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
-
AI
)
V
ol
.
10
, N
o.
2
,
J
une
202
1
, pp.
421
~
429
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
10
.i
2
.pp
421
-
429
421
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
C
l
ass
i
f
i
c
at
i
on
of
sk
i
n
c
an
c
e
r
i
m
age
s b
y ap
p
l
yi
n
g si
m
p
l
e
e
vol
vi
n
g
c
on
n
e
c
t
i
on
i
st
sys
t
e
m
Al
-
K
h
ow
ar
iz
m
i
1
, S
u
h
e
r
m
an
2
1
Department of Information
Technology, Universi
tas Muhammadiyah
Sumatera Utara, Indonesia
2
Department of Electrical
Engineering, Universitas Sumatera Utara, Indonesia
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
S
e
p
1
3
, 20
20
R
e
vi
s
e
d
M
a
r
2
1
, 20
21
A
c
c
e
pt
e
d A
pr
9
, 20
21
Simple
evolving
connectionist
system
(SECoS
)
is
on
e
of
data
mining
classifi
cation
techniqu
es
that
recognizin
g
data
based
on
the
tested
and
the
training
data
binding.
Data
recognition
is
achieve
d
by
aligning
testing data
to
trained
data
pattern.
SECoS
uses
a
feedfor
ward
neural
network
but
its
hidden
layer
evolves
so
that
each
input
layer
does
not
perform
epoch.
SECoS
distance
has
been
modified
with
the
normalized
Euclidean
distance
f
ormula
to
reduce
error
in
training.
This
paper
recognizes
skin
cancer
by
classifying
benign
malignant
skin
moles
images
using
SECoS
based
on
par
ameter
combinat
ions.
The
skin
cancer
classifi
cation
has
learning
rate
1
of
0.3,
learning
rate
2
of
0.3,
sensitivity
threshold
of
0.5,
error
threshold
of
0.1
and
MAPE
is
0.5184845
with
developing
hidden
node
of
23.
Skin
cancer
recog
nition
by
applying
modified
SECoS
algorithm
is
proven
more
acceptable.
Compared
to
other
methods
,
SECoS
is
more
robust
t
o
error
variations.
K
e
y
w
o
r
d
s
:
C
la
s
s
if
ic
a
ti
on
D
a
ta
m
in
in
g
S
E
C
oS
S
ki
n c
a
r
e
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
:
S
uhe
r
m
a
n
D
e
pa
r
tm
e
nt
of
E
le
c
tr
ic
a
l
E
nf
in
e
e
r
in
g
U
ni
ve
r
s
it
a
s
S
um
a
te
r
a
U
ta
r
a
J
a
la
n A
lm
a
m
a
te
r
K
a
m
pu
s
U
S
U
M
e
da
n, I
ndone
s
ia
E
m
a
il
:
s
uhe
r
m
a
n@
us
u.a
c
.i
d
1.
I
N
T
R
O
D
U
C
T
I
O
N
C
la
s
s
if
ic
a
ti
on
i
s
one
of
th
e
da
ta
m
in
in
g
a
lg
or
it
hm
s
th
a
t
a
im
s
to
e
na
bl
e
c
om
put
e
r
or
pr
oc
e
s
s
or
to
c
a
r
r
y
out
le
a
r
ni
ng
by
m
e
th
ods
a
nd
a
lg
or
it
hm
s
to
f
it
in
to
a
m
a
c
hi
ne
le
a
r
ni
ng
c
onc
e
pt
[
1]
.
C
la
s
s
if
ic
a
ti
on
in
vol
ve
s
t
r
a
in
in
g a
nd t
e
s
ti
ng da
ta
[
2
]
i
n
f
or
m
of
t
e
xt
, i
m
a
ge
s
, a
nd vide
os
[
3
]
-
[
5]
.
L
e
a
r
ni
ng pr
oc
e
s
s
i
n de
f
in
in
g
th
e
m
ode
l
m
a
y
in
vol
ve
da
ta
pr
ope
r
ti
e
s
th
a
t
r
e
f
e
r
r
e
d
to
a
s
d
a
ta
f
e
a
tu
r
e
s
,
s
o
th
a
t
le
a
r
ni
ng
pr
oc
e
s
s
e
xe
r
ts
id
e
nt
if
ic
a
ti
on,
de
c
is
io
n
s
uppor
t,
a
nd
m
e
th
od
im
pr
ove
m
e
nt
[
6]
.
C
la
s
s
if
ic
a
ti
on
ha
s
b
e
c
om
e
one
of
th
e
m
a
jo
r
f
oc
us
e
s
on c
om
put
e
r
s
c
i
e
nc
e
. D
a
ta
m
in
in
g a
lg
or
it
hm
i
s
a
bl
e
t
o
a
na
ly
z
e
c
om
pl
e
x da
ta
s
e
t
a
nd t
o pr
e
di
c
t
va
r
io
us
ne
w
ly
in
put
te
d
da
ta
by
not
m
a
ki
ng
g
e
ne
r
a
l
a
s
s
um
pt
io
ns
on
d
a
t
a
di
s
tr
ib
ut
io
n.
I
n
s
uppor
t
to
da
ta
m
in
in
g
s
tu
dy,
c
om
pl
e
x
da
ta
in
va
r
io
us
a
ppl
ic
a
ti
ons
h
a
ve
be
e
n
c
ol
le
c
te
d
f
or
c
l
a
s
s
if
ic
a
ti
on
s
tu
dy
pur
po
s
e
s
th
a
t
a
r
e
r
e
f
e
r
r
e
d
to
a
s
da
ta
s
e
t
s
[
7]
. D
a
ta
s
e
t
m
a
y i
n f
or
m
of
t
e
xt
r
e
c
or
ds
, l
it
e
r
a
tu
r
e
i
m
a
ge
s
, ol
d s
c
r
ip
tu
r
e
s
, s
pe
e
c
h r
e
c
o
r
ds
, bus
in
e
s
s
da
ta
, pha
r
m
a
c
e
ut
ic
a
l
da
ta
or
m
e
di
c
a
l
im
a
ge
s
.
S
ki
n
c
a
nc
e
r
is
th
e
pr
im
a
r
y
c
a
n
c
e
r
de
s
e
a
s
e
in
s
om
e
c
ount
r
ie
s
w
it
h
de
a
th
r
a
te
i
s
up
to
10%
[
8]
.
N
um
be
r
of
s
ki
n
c
a
n
c
e
r
pa
ti
e
nt
s
in
c
r
e
a
s
e
s
s
te
a
di
ly
e
ve
r
y
ye
a
r
.
S
ki
n
di
s
e
a
s
e
de
t
e
c
ti
on
a
nd
e
xa
m
in
a
ti
on
a
r
e
im
por
ta
nt
to
de
te
r
m
in
t
th
e
c
or
r
e
c
t
t
r
e
a
tm
e
nt
.
H
ow
e
ve
r
,
c
la
s
s
i
f
ic
a
ti
on
s
tu
dy
on
c
a
nc
e
r
im
a
ge
s
s
ti
ll
r
e
qui
r
e
s
a
c
c
ur
a
c
y
pr
e
di
c
ti
on
th
a
t
ha
s
be
e
n
r
e
s
e
a
r
c
h
c
h
a
ll
e
nge
s
,
in
c
lu
di
ng
r
e
s
e
a
r
c
h
on
s
ki
n
di
s
e
a
s
e
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
.
H
ow
e
ve
r
,
lo
w
le
ve
l
f
e
a
tu
r
e
e
xt
r
a
c
ti
ons
f
or
s
ki
n
de
s
e
a
s
e
s
ba
s
e
d
on
s
h
a
pe
,
te
xt
ur
e
,
a
nd
c
ol
or
pr
ope
r
ti
e
s
e
xt
r
a
c
te
d
f
r
om
th
e
le
s
io
n
im
a
ge
a
r
e
s
ti
ll
c
ons
id
e
r
e
d
a
s
la
c
k
of
c
li
ni
c
a
l
m
e
a
ni
ng
[
9
]
.
H
ig
h
le
ve
l
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
:
421
–
429
422
c
li
ni
c
a
l
or
ie
nt
e
d
f
e
a
tu
r
e
s
a
r
e
e
xpe
c
te
d
to
c
ont
r
ib
ut
e
m
or
e
by
a
ppl
yi
ng
va
r
io
us
c
ol
or
n
or
m
a
li
z
a
ti
on
a
ppr
oa
c
he
s
.
T
e
c
hni
que
s
s
uc
h
a
s
c
ont
e
nt
-
ba
s
e
d
im
a
ge
r
e
tr
ie
va
l
[
10]
a
nd
s
hor
t
te
r
m
of
le
s
io
ns
m
oni
to
r
in
g
a
r
e
e
xpe
c
te
d
to
pr
ovi
de
c
li
ni
c
a
l
m
e
a
ni
ngf
ul
da
ta
s
e
t
a
nd
to
e
nr
ic
h
th
e
c
la
s
s
if
ic
a
ti
on
di
a
gnos
ti
c
.
