I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
43
,
No
.
2
,
A
u
g
u
s
t
2
0
2
6
,
p
p
.
6
4
0
~
6
5
0
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/ijeecs.v
43
.i
2
.
pp
640
-
6
5
0
640
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
Adv
a
nced perso
n
a
l bank
ruptcy p
r
ediction usin
g
tre
e
-
ba
sed
deep learning
mo
dels
Nha
t
Ng
uy
en
M
inh
,
D
uy
Ng
o
H
o
a
ng
K
ha
nh
F
a
c
u
l
t
y
o
f
B
a
n
k
i
n
g
,
H
o
C
h
i
M
i
n
h
U
n
i
v
e
r
si
t
y
o
f
B
a
n
k
i
n
g
(
H
U
B
)
,
H
o
C
h
i
M
i
n
h
C
i
t
y
,
V
i
e
t
n
a
m
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
4
,
2
0
2
4
R
ev
is
ed
J
u
n
2
9
,
2
0
2
6
Acc
ep
ted
J
u
l 2
3
,
2
0
2
6
Du
e
to
th
e
u
n
sta
b
le
e
c
o
n
o
m
ic
c
o
n
d
it
i
o
n
s
,
wo
rse
n
e
d
b
y
th
e
p
o
st
-
COV
ID
-
1
9
e
n
v
iro
n
m
e
n
t
a
n
d
p
e
rsiste
n
t
f
o
re
i
g
n
wa
rs
in
2
0
2
4
,
fi
n
a
n
c
ial
i
n
stit
u
ti
o
n
s
h
a
v
e
g
ro
wi
n
g
d
iff
icu
lt
ies
in
a
c
c
u
ra
tely
p
re
d
icti
n
g
c
u
st
o
m
e
r
d
e
fa
u
lt
p
ro
b
a
b
i
li
ty
.
Th
is
stu
d
y
e
x
a
m
in
e
s
th
e
u
se
o
f
s
o
p
h
ist
ica
ted
tree
-
b
a
se
d
d
e
e
p
lea
rn
in
g
a
n
d
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
m
o
d
e
ls
fo
r
fo
re
c
a
stin
g
p
e
rso
n
a
l
b
a
n
k
r
u
p
tcy
.
T
h
is
re
se
a
rc
h
u
ti
li
se
s
a
d
a
tas
e
t
o
f
ro
u
g
h
ly
9
,
8
0
0
in
d
iv
i
d
u
a
ls
fr
o
m
Vie
tn
a
m
e
se
fin
a
n
c
ial
i
n
stit
u
ti
o
n
s,
sp
a
n
n
i
n
g
fr
o
m
2
0
1
2
to
2
0
2
2
,
t
o
e
v
a
l
u
a
te
th
e
e
ffi
c
a
c
y
o
f
m
o
d
e
ls
i
n
c
lu
d
in
g
n
e
u
ra
l
d
e
c
isio
n
tree
,
d
e
e
p
fo
re
st,
tab
u
lar
c
o
n
v
o
l
u
ti
o
n
a
l
n
e
u
ra
l
n
e
two
r
k
s
(TBCNN),
a
n
d
n
e
u
ra
l
o
b
li
v
i
o
u
s
d
e
c
isio
n
e
n
se
m
b
les
(NO
DE).
Th
e
re
su
lt
s
d
e
m
o
n
stra
t
e
th
a
t
t
h
e
De
e
p
F
o
re
st
m
o
d
e
l
fa
r
su
rp
a
ss
e
s
it
s
c
o
m
p
e
ti
t
o
rs,
p
r
o
v
i
d
in
g
n
e
a
rly
flaw
les
s
p
re
d
icte
d
a
c
c
u
ra
c
y
a
n
d
e
n
h
a
n
c
e
d
in
terp
re
tab
il
it
y
.
T
h
e
fin
d
in
g
s
h
ig
h
li
g
h
t
t
h
e
e
ffica
c
y
o
f
tree
-
b
a
se
d
d
e
e
p
lea
rn
in
g
a
n
d
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
m
o
d
e
ls
a
s
e
ffe
c
ti
v
e
i
n
stru
m
e
n
ts
fo
r
fin
a
n
c
ial
risk
m
a
n
a
g
e
m
e
n
t,
e
sp
e
c
ially
in
v
o
latil
e
a
n
d
u
n
p
re
d
icta
b
l
e
e
c
o
n
o
m
ic en
v
iro
n
m
e
n
ts.
K
ey
w
o
r
d
s
:
Dee
p
f
o
r
est
Neu
r
al
o
b
liv
io
u
s
d
ec
is
io
n
en
s
em
b
les
Per
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ictio
n
T
ab
u
lar
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
T
r
ee
-
b
ased
d
ee
p
lear
n
in
g
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Nh
at
Ng
u
y
en
Mi
n
h
Facu
lty
o
f
B
an
k
in
g
,
Ho
C
h
i M
in
h
Un
iv
er
s
ity
o
f
B
an
k
in
g
(
H
UB
)
3
6
T
o
n
T
h
at
Dam
s
tr
ee
t,
Ng
u
y
en
T
h
ai
B
in
h
war
d
,
D
is
tr
ict
1
Ho
C
h
i M
in
h
C
ity
,
Vietn
am
E
m
ail: n
h
atn
m
@
h
u
b
.
e
d
u
.
v
n
1.
I
NT
RO
D
UCT
I
O
N
Pre
cise
ly
f
o
r
e
ca
s
t
in
g
p
er
s
o
n
a
l
b
an
k
r
u
p
tc
y
is
c
r
u
ci
al
f
o
r
f
i
n
an
cia
l
o
r
g
an
is
at
io
n
s
,
s
i
n
c
e
it
p
r
o
f
o
u
n
d
l
y
im
p
ac
ts
r
is
k
m
a
n
ag
em
en
t
te
c
h
n
i
q
u
es
a
n
d
le
n
d
i
n
g
p
o
l
ici
es
[
1
]
.
C
o
n
v
e
n
ti
o
n
a
l
m
o
d
els
li
k
e
l
o
g
is
t
ic
r
e
g
r
ess
i
o
n
an
d
s
i
m
p
le
d
ec
is
i
o
n
t
r
ee
s
,
h
o
wev
er
c
o
m
m
o
n
,
f
r
e
q
u
e
n
tl
y
in
ad
eq
u
atel
y
a
d
d
r
ess
t
h
e
c
o
m
p
l
ex
i
ty
a
n
d
n
o
n
-
li
n
ea
r
co
r
r
el
ati
o
n
s
p
r
ese
n
t
i
n
c
o
n
te
m
p
o
r
a
r
y
f
i
n
a
n
c
ial
d
at
a
[
2
]
.
T
h
e
g
r
o
wi
n
g
c
o
m
p
le
x
it
y
o
f
co
n
s
u
m
er
f
i
n
a
n
ci
al
b
e
h
a
v
i
o
u
r
an
d
t
h
e
ac
c
ess
i
b
il
it
y
o
f
h
i
g
h
-
d
i
m
e
n
s
i
o
n
al
d
at
ase
ts
u
n
d
er
s
co
r
e
th
e
n
ee
d
f
o
r
in
c
r
ea
s
in
g
l
y
s
o
p
h
is
ti
ca
te
d
p
r
e
d
ic
ti
o
n
m
o
d
els
[
3
]
.
I
n
p
a
r
tic
u
la
r
,
th
e
p
r
e
d
i
cti
o
n
o
f
p
e
r
s
o
n
al
f
i
n
a
n
ci
al
d
is
t
r
ess
h
as
b
e
co
m
e
i
n
c
r
e
asi
n
g
l
y
im
p
o
r
t
an
t
i
n
le
n
d
i
n
g
e
n
v
ir
o
n
m
e
n
ts
w
h
e
r
e
b
o
r
r
o
w
er
ch
ar
ac
t
er
is
tics
,
r
e
p
a
y
m
e
n
t
c
ap
ac
i
ty
,
a
n
d
cr
e
d
it
b
e
h
a
v
i
o
u
r
in
t
er
ac
t
in
co
m
p
le
x
a
n
d
d
y
n
a
m
ic
w
a
y
s
.
I
n
s
u
c
h
s
e
tti
n
g
s
,
m
o
d
els
t
h
a
t
ca
n
ca
p
t
u
r
e
n
o
n
-
li
n
ea
r
s
tr
u
c
tu
r
es
m
o
r
e
ef
f
e
cti
v
e
ly
a
r
e
e
x
p
e
cte
d
t
o
im
p
r
o
v
e
t
h
e
q
u
al
it
y
o
f
r
is
k
ass
ess
m
e
n
t
a
n
d
cr
ed
it
-
r
ela
te
d
d
ec
is
i
o
n
-
m
a
k
i
n
g
[
4
]
.
Prio
r
s
tu
d
ies
h
a
v
e
e
x
am
in
ed
b
an
k
r
u
p
tcy
a
n
d
f
in
an
cial
d
is
tr
ess
p
r
ed
ic
tio
n
u
s
in
g
a
r
a
n
g
e
o
f
s
tatis
t
ical,
m
ac
h
in
e
lear
n
i
n
g
,
a
n
d
d
ee
p
l
ea
r
n
in
g
tec
h
n
iq
u
es.
T
r
ad
itio
n
al
ap
p
r
o
ac
h
es
h
av
e
s
h
o
wn
th
e
p
r
ac
tical
v
alu
e
o
f
s
tr
u
ctu
r
ed
p
r
e
d
ictio
n
m
o
d
els
in
id
en
tify
in
g
h
ig
h
-
r
is
k
b
o
r
r
o
wer
s
,
b
u
t
th
e
y
o
f
te
n
f
ac
e
lim
itatio
n
s
wh
en
r
elatio
n
s
h
ip
s
am
o
n
g
p
r
ed
icto
r
s
ar
e
h
ig
h
ly
n
o
n
-
lin
ea
r
o
r
wh
en
tab
u
lar
f
in
a
n
cial
d
ata
b
ec
o
m
e
m
o
r
e
co
m
p
lex
[
2
]
,
[
4
]
.
R
ec
en
t
a
d
v
an
ce
m
en
t
s
h
av
e
t
h
er
ef
o
r
e
s
h
if
ted
atten
t
io
n
to
war
d
m
o
r
e
s
o
p
h
is
ticated
m
ac
h
i
n
e
lear
n
in
g
m
o
d
els
ca
p
ab
le
o
f
lear
n
i
n
g
r
ich
er
p
atter
n
s
f
r
o
m
s
tr
u
ctu
r
e
d
d
atasets
.
Am
o
n
g
th
ese,
Dee
p
Fo
r
est
h
as
b
ee
n
h
ig
h
lig
h
ted
as
an
ef
f
ec
tiv
e
e
n
s
em
b
le
f
r
am
ewo
r
k
th
at
in
teg
r
ates
m
u
ltip
le
d
ec
is
io
n
tr
ee
s
an
d
ca
n
ac
h
iev
e
h
ig
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
A
d
va
n
ce
d
p
ers
o
n
a
l b
a
n
kru
p
tc
y
p
r
ed
ictio
n
u
s
in
g
tr
ee
-
b
a
s
ed
d
ee
p
lea
r
n
in
g
mo
d
els
(
N
h
a
t Ng
u
ye
n
Min
h
)
641
p
r
ed
ictiv
e
ac
cu
r
ac
y
in
co
m
p
lex
f
o
r
ec
asti
n
g
task
s
[
5
]
.
T
h
is
ap
p
r
o
ac
h
is
p
ar
ticu
lar
ly
r
el
ev
an
t
in
f
in
an
cia
l
ap
p
licatio
n
s
b
ec
au
s
e
it
ca
n
m
o
d
el
h
ier
ar
ch
ical
r
elatio
n
s
h
ip
s
in
th
e
d
ata
wh
ile
p
r
e
s
er
v
in
g
a
le
v
el
o
f
in
ter
p
r
etab
ilit
y
th
at
is
im
p
o
r
tan
t f
o
r
r
is
k
m
an
ag
e
m
en
t p
r
ac
tic
e
[
6
]
,
[
7
]
.
Alo
n
g
s
id
e
Dee
p
Fo
r
est,
tr
ee
-
b
ased
d
ee
p
lear
n
i
n
g
m
o
d
els
s
u
ch
as
th
e
Neu
r
al
Dec
is
i
o
n
T
r
ee
h
a
v
e
attr
ac
ted
atten
tio
n
f
o
r
co
m
b
in
i
n
g
th
e
in
ter
p
r
etab
ilit
y
o
f
co
n
v
en
tio
n
al
d
ec
is
io
n
tr
ee
s
with
th
e
lear
n
in
g
ca
p
ac
ity
o
f
d
ee
p
m
o
d
els
[
8
]
.
B
y
o
p
tim
is
in
g
d
ec
is
io
n
b
o
u
n
d
ar
ies
ac
r
o
s
s
lay
er
s
,
th
e
Neu
r
al
Dec
is
io
n
T
r
ee
ca
n
im
p
r
o
v
e
ac
cu
r
ac
y
an
d
g
e
n
er
aliza
t
io
n
[
9
]
,
wh
ile
r
etain
i
n
g
a
m
o
r
e
tr
an
s
p
ar
en
t
d
ec
is
io
n
p
r
o
ce
s
s
th
at
is
ea
s
ier
to
co
m
m
u
n
icate
to
m
an
a
g
er
s
,
s
tak
eh
o
ld
er
s
,
an
d
r
eg
u
lato
r
s
[
1
0
]
.
At
th
e
s
am
e
tim
e,
d
ee
p
lear
n
in
g
m
o
d
els
d
esig
n
ed
f
o
r
tab
u
la
r
d
ata,
s
u
ch
as
T
B
C
NN
an
d
NODE
,
h
av
e
also
b
ee
n
p
r
o
p
o
s
ed
as
p
r
o
m
i
s
in
g
alter
n
ativ
es
f
o
r
s
tr
u
ctu
r
ed
f
in
a
n
cial
d
atasets
[
1
1
]
.
