it
h s
uppor
t
v
e
c
to
r
m
a
c
h
in
e
s
c
la
s
s
if
ie
r
f
or
…
(
I
ls
y
a W
ir
as
at
i
)
435
pr
e
c
is
io
n
a
nd
r
e
c
a
ll
.
S
V
M
li
ne
a
r
is
in
s
e
c
ond
pl
a
c
e
f
or
pr
e
c
i
s
io
n,
w
hi
le
f
or
r
e
c
a
ll
S
V
M
pol
ynomi
a
l
is
in
s
e
c
ond pla
c
e
.
T
a
bl
e
1.
T
he
a
c
c
ur
a
c
y a
nd F
1
s
c
or
e
of
S
V
M
w
it
h
ke
r
ne
l
f
unc
ti
on
T
r
a
i
ni
n
g
D
a
t
a
A
c
c
ur
a
c
y
F
1
S
c
or
e
S
V
M
L
i
n
e
a
r
S
V
M
P
o
l
yn
o
m
i
a
l
S
V
M
G
a
u
s
s
i
a
n
S
V
M
L
i
n
e
a
r
S
V
M
P
o
l
yn
o
m
i
a
l
S
V
M
G
a
u
s
s
i
a
n
10%
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
20%
9
6.
6
7
1
00
.
00
1
00
.
00
9
6.
0
0
1
00
.
00
1
00
.
00
30%
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
40%
98
.00
9
6.
6
7
1
00
.
00
9
8.
0
0
9
6.
0
0
1
00
.
00
50%
9
8.
6
7
1
00
.
00
1
00
.
00
9
9.
0
0
1
00
.
00
1
00
.
00
60%
9
8.
8
9
9
7.
7
7
9
8.
8
9
9
9.
0
0
9
8.
0
0
9
9.
0
0
70%
9
7.
1
4
9
6
.1
9
1
00
.
00
9
8.
0
0
9
6.
0
0
1
00
.
00
80%
9
9.
1
6
9
7.
5
1
00
.
00
9
8.
0
0
1
00
.
00
1
00
.
00
90%
9
5.
5
6
9
3.
3
3
9
7.
7
8
9
7.
0
0
9
3.
0
0
9
9.
0
0
A
v
e
r
a
g
e
98
.23
97
.94
99
.
63
98.33
98.11
99.78
T
a
bl
e
2.
T
he
pr
e
c
is
io
n a
nd r
e
c
a
ll
of
S
V
M
w
it
h
ke
r
ne
l
f
unc
ti
on
T
r
a
i
ni
n
g
D
a
t
a
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
S
V
M
L
i
n
e
a
r
S
V
M
P
o
l
yn
o
m
i
a
l
S
V
M
G
a
u
s
s
i
a
n
S
V
M
L
i
n
e
a
r
S
V
M
P
o
l
yn
o
m
i
a
l
S
V
M
G
a
u
s
s
i
a
n
10%
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
20%
1
00
.
00
1
00
.
00
1
00
.
00
9
3.
0
0
1
00
.
00
1
00
.
00
30%
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
1
00
.
00
40%
1
00
.
00
1
00
.
00
1
00
.
00
9
5.
0
0
9
4.
0
0
1
00
.
00
50%
9
8.
0
0
1
00
.
00
1
00
.
00
9
8.
0
0
1
00
.
00
1
00
.
00
60%
1
00
.
00
1
00
.
00
1
00
.
00
9
8.
0
0
9
6.
0
0
9
8.
0
0
70%
9
5.
0
0
9
5.
0
0
1
00
.
00
9
8.
0
0
9
8.
0
0
1
00
.
00
80%
1
00
.
00
1
00
.
00
1
00
.
00
9
4.
0
0
9
6.
0
0
1
00
.
00
90%
1
00
.
00
9
7.
0
0
9
6.
0
0
9
5.
0
0
9
2.
0
0
1
00
.
00
A
v
e
r
a
g
e
99.22
99.11
99.56
96.78
97.33
99.78
A
ut
hor
s
a
ls
o
u
s
e
d
ot
h
e
r
m
a
c
hi
ne
l
e
a
r
ni
ng,
s
uc
h
a
s
K
N
N
w
it
h
k=
7
a
nd
r
a
ndom
f
or
e
s
t
.