M
or
e
ove
r
,
if
th
e
r
ic
h
vi
s
ua
l
f
e
a
tu
r
e
is
c
om
pl
e
m
e
nt
e
d
by
th
e
c
li
ni
c
a
l
m
e
ta
da
ta
,
di
a
gnos
ti
c
a
c
c
ur
a
c
y
m
a
y
in
c
r
e
a
s
e
a
nd
tr
e
a
tm
e
nt
m
a
y be
pr
ope
r
.
T
r
a
di
ti
ona
l
c
la
s
s
if
ie
r
s
uc
h
a
s
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
i
s
w
id
e
ly
us
e
d
to
c
la
s
s
if
y
s
ki
n
c
a
n
c
e
r
im
a
ge
s
.
F
or
in
s
ta
nc
e
,
A
lq
ur
a
n
e
t
al
.
[
9]
s
ki
n
di
s
e
a
s
e
a
na
l
ys
is
u
s
in
g
S
V
M
w
it
h
pr
in
c
ip
a
l
c
om
pone
nt
a
na
ly
s
is
(
P
C
A
)
to
c
la
s
s
if
y
s
ki
n
di
s
e
a
s
e
da
t
a
s
e
t.
F
e
a
tu
r
e
e
xt
r
a
c
ti
on
s
w
e
r
e
pe
r
f
or
m
e
d
in
f
our
s
te
p
s
(
R
G
B
im
a
ge
,
bi
na
r
y
im
a
ge
,
le
s
io
n
im
a
ge
a
nd
hi
s
to
gr
a
m
e
qu
a
li
z
e
d
im
a
ge
)
to
ge
t
kur
to
s
is
,
c
ont
r
a
s
t,
m
e
a
n,
s
ke
w
ne
s
s
,
e
ne
r
gy,
c
ir
c
ul
a
ti
on,
s
ta
nda
r
d
de
vi
a
ti
on,
hom
oge
ne
it
y
a
nd
c
or
r
e
la
ti
on
.
.
T
he
f
e
a
tu
r
e
s
w
e
r
e
th
e
n
f
e
d
in
to
P
C
A
,
be
f
or
e
S
V
M
c
la
s
s
if
ic
a
ti
on
w
a
s
pe
r
f
or
m
e
d.
T
he
f
iv
e
s
e
le
c
te
d
f
e
a
tu
r
e
e
xt
e
r
te
d
th
e
a
c
c
ur
a
c
y
of
92.1
%
.
M
e
a
nw
hi
le
,
C
hung
e
t
al
.
[
11]
us
e
d
S
V
M
w
it
h
th
e
he
lp
of
pe
r
s
is
te
nc
e
di
a
gr
a
m
s
f
r
om
pe
r
s
is
te
nt
hom
ol
ogy
a
lg
or
it
h
m
(
w
hi
c
h
is
pa
r
t
of
th
e
to
pol
ogi
c
a
l
da
ta
a
na
ly
s
is
(
T
D
A
)
)
f
or
s
e
gm
e
nt
in
g
im
a
ge
s
.
T
h
e
m
e
th
od
a
ls
o
c
om
bi
ne
s
th
e
pe
r
s
is
te
nt
hom
ol
ogy
w
it
h
th
e
pe
r
s
is
te
nc
e
s
ta
ts
w
it
h
th
e
B
e
tt
i
a
nd
e
nt
r
opy
c
ur
ve
s
be
f
or
e
c
la
s
s
if
ie
d
by
us
in
g
S
V
M
.
B
e
s
t
va
li
da
ti
on
s
c
or
e
w
a
s
a
c
hi
e
ve
d
w
he
n
X
Y
Z
c
ur
ve
s
,
X
Y
Z
s
ta
ts
,
R
G
B
s
ta
ts
w
e
r
e
us
e
d
a
s
s
e
le
c
te
d
f
e
a
tu
r
e
s
.
H
ow
e
ve
r
,
tr
a
di
ti
ona
l
c
la
s
s
if
ie
r
s
s
uc
h
a
s
S
V
M
,
ne
ur
a
l
ne
t
w
or
k
a
nd
de
c
is
io
n
tr
e
e
m
a
y
ha
ve
di
f
f
ic
ul
ti
e
s
in
de
a
li
ng
w
it
h
la
r
ge
d
a
ta
s
e
t
a
nd
f
r
e
que
nt
va
r
ia
ti
o
n
in
f
a
ul
t
pa
tt
e
r
n.
A
lt
hough
th
e
s
e
tr
a
di
ti
ona
l
c
la
s
s
if
ie
r
m
a
y
r
e
s
ul
t
hi
gh
a
c
c
ur
a
c
y
f
or
gi
ve
n
da
t
a
s
e
t,
f
ut
ur
e
c
o
m
pl
e
x
a
nd
r
ic
h
da
ta
s
e
t
r
e
qui
r
e
m
or
e
e
f
f
ic
ie
nt
c
la
s
s
if
ie
r
.
S
im
pl
e
e
vol
vi
ng
c
onne
c
ti
oni
s
t
s
y
s
te
m
(
S
E
C
oS
)
or
s
o
c
a
ll
e
d
a
s
e
vol
v
ed
m
ul
ti
la
ye
r
pe
r
c
e
pt
r
on
[
12]
is
th
e
s
m
a
ll
e
s
t
r
e
a
li
z
a
ti
on
of
E
C
oS
th
a
t
ha
s
c
a
pa
bi
li
ty
in
f
or
e
c
a
s
ti
ng
bi
g
da
ta
pa
tt
e
r
n.
T
hi
s
m
e
th
od
is
pow
e
r
f
ul
l
in
pr
oc
e
s
s
in
g
th
e
r
ic
h
da
ta
s
e
t.
S
E
C
oS
a
ppl
ie
s
th
e
E
C
oS
pr
in
c
ip
le
by
de
ve
lo
pi
ng
th
e
m
ul
ti
la
ye
r
pr
e
c
e
pt
r
on
(
M
L
P
)
m
e
th
od
[
13
]
.
S
E
C
oS
is
s
om
e
ti
m
e
s
r
e
f
e
r
r
e
d
to
a
s
th
e
e
vol
vi
ng
m
ul
t
il
a
ye
r
pe
r
c
e
pt
r
on
(
e
M
L
P
)
[
14]
.
S
E
C
oS
ha
s
th
r
e
e
la
ye
r
s
,
na
m
e
ly
in
put
,
hi
dde
n
a
n
d
out
put
la
ye
r
r
[
15]
.
T
he
in
put
la
ye
r
pr
oc
e
s
s
e
s
r
a
w
in
put
da
ta
f
or
te
s
ti
ng
a
nd
tr
a
in
in
g.
T
h
e
hi
dde
n
l
a
ye
r
c
a
lc
ul
a
te
s
th
e
S
E
C
oS
m
e
th
od
by
u
s
in
g
s
om
e
f
or
m
ul
a
s
th
a
t
ha
ve
be
e
n
de
ve
lo
pe
d
in
a
c
c
or
da
nc
e
to
th
e
E
C
oS
pr
in
c
ip
le
.
T
he
pr
oc
e
s
s
is
c
a
r
r
ie
d
out
e
vol
ve
ly
w
he
r
e
th
e
hi
dde
n
la
y
e
r
in
c
r
e
a
s
e
s
th
e
obj
e
c
t
pr
ope
r
ti
e
s
to
g
e
t
th
e
s
pe
c
if
ie
d
v
a
lu
e
.
T
h
e
out
put
la
ye
r
is
to
ge
t
c
la
s
s
if
ic
a
ti
on va
lu
e
[
16
]
-
[
18]
.
S
E
C
os
is
a
bl
e
to
p
e
r
f
or
m
da
ta
c
la
s
s
if
ic
a
ti
on
in
v
a
r
io
us
f
ie
ld
s
[
19]
.
T
he
w
or
k
in
[
20]
us
e
d
S
E
C
oS
to
c
la
s
s
if
y
s
tr
oke
da
ta
te
xt
s
to
he
lp
p
a
ti
e
nt
tr
e
a
tm
e
nt
by
de
t
e
r
m
in
in
g
how
m
uc
h
f
ib
r
in
oge
n
s
houl
d
be
c
ons
um
e
d.
T
he
c
l
a
s
s
if
ic
a
ti
on
pr
oc
e
s
s
is
ve
r
y
he
lp
f
ul
in
a
s
s
is
ti
ng
doc
to
r
s
a
nd
pa
ti
e
nt
s
.
I
n
a
ddi
ti
on,
th
e
m
e
th
od
ha
s
a
ls
o
be
e
n
us
e
d
to
c
la
s
s
if
y
im
a
ge
s
of
th
e
Q
ur
'
a
n
ve
r
s
e
s
to
he
lp
r
e
a
d
e
r
s
unde
r
s
ta
ndi
ng
th
e
ta
jwe
e
d
la
w
in
r
e
a
di
ng
th
e
Q
ur
'
a
n
[
21
]
.
S
in
c
e
da
ta
is
not
a
te
xt
ty
pe
,
f
e
a
tu
r
e
e
xt
r
a
c
ti
ons
w
e
r
e
pe
r
f
or
m
e
d
[
22
]
,
[
23]
.