T
h
ese
m
o
d
els
ar
e
ef
f
ec
ti
v
e
in
ca
p
tu
r
i
n
g
co
m
p
lex
n
o
n
-
lin
ea
r
d
ep
en
d
en
cies
in
b
o
r
r
o
wer
-
le
v
el
d
ata,
b
u
t
t
h
eir
p
r
ac
tical
u
s
e
i
n
f
in
a
n
cial
d
ec
is
io
n
-
m
ak
in
g
r
em
ain
s
c
o
n
s
tr
ain
ed
b
y
c
o
n
ce
r
n
s
o
v
er
o
p
ac
ity
an
d
lim
ited
in
ter
p
r
etab
ilit
y
[
1
2
]
.
Alth
o
u
g
h
p
r
io
r
s
tu
d
ies
h
av
e
s
h
o
wn
th
e
v
alu
e
o
f
b
o
th
tr
e
e
-
b
ased
d
ee
p
m
o
d
els
an
d
tab
u
lar
d
ee
p
n
e
u
r
al
m
o
d
els,
s
ev
er
al
is
s
u
es
r
em
ai
n
u
n
r
eso
lv
ed
.
Dir
ec
t
co
m
p
ar
i
s
o
n
s
o
f
th
ese
ad
v
a
n
ce
d
m
o
d
e
ls
with
in
th
e
s
am
e
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ictio
n
f
r
am
ewo
r
k
ar
e
s
till
lim
ited
,
esp
ec
ially
in
em
er
g
in
g
m
ar
k
et
s
s
u
ch
as
Vietn
am
,
wh
er
e
b
o
r
r
o
wer
b
eh
av
io
u
r
,
cr
ed
it
co
n
d
itio
n
s
,
a
n
d
d
ata
s
tr
u
ctu
r
es
m
ay
d
if
f
e
r
f
r
o
m
th
o
s
e
in
o
th
e
r
s
ettin
g
s
.
I
n
ad
d
itio
n
,
th
e
liter
atu
r
e
h
as
n
o
t
y
et
estab
lis
h
ed
wh
eth
er
t
h
e
g
r
ea
ter
c
o
m
p
lex
ity
o
f
n
eu
r
al
m
o
d
els
lead
s
to
a
p
r
ac
tically
m
ea
n
in
g
f
u
l
ad
v
a
n
tag
e
o
v
er
tr
ee
-
b
ased
m
o
d
els
wh
en
p
r
ed
ictiv
e
p
e
r
f
o
r
m
an
ce
an
d
in
ter
p
r
etab
ilit
y
ar
e
co
n
s
id
er
ed
to
g
eth
e
r
[
1
2
]
.
T
h
ese
g
ap
s
m
o
tiv
ate
th
e
p
r
ese
n
t stu
d
y
.
B
ased
o
n
th
e
ab
o
v
e
d
is
cu
s
s
i
o
n
s
,
th
is
s
tu
d
y
aim
s
to
in
v
esti
g
ate
th
e
co
r
e
q
u
esti
o
n
:
C
an
tr
ee
-
b
ased
d
ee
p
lear
n
in
g
m
o
d
els
ac
h
iev
e
p
r
ed
ictio
n
p
e
r
f
o
r
m
an
ce
c
o
m
p
ar
ab
le
to
o
r
b
etter
th
an
d
ee
p
n
eu
r
al
n
etwo
r
k
s
in
m
u
lti
-
class
p
er
s
o
n
al
b
an
k
r
u
p
t
cy
p
r
ed
ictio
n
wh
ile
r
etain
in
g
s
tr
o
n
g
er
in
ter
p
r
etab
ilit
y
?
B
o
th
tech
n
iq
u
es
h
a
v
e
p
r
o
v
e
n
h
ig
h
ly
u
s
ef
u
l
in
d
if
f
er
en
t
s
ettin
g
s
;
n
o
n
eth
eless
,
th
e
o
p
tim
al
ch
o
ice
lik
ely
h
in
g
es
o
n
th
e
ap
p
licatio
n
'
s
u
n
iq
u
e
r
e
q
u
ir
em
e
n
ts
wh
eth
er
th
e
em
p
h
asis
is
o
n
m
ax
im
is
in
g
ac
cu
r
ac
y
o
r
ass
u
r
in
g
in
ter
p
r
etab
ilit
y
.
Acc
o
r
d
in
g
ly
,
th
is
s
tu
d
y
co
m
p
ar
es
f
o
u
r
ad
v
an
ce
d
m
o
d
els
in
clu
d
in
g
Neu
r
al
Dec
is
io
n
T
r
ee
,
Dee
p
Fo
r
est,
T
B
C
NN
an
d
NODE
to
i
d
en
tify
th
e
m
o
d
el
th
at
p
r
o
v
id
es
th
e
m
o
s
t
s
u
itab
le
b
alan
ce
b
etwe
en
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
an
d
in
ter
p
r
etab
ili
ty
in
th
e
co
n
tex
t
o
f
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ictio
n
[
1
3
]
.
Mo
r
e
s
p
ec
if
ically
,
th
e
p
ap
er
co
n
tr
ib
u
tes
to
th
e
liter
atu
r
e
in
th
r
ee
way
s
.
First,
it
p
r
o
v
id
es
a
co
m
p
ar
ativ
e
ass
ess
m
en
t
o
f
tr
ee
-
b
ased
d
ee
p
lear
n
in
g
an
d
d
ee
p
n
eu
r
al
n
etwo
r
k
m
o
d
els
f
o
r
m
u
lti
-
class
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ictio
n
.
Seco
n
d
,
it
o
f
f
er
s
em
p
ir
ical
ev
i
d
e
n
ce
b
ase
d
o
n
a
d
ataset
s
u
p
p
o
r
ted
b
y
th
e
r
esear
ch
I
n
s
titu
te
o
f
Ho
C
h
i
Min
h
Un
iv
er
s
ity
o
f
B
an
k
in
g
,
co
n
s
is
tin
g
o
f
9
,
8
0
0
p
er
s
o
n
al
lo
an
cu
s
to
m
er
s
at
co
m
m
er
cial
b
an
k
s
a
n
d
cr
e
d
it
in
s
titu
tio
n
s
in
Vietn
am
d
u
r
in
g
th
e
p
e
r
io
d
f
r
o
m
2
0
1
2
t
o
2
0
2
2
.
T
h
ir
d
,
it
ev
alu
ates
m
o
d
el
s
u
itab
ilit
y
n
o
t
o
n
ly
in
ter
m
s
o
f
ac
cu
r
ac
y
an
d
F1
-
s
co
r
e,
b
u
t
also
f
r
o
m
th
e
p
er
s
p
ec
tiv
e
o
f
in
ter
p
r
etab
ilit
y
an
d
p
r
ac
tical
ap
p
licab
ilit
y
in
f
in
a
n
cial
r
is
k
m
an
ag
em
en
t
[
1
3
]
.
T
h
e
s
tu
d
y
was
co
n
d
u
cted
o
n
a
d
ataset
s
u
p
p
o
r
ted
b
y
t
h
e
r
esear
ch
I
n
s
titu
te
o
f
Ho
C
h
i
Min
h
Un
iv
er
s
ity
o
f
B
an
k
in
g
.
I
n
th
e
f
o
llo
win
g
s
ec
tio
n
s
o
f
th
e
s
tu
d
y
,
th
e
a
u
th
o
r
will
p
r
esen
t
th
e
m
eth
o
d
,
em
p
ir
ical
r
esu
lts
,
d
is
cu
s
s
io
n
o
f
th
e
f
in
d
i
n
g
s
,
an
d
c
o
n
clu
s
io
n
.
2.
M
E
T
H
O
D
Fig
u
r
e
1
illu
s
tr
ates
th
e
e
x
p
er
i
m
en
tal
wo
r
k
f
lo
w
a
d
o
p
te
d
in
t
h
is
s
tu
d
y
.
T
h
e
p
r
o
c
e
d
u
r
e
b
e
g
in
s
with
d
ata
p
r
ep
r
o
ce
s
s
in
g
,
f
o
llo
wed
b
y
f
ea
tu
r
e
p
r
e
p
ar
atio
n
,
d
ataset
s
p
litt
in
g
,
m
o
d
el
d
ev
elo
p
m
en
t,
h
y
p
er
p
ar
am
eter
o
p
tim
is
atio
n
,
an
d
o
u
t
-
of
-
s
am
p
le
ev
alu
atio
n
.
T
h
is
s
eq
u
en
tial
d
esig
n
was a
d
o
p
ted
to
en
s
u
r
e
t
h
at
th
e
co
m
p
ar
is
o
n
am
o
n
g
T
B
C
NN,
Neu
r
al
Dec
is
io
n
T
r
ee
,
NODE
,
an
d
Dee
p
Fo
r
est
was
co
n
d
u
cted
u
n
d
er
th
e
s
am
e
d
ata
co
n
d
itio
n
s
an
d
e
v
alu
atio
n
p
r
o
to
co
l,
th
er
e
b
y
allo
win
g
th
e
r
e
s
ea
r
ch
q
u
esti
o
n
s
tated
i
n
th
e
I
n
tr
o
d
u
ctio
n
t
o
b
e
ex
am
in
ed
in
a
tr
an
s
p
ar
e
n
t a
n
d
r
ep
r
o
d
u
cib
le
m
a
n
n
er
[
1
4
]
-
[
1
6
]
.
Fig
u
r
e
1
.
T
h
e
p
r
o
ce
s
s
o
f
th
e
r
e
s
ea
r
ch
m
eth
o
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
43
,
No
.
2
,
Au
g
u
s
t
20
26
:
640
-
6
5
0
642
2
.
1
.
P
re
pro
ce
s
s
ing
Pre
p
r
o
ce
s
s
in
g
is
th
e
in
itial
a
n
d
c
r
u
cial
p
h
ase
in
m
ac
h
in
e
lear
n
in
g
,
g
u
ar
a
n
teein
g
th
at
r
aw
d
ata
is
ef
f
ec
tiv
ely
co
n
v
er
ted
in
to
a
f
o
r
m
at
s
u
itab
le
f
o
r
m
o
d
el
tr
ai
n
in
g
.
Du
r
in
g
t
h
is
s
tep
,
elim
in
atin
g
d
u
p
licates
is
cr
u
cial
to
av
o
i
d
th
e
o
v
er
-
r
ep
r
esen
tatio
n
o
f
an
y
o
n
e
d
ata
p
o
in
t,
wh
ic
h
m
ay
d
is
to
r
t
th
e
f
in
d
i
n
g
s
.
Ad
d
r
ess
in
g
ab
s
en
t
d
ata
is
an
ess
en
tial
a
s
p
ec
t
o
f
p
r
ep
r
o
ce
s
s
in
g
.
A
p
r
ev
alen
t
m
eth
o
d
is
s
in
g
le
im
p
u
tatio
n
,
wh
en
ab
s
en
t
v
alu
es
ar
e
s
u
b
s
titu
ted
u
s
in
g
s
tatis
tical
m
etr
ics
s
u
ch
as
th
e
m
ea
n
,
m
e
d
ian
,
o
r
m
o
d
e,
c
o
n
tin
g
e
n
t
u
p
o
n
th
e
d
ata'
s
ch
ar
ac
ter
is
ti
cs
[
1
7
]
.
T
h
is
ap
p
r
o
a
ch
p
r
es
er
v
es
th
e
d
ataset's
in
teg
r
ity
wh
ile
r
ed
u
cin
g
th
e
in
f
lu
en
ce
o
f
ab
s
en
t d
ata
o
n
m
o
d
el
p
r
ec
is
io
n
.
I
n
th
is
s
tu
d
y
,
p
r
e
p
r
o
ce
s
s
in
g
was
co
n
d
u
cted
in
f
o
u
r
co
n
s
ec
u
tiv
e
s
tep
s
.
First,
d
u
p
licate
o
b
s
er
v
atio
n
s
w
er
e
r
em
o
v
ed
to
a
v
o
id
o
v
er
-
r
ep
r
esen
tin
g
r
ep
ea
ted
cu
s
to
m
er
r
ec
o
r
d
s
.
Seco
n
d
,
m
is
s
in
g
v
alu
es
wer
e
h
a
n
d
le
d
u
s
in
g
s
in
g
le
im
p
u
tatio
n
.
Nu
m
er
ical
v
ar
iab
les
wer
e
im
p
u
t
ed
u
s
in
g
th
e
m
e
d
ian
wh
e
n
t
h
e
d
is
tr
ib
u
tio
n
was
s
k
ewe
d
an
d
th
e
m
ea
n
w
h
en
th
e
d
is
tr
ib
u
tio
n
was
ap
p
r
o
x
i
m
ately
s
y
m
m
etr
ic,
wh
er
ea
s
c
ateg
o
r
ical
v
ar
iab
les
wer
e
im
p
u
ted
u
s
in
g
th
e
m
o
d
e
[
1
8
]
.
T
h
ir
d
,
co
n
s
tan
t f
ea
tu
r
es with
ze
r
o
v
ar
ian
ce
wer
e
r
em
o
v
ed
b
ec
au
s
e
th
e
y
d
o
n
o
t
co
n
t
r
ib
u
te
u
s
ef
u
l
d
is
cr
im
i
n
ato
r
y
i
n
f
o
r
m
atio
n
.
Fo
u
r
th
,
h
i
g
h
ly
co
r
r
elate
d
p
r
ed
icto
r
s
wer
e
s
cr
ee
n
ed
u
s
in
g
a
p
air
wis
e
Pear
s
o
n
co
r
r
elatio
n
th
r
esh
o
ld
o
f
0
.
9
0
,
an
d
o
n
e
v
ar
iab
le
f
r
o
m
ea
c
h
h
ig
h
ly
co
r
r
elate
d
p
air
was
r
em
o
v
ed
to
r
e
d
u
ce
r
e
d
u
n
d
an
c
y
an
d
im
p
r
o
v
e
m
o
d
el
s
tab
ilit
y
[
1
9
]
.