T
h
e
r
e
s
ul
t
i
s
90%
a
c
c
ur
a
c
y
f
r
om
K
N
N
a
nd
100%
a
c
c
ur
a
c
y
f
r
om
r
a
ndom
f
or
e
s
t
.
N
e
ve
r
th
e
le
s
s
,
S
V
M
w
it
h
s
om
e
of
th
e
ke
r
ne
l
f
unc
ti
ons
s
ti
ll
gi
ve
t
he
hi
ghe
s
t
a
c
c
ur
a
c
y, 100%
, s
o i
t
c
a
n
be
s
a
id
t
ha
t
S
V
M
pe
r
f
or
m
e
d t
he
be
s
t
m
a
c
hi
ne
le
a
r
ni
ng me
th
od t
o c
la
s
s
if
y t
ha
la
s
s
e
m
ia
.
4.
C
O
N
C
L
U
S
I
O
N
M
a
c
hi
ne
le
a
r
ni
ng
c
a
n
he
lp
m
e
di
c
a
l
s
ta
f
f
to
c
la
s
s
if
y
t
ha
la
s
s
e
m
i
a
di
s
e
a
s
e
pr
e
c
is
e
ly
.
I
f
e
a
r
ly
de
te
c
ti
on
is
done
, pa
ti
e
nt
s
c
a
n ge
t
th
e
r
ig
ht
t
r
e
a
tm
e
nt
.
I
t
he
lp
s
t
he
m
i
nc
r
e
a
s
e
t
he
ir
l
if
e
e
xpe
c
ta
nc
y a
nd r
e
duc
e
t
he
r
is
k of
th
a
la
s
s
e
m
ia
to
th
e
ne
xt
g
e
ne
r
a
ti
on.
I
n
th
is
r
e
s
e
a
r
c
h,
th
e
r
e
a
r
e
th
r
e
e
ke
r
ne
l
f
unc
ti
ons
us
e
d
in
S
V
M
w
it
h
li
ne
a
r
,
pol
ynomi
a
l,
a
nd
ga
us
s
ia
n
r
a
di
a
l
ba
s
i
s
f
unc
ti
on
ke
r
ne
l.
K
e
r
ne
l
f
unc
ti
on
c
a
n
he
lp
S
V
M
to
tr
a
n
s
f
or
m
in
put
s
pa
c
e
i
nt
o a
hi
ghe
r
-
di
m
e
ns
io
na
l
s
pa
c
e
w
h
e
r
e
w
e
c
a
n
w
or
k e
a
s
ie
r
.
F
r
om
th
is
r
e
s
e
a
r
c
h,
s
uppor
t
ve
c
to
r
m
a
c
hi
n
e
w
it
h
g
a
u
s
s
ia
n
R
B
F
k
e
r
ne
l
is
th
e
be
s
t
one
to
c
la
s
s
if
y
t
ha
la
s
s
e
m
ia
d
a
ta
f
r
om
H
a
r
a
pa
n
K
it
a
C
hi
ld
r
e
n
a
nd
W
om
e
n'
s
H
o
s
pi
ta
l,
I
ndone
s
ia
.
W
e
c
a
n
s
e
e
in
T
a
bl
e
1, e
a
c
h
ke
r
ne
l
pe
r
f
or
m
s
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y.
H
ow
e
ve
r
,
if
w
e
s
e
e
th
e
a
ve
r
a
ge
a
c
c
ur
a
c
y,
g
a
u
s
s
ia
n
R
B
F
i
s
th
e
b
e
s
t
one
w
it
h
a
n
a
c
c
ur
a
c
y
of
99.63%
.
T
he
s
e
c
ond
-
be
s
t
is
a
li
ne
a
r
ke
r
ne
l
w
it
h
98.23%
a
c
c
ur
a
c
y.
T
h
e
la
s
t
one
i
s
a
pol
ynomi
a
l
ke
r
ne
l
w
it
h
97.9%
a
c
c
ur
a
c
y.
B
e
s
id
e
s
th
a
t,
th
e
g
a
us
s
ia
n
r
a
di
a
l
ba
s
i
s
a
ls
o
gi
ve
s
th
e
hi
ghe
s
t
F
1
s
c
or
e
of
99.78%
.