S
E
C
oS
c
a
n
b
e
in
te
gr
a
te
d
in
to
a
s
in
gl
e
im
a
ge
pr
oc
e
s
s
in
g
m
odul
e
[
24
]
,
[
25]
,
f
or
in
s
ta
nc
e
,
to
is
ol
a
te
phon
e
m
e
r
e
c
ogni
ti
on
[
26]
w
he
r
e
la
r
ge
da
ta
pa
tt
e
n
c
a
n
be
m
e
m
or
iz
e
d
a
nd
ne
w
in
put
te
d
da
ta
c
a
n
be
a
da
pt
e
d.
S
E
C
oS
r
e
duc
e
s
a
la
r
ge
num
be
r
of
h
id
de
n
node
s
.
S
E
C
oS
ha
s
a
ls
o
be
e
n
us
e
d
f
or
a
da
p
ti
ve
s
pe
e
c
h
r
e
c
ogni
ti
on,
bus
in
e
s
s
a
na
ly
s
i
s
,
a
s
w
e
ll
a
s
he
a
lt
h a
nd pha
r
m
a
c
e
ut
ic
a
l
a
na
ly
s
i
s
[
27]
. T
hi
s
pa
pe
r
i
de
nt
if
ie
s
s
ki
n c
a
nc
e
r
f
r
om
a
ba
la
nc
e
d s
ki
n c
a
n
c
e
r
im
a
ge
s
da
ta
s
e
t
of
m
a
l
ig
na
nt
a
nd
be
ni
gn s
ki
n
im
a
ge
s.
2.
M
A
T
E
R
I
A
L
A
N
D
M
E
T
H
O
D
2.1.
D
at
as
e
t
T
he
e
m
pl
oye
d
da
ta
s
e
t
in
th
is
pa
pe
r
is
ba
s
e
d
on
pr
e
vi
ous
r
e
s
e
a
r
c
h
[
28]
th
a
t
ta
ke
n
f
r
om
a
ba
la
nc
e
d
s
ki
n
c
a
nc
e
r
im
a
ge
da
ta
s
e
t
of
m
a
li
gna
nt
a
nd
be
ni
gn
s
ki
n
im
a
g
e
s
.
T
hi
s
s
ki
n
c
a
nc
e
r
da
ta
s
e
t
is
a
c
c
e
s
s
ib
le
a
nd
a
va
il
a
bl
e
onl
in
e
a
t
ka
ggl
e
.
T
he
i
m
a
ge
s
w
e
r
e
ga
th
e
r
e
d by the
c
ol
la
bor
a
ti
on
of
s
ki
n i
m
a
ge
s
(
I
S
I
C
)
f
or
s
tu
dy a
nd
im
pl
e
m
e
nt
a
ti
on pur
pos
e
s
. T
he
r
e
a
r
e
360 im
a
ge
s
:
16.7%
f
a
ll
s
u
nde
r
t
he
c
a
te
gor
y of
m
a
li
gna
nt
m
e
la
nom
a
s
a
nd
83.3%
i
s
i
n be
n
ig
n c
a
te
gor
y.
T
he
c
la
s
s
if
ic
a
ti
on
is
pe
r
f
or
m
e
d
ba
s
e
d
on
th
e
224×
224
c
ol
or
e
d
im
a
ge
s
.
T
h
e
da
ta
s
e
t
it
s
e
lf
c
ont
a
in
s
1800
da
ta
s
e
t
s
,
w
he
r
e
136
da
t
a
s
e
t
s
w
it
h
100
tr
a
in
in
g
da
ta
a
nd
3
6
te
s
ti
ng
da
ta
w
e
r
e
s
e
le
c
te
d.
T
h
e
c
la
s
s
if
ic
a
ti
on
ta
r
ge
ts
a
r
e
“
be
ni
gn”
a
nd
“
m
a
li
gna
nt
”
.
I
n
or
de
r
to
f
a
s
t
e
n
a
lg
or
it
hm
e
xe
c
ut
io
n
ti
m
e
,
th
e
224
×
224
c
ol
or
e
d
im
a
ge
w
a
s
c
onv
e
r
te
d i
nt
o 28×
28 gr
a
ys
c
a
l
e
i
m
a
ge
.
2.2. S
E
C
oS
ar
c
h
it
e
c
t
u
r
e
F
ig
ur
e
1
s
how
s
th
e
c
om
m
on
a
r
c
hi
te
c
tu
r
e
of
S
E
C
oS
a
lg
or
it
hm
in
c
la
s
s
if
yi
ng
s
ki
n
c
a
nc
e
r
di
s
c
us
s
e
d
th
r
oughout
th
is
pa
pe
r
.
T
h
e
a
r
c
hi
te
c
tu
r
e
de
s
c
r
ib
e
s
s
te
ps
in
vol
v
e
d
in
c
la
s
s
if
ic
a
ti
on
s
u
c
h
a
s
gr
a
ys
c
a
li
ng,
f
e
a
tu
r
e
e
xt
r
a
c
ti
on, da
ta
t
r
a
in
in
g a
nd da
ta
t
e
s
ti
ng.
I
ni
ti
a
ll
y,
da
ta
s
e
t
is
s
pl
it
te
d
in
to
tr
a
in
in
g
a
nd
te
s
ti
ng
da
ta
w
he
r
e
th
e
di
s
tr
ib
ut
io
n
o
f
b
bot
h
b
e
ni
gn
a
nd
m
a
li
gna
nt
da
ta
a
r
e
pr
ovi
de
d.
T
h
e
s
e
im
a
ge
s
a
r
e
th
e
n
c
onve
r
te
d
in
to
gr
a
ys
c
a
le
im
a
ge
s
.
A
gr
a
y
s
c
a
le
im
a
ge
i
s
r
e
pr
e
s
e
nt
e
d by a
s
in
gl
e
s
a
m
pl
e
va
lu
e
i
n e
a
c
h pi
xe
l
to
r
e
pr
e
s
e
nt
t
he
a
m
ount
of
l
ig
ht
. G
r
a
ys
c
a
le
i
m
a
ge
c
a
n b
e
i
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
la
s
s
if
ic
at
io
n of
s
k
in
c
an
c
e
r
i
m
ag
e
s
by
apply
in
g
s
impl
e
e
v
ol
v
in
g c
onne
c
ti
oni
s
t
s
y
s
te
m
(
Al
-
K
how
ar
iz
m
i
)
423
f
or
m
of
bl
a
c
k/
w
hi
te
or
g
r
a
y
m
onoc
hr
om
e
im
a
ge
.
B
la
c
k
is
th
e
w
e
a
ke
s
t
li
ght
in
te
ns
it
y
a
nd
w
hi
te
is
th
e
s
tr
onge
s
t.
N
e
xt
is
to
do
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
w
it
h
th
e
or
ie
nt
e
d
gr
a
di
e
nt
hi
s
to
gr
a
m
s
(
H
O
G
)
.
H
O
G
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
de
te
r
m
in
e
s
th
e
f
r
e
que
nc
y
of
gr
a
di
e
nt
or
ie
nt
a
ti
on
i
n
a
por
ti
on
of
a
n
a
na
ly
z
e
d
im
a
ge
by
us
in
g
c
ont
r
a
s
t
nor
m
a
li
z
a
ti
on.
T
hi
s
gr
a
di
e
nt
f
ie
ld
c
a
pt
ur
e
s
th
e
im
a
ge
s
pa
ti
a
l
s
tr
uc
tu
r
e
.
H
O
G
de
s
c
r
ip
to
r
s
a
r
e
f
a
m
il
ia
r
to
be
us
e
d
on
c
om
put
e
r
vi
s
io
n
a
nd
r
e
c
ogni
ti
on.
H
O
G
de
s
c
r
ip
to
r
s
e
xt
r
a
c
t
f
e
a
tu
r
e
s
f
or
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
. H
O
G
f
e
a
tu
r
e
e
xt
r
a
c
ti
on e
xt
r
a
c
ts
gr
a
ys
c
a
le
i
m
a
ge
s
f
e
a
tu
r
e
s
t
o be
us
e
d by S
E
C
oS
a
lg
or
it
hm
.
F
ig
ur
e
1.
G
e
ne
r
a
l
a
r
c
hi
te
c
tu
r
e
D
a
ta
i
s
t
he
n t
r
a
in
e
d i
n S
E
C
oS
a
nd t
he
w
e
ig
ht
i
s
upda
te
d. I
f
t
he
w
e
ig
ht
i
s
out
of
s
c
ope
, t
he
n i
t
r
e
tu
r
ns
to
S
E
C
oS
,
but
if
it
is
c
or
r
e
c
t
th
e
n
n
e
xt
s
t
e
p
is
pe
r
f
or
m
e
d.
D
a
t
a
te
s
ti
ng
in
S
E
C
oS
is
p
e
r
f
or
m
e
d
by
upda
ti
ng
w
e
ig
ht
s
.
I
f
th
e
w
e
ig
ht
i
s
c
or
r
e
c
t,
th
e
n
e
xt
s
te
p
is
p
e
r
f
or
m
e
d.