Fu
r
th
er
m
o
r
e
,
th
e
elim
in
atio
n
o
f
co
n
s
tan
t
an
d
s
tr
o
n
g
ly
c
o
r
r
elate
d
f
ea
tu
r
es
is
e
x
ec
u
t
ed
to
d
im
in
is
h
n
o
is
e
in
th
e
d
ataset
an
d
p
r
ev
e
n
t
co
m
p
licatio
n
s
lik
e
m
u
ltico
llin
ea
r
ity
,
wh
ich
ca
n
s
k
ew
m
o
d
el
p
r
ed
ictio
n
s
[
2
0
]
.
Mu
ltico
llin
ea
r
ity
ar
is
es
wh
en
in
d
ep
en
d
en
t
v
a
r
iab
les
ex
h
i
b
it
s
ig
n
if
ican
t
co
r
r
elatio
n
,
h
en
ce
co
m
p
r
o
m
is
in
g
th
e
p
r
ec
is
io
n
o
f
f
ea
tu
r
e
s
i
g
n
if
ica
n
ce
an
d
in
f
latin
g
m
o
d
el
co
ef
f
icien
ts
.
T
h
is
p
h
ase
g
u
ar
an
tees
th
at
th
e
m
o
d
els
co
n
ce
n
tr
ate
o
n
s
ig
n
if
ican
t a
n
d
n
o
n
-
r
e
d
u
n
d
an
t d
ata
item
s
.
Mu
ltico
llin
ea
r
ity
ca
n
b
e
q
u
an
t
itativ
ely
r
ep
r
esen
ted
b
y
th
e
v
a
r
ian
ce
in
f
latio
n
f
ac
to
r
(
VI
F):
=
1
1
−
2
wh
er
e
2
is
th
e
co
ef
f
icien
t
o
f
d
e
ter
m
in
atio
n
o
f
th
e
r
e
g
r
ess
io
n
m
o
d
el
th
at
p
r
ed
icts
th
e
ℎ
f
ea
tu
r
e
u
s
in
g
all
th
e
o
th
er
f
ea
tu
r
es.
A
h
ig
h
e
r
in
d
icate
s
a
h
ig
h
co
r
r
elatio
n
,
s
u
g
g
esti
n
g
th
e
f
ea
tu
r
e
s
h
o
u
ld
b
e
r
e
m
o
v
ed
[
2
1
]
.
Af
ter
th
e
in
itial
co
r
r
elatio
n
s
cr
ee
n
in
g
,
m
u
ltico
llin
ea
r
ity
w
as
f
u
r
th
er
ass
ess
ed
u
s
in
g
th
e
Var
ian
ce
I
n
f
latio
n
Facto
r
.
Var
iab
les
w
ith
a
VI
F
g
r
ea
ter
th
an
1
0
w
er
e
ex
clu
d
ed
f
r
o
m
th
e
f
in
al
f
ea
tu
r
e
s
et,
as
th
is
th
r
esh
o
ld
is
co
m
m
o
n
ly
u
s
ed
t
o
in
d
icate
s
er
io
u
s
m
u
ltico
llin
ea
r
ity
.
T
h
is
two
-
s
tag
e
p
r
o
ce
d
u
r
e
h
elp
ed
en
s
u
r
e
th
at
th
e
r
etain
ed
v
a
r
iab
les we
r
e
b
o
t
h
in
f
o
r
m
ativ
e
an
d
s
u
f
f
icien
tly
in
d
ep
en
d
en
t f
o
r
r
o
b
u
s
t m
o
d
el
tr
ain
in
g
.
2
.
2
.
Da
t
a
prepa
ra
t
i
o
n
T
h
e
d
ata
p
r
e
p
ar
atio
n
p
h
ase
e
n
tails
s
ca
lin
g
an
d
n
o
r
m
alis
in
g
f
ea
tu
r
es
to
g
u
ar
a
n
tee
th
at
a
ll
v
ar
iab
les
co
n
tr
ib
u
te
u
n
if
o
r
m
ly
t
o
th
e
m
o
d
el
[
2
0
]
.
I
n
t
h
e
ab
s
en
ce
o
f
s
ca
lin
g
,
c
h
ar
ac
ter
is
tics
with
b
r
o
a
d
er
r
an
g
es
(
e.
g
.
,
i
n
co
m
e)
m
ay
o
v
e
r
s
h
a
d
o
w
th
o
s
e
with
n
ar
r
o
wer
r
a
n
g
es
(
e.
g
.
,
a
g
e)
,
r
esu
ltin
g
in
s
k
ewe
d
m
o
d
el
p
er
f
o
r
m
an
ce
.
Min
-
Ma
x
s
ca
lin
g
is
a
p
r
ev
alen
t
m
eth
o
d
in
wh
i
ch
ea
ch
f
ea
tu
r
e
is
n
o
r
m
alis
ed
to
a
r
an
g
e
o
f
[
0
,
1
]
o
r
[
-
1
,
1
]
.
T
h
is
is
d
en
o
te
d
b
y
th
e
eq
u
atio
n
:
′
=
−
−
wh
er
e
r
ep
r
esen
ts
th
e
o
r
ig
in
al
v
alu
e,
an
d
an
d
d
en
o
te
th
e
m
in
im
u
m
an
d
m
a
x
im
u
m
v
al
u
es
o
f
th
e
f
ea
tu
r
e,
r
esp
ec
tiv
ely
[
2
2
]
.
C
ateg
o
r
ical
d
ata
m
u
s
t
b
e
c
o
n
v
er
ted
u
s
in
g
o
n
e
-
h
o
t
en
co
d
in
g
o
r
lab
el
en
co
d
in
g
,
b
ased
o
n
th
e
ch
ar
ac
ter
is
tics
o
f
th
e
d
ata.
On
e
-
ho
t
en
c
o
d
in
g
tr
an
s
f
o
r
m
s
ca
teg
o
r
ical
v
ar
ia
b
les
in
to
b
in
a
r
y
v
ec
to
r
s
,
with
ea
c
h
ca
teg
o
r
y
r
ep
r
esen
te
d
as
a
d
is
tin
ct
co
lu
m
n
with
v
alu
es
o
f
0
o
r
1
,
h
en
ce
f
ac
ilit
atin
g
t
h
e
ef
f
e
ctiv
e
u
tili
s
atio
n
o
f
ca
teg
o
r
ical
d
ata
b
y
m
ac
h
i
n
e
le
ar
n
in
g
m
o
d
els.
I
n
th
is
s
tu
d
y
,
n
u
m
er
ical
f
ea
tu
r
es
wer
e
s
ca
led
to
th
e
r
an
g
e
[
0
,
1
]
u
s
in
g
Min
-
Ma
x
n
o
r
m
alis
atio
n
.
No
m
in
al
ca
teg
o
r
ical
v
ar
ia
b
les
wer
e
tr
an
s
f
o
r
m
ed
u
s
in
g
o
n
e
-
h
o
t
e
n
co
d
i
n
g
,
wh
ile
o
r
d
in
al
v
ar
iab
les,
wh
en
p
r
esen
t,
wer
e
en
co
d
ed
u
s
in
g
lab
el
en
c
o
d
in
g
.
T
o
av
o
id
d
ata
le
a
k
ag
e,
th
e
s
ca
lin
g
an
d
e
n
co
d
i
n
g
tr
an
s
f
o
r
m
er
s
wer
e
f
itted
o
n
ly
o
n
th
e
tr
ain
in
g
d
ata
a
n
d
th
e
n
ap
p
lied
to
th
e
v
alid
atio
n
a
n
d
test
d
ata
u
s
in
g
th
e
f
itted
p
ar
am
ete
r
s
.
Mitig
atin
g
class
im
b
alan
ce
is
ess
en
tial
at
th
is
p
h
ase,
p
ar
ticu
lar
ly
in
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
e
d
ictio
n
,
b
ec
au
s
e
th
e
m
ajo
r
it
y
o
f
p
er
s
o
n
s
ar
e
u
n
lik
ely
to
f
ail,
r
en
d
e
r
in
g
th
e
m
i
n
o
r
ity
class
(
th
o
s
e
w
h
o
d
o
d
ef
a
u
lt)
m
o
r
e
ch
allen
g
in
g
to
i
d
en
tify
.
B
o
r
d
e
r
lin
e
s
y
n
th
etic
m
in
o
r
ity
o
v
e
r
-
s
am
p
lin
g
tec
h
n
iq
u
e
(
SMOT
E
)
g
en
er
ates
s
y
n
th
etic
d
ata
p
o
in
ts
f
o
r
th
e
m
in
o
r
ity
cl
ass
,
co
n
ce
n
tr
atin
g
o
n
s
am
p
les
s
itu
ated
ar
o
u
n
d
th
e
d
ec
is
io
n
b
o
r
d
er
as
s
h
o
wn
i
n
Fig
u
r
e
2
.
T
h
e
m
o
d
el
is
m
o
r
e
p
r
o
n
e
t
o
m
is
class
if
y
in
g
th
ese
b
o
r
d
er
li
n
e
s
itu
atio
n
s
;
h
en
ce
,
o
v
er
s
am
p
lin
g
th
e
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
A
d
va
n
ce
d
p
ers
o
n
a
l b
a
n
kru
p
tc
y
p
r
ed
ictio
n
u
s
in
g
tr
ee
-
b
a
s
ed
d
ee
p
lea
r
n
in
g
mo
d
els
(
N
h
a
t Ng
u
ye
n
Min
h
)
643
en
h
an
ce
s
th
e
m
o
d
el'
s
ca
p
ac
ity
to
d
if
f
er
en
tiate
b
etwe
en
cla
s
s
es
[
2
3
]
.
T
h
e
s
af
e
-
lev
el
v
ar
iatio
n
o
f
B
o
r
d
er
lin
e
SMOT
E
en
h
an
ce
s
th
is
m
eth
o
d
b
y
cr
ea
tin
g
s
y
n
t
h
etic
p
o
in
t
s
in
ar
ea
s
with
a
g
r
ea
ter
p
r
o
b
ab
ilit
y
o
f
ac
cu
r
ate
class
if
icatio
n
,
h
en
ce
r
ed
u
cin
g
th
e
d
an
g
er
o
f
n
o
is
e
in
tr
o
d
u
ct
io
n
.
T
h
is
ap
p
r
o
ac
h
en
h
a
n
ce
s
d
ataset
b
alan
ce
an
d
im
p
r
o
v
es th
e
m
o
d
el'
s
s
en
s
itiv
i
ty
to
p
r
e
d
ictio
n
s
o
f
th
e
m
in
o
r
ity
class
[
2
4
]
.
B
ec
au
s
e
th
e
tar
g
et
v
ar
ia
b
le
i
n
th
is
s
tu
d
y
c
o
n
s
is
ts
o
f
f
o
u
r
p
r
o
g
r
ess
iv
e
class
es
o
f
f
in
a
n
ci
al
d
is
tr
ess
,
class
im
b
alan
ce
wa
s
ad
d
r
ess
e
d
u
s
in
g
B
o
r
d
er
lin
e
-
SMOT
E
ap
p
lied
o
n
ly
to
th
e
tr
ain
in
g
d
at
a.
T
h
e
o
v
er
s
am
p
lin
g
p
r
o
ce
d
u
r
e
was
co
n
f
ig
u
r
ed
with
5
n
ea
r
est
n
eig
h
b
o
u
r
s
f
o
r
s
y
n
th
etic
in
s
tan
ce
g
en
er
atio
n
an
d
1
0
n
ea
r
est
n
eig
h
b
o
u
r
s
f
o
r
id
en
tif
y
in
g
b
o
r
d
er
lin
e
s
am
p
les.
E
ac
h
m
in
o
r
ity
class
in
t
h
e
tr
ai
n
in
g
f
o
ld
wa
s
o
v
er
s
am
p
led
u
n
til
it
m
atch
ed
t
h
e
s
ize
o
f
th
e
m
ajo
r
ity
class
in
th
at
f
o
l
d
.
A
f
ix
ed
r
an
d
o
m
s
ee
d
o
f
4
2
was
u
s
ed
to
en
s
u
r
e
r
ep
r
o
d
u
cib
ilit
y
.
T
h
is
s
eq
u
en
c
in
g
was
im
p
o
r
tan
t
b
ec
a
u
s
e
r
esam
p
lin
g
th
e
f
u
ll
d
ataset
b
e
f
o
r
e
s
p
litt
in
g
c
o
u
ld
in
tr
o
d
u
ce
i
n
f
o
r
m
atio
n
leak
a
g
e
an
d
o
v
er
s
tate
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
Fig
u
r
e
2
.
Sp
ec
if
ic
g
r
o
u
p
s
o
f
i
m
b
alan
ce
d
d
ata
in
th
e
B
o
r
d
er
l
in
e
-
SMOT
E
2
.
3
.
Sp
litt
ing
t
he
da
t
a
s
et
Du
r
in
g
th
is
s
tep
,
th
e
d
ataset
i
s
d
iv
id
ed
in
to
tr
ain
in
g
an
d
test
in
g
s
u
b
s
ets,
co
m
m
o
n
ly
ac
co
r
d
in
g
to
an
8
0
/2
0
r
atio
.
T
h
e
tr
ain
in
g
s
et
is
u
tili
s
ed
to
ca
lib
r
ate
th
e
m
o
d
el,
wh
ils
t
th
e
test
in
g
s
et
a
s
s
es
s
es
th
e
m
o
d
el'
s
ca
p
ac
ity
to
g
en
e
r
alis
e
to
n
o
v
e
l
d
ata.
T
h
is
m
e
th
o
d
g
u
ar
a
n
tees
th
at
th
e
m
o
d
el
is
n
o
t
ex
ce
s
s
iv
ely
tailo
r
e
d
to
th
e
tr
ain
in
g
d
ata
a
n
d
ca
n
ex
ce
l w
it
h
f
r
esh
,
r
ea
l
-
wo
r
ld
d
ata.