A
ls
o,
in
T
a
bl
e
2,
th
e
g
a
us
s
ia
n
R
B
F
ke
r
ne
l
h
a
s
th
e
hi
ghe
s
t
a
ve
r
a
ge
of
pr
e
c
is
io
n
a
nd
r
e
c
a
ll
w
it
h
99.56%
a
nd
99.78%
r
e
s
p
e
c
ti
ve
ly
.
F
or
f
ut
ur
e
r
e
s
e
a
r
c
h,
us
e
a
la
r
ge
r
d
a
ta
s
e
t
is
r
e
c
om
m
e
nde
d
to
g
e
ne
r
a
te
hi
ghe
r
a
c
c
ur
a
c
ie
s
in
e
a
c
h
m
e
th
od.
A
ls
o,
w
e
be
li
e
ve
th
a
t
f
ut
ur
e
r
e
s
e
a
r
c
h
c
a
n
de
ve
lo
p
th
is
m
e
th
od
to
g
iv
e
th
e
be
s
t
a
c
c
ur
a
c
y f
or
pr
e
di
c
ti
ng or
c
la
s
s
if
yi
ng othe
r
di
s
e
a
s
e
s
.
A
C
K
N
O
WL
E
D
G
E
M
E
N
T
S
T
hi
s
r
e
s
e
a
r
c
h
f
ul
l
s
uppor
te
d
f
in
a
nc
ia
ll
y
by
U
ni
ve
r
s
it
y
of
I
ndon
e
s
ia
,
w
it
h
a
P
U
T
I
S
A
I
N
T
E
K
E
S
2020
r
e
s
e
a
r
c
h
gr
a
nt
s
c
he
m
e
(
I
D
num
be
r
N
K
B
-
2408/UN2.R
S
T
/HK
P
.05.00/2020)
.
A
nd
a
l
s
o,
a
ut
hor
s
f
e
lt
gr
a
te
f
ul
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
10
, N
o.
2
,
J
une
20
2
1
:
430
–
437
436
a
nd w
a
nt
t
o t
ha
nk t
he
H
a
r
a
pa
n K
it
a
C
hi
ld
r
e
n a
nd W
om
e
n'
s
H
os
pi
ta
l
, I
ndone
s
ia
, f
or
pr
ovi
di
ng
t
he
t
ha
la
s
s
e
m
ia
da
ta
s
e
t.
R
E
F
E
R
E
N
C
E
S
[1]
R.
Risoluti
et
al
.
,
“Update
on
thalassem
ia
diagnosi
s:
New
insigh
ts
and
methods
,
”
in
Talanta
,
vol.
183,
pp.
216
-
222,
2018
,
doi:
10.1016/j.talanta.2018.02.071
.
[2]
A
.
T
Taher,
D
.
J
.
Weatherall,
and
M
.
D
.
Cappellini
,
“Thalassemia
,
”
in
The
Lancet
,
vol.
391,
pp.157
-
167,
2018
,
doi
:
10.1016/S0140
-
6736(17)31822
-
6
.
[3]
“Minist
ry of heal
th republ
ic of Ind
onesia,”
http://www.
kemk
es.go.i
d/
[
accessed
2020 Feb 11].
[4]
Z.
Rustam
et
al
.
,
“
Comparison
of
Fuzzy
C
-
Means,
Fuzzy
Kernel
C
-
Means,
and
Fuzzy
Kernel
Robust
C
-
Means
t
o
Classify
Thalasse
mia
Data
,
”
in
Internati
onal
Journal
on
Advan
ced
Science,
Engineering
and
Informati
on
Techno
logy
, vol.
9
, no. 4, pp.
1205
-
1210, 2019
,
d
oi
:
10.18517/ijaseit.9.4.9580
.
[5]
N
.
C
.
Egejuru
et
al
.
,
“
Using
Data
Mining
Algorithms
for
Thalassemia
Risk
Prediction
,
”
in
Internati
onal
Journal
of
Biomedical Science and Engineering
,
vol.
7
, no.
2
, pp.