M
A
P
E
is
c
a
lc
ul
a
t
e
d
by
u
s
in
g
(
1
)
,
w
he
r
e
a
c
tu
a
l
va
lu
e
, c
la
s
s
if
ic
a
ti
on va
lu
e
a
nd numbe
r
of
da
ta
a
r
e
r
e
pr
e
s
e
nt
e
d b
y a
, b, a
nd n s
ubs
e
que
nt
ly
.
=
∑
|
−
|
=
×
%
(
1)
2.3. S
E
C
oS
d
e
ve
lo
p
m
e
n
t
on
n
or
m
al
iz
e
d
e
u
c
li
d
e
an
d
is
t
an
c
e
S
E
C
oS
ha
s
a
ne
ur
a
l
ne
twor
k
th
a
t
e
xt
e
nds
to
th
e
hi
dde
n
la
ye
r
a
nd
th
e
out
put
la
ye
r
a
s
de
s
c
r
ib
e
d
in
F
ig
ur
e
2
(
a
)
.
F
ig
ur
e
2(
a
)
s
how
s
th
a
t
th
r
e
e
th
r
e
e
la
ye
r
s
of
ne
ur
on
pr
oc
e
s
s
e
s
.
T
h
e
f
ir
s
t
la
ye
r
is
th
e
in
put
la
ye
r
,
f
ol
lo
w
e
d
by
th
e
e
vol
vi
ng
or
c
ons
tr
uc
ti
ve
s
ta
ge
a
s
th
e
s
e
c
ond
la
ye
r
a
nd
out
put
la
ye
r
a
s
th
e
f
ia
l
s
ta
g
e
.
T
he
f
ir
s
t
la
ye
r
c
or
r
e
la
te
s
to
a
ll
s
e
c
ond
la
ye
r
.
T
he
s
e
c
ond
la
ye
r
a
dd
s
ne
ur
ons
a
nd
pr
oc
e
e
ds
l
e
a
r
ni
ng.
T
he
e
vol
v
ed
l
a
ye
r
ne
ur
ons
a
r
e
tu
r
n
e
d
on
a
c
c
or
di
ng
to
di
s
ta
nc
e
of
E
uc
le
di
a
n
be
tw
e
e
n
th
e
in
put
pa
r
t
of
S
E
C
oS
a
nd
th
e
e
vol
vi
ng
la
ye
r
.
F
in
a
ll
y,
th
e
r
e
s
ul
t
s
go
to
out
put
la
ye
r
.
S
E
C
oS
is
f
ul
ly
c
onne
c
te
d
ne
twor
k
w
he
r
e
th
e
c
onn
e
c
ti
ons
num
be
r
f
r
om
in
put
to
e
vol
vi
ng
la
ye
r
e
qua
ls
to
th
e
num
be
r
of
i
nput
ne
ur
ons
.
B
y
us
in
g
th
e
te
s
te
d
p
a
ti
e
nt
da
ta
of
-
455
G
/A
pol
im
o
r
phi
s
m
of
be
ta
f
ib
r
in
oge
n
ge
ne
,
th
e
no
r
m
a
li
z
e
d
E
uc
li
de
a
n
ha
s
th
e
s
m
a
ll
e
s
t
M
A
P
E
c
om
pa
r
e
d
to
H
a
m
m
in
g
a
nd
M
a
nha
tt
a
n
di
s
ta
nc
e
s
a
s
s
how
n
in
F
ig
ur
e
2
(
b)
.
T
he
pe
r
f
or
m
a
nc
e
di
f
f
e
r
e
nc
e
s
s
how
th
a
t
th
e
nor
m
a
li
z
e
d
E
uc
li
de
a
n
di
s
ta
nc
e
ha
s
th
e
s
m
a
ll
e
s
t
M
A
P
E
.
A
lt
hough
di
f
f
e
r
e
nc
e
is
lo
w
e
r
th
a
n
1.8%
w
hi
c
h
m
e
a
n
s
th
a
t
di
s
ta
nc
e
f
or
m
ul
a
e
xe
r
ts
in
s
ig
ni
f
ic
a
nt
e
f
f
e
c
t
o
n
th
e
tr
a
in
in
g
r
e
s
ul
t
s
a
c
c
ur
a
c
y
,
E
uc
li
de
a
n
w
il
l
be
e
m
pl
oye
d
in
th
is
r
e
a
s
e
a
r
c
h.
T
h
e
r
e
f
or
e
th
e
E
uc
li
de
a
n
di
s
ta
nc
e
c
a
n
be
di
r
e
c
tl
y
c
a
lc
ul
a
te
d.
S
in
c
e
di
s
ta
nc
e
in
th
e
r
a
nge
of
[
0, 1]
, i
nput
da
ta
s
houl
d be
nor
m
a
li
z
e
d [
29]
.
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
:
421
–
429
424
(
a
)
S
E
C
oS
l
a
ye
r
s
(
b)
N
or
m
a
li
z
e
d di
s
ta
nc
e
pe
r
f
or
m
a
nc
e
c
om
pa
r
is
on
F
ig
ur
e
2.
S
E
C
oS
ne
ur
a
l
ne
twor
k a
nd dis
ta
nc
e
c
om
pa
r
is
on
[
30]
I
n
ge
ne
r
a
l,
f
e
e
df
or
w
a
r
d
ne
twor
ks
ha
v
e
a
n
a
c
ti
va
ti
on
f
unc
ti
on.
B
ut
,
S
E
C
oS
us
e
s
an
a
c
ti
va
ti
on
f
unc
ti
on
th
a
t
is
l
in
e
a
r
a
nd
c
a
ll
e
d a
s
a
c
ti
va
ti
on A
a
t
th
e
e
vol
vi
ng
la
ye
r
node
n t
ha
t
is
de
te
r
m
in
e
d by
(
2
)
.
A
n
=
1
−
D
n
(
2)
A
n
is
a
n
a
c
ti
va
ti
on
f
unc
ti
on
a
t
node
s
n
a
nd
D
n
w
hi
c
h
i
s
d
e
te
r
m
in
e
d
by
us
in
g
th
e
nor
m
a
li
z
e
d
c
a
lc
ul
a
te
d
di
s
t
a
nc
e
th
a
t
li
e
s
in
be
tw
e
e
n
of
in
put
w
e
ig
ht
a
n
d
th
e
out
put
w
e
ig
ht
a
t
th
e
d
e
s
ti
na
ti
on
no
de
.
G
e
ne
r
a
ll
y,
D
n
is
a
no
r
m
a
li
z
e
d
H
a
m
m
in
g
di
s
ta
nc
e
,
but
th
is
p
a
pe
r
r
e
f
e
r
s
to
[
31
]
w
hi
c
h
ha
s
c
a
r
r
ie
d
out
a
n
opt
im
iz
a
ti
on
by
c
ha
ngi
ng
th
e
nor
m
a
li
z
e
d
H
a
m
m
in
g
di
s
ta
nc
e
to
nor
m
a
li
z
e
d
E
uc
li
de
a
n
di
s
ta
nc
e
w
hi
c
h
ha
s
be
tt
e
r
r
e
s
ul
ts
i
n l
e
a
r
ni
ng. T
he
nor
m
a
li
z
e
d E
uc
li
de
a
n di
s
ta
nc
e
f
o
r
m
ul
a
i
s
gi
ve
n by
(
3
)
, w
he
r
e
K
,
node
s
num
be
r
a
t
th
e
ti
m
e
of
S
E
C
oS
ne
twor
k
in
put
,
I
is
th
e
va
lu
e
th
a
t
is
in
put
te
d
by
in
c
om
in
g
ve
c
to
r
,
W
is
th
e
m
a
tr
ix
of
th
e
e
vol
vi
ng l
a
ye
r
w
e
ig
ht
.
D
n
=
√
∑
|
I
i
−
W
i
|
2
k
K
i
(
3)
F
ur
th
e
r
,
le
a
r
ni
ng
w
a
s
pe
r
f
or
m
e
d
by
a
na
ly
z
in
g
im
a
ge
s
on
t
he
s
ki
n
c
a
nc
e
r
da
ta
s
e
t.
T
he
S
e
C
O
S
le
a
r
ni
ng
pr
oc
e
s
s
in
je
c
ts
ne
ur
ons
to
th
e
e
vol
ve
d
la
y
e
r
onl
y
if
th
e
va
lu
e
of
th
e
li
ne
a
r
a
c
ti
va
ti
on
is
s
m
a
ll
e
r
th
a
n
th
e
th
r
e
s
hol
d
of
s
e
ns
it
iv
it
y
or
if
th
e
c
a
lc
ul
a
te
d
e
r
r
or
is
m
uc
h
hi
ghe
r
th
a
n
th
e
th
r
e
s
hol
d.
T
he
out
goi
ng
a
nd
in
c
om
in
g
c
onne
c
ti
on
w
e
ig
ht
of
th
e
in
tr
oduc
e
d
ne
ur
on
is
r
e
la
te
d
to
th
e
ve
c
to
r
s
in
put
a
nd
out
put
.