I
n
th
is
s
tu
d
y
,
t
h
e
clea
n
ed
d
at
aset
was
d
iv
id
ed
in
to
tr
ain
in
g
an
d
test
in
g
s
u
b
s
ets
u
s
in
g
a
s
tr
atif
ied
8
0
/2
0
s
p
lit,
with
th
e
s
am
e
cla
s
s
p
r
o
p
o
r
tio
n
s
p
r
eser
v
ed
ac
r
o
s
s
b
o
th
s
u
b
s
ets.
T
h
e
r
an
d
o
m
s
e
ed
was
s
et
to
4
2
to
en
s
u
r
e
th
at
th
e
p
ar
titi
o
n
c
o
u
ld
b
e
r
ep
r
o
d
u
ce
d
e
x
ac
tly
[
2
4
]
.
T
h
e
tr
ain
in
g
s
u
b
s
et
was
u
s
ed
f
o
r
m
o
d
el
f
itti
n
g
an
d
h
y
p
er
p
ar
am
eter
tu
n
in
g
,
wh
er
e
as th
e
test
s
u
b
s
et
was r
eser
v
ed
s
tr
ictly
f
o
r
f
in
al
o
u
t
-
of
-
s
am
p
l
e
ev
alu
atio
n
.
C
r
o
s
s
-
v
alid
atio
n
is
im
p
lem
en
ted
to
au
g
m
en
t
m
o
d
el
r
esil
ien
ce
.
I
n
k
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
,
th
e
d
ataset
is
p
ar
titi
o
n
ed
in
to
k
e
q
u
ally
s
ized
s
eg
m
en
ts
.
T
h
e
m
o
d
el
is
tr
ain
ed
u
s
in
g
k
-
1
f
o
ld
s
a
n
d
v
er
if
ie
d
o
n
t
h
e
r
em
ain
in
g
f
o
l
d
.
T
h
is
p
r
o
ce
d
u
r
e
is
ex
ec
u
ted
k
tim
es,
an
d
th
e
o
u
tco
m
es
ar
e
av
er
ag
e
d
to
g
et
a
m
o
r
e
d
ep
en
d
ab
le
ass
es
s
m
en
t
o
f
t
h
e
m
o
d
el'
s
ef
f
icac
y
.
T
h
is
s
tr
ateg
y
is
cr
u
cial
f
o
r
allev
iatin
g
o
v
er
f
itti
n
g
an
d
g
u
ar
an
tees
co
n
s
is
ten
t m
o
d
el
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
v
ar
io
u
s
d
ata
s
u
b
s
et
s.
W
ith
in
th
e
tr
ain
in
g
s
u
b
s
et,
s
t
r
atif
ied
5
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
was
u
s
ed
d
u
r
in
g
m
o
d
el
s
elec
tio
n
an
d
h
y
p
er
p
ar
am
eter
o
p
tim
is
atio
n
.
I
n
ea
ch
iter
atio
n
,
f
o
u
r
f
o
ld
s
wer
e
u
s
ed
f
o
r
tr
ain
in
g
an
d
o
n
e
f
o
ld
was
u
s
ed
f
o
r
v
alid
atio
n
,
a
n
d
t
h
e
av
er
ag
e
p
e
r
f
o
r
m
a
n
ce
ac
r
o
s
s
th
e
f
iv
e
f
o
ld
s
was
u
s
ed
to
ass
ess
a
ca
n
d
id
ate
h
y
p
er
p
ar
am
eter
co
n
f
ig
u
r
atio
n
.
Stra
tific
atio
n
w
as
m
ain
tain
ed
t
h
r
o
u
g
h
o
u
t
th
e
cr
o
s
s
-
v
alid
atio
n
p
r
o
ce
s
s
in
o
r
d
er
to
p
r
eser
v
e
th
e
class
d
is
tr
ib
u
tio
n
o
f
th
e
m
u
lti
-
class
b
an
k
r
u
p
tcy
v
ar
iab
le.
2
.
4
.
M
o
del t
ra
ini
ng
a
nd
ev
a
l
ua
t
io
n
Fo
u
r
ad
v
a
n
ce
d
m
o
d
els
wer
e
s
elec
ted
f
o
r
co
m
p
a
r
is
o
n
:
T
B
C
N
N,
Neu
r
al
Dec
is
io
n
T
r
ee
,
NODE
,
an
d
Dee
p
Fo
r
est.
T
h
ese
m
o
d
els
wer
e
ch
o
s
en
b
ec
au
s
e
th
e
y
r
ep
r
esen
t
two
r
elev
a
n
t
m
eth
o
d
o
lo
g
ical
f
am
ilies
d
is
cu
s
s
ed
in
th
e
I
n
tr
o
d
u
ctio
n
:
d
ee
p
n
e
u
r
al
m
o
d
els
d
esig
n
ed
f
o
r
tab
u
la
r
lear
n
in
g
a
n
d
tr
ee
-
b
ased
d
ee
p
ar
ch
itectu
r
es
th
at
m
ay
p
r
o
v
id
e
a
b
etter
b
alan
ce
b
etwe
en
p
r
ed
ictiv
e
p
er
f
o
r
m
a
n
ce
an
d
in
ter
p
r
etab
ilit
y
.
All
m
o
d
els
wer
e
tr
ain
ed
o
n
t
h
e
s
am
e
p
r
ep
r
o
ce
s
s
ed
d
ata
an
d
ev
alu
ated
u
n
d
er
th
e
s
am
e
p
r
o
to
co
l
s
o
th
at
th
eir
r
el
ativ
e
p
er
f
o
r
m
an
ce
c
o
u
ld
b
e
co
m
p
ar
ed
f
air
ly
.
2
.
4
.
1
.
T
a
bu
la
r
co
nv
o
lutio
na
l
neura
l net
wo
rk
(
T
B
CNN
)
T
h
is
m
o
d
el
em
p
lo
y
s
co
n
v
o
l
u
tio
n
al
lay
er
s
o
n
tab
u
lar
d
ata,
u
tili
s
in
g
co
n
v
o
lu
tio
n
p
r
o
ce
s
s
es
n
o
r
m
ally
ap
p
lied
to
p
ictu
r
e
d
ata.
T
h
e
co
n
v
o
lu
tio
n
f
u
n
ctio
n
in
T
B
C
NN
is
d
en
o
ted
as:
ℎ
=
(
∑
+
,
+
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
43
,
No
.
2
,
Au
g
u
s
t
20
26
:
640
-
6
5
0
644
wh
er
e
ℎ
is
th
e
o
u
p
u
t
f
ea
tu
r
e
m
ap
,
r
ep
r
esen
ts
th
e
c
o
n
v
o
l
u
tio
n
al
weig
h
ts
,
an
d
is
th
e
ac
tiv
atio
n
f
u
n
ctio
n
[
2
5
]
.
I
n
th
e
p
r
esen
t
s
tu
d
y
,
T
B
C
NN
was
im
p
lem
en
ted
as
a
n
eu
r
al
b
en
ch
m
ar
k
f
o
r
m
o
d
ellin
g
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
in
s
tr
u
ctu
r
ed
b
o
r
r
o
wer
-
lev
el
d
ata.
T
h
e
T
B
C
N
N
h
y
p
er
p
a
r
am
eter
s
tu
n
ed
b
y
Op
tu
n
a
in
clu
d
e
d
th
e
lear
n
in
g
r
ate
(
10
−
4
to
10
−
2
)
,
b
atch
s
ize
{3
2
,
6
4
,
1
2
8
}
,
n
u
m
b
er
o
f
h
id
d
en
u
n
its
{6
4
,
1
2
8
,
2
5
6
},
d
r
o
p
o
u
t
r
ate
(
0
.
10
to
0
.
50
)
,
an
d
n
u
m
b
er
o
f
e
p
o
ch
s
u
p
t
o
5
0
.
T
h
e
A
d
am
o
p
tim
is
er
a
n
d
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
wer
e
u
s
ed
,
an
d
ea
r
ly
s
to
p
p
i
n
g
with
a
p
atien
ce
o
f
1
0
e
p
o
ch
s
was a
p
p
lied
to
r
e
d
u
ce
o
v
er
f
itti
n
g
.
2
.
4
.
2
.
Neura
l
decisi
o
n t
r
ee
T
h
is
ap
p
r
o
ac
h
in
teg
r
ates
th
e
ar
ch
itectu
r
e
o
f
d
ec
is
io
n
tr
ee
s
with
th
e
ad
ap
tab
ilit
y
o
f
n
eu
r
a
l
n
etwo
r
k
s
.
I
t su
b
s
titu
tes r
ig
id
d
iv
id
es with
f
lex
ib
le,
p
r
o
b
ab
ilis
tic
d
iv
is
io
n
s
.
T
h
e
s
o
f
t d
ec
is
io
n
f
u
n
ctio
n
is
r
ep
r
esen
ted
as:
(
)
=
1
1
+
−
w
h
er
e
(
)
is
th
e
p
r
o
b
a
b
ilit
y
o
f
s
e
lectin
g
a
b
r
an
c
h
,
a
n
d
−
is
th
e
d
o
t
p
r
o
d
u
ct
o
f
th
e
weig
h
t
v
ec
t
o
r
an
d
th
e
in
p
u
t f
ea
tu
r
es.
Neu
r
al
d
ec
is
io
n
tr
ee
was
in
clu
d
ed
b
ec
au
s
e
it
co
m
b
i
n
es
h
ier
a
r
ch
ical
d
ec
is
io
n
lo
g
ic
with
d
if
f
er
en
tiab
le
lear
n
in
g
.
T
h
e
tu
n
e
d
h
y
p
er
p
ar
am
eter
s
f
o
r
th
is
m
o
d
el
in
cl
u
d
ed
tr
ee
d
e
p
th
{
3
,
4
,
5
,
6
,
7
}
,
h
i
d
d
en
d
im
en
s
io
n
{3
2
,
6
4
,
1
2
8
}
,
lear
n
i
n
g
r
ate
(
10
−
4
to
10
−
2
)
,
b
atch
s
ize
{6
4
,
1
2
8
,
2
5
6
}
,
an
d
d
r
o
p
o
u
t
r
ate
(
0
.
10
to
0
.
50
)
.
T
h
e
m
o
d
el
was
tr
ain
ed
f
o
r
a
m
ax
im
u
m
o
f
5
0
ep
o
c
h
s
,
with
ea
r
ly
s
to
p
p
in
g
af
ter
1
0
ep
o
ch
s
with
o
u
t
v
alid
atio
n
im
p
r
o
v
em
e
n
t.
2
.
4
.
3
.
Neura
l
o
bli
v
io
us
decisi
o
n e
ns
em
bles
(
NO
DE
)
T
h
is
m
o
d
el
em
p
lo
y
s
an
en
s
e
m
b
le
o
f
d
ec
is
io
n
tr
ee
s
in
wh
i
ch
all
n
o
d
es
at
th
e
s
am
e
lev
el
u
tili
s
e
th
e
id
en
tical
ch
ar
ac
ter
is
tic
f
o
r
p
ar
t
itio
n
in
g
.
T
h
e
u
ltima
te
f
o
r
ec
ast is
ca
lcu
la
ted
as f
o
llo
ws:
=
∑
(
)
=
1
w
h
e
r
e
is
t
h
e
f
i
n
al
o
u
t
p
u
t
,
is
t
h
e
w
ei
g
h
t
f
o
r
e
ac
h
tr
ee
,
a
n
d
(
)
is
t
h
e
p
r
ed
ict
io
n
o
f
t
h
e
-
th
t
r
e
e
[
2
6
]
.
NODE
was
s
elec
ted
b
ec
au
s
e
it
is
s
p
ec
if
ically
d
esig
n
ed
f
o
r
t
ab
u
lar
d
ata
an
d
h
as
b
ee
n
r
ep
o
r
ted
as
an
ef
f
ec
tiv
e
d
ee
p
lear
n
in
g
ap
p
r
o
ac
h
f
o
r
s
tr
u
ctu
r
ed
p
r
e
d
ictio
n
task
s
.
T
h
e
tu
n
ed
h
y
p
er
p
a
r
a
m
eter
s
in
clu
d
ed
th
e
n
u
m
b
er
o
f
en
s
em
b
le
la
y
er
s
{
2
,
3
,
4
}
,
n
u
m
b
er
o
f
tr
ee
s
{
1
2
8
,
2
5
6
,
5
1
2
}
,
tr
ee
d
ep
th
{4
,
6
,
8
}
,
lear
n
in
g
r
ate
(
10
−
4
to
10
−
2
)
,
b
atch
s
ize
{3
2
,
6
4
,
1
2
8
}
,
a
n
d
d
r
o
p
o
u
t
r
ate
(
0
.
10
to
0
.
40
)
.
As
with
th
e
o
th
er
n
e
u
r
al
m
o
d
els,
tr
ain
in
g
was lim
ited
to
5
0
ep
o
ch
s
with
ea
r
ly
s
to
p
p
in
g
p
atien
ce
s
et
to
1
0
ep
o
ch
s
.
2
.
4
.
4
.
Dee
p
f
o
re
s
t
(
G
CF
o
re
s
t
)
T
h
is
m
o
d
el
co
n
s
tr
u
cts
a
n
e
n
s
em
b
le
o
f
d
ec
is
io
n
tr
ee
s
an
d
im
p
lem
en
ts
th
em
in
a
ca
s
ca
d
in
g
ar
ch
itectu
r
e.
E
v
e
r
y
lay
e
r
en
h
an
ce
s
th
e
p
r
ed
ictio
n
s
o
f
th
e
p
r
ec
ed
in
g
lay
er
,
a
u
g
m
en
ti
n
g
ac
cu
r
ac
y
with
ea
ch
iter
atio
n
.