33
-
44
, 2019
,
doi:
10.11648/j.ijbse.20190702.12
.
[6]
S.
Fuchar
oen
and
P.
Winichago
on
,
"
Thalassemia
and
abnormal
hemoglobin
,
"
in
Internati
onal
journal
of
hematology
vol.
76
, no.
2
, pp.
83
-
89
, 2002
,
d
oi
:
10.1007/BF03165094
.
[7]
H
.
Nawal
et
al
.
,
"
Thalassemia
review:
features,
dental
consideration
s
and
management
,
"
in
Ele
ctronic
physician
,
vol.
9
,
no.
3
, pp.
4003
-
4005, 2017
,
doi:
10.19082/4003
.
[8]
V.
Viprakasit
and
S.
Ekwattanakit,
“Clinical
classification,
scre
ening
and
diagnosis
for
thalassemia
,
”
in
Hematology/Oncology C
linics
, vol.
32, no.
2 pp.
193
-
211, 2018.
d
oi
:
10.1016/j.hoc.2017.11.006
[9]
H.
H.
Zhou,
“
Kernel
Methods
in
the
Analysis
of
Big
and
Complex
Data:
A
Modern
Statistical
Challenge
,
”
Ph.D
dissertation,
Wisconsi
n
-
Madison
Univ.
,
Madison,
United
State,
2019.
[Online].
Available:
https://search.proquest.com/openview/0b85eaa4c
7718832bf36bb357ec7f1ac/1?pq
-
origsite=
gscholar&cbl=
18750&diss=
y
[10]
Wongseree
et
al
.
,
“
Thalassemia
classification
by
neural
networks
and
genetic
programming
,
”
in
Informati
on
Sciences
, vol.
177
, no.
3, pp.
771
-
786
,
2007
,
doi:
10.1016/j.ins.2006.07.009
.
[11]
F
.
Yousefian,
T
.
Banirosta
m,
and
A
.
Azarkeivan
,
“
Predic
tion
T
halass
emia
Based
on
Artific
ial
Intelli
gence
Techniques:
A
Survey
,
”
in
Internati
onal
Journal
of
Advanced
R
esearch
in
Computer
and
Communica
tion
Engineering
,
v
ol.
6,
no.
6
, pp. 281
-
287,
2017
,
d
oi
:
10.17148/IJARCCE.2017.6847
.
[12]
S.
R.
Amendolia
et
al
.
,
“A
Real
-
Time
Classification
System
of
Thalas
semic
Patholog
ies
Based
on
Artific
ial
Neura
l
Networks,”
in
Medical Decisi
on Making
,
vol. 22, no. 1,
pp. 18
-
26, 2002
,
doi:
10.1177/0272989x0202200102
.
[13]
D.
Setsirichok
et
al
.
,
“Classi
fication
of
complete
blood
count
and
haemoglo
bin
typing
data
by
a
C4.5
decision
tree,
a Naïve Bay
es classi
fier
and a Mu
ltilay
er
Perceptro
n for
Thalassaem
ia
screening
,”
in
Biomedical
Signal Processing
and
Control
, vol. 7,
n
o. 2, pp. 202
-
212, 2012
,
do
i:
10.1016/j.bspc.2011.03.007
.
[14]
V.
Laengsri
et
al
.
,
“ThalPred:
a
web
-
based
prediction
tool
for
discri
minating
thalassemia
trait
and
iron
deficiency
anemia
,
”
BMC
medical
informatics
and
decision
making
,
vol.
19,
2
019
,
Art.
n
o
212,
doi:
10.1186/s12911
-
019
-
0929
-
2
.
[15]
A
.
Muzaffer
et
al
.
,
“
Kernel
Sele
ction
for
Support
Vector
Machines
fo
r
System
Identification
of
a
CNC
Machining
Center
,
” in
IFAC
-
PapersOnLine
, vol. 52,
no.
29, pp.
192
-
198, 2019
, doi:
10.1016/j.ifacol.2019.12.643
.
[16]
A
.
Tharwat
,
“
Parame
ter
investig
ation
of
support
vector
ma
chine
cla
ssifier
with
kerne
l
functi
ons
,
”
in
Knowledge
and Information Systems,
vol. 6
1,
pp.