O
th
e
r
w
is
e
,
th
e
w
e
ig
ht
of
c
onne
c
ti
on
is
a
dj
us
te
d
to
e
ns
ur
e
th
a
t
th
e
a
na
ly
z
e
d
ne
ur
on
a
ppr
oxi
m
a
te
s
th
e
in
put
ve
c
to
r
.
T
he
jo
ur
ne
y
of
S
E
C
oS
is
out
li
ne
d
by
[
30
]
.
T
he
ve
c
to
r
of
in
put
I
is
f
e
d
in
to
th
e
f
e
e
df
or
w
a
r
d
m
a
t
r
ix
,
if
th
e
a
c
ti
va
ti
on
va
lu
e
r
e
a
c
he
s
m
a
xi
m
um
of
A
m
a
x
in
th
e
node
ha
s
va
lu
e
s
m
a
ll
e
r
th
a
n
th
e
th
r
e
s
hol
d
of
s
e
ns
it
iv
it
y
(S
t
hr
)
,
th
e
n
a
ne
w
node
w
il
l
be
a
dd
e
d
to
th
e
c
ov
e
r
e
d
la
ye
r
.
O
th
e
r
w
is
e
,
S
E
C
oS
c
a
lc
ul
a
te
s
th
e
di
f
f
e
r
e
nc
e
va
lu
e
of
th
e
in
put
a
n
d
out
put
ve
c
to
r
s
,
O
c
a
nd
O
d
.
I
f
th
e
di
f
f
e
r
e
nt
is
hi
ghe
r
th
a
n
th
e
th
r
e
s
hol
d
e
r
r
or
c
oe
f
f
ic
ie
nt
(
E
t
r
h
)
or
if
th
e
e
xpe
c
te
d
out
put
node
is
una
va
il
a
bl
e
,
th
e
n
S
E
C
oS
a
dd
s
a
ne
w
node
to
th
e
hi
dde
n
la
y
e
r
.
O
th
e
r
w
is
e
,
w
in
ni
ng
node
on
hi
dde
n
la
ye
r
s
houl
d
upda
te
it
s
c
on
n
e
c
ti
on
w
e
ig
ht
.
T
he
a
f
or
m
e
nt
io
ne
d
s
te
p
i
s
pe
r
f
or
m
e
d
f
or
e
a
c
h i
nput
w
it
hout
a
n e
poc
h.
A
node
s
houl
d
be
a
dd
e
d
onl
y
if
th
e
c
onne
c
ti
on
w
e
ig
ht
ha
s
a
va
lu
e
f
ol
lo
w
in
g
th
e
in
put
ve
c
to
r
I
.
O
ut
put
ve
c
to
r
w
e
ig
th
is
s
e
t
f
ol
ow
in
g
to
th
e
out
put
ve
c
to
r
O
d.
T
he
pr
oc
e
s
s
of
w
f
e
e
df
or
w
a
r
d
f
r
om
th
e
c
ove
r
e
d
la
ye
r
to
th
e
out
put
is
pe
r
f
or
m
e
d
in
two
di
f
f
e
r
e
nt
w
a
ys
.
T
he
f
ir
s
t
w
a
y
i
s
by
u
s
in
g
th
e
pr
opa
ga
ti
on
m
e
th
od
of
O
ne
-
of
-
N
.
T
hi
s
pr
opa
ga
ti
on
is
e
xe
c
ut
e
d
onl
y
by
th
e
c
ove
r
e
d
node
w
it
h
ha
s
th
e
m
a
xi
m
um
va
lu
e
.
T
he
a
dj
us
tm
e
nt
of
t
he
w
in
ni
ng node
e
nt
r
y w
e
ig
ht
j
i
s
pe
r
f
or
m
e
d by us
in
g (
4)
.
,
(
+
1)
=
,
(
)
+
1
−
,
(
)
(
4)
T
he
in
put
te
d
w
e
ig
ht
a
t
is
de
not
e
d
a
s
,
(
)
a
nd
,
(
+
1)
is
th
e
w
e
ig
ht
a
t
(
+
1)
.
M
e
a
nw
hi
le
,
1
is
th
e
le
a
r
ni
ng
r
a
te
1
va
lu
e
a
nd
is
th
e
i
v
a
lu
e
of
in
put
ve
c
to
r
I
.
T
he
a
dj
us
tm
e
nt
of
th
e
out
put
w
e
ig
ht
i
n node
j
i
s
pe
r
f
or
m
e
d ba
s
e
d on (
5)
.
(
+
1)
=
,
(
)
+
2
×
(
5)
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
la
s
s
if
ic
at
io
n of
s
k
in
c
an
c
e
r
i
m
ag
e
s
by
apply
in
g
s
impl
e
e
v
ol
v
in
g c
onne
c
ti
oni
s
t
s
y
s
te
m
(
Al
-
K
how
ar
iz
m
i
)
425
W
he
r
e
,
(
)
is
th
e
va
lu
e
of
th
e
e
xi
t
w
e
ig
ht
a
t
,
w
he
n
(
)
,
,
(
+
1)
is
th
e
va
lu
e
of
th
e
w
e
ig
ht
e
nt
e
r
e
d
a
t
,
w
he
n
(
+
1)
,
2
is
th
e
le
a
r
n
in
g
r
a
te
2,
is
th
e
a
c
ti
va
ti
on
va
lu
e
of
th
e
node
.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
S
A
t
th
is
s
ta
ge
, t
he
s
ki
n c
a
nc
e
r
c
la
s
s
if
ic
a
ti
on pr
oc
e
s
s
f
or
pa
ti
e
nt
s
w
it
h m
a
li
gna
nt
a
nd be
ni
gn mol
e
s
ha
s
be
e
n
c
a
r
r
ie
d
out
us
in
g
th
e
m
odi
f
ie
d
S
E
C
oS
m
e
th
od.
T
he
in
it
ia
l
pr
oc
e
s
s
in
c
lu
de
s
da
ta
s
e
t
c
ol
le
c
ti
on
in
f
o
r
m
of
im
a
ge
s
of
m
a
li
gna
nt
a
nd
be
ni
gn
s
ki
n
m
ol
e
s
.
F
ig
ur
e
3
s
how
s
t
he
pr
oc
e
s
s
of
gr
a
ys
c
a
le
im
a
ge
c
onv
e
r
ti
on
a
nd
H
O
G
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
of
m
a
li
gna
nt
a
nd
be
ni
gn
s
ki
n
m
ol
e
s
b
e
f
or
e
f
e
d
in
to
S
E
C
oS
c
la
s
s
if
ie
r
.
T
he
gr
a
ys
c
a
le
of
da
ta
s
e
t
im
a
ge
s
is
a
dj
us
t
ed
in
to
th
e
d
e
s
ir
e
d
c
ol
or
,
s
uc
h
a
s
m
onoc
hr
om
a
ti
c
c
ol
or
s
f
r
om
bl
a
c
k
to
w
hi
te
.
B
y
doi
ng
th
is
,
th
e
gr
a
ys
c
a
le
im
a
g
e
onl
y
ha
s
gr
a
y
a
nd
is
c
ol
or
le
s
s
.
T
he
f
e
a
tu
r
e
e
xt
r
a
c
ti
ons
w
e
r
e
th
e
n
pe
r
f
or
m
e
d
by us
in
g t
he
H
O
G
f
e
a
tu
r
e
e
xt
r
a
c
to
r
w
it
h t
he
he
lp
of
t
ool
s
c
r
e
a
te
d w
it
h python s
yn
ta
x.
F
ig
ur
e
3. H
O
G
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
F
ig
ur
e
4
s
how
s
s
a
m
pe
l
of
m
a
li
gna
nt
s
ki
n
m
ol
e
d
a
nd
be
ni
gn
s
ki
n
m
ol
e
s
a
nd
th
e
H
O
G
of
th
os
e
im
a
ge
s
.
F
ig
ur
e
4(
a
)
a
nd
4(
c
)
s
how
e
xa
m
pl
e
s
of
m
a
li
gna
nt
a
nd
be
ni
gn
s
ki
n
m
ol
e
im
a
ge
s
us
e
d
in
th
e
e
xpe
r
im
e
nt
.
T
he
s
e
im
a
ge
s
de
s
c
r
ib
e
br
us
he
s
c
a
u
s
e
d
by
s
ki
n
c
a
nc
e
r
.
T
he
s
e
im
a
ge
s
a
r
e
th
e
n
c
onve
r
te
d
in
to
H
O
G
im
a
ge
s
a
s
s
how
n
in
F
ig
ur
e
4(
b)
a
nd
4(
d)
,
w
hi
c
h
pos
e
m
or
e
uni
f
or
m
pa
tt
e
r
ns
.
T
he
H
O
G
im
a
ge
s
a
s
S
E
C
oS
in
put
ha
ve
nor
m
a
li
z
e
d
va
lu
e
s
in
r
a
nge
of
0
to
1.
I
n
te
s
ti
ng
s
ta
ge
,
th
e
pr
oc
e
s
s
is
s
im
il
a
r
to
th
e
c
a
lc
ul
a
ti
on
in
S
E
C
oS
.