T
h
e
r
esu
lt
o
f
th
e
ca
s
ca
d
e
lay
er
is
:
̂
(
)
=
1
∑
ℎ
(
)
(
)
=
1
w
h
er
e
̂
(
)
is
th
e
o
u
tp
u
t f
r
o
m
lay
e
r
ll
,
an
d
ℎ
(
)
(
)
is
th
e
p
r
e
d
ictio
n
f
r
o
m
tr
ee
in
lay
er
[
2
6
]
.
Dee
p
Fo
r
est
was
in
clu
d
ed
as
th
e
p
r
in
cip
al
tr
ee
-
b
ased
en
s
e
m
b
le
b
en
c
h
m
ar
k
b
ec
a
u
s
e
it
c
an
m
o
d
el
co
m
p
lex
n
o
n
-
lin
ea
r
s
tr
u
ctu
r
es
th
r
o
u
g
h
lay
er
ed
tr
ee
en
s
em
b
l
es
wh
ile
r
etain
in
g
a
c
o
m
p
ar
at
iv
ely
in
ter
p
r
etab
le
d
ec
is
io
n
p
r
o
ce
s
s
.
T
h
e
tu
n
ed
h
y
p
e
r
p
ar
am
ete
r
s
f
o
r
Dee
p
Fo
r
est
in
clu
d
ed
th
e
n
u
m
b
e
r
o
f
tr
ee
s
p
e
r
f
o
r
est
{1
0
0
,
2
0
0
,
3
0
0
},
m
a
x
im
u
m
tr
ee
d
ep
th
{
5
,
1
0
,
1
5
}
,
m
in
im
u
m
s
am
p
les
p
er
leaf
{
1
,
2
,
5
}
,
an
d
th
e
n
u
m
b
er
o
f
ca
s
ca
d
e
lay
er
s
{2
,
3
,
4
}
.
T
h
e
f
in
al
p
r
ed
ictio
n
was
o
b
tain
ed
f
r
o
m
th
e
b
est
-
p
er
f
o
r
m
in
g
ca
s
ca
d
e
co
n
f
i
g
u
r
atio
n
s
elec
ted
o
n
th
e
tr
ain
in
g
f
o
l
d
s
.
Op
tu
n
a
is
u
tili
s
ed
to
o
p
tim
is
e
th
e
h
y
p
e
r
p
ar
am
ete
r
s
o
f
th
ese
m
o
d
els,
em
p
lo
y
in
g
th
e
tr
ee
-
s
tr
u
ctu
r
ed
p
ar
ze
n
esti
m
ato
r
(
T
PE)
to
ef
f
ec
tiv
ely
n
av
ig
ate
t
h
e
h
y
p
er
p
ar
am
eter
s
p
ac
e.
T
h
e
o
b
je
ctiv
e
f
u
n
ctio
n
f
o
r
h
y
p
er
p
ar
am
eter
o
p
tim
is
atio
n
m
ay
b
e
r
e
p
r
esen
ted
as:
(
)
=
1
−
(
)
w
h
er
e
d
en
o
tes
th
e
h
y
p
er
p
ar
a
m
eter
co
n
f
ig
u
r
atio
n
.
Op
tu
n
a'
s
d
y
n
am
ic
s
am
p
lin
g
a
n
d
p
r
u
n
i
n
g
f
ea
tu
r
es
e
n
ab
le
th
e
f
r
am
ewo
r
k
to
r
ap
id
l
y
co
n
v
er
g
e
o
n
o
p
tim
al
h
y
p
er
p
ar
a
m
eter
s
,
en
h
an
ci
n
g
m
o
d
el
p
e
r
f
o
r
m
an
ce
with
o
u
t
n
ec
ess
itatin
g
lab
o
r
io
u
s
s
ea
r
ch
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
A
d
va
n
ce
d
p
ers
o
n
a
l b
a
n
kru
p
tc
y
p
r
ed
ictio
n
u
s
in
g
tr
ee
-
b
a
s
ed
d
ee
p
lea
r
n
in
g
mo
d
els
(
N
h
a
t Ng
u
ye
n
Min
h
)
645
I
n
th
is
s
tu
d
y
,
Op
tu
n
a
was
r
u
n
f
o
r
1
0
0
tr
ials
p
er
m
o
d
el
u
s
in
g
th
e
T
PE
s
am
p
ler
with
a
f
ix
ed
r
an
d
o
m
s
ee
d
o
f
4
2
.
T
h
e
o
p
tim
is
atio
n
o
b
jectiv
e
was
th
e
m
ea
n
cr
o
s
s
-
v
alid
ated
ac
cu
r
ac
y
ac
r
o
s
s
th
e
f
iv
e
tr
ain
in
g
f
o
ld
s
,
wh
ich
is
eq
u
iv
alen
t
to
m
in
i
m
is
in
g
1
−
Acc
u
r
ac
y
(
)
.
A
m
ed
ian
p
r
u
n
in
g
s
tr
ateg
y
w
as
u
s
ed
to
ter
m
in
ate
u
n
p
r
o
m
is
in
g
tr
ials
ea
r
ly
an
d
r
ed
u
ce
co
m
p
u
tatio
n
al
c
o
s
t.
T
h
is
o
p
tim
is
atio
n
p
r
o
to
c
o
l
was
ad
o
p
ted
t
o
en
s
u
r
e
th
at
ea
ch
m
o
d
el
was e
v
alu
ated
u
n
d
er
a
co
m
p
ar
ab
le
a
n
d
s
y
s
tem
atic
tu
n
in
g
p
r
o
ce
d
u
r
e.
2
.
5
.
P
er
f
o
r
m
a
nce
ev
a
lua
t
io
n a
nd
m
o
del c
o
m
pa
ris
on
Af
ter
h
y
p
e
r
p
ar
am
ete
r
o
p
tim
is
atio
n
,
th
e
b
est
v
er
s
io
n
o
f
ea
ch
m
o
d
el
was
r
etr
ain
e
d
o
n
th
e
f
u
ll
tr
ain
in
g
s
et
an
d
ev
alu
ated
o
n
th
e
h
o
ld
-
o
u
t
test
s
et.
B
ec
au
s
e
th
e
s
tu
d
y
aim
s
to
co
m
p
ar
e
th
e
p
r
ed
ic
tiv
e
ab
ilit
y
o
f
f
o
u
r
alter
n
ativ
e
m
o
d
els,
th
e
f
in
al
an
aly
s
is
f
o
cu
s
ed
o
n
m
o
d
el
-
by
-
m
o
d
el
o
u
t
-
of
-
s
am
p
le
p
er
f
o
r
m
an
ce
r
at
h
er
t
h
an
a
weig
h
ted
en
s
em
b
le.
Per
f
o
r
m
a
n
ce
was
ass
e
s
s
ed
u
s
in
g
Acc
u
r
ac
y
,
Pre
cisi
o
n
,
Sen
s
itiv
ity
,
Sp
ec
if
icity
,
F1
-
s
co
r
e,
an
d
AUC;
f
o
r
th
e
m
u
lti
-
class
s
ettin
g
,
th
ese
m
etr
ics
wer
e
co
m
p
u
ted
u
s
in
g
m
a
cr
o
-
a
v
er
ag
i
n
g
,
wh
ile
AUC
was
esti
m
ated
th
r
o
u
g
h
a
o
n
e
-
vs
-
r
e
s
t
ap
p
r
o
ac
h
ac
r
o
s
s
th
e
f
o
u
r
b
an
k
r
u
p
tcy
class
es.
C
o
n
f
u
s
io
n
m
atr
ices
an
d
class
-
wis
e
R
OC
cu
r
v
es
wer
e
also
e
x
am
in
ed
to
p
r
o
v
id
e
a
m
o
r
e
d
etailed
ev
alu
atio
n
o
f
h
o
w
ef
f
e
ctiv
ely
ea
ch
m
o
d
e
l
d
is
tin
g
u
is
h
ed
am
o
n
g
th
e
f
o
u
r
lev
els
o
f
f
i
n
an
cial
d
is
tr
ess
.
T
h
is
ev
al
u
atio
n
d
esig
n
en
s
u
r
es
a
co
n
s
is
ten
t
co
m
p
ar
is
o
n
b
etwe
en
tr
ee
-
b
a
s
ed
d
ee
p
m
o
d
els
a
n
d
d
ee
p
n
eu
r
al
m
o
d
els
u
n
d
er
th
e
s
am
e
p
r
e
p
r
o
ce
s
s
in
g
,
r
esam
p
lin
g
,
an
d
tu
n
i
n
g
co
n
d
iti
o
n
s
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Da
t
a
re
s
ea
rch
T
h
e
d
ataset
f
o
r
th
is
s
tu
d
y
o
n
f
o
r
ec
asti
n
g
p
er
s
o
n
al
d
ef
a
u
lts
co
m
p
r
is
es
in
f
o
r
m
atio
n
f
r
o
m
9
,
8
0
0
clien
ts
o
f
s
ev
er
al
co
m
m
er
cial
b
a
n
k
s
an
d
c
r
ed
it
in
s
titu
tio
n
s
i
n
Viet
n
am
,
c
o
v
er
in
g
t
h
e
p
er
io
d
f
r
o
m
2
0
1
2
to
2
0
2
2
.
All
clien
t
d
ata
was
en
c
r
y
p
ted
d
u
r
in
g
c
o
llectio
n
in
co
m
p
lian
ce
with
r
ig
o
r
o
u
s
d
ata
p
r
o
tectio
n
r
eg
u
latio
n
s
.
T
h
e
tar
g
et
v
ar
iab
le,
B
an
k
r
u
p
tc
y
,
is
class
if
ied
in
to
f
o
u
r
p
r
o
g
r
ess
iv
e
tier
s
o
f
f
in
an
cial
d
is
tr
ess
:
−
C
las
s
0
:
Den
o
tes
d
eb
ts
th
at
a
r
e
eith
er
in
g
o
o
d
s
tan
d
in
g
o
r
o
v
er
d
u
e
b
y
f
ewe
r
t
h
an
1
0
d
ay
s
,
h
o
wev
e
r
d
ee
m
ed
r
ec
o
v
er
a
b
le.
−
C
las
s
1
:
Deb
ts
th
at
ar
e
o
v
er
d
u
e
f
o
r
a
p
er
io
d
r
a
n
g
in
g
f
r
o
m
1
0
to
1
8
0
d
a
y
s
o
r
h
av
e
b
ee
n
r
estru
ctu
r
ed
f
o
r
th
e
in
itial
in
s
tan
ce
,
s
ig
n
if
y
in
g
p
r
elim
in
ar
y
in
d
icato
r
s
o
f
f
in
an
cial
d
is
tr
ess
.
T
h
is
g
r
o
u
p
en
co
m
p
ass
es
b
o
r
r
o
we
r
s
wh
o
m
ay
en
c
o
u
n
te
r
ch
allen
g
es
y
et
ar
e
s
till
ex
p
e
cted
to
r
ec
u
p
e
r
ate,
n
ec
ess
itati
n
g
in
c
r
ea
s
ed
o
v
er
s
ig
h
t b
y
len
d
e
r
s
.
−
C
las
s
2
:
Per
tain
s
to
lo
an
s
th
at
ar
e
o
v
er
d
u
e
f
o
r
a
d
u
r
atio
n
o
f
1
8
1
to
3
6
0
d
ay
s
o
r
th
o
s
e
th
at
h
av
e
u
n
d
er
g
o
n
e
r
estru
ctu
r
in
g
f
o
r
th
e
s
ec
o
n
d
tim
e.
T
h
is
C
la
s
s
in
d
icate
s
s
ig
n
if
ican
t
f
in
an
cial
d
if
f
icu
lty
,
ch
ar
ac
ter
is
ed
b
y
u
n
ce
r
tain
r
ec
o
v
er
y
an
d
b
o
r
r
o
wer
s
'
d
if
f
icu
lties
in
f
u
lf
illi
n
g
t
h
eir
f
in
an
cial
co
m
m
itm
en
ts
.
A
m
o
r
e
s
tr
in
g
en
t
r
is
k
m
an
ag
e
m
en
t stra
teg
y
is
r
eq
u
ir
e
d
at
th
i
s
ju
n
ctu
r
e.
−
C
las
s
3
:
Den
o
tes
d
eb
ts
th
at
ar
e
o
v
er
d
u
e
f
o
r
o
v
e
r
3
6
0
d
a
y
s
,
ca
teg
o
r
is
ed
as
p
r
o
b
ab
ly
i
r
r
ec
o
v
er
a
b
le.
B
o
r
r
o
wer
s
in
th
is
ca
teg
o
r
y
h
av
e
co
n
s
id
er
ab
le
f
in
a
n
cial
in
s
tab
ilit
y
,
r
en
d
er
in
g
r
ec
o
v
er
y
ex
ce
ed
in
g
ly
u
n
lik
ely
,
th
e
r
ef
o
r
e
r
eq
u
ir
in
g
s
u
b
s
tan
tial a
s
s
is
tan
ce
o
r
wr
ite
-
o
f
f
s
b
y
le
n
d
er
s
.
T
h
is
r
ev
is
ed
ca
teg
o
r
is
atio
n
p
r
o
v
id
es
a
m
o
r
e
d
etailed
r
ep
r
esen
tatio
n
o
f
b
o
r
r
o
wer
s
’
f
in
an
cial
d
if
f
icu
lties
,
wh
ich
is
im
p
o
r
ta
n
t
f
o
r
d
ev
elo
p
i
n
g
ac
cu
r
ate
p
r
ed
ictio
n
m
o
d
els
a
n
d
s
tr
o
n
g
er
r
is
k
ass
ess
m
en
t
f
r
am
ewo
r
k
s
.