1269
-
1302
,
2019
, doi: 1
0.1007/s10115
-
019
-
01335
-
4
.
[17]
X.
Zhang,
J.
Xiao
and
F.
Gu,
"
Applying
Support
Vector
Machin
e
to
Electronic
Health
Records
for
Cancer
Classific
ation,"
in
Spring
Simulation
Conference
(S
pringSim)
,
pp.
1
-
9
,
2019
,
doi:
10.23919/SpringSim.2019.8732906
.
[18]
H
.
Zheng
et al
.
,
“
Online
Support
Vector
Machine with Adaptive
Kern
el Functions
,
” in
Internati
onal
Conference on
Computer
Informatio
n Analytic
s and Inte
lligent Sy
stems
,
2019
, doi:
10.25236/ciais.2019.029
.
[19]
Z.
Rustam
and
N.
P.
N.
A.
Ariantari
,
“Support
Vector
Machines
fo
r
Classifying
Policyholders
Satisfactorily
i
n
Automobile
Insurance
,
”
in
Journal
of
Physics:
Conferen
ce
Serie
s
1028
012005
,
2018
,
doi:
10.1088/1742
-
6596/1028/1/012005
.
[20]
Z
.
Rustam
and
D
.
Zahras
,
“
Comparison
betwee
n
Support
Vector
M
achine
and
Fuzzy
C
-
Means
as
Classifier
for
Intrusion
Detection S
ystem
,
” in
J. Phys.:
Conf.
Ser. 102
8 01222
7
,
2018
, doi:
10.1088/1742
-
6596/1028/1/012227
.
[21]
M
.
Vatankhah,
V
.
Asadpour
and
R
.
Fazel
-
Rezai
,
“
Perce
ptual
pain
classif
ication
using
ANFIS
adapte
d
RBF
kerne
l
support vec
tor machine
for ther
apeutic
usage
,
” in
Applied
Soft Comput
ing
, vol.
13,
no.
5, pp. 2537
-
2546,
2013
, doi
:
10.1016/j.asoc.2012.11.032
.
[22]
J
.
Abukhait
and
M
.
Obeidat
,
“
Classific
ation
based
on
Gaussian
-
ker
nel
Support
Vector
Machine
with
Adaptive
Fuzzy
Inference
System
,
”
in
Przegląd
Elektrotechniczny
,
v
ol.
1
,
no.
5
pp.
16
-
24
,
2018
,
doi:
10.15199/48.2018.05.03
.
[23]
A
R.
Bagasta
et
al
.
,
“
Comparison
of
Cubic
SVM
with
Gaussian
SV
M:
Classific
ation
of
Infar
ction
for
detectin
g
Ischemic
Stroke
,
”
in
IOP
Conf.
Series:
Materials
Science
and
Engine
ering
546
052016
,
2019
,
doi:
10.1088/1757
-
899X/546/5/052016
[24]
M.
A
Ohemeng
,
"
A
Comparative
Study
on
Support
Vector
Machin
es
,
"
Ph.D
dissertation,
Mathematics
,
Nevada
Univ.,
Reno
, 201
7.
[Online]. Avai
lable:
https://scholarworks.unr.edu/handle/11714/2816
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
C
om
par
is
on
s
om
e
of
k
e
r
ne
l
fu
nc
ti
ons
w
it
h s
uppor
t
v
e
c
to
r
m
a
c
h
in
e
s
c
la
s
s
if
ie
r
f
or
…
(
I
ls
y
a W
ir
as
at
i
)
437
[25]
A.
Apsemidis
, S.
Psarak
is, and
J
.
M. Moguerza
. "
A review of machine lea
rning kernel methods in statistical process
monitoring,"
in
Computers
&
Industrial
Enginee
ring
,
vol.
142
,
2020
,
Art.
no.
106376
,
doi:
10.1016/j.cie.2020.106376
.
[26]
D. Zhao
et al
.
,
“
Whale opt
imized
mixed
kernel fun
ction
of
support ve
ctor machine f
or colorecta
l cancer dia
gnosis
,
”
in
Journal
of Bi
omedical
Informat
ics
, vol. 92, 2019
, Art. no.