F
ir
s
t,
c
a
l
c
ul
a
te
a
ll
pa
r
a
m
e
te
r
va
lu
e
s
f
or
b
ot
h
le
a
r
ni
ng
r
a
te
s
,
th
r
e
s
hol
d
of
s
e
ns
it
iv
it
y
a
nd
e
r
r
or
a
nd
th
e
lo
t
o
f
th
e
tr
a
in
in
g
da
ta
.
T
he
n
th
e
pr
opa
ga
ti
on
is
p
e
r
f
or
m
e
d
s
ta
r
ts
f
r
om
in
pu
t
to
hi
dde
n
la
ye
r
s
by
de
te
r
m
in
i
ng
th
e
va
lu
e
s
of
a
c
ti
va
ti
ng
f
unc
ti
on
in
c
lu
di
ng
th
e
n
or
m
a
li
z
e
d
E
uc
li
de
a
n
va
lu
e
.
T
h
e
pr
opa
ga
ti
on
s
ta
r
ts
f
r
om
hi
dde
n t
o t
he
out
put
l
a
ye
r
i
s
pe
r
f
or
m
e
d a
f
te
r
w
a
r
d. T
he
c
a
lc
ul
a
te
d M
A
P
E
va
lu
e
i
s
de
t
e
r
m
in
e
d f
r
om
th
e
t
r
a
in
in
g da
ta
ba
s
e
d on the
pa
tt
e
r
n t
ha
t
ha
s
be
e
n
obt
a
in
e
d. B
ot
h t
r
a
in
in
g a
nd t
e
s
ti
ng pr
oc
e
s
s
e
s
pe
r
f
or
m
e
d i
n
th
is
pa
pe
r
us
e
136
r
a
ndomi
z
e
d
im
a
ge
s
da
ta
on
m
a
li
gna
nt
a
nd
be
ni
gn
m
ol
e
s
im
a
ge
s
.
A
bout
100
te
s
ti
ng
da
ta
a
nd 36 tr
a
in
in
g da
ta
a
r
e
de
pi
c
te
d i
n F
ig
ur
e
5.
(
a
)
(
b)
(
c
)
(
d)
F
ig
ur
e
4.
S
a
m
pl
e
of
s
ki
n c
a
nc
e
r
i
m
a
ge
s
a
nd t
he
ir
H
O
G
s
, (
a
)
m
a
li
gna
nt
s
ki
n m
ol
e
s
, (
b)
H
O
G
of
m
a
li
gna
nt
s
ki
n m
ol
e
s
, (
c
)
be
ni
gn s
ki
n m
ol
e
s
, (
d)
H
O
G
of
be
ni
gn s
ki
n m
ol
e
s
F
ig
ur
e
5
s
how
s
tr
a
in
in
g
pr
oc
e
s
s
.
T
h
e
r
e
is
a
da
ta
s
e
t
on
th
e
upp
e
r
le
f
t
w
hi
c
h
is
be
ni
gn
a
nd
m
a
li
gn
a
n
t
s
ki
n
im
a
ge
s
da
ta
,
th
e
lo
w
e
r
le
f
t
da
ta
s
e
t
i
s
nor
m
a
li
z
e
d
d
a
ta
f
or
be
ni
gn
a
nd
m
a
li
gna
nt
s
ki
n
m
ol
e
s
w
it
h
a
r
a
nge
of
0 t
o 1 a
nd t
he
da
ta
s
e
t
li
s
t
on t
he
r
ig
ht
i
s
t
he
r
e
s
ul
t
of
pr
opa
ga
ti
on f
r
om
hi
dde
n l
a
ye
r
t
o
th
e
out
put
l
a
ye
r
. T
he
out
put
la
ye
r
a
ls
o
r
e
s
ul
t
s
th
e
c
a
lc
ul
a
te
d
M
A
P
E
va
lu
e
f
r
om
th
e
a
m
ount
of
da
ta
be
in
g
tr
a
in
e
d.
F
ig
ur
e
6
s
how
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
:
421
–
429
426
th
e
vi
r
tu
a
l
r
e
s
ul
ts
of
th
e
te
s
ti
ng
da
ta
.
T
he
ge
ne
r
a
te
d
M
A
P
E
pr
oduc
e
d
th
e
m
in
im
um
r
e
s
ul
ts
w
h
e
n
in
put
pa
r
a
m
e
te
r
va
lu
e
of
le
a
r
ni
ng
r
a
te
1
is
0.3,
le
a
r
ni
ng
r
a
te
2
is
0.3
,
th
e
s
e
ns
it
iv
it
y
th
r
e
s
hol
d
is
0.5
a
nd
th
e
e
r
r
or
th
r
e
s
hol
d
is
0.1,
pr
oduc
in
g
M
A
P
E
of
0.5184845
a
s
s
how
n
in
F
ig
ur
e
6(
a
)
.
M
e
a
nw
hi
le
,
th
e
num
be
r
of
hi
dde
n
node
s
i
s
23 f
or
gi
ve
n m
in
im
um
M
A
P
E
a
s
de
pi
c
te
d i
n F
ig
ur
e
6(
b)
.
F
ig
ur
e
5. S
ki
n c
a
nc
e
r
pa
tt
e
r
n t
r
a
in
in
g
(
a
)
(
b)
F
ig
ur
e
6. V
is
ua
l
r
e
s
ul
ts
of
s
ki
n c
a
nc
e
r
t
e
s
ti
ng
, (
a
)
M
A
P
E
vs
e
r
r
or
t
hr
e
s
hol
d a
nd s
e
ns
it
iv
it
y t
hr
e
s
hol
d
,
(
b)
M
A
P
E
vs
N
um
be
r
of
hi
dde
n node
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
la
s
s
if
ic
at
io
n of
s
k
in
c
an
c
e
r
i
m
ag
e
s
by
apply
in
g
s
impl
e
e
v
ol
v
in
g c
onne
c
ti
oni
s
t
s
y
s
te
m
(
Al
-
K
how
ar
iz
m
i
)
427
T
a
bl
e
1
s
how
s
th
e
c
om
bi
na
ti
on
of
pa
r
a
m
e
te
r
s
in
th
e
m
odi
f
ie
d
S
E
C
oS
m
e
th
od
w
he
r
e
th
e
a
c
ti
va
ti
on
f
unc
ti
on
in
vol
ve
s
th
e
nor
m
a
li
z
e
d
E
uc
li
de
a
n
di
s
ta
nc
e
f
or
m
ul
a
.
T
he
le
a
r
ni
ng
r
a
te
va
r
ie
s
f
r
om
0.3
to
0.9,
th
e
s
e
ns
it
iv
it
y
th
r
e
s
hol
d
is
f
r
om
0.5
to
0.99
a
nd
e
r
r
or
th
r
e
s
hol
d
is
f
r
om
0.05
to
0.1.
T
he
num
be
r
of
hi
dde
n
node
s
is
c
ha
ngi
ng
dyna
m
ic
a
ll
y
a
s
S
E
C
oS
a
lg
or
it
hm
c
ont
in
uous
ly
le
a
r
ns
f
r
om
th
e
in
put
r
e
c
e
iv
e
d
a
nd
pe
r
f
or
m
s
a
da
pt
iv
e
l
e
a
r
ni
ng.
T
a
bl
e
1. C
la
s
s
if
ic
a
ti
on r
e
s
ul
t
w
it
h m
odi
f
ie
d S
E
C
oS
L
e
a
r
ni
ng R
a
t
e
1
L
e
a
r
ni
ng R
a
t
e
2
S
e
ns
. T
hr
e
s
hol
d
E
r
r
or
T
hr
e
s
hol
d
M
A
P
E
N
ode
0.3
0.3
0.5
0.1
0.614285
28
0.3
0.6
0.5
0.1
0.671923
33
0.3
0.9
0.5
0.1
0.645498
37
0.6
0.3
0.5
0.1
0.585737
40
0.6
0.6
0.5
0.1
0.622073
45
0.6
0.9
0.5
0.1
0.611276
47
0.9
0.3
0.5
0.1
0.599489
45
0.9
0.6
0.5
0.1
0.595727
49
0.9
0.9
0.5
0.1
0.616414
51
0.3
0.3
0.5
0.05
0.649967
32
0.3
0.6
0.5
0.05
0.686454
40
0.3
0.9
0.5
0.05
0.676481
41
0.6
0.3
0.5
0.05
0.627259
45
0.6
0.6
0.5
0.05
0.649168
48
0.6
0.9
0.5
0.05
0.615139
48
0.9
0.3
0.5
0.05
0.614936
48
0.9
0.6
0.5
0.05
0.636954
54
0.9
0.9
0.5
0.05
0.599240
54
0.3
0.3
0.5
0.1
0.518484
23
0.3
0.6
0.5
0.1
0.592122
32
0.3
0.9
0.5
0.1
0.564593
32
0.6
0.3
0.99
0.1
0.527925
24
0.6
0.6
0.99
0.1
0.589566
32
0.6
0.9
0.99
0.1
0.564456
33
0.9
0.3
0.99
0.1
0.539383
24
0.9
0.6
0.99
0.1
0.582916
32
0.9
0.9
0.99
0.1
0.557999
34
0.3
0.3
0.99
0.05
0.567959
28
0.3
0.6
0.99
0.05
0.585832
33
0.3
0.9
0.99
0.05
0.579936
34
0.6
0.3
0.99
0.05
0.556921
28
0.6
0.6
0.99
0.05
0.597425
33
0.6
0.9
0.99
0.05
0.577321
36
0.9
0.3
0.99
0.05
0.550236
28
0.9
0.6
0.99
0.05
0.591618
35
0.9
0.9
0.99
0.05
0.556443
37
S
E
C
oS
pe
r
f
or
m
s
be
s
t
ove
r
ot
he
r
m
e
th
ods
s
uc
h
a
s
pr
oba
bi
li
s
ti
c
ne
ur
a
l
ne
twor
k
(
P
N
N
)
a
nd
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
in
te
r
m
of
f
a
ls
e
ne
ga
ti
ve
a
s
S
E
C
oS
is
a
bl
e
to
l
e
a
r
n
a
nd
to
a
da
pt
in
put
pa
tt
e
r
n
c
ont
in
uous
ly
.