Alo
n
g
s
id
e
th
e
t
ar
g
et
v
a
r
iab
le,
t
h
e
d
ataset
in
clu
d
es
ex
p
la
n
ato
r
y
f
ac
to
r
s
s
u
ch
as
l
o
an
s
tatu
s
,
o
u
ts
tan
d
in
g
am
o
u
n
ts
,
lo
an
d
u
r
atio
n
,
in
co
m
e
,
jo
b
h
is
to
r
y
,
h
o
m
e
o
wn
er
s
h
ip
,
lo
an
p
u
r
p
o
s
e,
m
o
n
t
h
ly
d
eb
t
co
m
m
itm
en
ts
,
an
d
cr
ed
it
h
is
to
r
y
.
Usi
n
g
th
ese
f
in
an
cial
ch
a
r
ac
ter
is
tics
,
th
e
m
o
d
els
aim
to
p
r
ed
ict
d
e
f
au
lt
r
is
k
m
o
r
e
ac
c
u
r
a
tely
an
d
s
u
p
p
o
r
t
p
r
o
ac
tiv
e
r
is
k
m
itig
atio
n
b
y
f
in
an
cial
in
s
titu
tio
n
s
.
I
m
p
o
r
ta
n
tly
,
th
e
f
o
u
r
-
class
s
tr
u
ctu
r
e
ca
p
tu
r
es
n
o
t
o
n
ly
wh
eth
er
a
b
o
r
r
o
we
r
is
d
is
tr
ess
e
d
,
b
u
t
also
th
e
s
ev
e
r
ity
o
f
th
a
t
d
is
tr
ess
,
wh
ich
is
p
ar
ticu
lar
ly
u
s
ef
u
l f
o
r
d
ec
is
io
n
s
r
elate
d
to
m
o
n
ito
r
in
g
,
r
estr
u
ctu
r
in
g
,
an
d
r
ec
o
v
e
r
y
s
tr
ateg
i
es.
3
.
2
.
Co
m
pa
riso
n r
esu
lt
s
o
n t
he
predict
iv
e
a
bil
it
y
o
f
t
he
mo
dels
T
h
e
o
u
t
-
of
-
s
am
p
le
test
r
esu
lts
o
f
th
e
t
r
ee
-
b
ased
d
ee
p
-
lear
n
in
g
a
n
d
d
ee
p
n
eu
r
al
n
etwo
r
k
s
m
o
d
els
in
clu
d
in
g
T
B
C
NN,
Neu
r
al
Dec
is
io
n
T
r
ee
,
NODE
,
an
d
Dee
p
Fo
r
est
ar
e
p
r
esen
ted
in
d
etail
in
T
ab
le
1.
T
h
e
p
er
f
o
r
m
a
n
ce
m
ea
s
u
r
es
i
n
th
e
tab
le
d
em
o
n
s
tr
ate
n
o
tab
l
e
d
is
cr
ep
an
cies
ac
r
o
s
s
th
e
m
o
d
els
in
f
o
r
ec
asti
n
g
p
er
s
o
n
al
b
an
k
r
u
p
tcy
.
T
B
C
NN
d
em
o
n
s
tr
ates
a
r
elativ
ely
l
o
w
ac
cu
r
a
c
y
o
f
0
.
6
5
7
5
,
ac
c
o
m
p
an
ied
b
y
s
u
b
p
ar
p
r
ec
is
io
n
an
d
r
ec
all
s
co
r
es,
r
ef
lectin
g
its
in
ad
eq
u
ate
ca
p
a
city
to
ac
cu
r
ately
ca
teg
o
r
ize
b
o
th
p
o
s
itiv
e
an
d
n
eg
ativ
e
in
s
tan
ce
s
.
T
h
e
Neu
r
al
Dec
is
io
n
T
r
ee
m
o
d
el
ex
h
ib
its
s
u
b
p
ar
p
er
f
o
r
m
a
n
ce
,
ev
id
en
ce
d
b
y
an
ac
cu
r
ac
y
o
f
0
.
6
2
3
3
,
a
p
r
ec
is
io
n
o
f
0
.
8
1
1
6
,
an
d
a
n
F1
s
co
r
e
o
f
0
.
6
4
2
9
,
in
d
icatin
g
an
im
b
alan
ce
in
it
s
ca
p
ac
ity
to
d
etec
t
g
en
u
in
e
p
o
s
itiv
es
wh
ile
m
in
im
izin
g
f
alse
n
eg
ativ
es.
C
o
n
v
er
s
ely
,
NODE
ex
h
ib
its
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
co
m
p
ar
ed
t
o
th
e
Neu
r
al
Dec
is
io
n
T
r
ee
an
d
T
B
C
NN,
attain
i
n
g
an
ac
cu
r
ac
y
o
f
0
.
7
9
6
7
a
n
d
an
AUC
o
f
0
.
7
3
0
2
,
in
d
icatin
g
a
m
o
d
er
ate
ca
p
ac
i
ty
f
o
r
class
d
if
f
er
e
n
tiatio
n
.
No
n
eth
eless
,
its
F1
s
co
r
e
o
f
0
.
8
5
3
1
s
ig
n
if
ies
a
r
ea
s
o
n
ab
le
b
alan
ce
b
etwe
en
a
cc
u
r
ac
y
a
n
d
r
ec
all,
alth
o
u
g
h
t
h
er
e
r
em
ain
s
r
o
o
m
f
o
r
im
p
r
o
v
em
en
t
in
p
r
ed
ict
io
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
43
,
No
.
2
,
Au
g
u
s
t
20
26
:
640
-
6
5
0
646
task
s
.
T
h
e
p
r
ee
m
in
en
t
m
o
d
el
is
d
ef
in
itely
Dee
p
Fo
r
est,
wh
ic
h
s
u
r
p
ass
es
th
e
o
th
er
m
o
d
els
with
an
ac
cu
r
ac
y
o
f
0
.
9
9
2
9
a
n
d
n
ea
r
ly
f
lawless
p
r
ec
is
io
n
an
d
r
ec
all.
T
h
e
F1
s
co
r
e
o
f
0
.
9
9
5
3
a
n
d
AUC
o
f
0
.
9
9
0
5
in
d
icate
a
n
ef
f
ec
tiv
e
b
alan
ce
b
etwe
en
f
a
ls
e
p
o
s
itiv
es
an
d
f
alse
n
eg
at
iv
es,
wh
ile
p
r
eser
v
in
g
ex
ce
ll
en
t
d
is
cr
im
in
ato
r
y
ca
p
ab
ilit
y
.
T
h
is
o
u
tco
m
e
h
i
g
h
lig
h
ts
th
e
s
u
p
er
io
r
ity
o
f
tr
e
e
-
b
ased
en
s
em
b
le
m
o
d
els
s
u
ch
as
Dee
p
Fo
r
est,
esp
ec
ially
in
m
an
ag
in
g
th
e
in
t
r
icate
,
n
o
n
-
lin
ea
r
in
ter
ac
tio
n
s
in
h
er
en
t in
f
in
an
cial
in
f
o
r
m
atio
n
.
T
ab
le
1
.
Per
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ictio
n
r
esu
lts
o
f
m
o
d
els o
n
o
u
t
-
of
-
s
am
p
le
d
atasets
M
o
d
e
l
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
S
e
n
s
i
t
i
v
i
t
y
S
p
e
c
i
f
i
c
i
t
y
F
1
sc
o
r
e
AUC
TB
C
N
N
0
.
6
5
7
5
0
.
8
2
3
9
0
.
7
6
9
6
0
.
3
0
8
8
0
.
6
9
8
9
0
.
5
3
9
2
N
e
u
r
a
l
d
e
c
i
si
o
n
t
r
e
e
0
.
6
2
3
3
0
.
8
1
1
6
0
.
7
5
0
3
0
.
2
5
0
9
0
.
6
4
2
9
0
.
5
0
0
6
NODE
0
.
7
9
6
7
0
.
8
9
0
3
0
.
8
6
5
6
0
.
5
9
4
8
0
.
8
5
3
1
0
.
7
3
0
2
D
e
e
p
f
o
r
e
s
t
0
.
9
9
2
9
0
.
9
9
5
3
0
.
9
9
5
3
0
.
9
8
5
8
0
.
9
9
5
3
0
.
9
9
0
5
T
h
ese
f
in
d
in
g
s
d
ir
ec
tly
a
d
d
r
e
s
s
th
e
m
ain
r
esear
ch
o
b
jectiv
e
o
f
th
e
s
tu
d
y
,
wh
ich
was
to
d
eter
m
in
e
wh
eth
er
tr
ee
-
b
ased
d
ee
p
lear
n
in
g
m
o
d
els
co
u
ld
ac
h
ie
v
e
p
r
e
d
ictiv
e
p
er
f
o
r
m
an
ce
co
m
p
ar
ab
le
to
,
o
r
b
etter
th
an
,
d
ee
p
n
eu
r
al
alter
n
ativ
es
in
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
e
d
ictio
n
.
T
h
e
r
esu
lts
s
h
o
w
th
at
th
is
is
i
n
d
ee
d
th
e
ca
s
e,
with
Dee
p
Fo
r
est
co
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
in
g
T
B
C
NN,
Neu
r
al
Dec
is
io
n
T
r
ee
,
an
d
NODE
ac
r
o
s
s
all
ev
alu
atio
n
cr
iter
ia.
T
h
is
s
u
g
g
ests
th
at,
f
o
r
s
tr
u
ctu
r
ed
b
o
r
r
o
wer
-
lev
el
f
in
an
cial
d
ata,
a
lay
er
e
d
tr
ee
en
s
em
b
le
is
m
o
r
e
ef
f
ec
tiv
e
in
ca
p
tu
r
in
g
co
m
p
lex
in
ter
ac
tio
n
s
an
d
n
o
n
-
lin
e
ar
d
ec
is
io
n
b
o
u
n
d
ar
ies
th
an
th
e
test
ed
n
eu
r
a
l
ar
ch
itectu
r
es.
Fro
m
a
p
r
ac
tic
al
p
er
s
p
ec
tiv
e,
th
e
r
esu
lts
r
e
p
o
r
ted
in
T
ab
le
1
a
r
e
also
i
m
p
o
r
tan
t
b
ec
au
s
e
ef
f
ec
tiv
e
cr
ed
it
-
r
is
k
p
r
ed
ictio
n
r
eq
u
i
r
es
n
o
t
o
n
ly
h
i
g
h
s
e
n
s
itiv
ity
,
s
o
th
at
d
is
tr
ess
ed
b
o
r
r
o
we
r
s
ca
n
b
e
id
en
tifie
d
ea
r
ly
,
b
u
t
also
s
u
f
f
i
cien
t
s
p
ec
if
icity
to
av
o
id
e
x
ce
s
s
iv
e
f
alse
alar
m
s
an
d
in
ef
f
icien
t
in
ter
v
en
tio
n
.
I
n
th
is
r
esp
ec
t,
T
B
C
N
N
an
d
Neu
r
al
Dec
is
io
n
T
r
ee
p
er
f
o
r
m
wea
k
ly
d
u
e
to
th
ei
r
l
o
w
s
p
ec
if
icity
,
wh
ile
NODE
o
f
f
er
s
a
clea
r
im
p
r
o
v
em
en
t
b
u
t
s
till
r
em
ain
s
in
f
er
io
r
to
Dee
p
Fo
r
est.
T
h
er
ef
o
r
e,
th
e
m
ain
im
p
licatio
n
o
f
T
ab
le
1
is
n
o
t
s
im
p
ly
t
h
at
De
ep
Fo
r
est
ac
h
iev
es
th
e
h
ig
h
est
ac
cu
r
ac
y
,
b
u
t
th
at
it
p
r
o
v
i
d
es
th
e
m
o
s
t
b
alan
ce
d
an
d
o
p
er
atio
n
a
lly
c
r
ed
ib
le
p
e
r
f
o
r
m
an
ce
f
o
r
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
e
d
ictio
n
.
An
an
aly
s
is
o
f
th
e
co
n
f
u
s
io
n
m
atr
ices
f
o
r
Dee
p
Fo
r
est,
NO
DE
,
Neu
r
al
Dec
is
io
n
T
r
ee
,
an
d
T
B
C
NN
in
Fig
u
r
e
3
in
d
icate
s
n
o
tab
le
d
if
f
er
en
ce
s
in
m
o
d
el
p
er
f
o
r
m
an
ce
,
with
Dee
p
Fo
r
est
id
en
tifie
d
as
th
e
m
o
s
t
p
r
ec
is
e
an
d
eq
u
itab
le
m
o
d
el.
I
t
ac
cu
r
ately
i
d
en
tifie
s
8
4
3
n
o
n
-
b
an
k
r
u
p
t
p
er
s
o
n
s
a
n
d
8
0
0
b
a
n
k
r
u
p
t
in
d
i
v
id
u
als,
with
n
eg
lig
ib
le
m
is
class
if
icat
io
n
s
,
m
er
ely
2
0
er
r
o
n
e
o
u
s
p
o
s
itiv
es
an
d
2
6
f
alse
n
eg
ativ
es.
T
h
is
d
eg
r
ee
o
f
p
r
ed
ictio
n
ac
cu
r
ac
y
d
em
o
n
s
tr
ates
th
at
De
ep
Fo
r
est
i
s
ex
ce
p
tio
n
ally
p
r
o
f
icien
t
at
d
etec
tin
g
p
o
ten
tial
d
ef
au
lter
s
wh
ile
m
in
im
is
in
g
th
e
m
is
class
if
icatio
n
o
f
n
o
n
-
b
an
k
r
u
p
t p
er
s
o
n
s
.
T
h
e
m
o
d
el'
s
ca
p
ac
ity
to
r
ed
u
ce
b
o
t
h
m
is
tak
e
ty
p
es
u
n
d
er
s
co
r
es
its
r
esil
ien
c
e,
r
en
d
er
in
g
it
esp
ec
ially
ap
p
r
o
p
r
iate
f
o
r
f
i
n
a
n
cial
ap
p
licatio
n
s
wh
er
e
p
r
ec
is
io
n
in
r
is
k
d
if
f
er
e
n
tiatio
n
is
ess
en
tial.