103124
, doi:
10.1016/j.jbi.2019.103124
.
[27]
Z
.
Liu
and
H
.
Xu
,
“
Kernel
Parameter
Selection
for
Support
Vec
tor
Machine
Classific
ation
,
”
in
Journal
of
Algorithms & Comput
ational Technology
, vol. 8, no. 2, pp. 163
-
177,
2013
, doi:
10.1260/1748
-
3018.8.2.163
.
[28]
T
.
V
.
Rampisela
and
Z
Rustam
,
“
Classific
ation
of
Schizophr
enia
Data
Using
Support
Vector
Machine
(SVM)
,
”
in
J.
Phys.: Conf. Ser. 1108 012044
,
2018
, doi:
10.1088/1742
-
6596/1108/1/012044
.
[29]
S
.
Hartini
and
Z
.
Rustam
,
“
Hierarchical
Clustering
Algorithm
Based
on
Density
Peaks
using
Kernel
Function
for
Thalassemia Classification
,
” in
J. Phys.:
Conf.
Ser. 141
7 01201
6
,
2019
, doi:
10.1088/1742
-
6596/1417/1/012016
.
[30]
J.
Moeller,
“
Kernels
and
Geometry
of
Machine
Learning
,
”
Ph.D
dis
sertation,
Utah
Univ.,
Utah,
2017
.
[Online].
Available:
https://arxiv.org/abs/2010.15110
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Ilsya
Wirasati
is
a
f
inal
year
student
in
the
Depart
ment
of
M
athematics,
University
of
Indonesia
,
who
is
currently
working
on
her
thesis
.
Her
research
is
firmly
about
applied
mathematics
using
machine
learning
in
medical
field
.
Ms.
Ilsya’s
specialties
in
research
are
mostly about m
achine learn
ing, m
athemati
cal model
ing, and
data mi
ning.
Zuherman
Rustam
is
an
Associate
Professor
and
a
lecturer
of
the
i
ntelligence
computation
at
the
Department
of
Mathematics,
University
of
Indonesia.
He
obtaine
d
his
Master
of
Science
in
1989
in
informatics
,
Paris
Diderot
University,
French,a
nd
complete
d
his
Ph.D.
in
2006
from
computer
science,
Universi
ty
of
Indonesi
a.
Assoc.
Prof.
Dr.
Rustam
is
a
member
of
IEEE
who
is
actively
researching
machine l
earning,
pattern recognition, neural networ
k,
artificial
intelligence
.
Jane
Eva
Aurelia
was
born
in
Jakarta,
19
June
1998.
She
is
a
f
inal
year
student
in
the
Depart
ment
of
Mathematics,
University
of
Indonesia.
She
is
current
ly
working
on
her
thesis,
which
is
firmly
about
applied
mathematics
usin
g
machine
learning.
A
lso,
Ms.
Jane’s
specialties
in resea
rch are
mostly about ma
chine lea
rning, mathe
matical mode
ling, and da
ta mining.
Sri
Hartini
is
a
Bachelor
of
Science
from
the
Department
of
M
athematics,
University
of
Indonesia,
who
is
also
completin
g
the
Master
of
Science
at
the
Uni
versity
of
Indonesia
and
is
currently
pursuin
g
a
Ph.D.
in
intell
igence
computat
ion.
Ms.
Hartini
i
s
passion
ately
re
searching
machine le
arning, c
omputer vision,
neural ne
tworks and
deep lea
rning in var
ious fields.
Glori
Stephani
Saragih
was
born
in
Medan,
17
January
1997.
She
is
a
Bachelor
of
Science
from
Department
of
Mathematics,
Universitas
Indonesia,
who
is
co
mpleting
the
Master
of
Scienc
e
at
Univer
sitas
Indone
sia
and
is
curre
ntly
pursuin
g
a
Ph.D.
in
intellige
nce
computa
tion.
Ms.
Glori
is
currently
a
Process
Improvement
Manager
in
PT.
Apl
ikasi
Karya
Anak
Bangsa
(Gojek).
Her
current
research
is
machine
on
machin
e
learning
and
neural
network
in
various
fields, esp
ecially medi
cal and finance.
Evaluation Warning : The document was created with Spire.PDF for Python.