E
ve
n
dough
it
tr
ue
pos
it
iv
e
i
s
le
s
s
th
a
n
S
V
M
.
S
E
C
oS
is
m
or
e
r
obus
t
in
de
a
li
ng
w
it
h
c
ha
ng
e
s
in
e
r
r
or
pa
tt
e
r
ns
F
ig
ur
e
7 [
31]
.
F
ig
ur
e
7. S
E
C
oS
a
nd
ot
he
r
c
la
s
s
if
ie
r
s
c
om
pa
r
is
on
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
:
421
–
429
428
4.
C
O
N
C
L
U
S
I
O
N
I
n
s
um
m
a
r
y,
th
is
pa
pe
r
ha
s
di
s
c
us
s
e
d
th
e
c
la
s
s
if
ic
a
ti
on
of
s
ki
n
c
a
nc
e
r
w
it
h
a
n
im
a
ge
da
ta
s
e
t
in
th
e
f
or
m
o
f
m
a
li
gna
nt
a
nd
be
ni
gn s
ki
n i
m
a
ge
s
by us
in
g
S
E
C
oS
a
lg
or
it
hm
. I
n
or
de
r
t
o p
r
e
pa
r
e
t
he
c
la
s
s
if
ie
r
i
nput
,
gr
a
ys
c
a
li
ng
a
nd
H
O
G
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
w
e
r
e
p
e
r
f
or
m
e
d.
T
he
nor
m
a
li
z
e
d
E
uc
li
de
a
n
di
s
t
a
nc
e
w
a
s
e
m
pl
oye
d
a
s
it
s
pe
r
f
or
m
a
nc
e
is
be
tt
e
r
th
a
n
H
a
m
m
in
g
a
nd
M
a
nha
tt
a
n
di
s
ta
nc
e
s
.
T
he
m
odi
f
ie
d
S
E
C
oS
w
or
ks
w
e
ll
i
n
pe
r
f
or
m
in
g
c
ont
in
uous
le
a
r
ni
ng
of
th
e
r
e
c
e
iv
e
d
a
nd
it
c
ha
nge
s
hi
dde
n
node
a
da
pt
iv
e
ly
.
T
he
e
va
lu
a
ti
on
pr
oc
e
s
s
of
th
e
s
ki
n
c
a
nc
e
r
di
s
e
a
s
e
w
a
s
c
a
r
r
ie
d
out
on
136
da
ta
s
e
ts
w
it
h
100
tr
a
in
in
g
a
nd
36
te
s
ti
ng
da
ta
.
C
om
bi
na
ti
on
pr
oc
e
s
s
of
pa
r
a
m
e
te
r
s
in
th
e
c
la
s
s
if
ic
a
ti
on
obt
a
in
e
d
th
e
s
m
a
ll
e
s
t
e
r
r
or
in
th
e
le
a
r
ni
ng
r
a
te
1
of
0.3,
le
a
r
ni
ng
r
a
te
2
of
0.3,
w
he
n
th
e
s
e
ns
it
iv
it
y
th
r
e
s
hol
d
is
0.5
a
nd
th
e
e
r
r
or
th
r
e
s
hol
d
is
0.1.
T
he
m
in
im
um
obt
a
in
e
d
M
A
P
E
is
0.5184845
a
nd
th
e
num
be
r
of
hi
dde
n
node
s
is
23
node
s
.
E
ve
n
dough
tr
ue
pos
it
iv
e
of
S
E
c
oS
is
s
m
a
ll
e
r
th
a
n
S
V
M
;
it
s
f
a
l
s
e
ne
g
a
ti
ve
is
m
uc
h
be
tt
e
r
w
hi
c
h
pr
ove
s
th
a
t
S
E
C
oS
i
s
m
or
e
r
obus
t
in
ha
ndl
in
g f
r
e
que
nt
i
nput
pa
tt
e
r
n va
r
ia
ti
ons
.
A
C
K
N
O
WL
E
D
G
E
M
E
N
T
S
W
e
w
oul
d
li
ke
to
th
a
nk
to
th
e
U
ni
v
e
r
s
it
y
of
M
uh
a
m
m
a
d
iy
a
h
S
um
a
te
r
a
U
ta
r
a
(
U
M
S
U
)
a
nd
U
ni
ve
r
s
it
a
s
S
um
a
te
r
a
U
ta
r
a
(
U
S
U
)
f
or
t
he
s
uppor
t
pr
ovi
de
d dur
in
g t
hi
s
r
e
s
e
a
r
c
h.
R
E
F
E
R
E
N
C
E
S
[1]
S.
S.
Nikam,
“A
Comparative
Study
of
Classification
Techniques
in
Data
Mining
Algorithms,”
Int.
Journal
Mod.
Trends E
ng. Re
s
., vol. 4, no. 7, pp. 58
-
63, 2017, doi: 10.21884/ijmter.2017.4211.vxayk.
[2]
E.
M.
Hassib,
A.
I.
El
-
Desouky,
L.
M.
Labib,
and
E.
-
S.
M.
El
-
kena
wy,
“WOA
+
BRNN:
An
imbalanced
big
data
classifi
cation
framework
using
Whale
optimi
zation
and
deep
neural
n
etwork,”
Soft
Comput
.,
v.
24,
no.
8,
pp.
5573
-
5592, 2020, doi: 10.1007/s00500
-
019
-
03901
-
y.
[3]
A.
Lu
bis,
M.
Lubis,
and
A.
-
Khowarizmi,
“Optimization
of
distanc
e
formula
in
K
-
Nearest
Neighbor
method,
”
Bulletin. Elec. Eng. Informati
cs
, vol. 9, no. 1, pp. 326
-
338, 2020, doi: 10.11591/eei.v9i1.1464.
[4]
S. Pray
udani,
A. Hiz
riadi,
Y. La
se, Y.
Fatmi, a
nd A. K
howa
rizmi, “An
alysis Accuracy
of Forecastin
g Measurement
Technique
on
Random
K
-
Nearest
Neighbor
using
MAPE
and
MSE,”
J.
Physic
Conf.
Ser
.,
vol.
1361,
no.
1,
pp.
0
-
8,
2019, doi: 10.1088/1742
-
6596/1361/1/012089.
[5]
V.
Lopez
-
vazquez,
J.
Lopez
-
guede,
S.
Marini,
E.
F
anelli,
E.
Johnsen
,
and
J.
Aguzzi,
“Video
image
enhancement
and machi
ne learnin
g pipel
ine
for underw
ater
animal
detection
and
cla
ss. at
cabled obs
ervatories,
”
Sensors
, vol.
20,
no. 3, 2020, doi: 10.3390/s20030726.
[6]
A.
P.
Gopi,
R.
Jyothi,
V.
L.
Narayana,
an
d
K.
Sandeep,
“Classific
ation
of
tweets
data
based
on
pol
arity
using
improved RBF ker
nel of SVM,”
In. Jour. Inf
o. Technol
.
, 2020, doi: 10.1007/s41870
-
019
-
00409
-
4.
[7]
A.
E.
Maxwell,
T.
A.
Warner,
and
F.
Fang,
"
Implementation
of
machine
learning
classification
in
remote
sensing:
An
applied
review,"
Int.
J.
of
Remote
Sensing
,
vol.
3
9,
no.
9,
pp.
2784
-
2817,
2018,
doi:
10.1080/01431161.2018.1433343
[8]
American
Cancer
Society,
Cancer
Facts
&
Figures
2020,
"
Spe
cial
Section:
Cancer
in
Adolescents
and
Young
Adults,"
2020, [Online s
erial]. Available: ht
tp://
www.cancer.org. [Ac
cessed Sept. 2, 20
20].
[9]
M.
E.
Celebi,
N.
Codella,
and
A.
Halpern,
"
Dermoscopy
image
ana
lysis:
overview
and
future
directions,"
IEEE
jour. of biome
d. and health
informatics
, vol. 23, no. 2, pp. 474
-
478, 2019, doi: 10.1109/JBHI.2019.2895803.
[10]
M.
N.
Munjal
and
S.