C
o
n
v
er
s
ely
,
NODE
h
as
co
n
s
id
er
ab
le
d
if
f
icu
lties
,
p
ar
ticu
lar
ly
in
ca
teg
o
r
is
in
g
n
o
n
-
b
an
k
r
u
p
t
p
er
s
o
n
s
.
NODE
'
s
s
h
o
r
tco
m
in
g
s
in
d
is
tin
g
u
is
h
in
g
b
etwe
en
d
ef
a
u
lt
an
d
n
o
n
-
d
ef
a
u
lt
in
s
tan
ce
s
ar
e
clea
r
,
with
6
3
m
is
class
if
ied
n
o
n
-
b
an
k
r
u
p
t
p
e
r
s
o
n
s
an
d
o
n
ly
5
ac
c
u
r
ately
r
ec
o
g
n
is
ed
.
NODE
'
s
s
tr
en
g
th
r
esid
es
in
its
ab
ilit
y
t
o
r
eliab
ly
i
d
en
tify
i
n
s
o
lv
en
t
p
er
s
o
n
s
,
ev
id
en
ce
d
b
y
its
8
7
3
r
ig
h
t
class
if
icatio
n
s
.
T
h
e
elev
ated
in
cid
e
n
ce
o
f
m
is
class
if
icatio
n
s
in
non
-
b
an
k
r
u
p
t
in
s
tan
ce
s
d
im
in
is
h
es
its
o
v
er
all
tr
u
s
two
r
th
in
ess
,
esp
ec
ially
in
p
r
ac
tical
s
itu
atio
n
s
wh
en
f
alse
p
o
s
itiv
es
m
ay
r
esu
lt
in
d
etr
im
e
n
tal
lo
a
n
c
h
o
ic
es.
Neu
r
al
Dec
is
io
n
T
r
ee
d
em
o
n
s
tr
ate
a
s
ig
n
if
ic
an
t
b
ias,
class
if
y
in
g
th
e
m
a
jo
r
ity
o
f
in
s
tan
ce
s
as
b
a
n
k
r
u
p
t
an
d
c
o
r
r
ec
tly
r
ec
o
g
n
is
in
g
o
n
ly
th
r
ee
n
o
n
-
b
an
k
r
u
p
t
p
er
s
o
n
s
.
T
h
is
b
iass
ed
class
if
icatio
n
s
ig
n
if
ican
tly
p
r
io
r
itis
es
th
e
id
en
tific
atio
n
o
f
b
a
n
k
r
u
p
t
ca
s
es,
as
d
em
o
n
s
tr
ated
b
y
8
6
0
ac
cu
r
ate
class
if
icatio
n
s
.
T
h
i
s
p
r
o
n
o
u
n
ce
d
b
ias
to
war
d
s
a
s
in
g
le
ca
teg
o
r
y
u
n
d
er
m
in
es
th
e
m
o
d
el'
s
ca
p
ac
ity
to
d
eliv
er
b
alan
ce
d
p
r
e
d
ictio
n
s
,
r
en
d
er
in
g
it
less
ap
p
r
o
p
r
iate
f
o
r
p
r
ac
tical
ap
p
l
icatio
n
s
wh
er
e
d
if
f
er
en
tiatin
g
b
etwe
en
th
e
two
ca
teg
o
r
ies
is
eq
u
ally
cr
itical.
L
ik
ewise,
T
B
C
N
N
en
co
u
n
ter
s
d
if
f
icu
lties
in
m
is
class
if
y
in
g
n
o
n
-
b
an
k
r
u
p
t
p
er
s
o
n
s
,
e
r
r
o
n
eo
u
s
ly
ca
teg
o
r
is
in
g
4
3
n
o
n
-
b
an
k
r
u
p
t in
s
tan
ce
s
as
b
an
k
r
u
p
t.
Desp
ite
attain
in
g
a
m
o
d
est s
u
cc
ess
r
ate
in
d
etec
ti
n
g
b
an
k
r
u
p
t p
e
r
s
o
n
s
,
ev
id
en
ce
d
b
y
8
2
8
ac
c
u
r
ate
c
lass
if
icatio
n
s
,
th
e
m
o
d
el
'
s
f
ai
lu
r
e
to
d
is
tin
ctly
d
if
f
er
en
tiate
b
etwe
en
th
e
two
g
r
o
u
p
s
co
n
s
tr
ain
s
its
u
tili
ty
i
n
co
n
tex
ts
n
ec
ess
itatin
g
p
r
ec
is
e
r
is
k
ev
alu
atio
n
.
Dee
p
Fo
r
e
s
t
d
em
o
n
s
tr
ates
th
e
h
ig
h
est
r
eliab
ilit
y
,
ef
f
ec
tiv
ely
d
is
tin
g
u
is
h
in
g
b
etwe
en
b
an
k
r
u
p
t
an
d
n
o
n
-
b
a
n
k
r
u
p
t
p
eo
p
le
wh
ile
s
u
s
ta
in
in
g
a
m
in
im
al
m
is
class
if
icatio
n
r
ate.
T
h
e
al
ter
n
ativ
e
m
o
d
els
-
NO
DE
,
Neu
r
al
Dec
is
io
n
T
r
ee
,
a
n
d
T
B
C
NN
-
ex
h
ib
it
co
n
s
id
er
ab
le
d
ef
icien
cies,
esp
ec
ially
in
d
if
f
er
en
tiatin
g
n
o
n
-
b
an
k
r
u
p
t
p
er
s
o
n
s
,
h
en
ce
r
e
d
u
cin
g
th
eir
u
s
e
in
f
in
an
cial
co
n
te
x
ts
.
T
h
is
co
m
p
ar
is
o
n
h
ig
h
lig
h
ts
Dee
p
Fo
r
est'
s
ad
v
an
tag
e
in
d
eliv
er
in
g
a
m
o
r
e
th
o
r
o
u
g
h
a
n
d
p
r
ec
is
e
r
is
k
ass
ess
m
en
t,
es
s
en
tial f
o
r
p
r
e
d
ictin
g
p
e
r
s
o
n
al
b
a
n
k
r
u
p
tcy
.
Fig
u
r
e
3
s
h
o
ws
m
o
r
e
th
a
n
a
s
im
p
le
r
an
k
in
g
o
f
m
o
d
els;
it
also
r
ev
ea
ls
th
e
ty
p
es
o
f
e
r
r
o
r
s
e
ac
h
m
o
d
el
ten
d
s
to
m
ak
e.
T
h
is
d
is
tin
ctio
n
is
im
p
o
r
tan
t
b
ec
au
s
e
f
al
s
e
p
o
s
itiv
es
an
d
f
alse
n
eg
ativ
es
h
av
e
d
if
f
er
en
t
m
an
ag
er
ial
co
n
s
eq
u
en
ce
s
.
E
x
ce
s
s
iv
e
f
alse
p
o
s
itiv
es
m
ay
lead
in
s
titu
tio
n
s
to
class
if
y
r
elativ
ely
h
ea
lth
y
b
o
r
r
o
we
r
s
as
r
is
k
y
,
r
esu
ltin
g
in
u
n
n
ec
ess
ar
y
r
estrictio
n
s
o
r
m
o
n
ito
r
in
g
co
s
ts
,
wh
e
r
e
as
ex
ce
s
s
iv
e
f
alse
n
eg
ativ
es
m
ay
d
elay
in
ter
v
e
n
tio
n
f
o
r
b
o
r
r
o
wer
s
wh
o
ar
e
g
e
n
u
in
ely
d
eter
io
r
atin
g
.
Dee
p
F
o
r
est
is
p
ar
ticu
lar
ly
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
A
d
va
n
ce
d
p
ers
o
n
a
l b
a
n
kru
p
tc
y
p
r
ed
ictio
n
u
s
in
g
tr
ee
-
b
a
s
ed
d
ee
p
lea
r
n
in
g
mo
d
els
(
N
h
a
t Ng
u
ye
n
Min
h
)
647
attr
ac
tiv
e
b
ec
au
s
e
it
r
ed
u
ce
s
b
o
th
ty
p
es
o
f
er
r
o
r
s
im
u
ltan
eo
u
s
ly
,
m
ak
i
n
g
it
n
o
t
o
n
l
y
s
tatis
tically
s
u
p
er
io
r
b
u
t
also
m
o
r
e
u
s
ef
u
l i
n
p
r
ac
tical
l
en
d
in
g
a
n
d
c
o
lle
ctio
n
s
ettin
g
s
.
Fig
u
r
e
3
.
Pre
d
ictio
n
r
esu
lts
o
f
m
o
d
els o
n
th
e
co
n
f
u
s
io
n
m
atr
i
x
Fig
u
r
e
4
d
ep
icts
a
d
is
tin
ct
p
er
f
o
r
m
an
ce
d
if
f
e
r
en
tial
ac
r
o
s
s
T
B
C
NN,
Neu
r
al
Dec
is
io
n
T
r
ee
,
NODE
,
an
d
Dee
p
Fo
r
est
ac
r
o
s
s
s
ev
er
al
class
e
s
,
as
s
ee
n
b
y
th
e
R
OC
cu
r
v
es.
Dee
p
Fo
r
est
d
is
ti
n
g
u
is
h
es
its
elf
with
ex
ce
p
tio
n
al
ca
te
g
o
r
is
atio
n
,
wi
th
an
AUC
o
f
0
.
9
9
8
0
o
r
ab
o
v
e
ac
r
o
s
s
all
ca
teg
o
r
ies,
i
n
clu
d
i
n
g
a
f
lawless
AUC
o
f
1
.
0
0
0
i
n
C
lass
3
.
T
h
is
p
er
f
o
r
m
an
ce
illu
s
tr
ates
its
s
tr
o
n
g
ca
p
ac
ity
to
p
r
ec
is
ely
d
if
f
er
e
n
tiate
b
etwe
e
n
d
ef
au
lt
an
d
n
o
n
-
d
ef
au
lt
s
ce
n
ar
i
o
s
.
C
o
n
v
er
s
ely
,
T
B
C
NN
co
n
tin
u
o
u
s
ly
ex
h
ib
its
wo
r
s
e
p
er
f
o
r
m
an
ce
,
with
AUC
v
alu
es
s
p
an
n
in
g
f
r
o
m
0
.
3
0
8
1
to
0
.
5
1
4
6
,
s
ig
n
if
y
in
g
lim
ited
d
is
cr
im
in
ato
r
y
ca
p
ab
ilit
y
,
esp
ec
ially
in
C
las
s
0
,
wh
er
e
it
f
ails
to
ad
eq
u
ately
d
if
f
er
e
n
tiate
b
etwe
en
tr
u
e
an
d
f
alse
p
o
s
itiv
es.
Neu
r
al
Dec
i
s
io
n
T
r
ee
p
r
o
v
id
e
m
a
r
g
in
al
en
h
an
ce
m
e
n
t
co
m
p
ar
ed
to
T
B
C
N
N,
alth
o
u
g
h
th
ey
co
n
tin
u
e
to
d
is
p
lay
s
u
b
p
ar
p
er
f
o
r
m
an
ce
with
AUCs
r
an
g
in
g
f
r
o
m
0
.
4
0
0
0
to
0
.
4
7
5
7
,
r
ef
lectin
g
a
co
n
s
tr
ain
e
d
ca
p
ac
ity
to
elu
cid
ate
th
e
u
n
d
er
l
y
i
n
g
d
ata
s
tr
u
ct
u
r
e,
p
ar
ticu
lar
ly
in
C
lass
3
.
NODE
h
as
d
ec
en
t
p
er
f
o
r
m
a
n
ce
,
ac
h
i
ev
in
g
AUC
v
alu
es
o
f
0
.
7
7
2
6
i
n
C
lass
3
,
h
o
wev
er
it
is
co
n
tin
u
o
u
s
ly
s
u
r
p
ass
ed
b
y
Dee
p
Fo
r
est
ac
r
o
s
s
all
class
es.
T
h
e
s
ig
n
if
ican
t
d
is
p
ar
ity
in
AUC
v
alu
es
b
etwe
en
Dee
p
Fo
r
est
an
d
th
e
o
th
er
m
o
d
els
u
n
d
er
s
co
r
es
its
ex
ce
p
tio
n
al
ca
p
ac
ity
to
g
e
n
er
alis
e
ac
r
o
s
s
all
class
es,
estab
li
s
h
in
g
it
as
th
e
m
o
s
t
d
e
p
en
d
a
b
le
m
o
d
el
f
o
r
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ict
io
n
.
T
h
e
m
a
r
k
ed
l
y
r
ed
u
ce
d
AUCs
o
f
T
B
C
NN
a
n
d
Neu
r
al
Dec
is
io
n
T
r
ee
in
d
i
ca
te
th
eir
in
ad
e
q
u
ac
y
in
m
a
n
ag
in
g
th
e
d
ataset's
co
m
p
lex
ity
,
h
ig
h
lig
h
tin
g
th
e
n
ee
d
f
o
r
m
o
r
e
s
o
p
h
is
ticated
m
o
d
els
s
u
ch
as
Dee
p
Fo
r
est
to
attain
elev
ated
p
r
ed
ictio
n
ac
c
u
r
ac
y
.
T
h
e
R
OC
ev
id
en
ce
is
p
ar
ticu
l
ar
ly
in
f
o
r
m
ativ
e
b
ec
au
s
e
th
e
t
ar
g
et
v
ar
ia
b
le
is
d
ef
in
e
d
in
f
o
u
r
lev
els
o
f
d
is
tr
ess
r
ath
er
th
an
as
a
s
im
p
le
b
in
ar
y
o
u
tco
m
e.
I
n
th
is
co
n
tex
t,
a
s
tr
o
n
g
av
e
r
ag
e
m
e
tr
ic
is
n
o
t
e
n
o
u
g
h
;
th
e
m
o
d
el
m
u
s
t
also
p
r
eser
v
e
d
is
cr
im
in
ativ
e
ab
ilit
y
ac
r
o
s
s
a
ll
b
an
k
r
u
p
tcy
class
es.