Bhatia,
"
A
Novel
Technique
for
Effective
Image
Gallery
Search
using
Content
Based
Image
Retrieva
l
Sys
tem,"
2019
Int.
Conf.
on
Machine
Learning,
Big
Data,
Cloud
and
Parallel
Computing
,
Faridabad,
India, pp.
25
-
29, 2019, doi: 10.1109/COMITCon.2019.8862206.
[11]
Y.
M.
Chung,
C.
S.
Hu,
A.
Lawson,
and
C.
Smyth,
"
Topologic
al
ap
proaches
to
skin
disease
image
analysis
".
In
2018 IEEE Int. Conf. on Big Data (Big Data)
, IEEE, pp. 100
-
105, 2008, doi: 10.1109/BigData.2018.8622175
[12]
A.
Khowarizmi,
I.
Nasution,
M.
Lubis,
and
A.
Lubis,
"
The
effect
o
f
a
SECoS
in
crude
palm
oil
forecasting
to
improve
business
intelligence
,"
Bull.
of
El.
Eng.
and
Informatics
,
vol.
9,
no.
4,
pp.
1604
-
1611,
2020,
doi:
10.11591/eei.v9i4.2388
[13]
M.
J.
Watts,
“Fuzzy
Rule
Extraction
From
Simple
Evolving
Conn
ectionist
Systems,”
Int.
Journal
on
Comput
.
Intell. Ap
pl.,
vol. 04, no. 03, pp. 299
-
308, 2004, doi:
10.1142/s146902680400132x.
[14]
S.
Silva,
P.
Costa,
M.
Gouvea,
A.
Lacerda,
F.
Alves,
and
D.
Leite,
“H
igh
impedance
fault
detection
in
power
distr.
systems
using
wavelet
transform
and
evolving
neural
network,”
El
e
ctr.
Pow.
Syst.
Res.,
vol.
154,
pp.
474
-
483,
201
8, doi: 10.1016/j.epsr.2017.08.039.
[15]
N.
K.
Kasabov,
“Evolving
connectionist
systems
for
adaptive
lear
ning
and
knowledge
discovery:
Trends
and
directions,”
Knowledge
-
B. Syst
., v. 80, pp. 24
-
33, 2015, doi: 10.1016/j.knosys.2014.12.032.
[16]
M.
J.
Watts,
“Evolving
Connectio
nist
Systems
can
predic
t
outbrea
ks
of
the
aphid
Rhopalosiph
um
padi,”
Int.
Conf.
Neural Netwo
rks
, pp. 646
-
650, 2014, doi: 10.1109/IJCNN.2014.6889755.
[17]
M.
J.
Watts,
“Sleep
learning
and
Max
-
Min
aggregation
of
Evolving
Connectionist
Systems,”
Proc.
In
t.
Jt.
Conf.
Neural Netwo
rks
, vol. 2016
-
Octob, pp. 4325
-
4330, 2016, doi: 10.1109/IJCNN.2016.7727764.
[18]
N.
Kasabov,
“Evolving
connectionist
systems:
The
knowledge
engi
neering
approach,”
Evol.
Connect.
Syst.
Kn.
Eng. Approach
(Second Ed.),
pp. 1
-
457, 2007, doi
: 10.1007/978
-
1
-
84628
-
347
-
5.
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
la
s
s
if
ic
at
io
n of
s
k
in
c
an
c
e
r
i
m
ag
e
s
by
apply
in
g
s
impl
e
e
v
ol
v
in
g c
onne
c
ti
oni
s
t
s
y
s
te
m
(
Al
-
K
how
ar
iz
m
i
)
429
[19]
S.
Agrawal
and
J.
Agrawal,
“Survey
on
anomaly
detection
using
dat
a
mining
techniques,”
Procedia
Comp..
Sci
.,
vol. 60, no. 1, pp. 708
-
713, 2015, doi: 10.1016/j.procs.2015.08.220.
[20]
A.
Khowarizmi,
“Modified
of
Simple
Evolving
Con
nectionist
System
using
Distance
Formulas,”
Universitas
Sumatera Utara
, 2017.
[21]
A.
Khowarizmi,
A.
M.
Lubis,
and
A.
R.
Lubis,
“Classification
of
Taj
weed
Al
-
Qur’an
on
Images
Applied
Varying
Normalized Distance Fo
rmulas,” no. 3, p
p. 21
-
25, 2020, doi: 10.1145/3
396730.3396739.
[22]
I.
Mierswa,
K.
Morik,
"
Automatic
feature
extraction
for
classifying
audio
data"
,
Machine
learning
,
Vol.
58,
No.2,
p.127
-
149, doi: 10.1007/s10994
-
005
-
5824
-
7.
[23]
G.
Forman
and
E.
Kirshenbaum,
“Extremely
fast
text
feature
extracti
on
for
classif
ication
and
indexing,”
Int.
Conf.
Inf. Know
l. Manag.
Proc
., pp. 1221
-
1230, 2008, doi: 10.1145/1458082.1458243.
[24]
L.
Benuskova,
N.
Kasabov,
L.
Benuskova,
and
N.
Kasabov,
"
Evolvin
g
Connectionist
Systems
(ECOS),"
Springer
,
2007, doi: 10.1007/978
-
0
-
387
-
48355
-
9_5.
[25]
A.
Sebbar,
S.
Hedd
am,
O.
Kisi,
L.
Djemili,
and
L.
Houichi,
“Com
parison
of
Evolving
Conne
ctionist
Systems
(EC
oS)
and
Neural
Networks
for
Modelling
Daily
Pan
Evaporation
f
rom
Algerian
Dam
Reservoirs,”
Heidelberg:
Springe
r Ber
lin Hei
del., p
p. 1
-
19, doi: 10.1007/698_2020_527.
[26]
A.
Khowarizmi,
I.
R.
Nasution,
M.
Lubis,
and
A.
R.
Lubis,
“The
effe
ct
of
a
secos
in
crude
palm
oil
forecasting
to
improve
business
intelligence
,”
Bull.
Electr.
Eng.
Informatics
,
vol.
9,
no.
4,
pp.
1604
-
1611,
2020,
doi
:
10.11591/eei.v9i4.2388.
[27]
N. K. Kasabov, L.
Goh, and
M. Sulli
van,
“Integrated Prognost
ic Profil
es: Combini
ng Clinical and
Gene E
xpressio
n
Information
through
Evolving
Connection
ist
Approach,”
in
Inf.
Proc.
and
Liv.
Systems,
vol.
2
,
pp.
695
-
707,
2005,
doi: 10.1142/9781860946882_0010.
[28]
S. Kalo
uche,
“Visio
n
-
Based Cla
ssificati
on of Skin
Cancer
using De
ep Lea
rning,”
Stanford
, 2016.
[29]
R.
F.
Rahmat,
A.
Rizki,
A.
F.
Alharthi,
and
R.
Budiarto,
“Big
data
forecasting
using
evolving
multi
-
layer
perceptron,”
2016
4th
Saudi
Int.
C.
Inf.
Technol.
(Big
D
ata
Anal.
KACSTIT)
2016,
2016,
doi:
10.1109/KACSTIT.2016.7756069.
[30]
A.
Khowarizmi,
O.
S.
Sitompul,
S.
Suherman,
and
E.
B.
Nababan,
“
Measuring
the
Accuracy
of
Simple
Evolving
Connectio
nist
System
with
Varying
Distance
Formulas,”
in
J.
of
Ph
y.:
Conference
Series
,
2017,
vol.
930,
no.
1,
doi: 10.1088/1742
-
6596/930/1/012004.
[31]
S.
Silva,
"
High
impedance
fault
detection
in
power
distribution
sys
tems
using
wavelet
transform
and
evolvin
g
neural network."
Elec. Pow. Syst. Research
, vol. 154, pp. 474
-
483
,
201
8, doi: 10.1016/j.epsr.2017.08.039.
B
I
O
G
R
A
P
H
I
E
S
O
F
A
U
T
H
O
R
S
Al
-
Khowarizmi
is
the
head
of
Information
Technology
Department
within
CSIT
Faculty
in
Universitas
Muhammadiyah
Sumatera
Utara
(UMSU),
Medan
Indon
esia.
He
obtained
his
MSc
in
Information
Technology
from
University
of
Sumatera
Utara
in
201
7.
His
research
interest
lies
on
artificial
intell
igence and
machine l
earning, d
ata science
and busi
ness appl
ications
.
Suherma
n
is
the
head
of
master
degree
program
in
Electrical
Engi
neering
within
Universitas
Sumater
a
Utara
.
He
obtaine
d
his
B.Eng,
MSc
and
PhD
from
Un
iversit
as
Sumater
a
Utara
-
Indonesia,
RMIT
-
Australia
and
De
Montfort
University
-
UK
subseq
uently
in
2000,
2009
and
201
3.
He
joined
some
courses
in
Melbourne
University,
Ilmenau
Uni
versity
of
Technology
and
European
Bioinformatics
Institute
from
2008
to
2012.
His
researc
h
interests
are
in
applied
communi
cation s
ystem,
renewable energ
y, and co
mputer s
cience.
Evaluation Warning : The document was created with Spire.PDF for Python.