Fig
u
r
e
4
s
h
o
ws
th
at
Dee
p
Fo
r
est
r
em
ain
s
co
n
s
is
ten
tly
s
tr
o
n
g
f
o
r
e
v
er
y
class
,
in
clu
d
i
n
g
th
e
m
o
s
t
s
ev
er
e
class
,
wh
ich
is
p
ar
ticu
lar
ly
im
p
o
r
tan
t
f
o
r
in
s
titu
tio
n
s
th
at
n
ee
d
r
eliab
le
id
e
n
tific
atio
n
o
f
h
ig
h
-
r
is
k
b
o
r
r
o
wer
s
.
T
h
is
r
e
s
u
lt
s
tr
en
g
th
en
s
th
e
in
ter
p
r
etatio
n
f
r
o
m
T
a
b
le
1
an
d
Fig
u
r
e
3
:
t
h
e
ad
v
an
tag
e
o
f
D
ee
p
Fo
r
est
is
n
o
t
co
n
f
in
ed
to
o
n
e
ca
teg
o
r
y
o
r
o
n
e
m
etr
ic,
b
u
t is v
is
ib
le
th
r
o
u
g
h
o
u
t th
e
en
tire
class
if
icatio
n
s
tr
u
ctu
r
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
5
2
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
43
,
No
.
2
,
Au
g
u
s
t
20
26
:
640
-
6
5
0
648
Fig
u
r
e
4
.
Pre
d
ictio
n
r
esu
lts
o
f
m
o
d
els o
n
R
OC
C
u
r
v
es
3
.
3
.
Dis
cus
s
io
n a
nd
im
pli
ca
t
io
ns
T
h
e
f
in
d
i
n
g
s
in
d
icate
th
at
Dee
p
Fo
r
est
is
th
e
m
o
s
t
ef
f
ec
tiv
e
m
o
d
el
f
o
r
m
u
lti
-
class
p
er
s
o
n
al
b
an
k
r
u
p
tcy
p
r
ed
ictio
n
am
o
n
g
th
e
f
o
u
r
e
v
alu
ated
a
p
p
r
o
ac
h
es
.
T
h
is
s
u
g
g
ests
th
at,
f
o
r
b
o
r
r
o
wer
-
lev
el
f
in
a
n
cial
d
ata,
a
ca
s
ca
d
e
en
s
em
b
le
o
f
tr
ee
s
is
b
etter
ab
le
to
ca
p
tu
r
e
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
,
th
r
e
s
h
o
ld
ef
f
ec
t
s
,
an
d
v
ar
iab
le
in
ter
ac
tio
n
s
th
a
n
th
e
t
ested
n
eu
r
al
ar
ch
itectu
r
es.
T
h
e
r
esu
lt is
im
p
o
r
tan
t n
o
t
o
n
ly
b
e
ca
u
s
e
Dee
p
Fo
r
est
ac
h
iev
es
th
e
b
est
p
r
e
d
ictiv
e
p
er
f
o
r
m
a
n
ce
,
b
u
t
also
b
ec
au
s
e
its
s
tr
u
ctu
r
e
ap
p
ea
r
s
to
b
e
m
o
r
e
co
m
p
atib
le
wit
h
th
e
ch
ar
ac
ter
is
tics
o
f
f
i
n
an
cia
l
d
ata
an
d
th
e
p
r
ac
tical
r
eq
u
i
r
em
en
ts
o
f
r
is
k
ass
ess
m
en
t.
Alth
o
u
g
h
th
e
in
itial
ex
p
ec
tatio
n
was
th
at
tr
ee
-
b
a
s
ed
d
ee
p
m
o
d
els
wo
u
ld
b
e
co
m
p
etitiv
e
with
d
ee
p
n
eu
r
a
l
alter
n
ativ
es,
th
e
ev
id
en
ce
r
ev
ea
ls
a
s
tr
o
n
g
er
p
atter
n
,
with
Dee
p
Fo
r
est
clea
r
ly
o
u
tp
e
r
f
o
r
m
in
g
th
e
o
th
er
m
o
d
els
an
d
NODE
s
h
o
win
g
o
n
l
y
in
ter
m
e
d
iate
p
e
r
f
o
r
m
a
n
ce
.
Nev
er
t
h
eless
,
th
e
m
ag
n
itu
d
e
o
f
th
is
g
a
p
s
h
o
u
ld
b
e
in
ter
p
r
ete
d
with
ca
u
tio
n
,
as it m
ay
p
ar
tly
r
e
f
lect
f
ea
tu
r
es o
f
th
e
p
r
esen
t d
atase
t th
at
f
av
o
u
r
h
ier
a
r
ch
ical
tr
ee
-
b
ased
p
ar
titi
o
n
in
g
.
T
h
is
s
tu
d
y
co
n
tr
ib
u
tes
em
p
ir
ical
ev
id
en
ce
f
r
o
m
a
r
elativ
ely
lar
g
e
b
o
r
r
o
wer
-
lev
el
d
ataset
co
llected
f
r
o
m
co
m
m
e
r
cial
b
an
k
s
an
d
cr
ed
it
in
s
titu
tio
n
s
in
Vietn
a
m
o
v
er
th
e
p
er
io
d
2
0
1
2
to
2
0
2
2
an
d
ev
alu
ates
m
u
ltip
le
ad
v
an
ce
d
m
o
d
els
u
n
d
er
a
co
m
m
o
n
ex
p
er
im
en
tal
d
esig
n
.
Ho
wev
er
,
s
ev
er
al
lim
i
tatio
n
s
r
em
ain
.
T
h
e
ev
id
en
ce
is
d
r
awn
f
r
o
m
a
s
in
g
le
n
atio
n
al
co
n
tex
t,
t
h
e
an
al
y
s
is
f
o
cu
s
es
o
n
p
r
ed
ictiv
e
co
m
p
ar
is
o
n
r
ath
er
t
h
an
ex
ter
n
al
v
alid
atio
n
,
an
d
f
u
r
th
er
wo
r
k
is
n
ee
d
ed
to
ex
am
in
e
th
e
s
tab
ilit
y
a
n
d
d
r
iv
e
r
s
o
f
t
h
e
p
r
ed
ictio
n
s
o
v
er
tim
e.
Ov
er
all,
th
e
r
esu
lts
s
u
g
g
est
th
at
Dee
p
Fo
r
est
is
th
e
m
o
s
t
s
u
itab
le
m
o
d
el
in
th
i
s
s
ettin
g
b
ec
au
s
e
it
co
m
b
in
es
s
tr
o
n
g
d
is
cr
im
in
atio
n
,
lo
w
m
is
class
if
icatio
n
,
an
d
p
r
ac
tical
r
elev
an
ce
f
o
r
f
in
an
ci
al
r
is
k
ass
ess
m
en
t.
Fu
tu
r
e
r
esear
ch
s
h
o
u
ld
v
alid
at
e
th
ese
f
i
n
d
in
g
s
o
n
ex
ter
n
al
d
a
tasets
,
co
m
p
ar
e
Dee
p
Fo
r
est with
o
th
er
ad
v
a
n
ce
d
tab
u
lar
-
lear
n
in
g
m
o
d
els,
an
d
in
co
r
p
o
r
ate
ex
p
lain
ab
ilit
y
t
ec
h
n
iq
u
es
to
id
en
tify
th
e
k
e
y
d
eter
m
in
an
ts
o
f
b
o
r
r
o
we
r
d
is
tr
ess
.
4.
CO
NCLU
SI
O
N
T
h
e
Dee
p
Fo
r
est
m
o
d
el
h
a
s
g
r
ea
ter
e
f
f
icac
y
in
m
u
lti
-
class
p
er
s
o
n
al
b
a
n
k
r
u
p
tcy
p
r
ed
ictio
n
,
s
u
r
p
ass
in
g
ex
is
tin
g
tr
ee
-
b
ased
d
ee
p
-
lear
n
in
g
a
n
d
d
ee
p
n
e
u
r
al
n
etwo
r
k
s
m
o
d
els
in
clu
d
i
n
g
T
B
C
NN,
Neu
r
al
Dec
is
io
n
T
r
ee
,
an
d
NODE
.
Dee
p
Fo
r
est
h
as
p
r
o
f
icien
cy
in
ca
p
tu
r
in
g
in
tr
icate
,
n
o
n
-
lin
ea
r
p
atter
n
s
ac
r
o
s
s
m
an
y
class
es,
as
s
ee
n
b
y
its
n
ea
r
ly
p
e
r
f
ec
t
AUC
r
atin
g
s
,
esp
ec
ially
its
im
p
ec
ca
b
le
p
er
f
o
r
m
an
ce
in
C
lass
3
.
T
h
e
en
s
em
b
le
-
b
ased
d
esig
n
o
f
th
is
m
o
d
el,
wh
ich
en
h
an
ce
s
p
r
ed
ictio
n
s
in
cr
em
e
n
tally
,
en
ab
les
s
u
p
er
io
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2502
-
4
7
5
2
A
d
va
n
ce
d
p
ers
o
n
a
l b
a
n
kru
p
tc
y
p
r
ed
ictio
n
u
s
in
g
tr
ee
-
b
a
s
ed
d
ee
p
lea
r
n
in
g
mo
d
els
(
N
h
a
t Ng
u
ye
n
Min
h
)
649
g
en
er
alis
a
tio
n
to
v
a
r
ied
an
d
co
m
p
lex
f
in
an
cial
f
ac
ts
.
Fu
r
th
er
m
o
r
e
,
th
e
u
s
e
o
f
Op
tu
n
a
f
o
r
h
y
p
e
r
p
ar
a
m
eter
o
p
tim
is
atio
n
s
ig
n
if
ican
tly
im
p
r
o
v
ed
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
c
e
b
y
e
n
ab
lin
g
th
e
p
r
ec
is
e
ad
j
u
s
tm
en
t
o
f
c
r
itical
p
ar
am
eter
s
to
g
et
o
p
tim
al
p
r
ed
icted
ac
cu
r
ac
y
.
Op
tu
n
a'
s
e
f
f
icac
y
in
n
a
v
ig
atin
g
th
e
h
y
p
er
p
ar
am
eter
s
p
ac
e
g
u
ar
an
teed
th
at
Dee
p
Fo
r
est
was
o
p
tim
is
ed
f
o
r
s
u
p
er
io
r
m
u
lti
-
class
d
if
f
er
en
tiatio
n
co
m
p
ar
ed
to
o
th
er
m
o
d
els.
T
h
e
in
teg
r
atio
n
o
f
Dee
p
Fo
r
est
'
s
en
s
em
b
le
ef
f
icac
y
an
d
Op
tu
n
a'
s
o
p
tim
is
atio
n
p
r
o
wess
u
n
d
er
s
co
r
es
t
h
e
m
o
d
el'
s
v
er
s
atility
an
d
p
r
ec
is
io
n
,
estab
lis
h
in
g
it
as
t
h
e
m
o
s
t
d
ep
e
n
d
ab
le
in
s
tr
u
m
en
t
f
o
r
i
n
tr
icate
,
m
u
lti
-
class
p
r
ed
ictio
n
s
in
f
i
n
an
cial
r
is
k
ev
alu
atio
n
s
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
e
au
th
o
r
s
th
an
k
all
lectu
r
er
s
an
d
m
an
a
g
er
s
o
f
Ho
C
h
i
M
in
h
Un
iv
er
s
ity
o
f
B
an
k
in
g
(
H
UB
)
.
T
h
is
ar
ticle
is
s
u
p
p
o
r
ted
an
d
f
u
n
d
e
d
b
y
HUB,
Vietn
am
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
tr
ib
u
to
r
R
o
les
T
ax
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
id
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
Nh
at
Ng
u
y
en
Mi
n
h
✓
✓
✓
✓
✓
✓
✓
✓
Du
y
Ng
o
Ho
a
n
g
Kh
an
h
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
Au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
t
er
est.
DATA AV
AI
L
AB
I
L
I
T
Y
Der
iv
ed
d
ata
s
u
p
p
o
r
tin
g
th
e
f
i
n
d
in
g
s
o
f
t
h
is
s
tu
d
y
ar
e
av
aila
b
le
f
r
o
m
t
h
e
co
r
r
esp
o
n
d
i
n
g
au
th
o
r
Nh
at
M.
Ng
u
y
en
o
n
r
eq
u
est.
RE
F
E
R
E
NC
E
S
[
1
]
J.
-
P
.
L
a
i
,
Y
.
-
L
.
L
i
n
,
H
.
-
C
.
Li
n
,
C
.
-
Y
.
S
h
i
h
,
Y
.
-
P
.
W
a
n
g
,
a
n
d
P
.
-
F
.
P
a
i
,
“
Tr
e
e
-
b
a
se
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
s
w
i
t
h
o
p
t
u
n
a
i
n
p
r
e
d
i
c
t
i
n
g
i
mp
e
d
a
n
c
e
v
a
l
u
e
s
f
o
r
c
i
r
c
u
i
t
a
n
a
l
y
si
s
,
”
Mi
c
r
o
m
a
c
h
i
n
e
s
,
v
o
l
.
1
4
,
n
o
.
2
,
p
.
2
6
5
,
J
a
n
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
mi
1
4
0
2
0
2
6
5
.
[
2
]
F
.
B
a
r
b
o
z
a
,
H
.
K
i
m
u
r
a
,
a
n
d
E.
A
l
t
ma
n
,
“
M
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
s
a
n
d
b
a
n
k
r
u
p
t
c
y
p
r
e
d
i
c
t
i
o
n
,
”
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
.
8
3
,
p
p
.
4
0
5
–
4
1
7
,
O
c
t
.
2
0
1
7
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
1
7
.
0
4
.
0
0
6
.
[
3
]
S
.
S
mi
t
i
a
n
d
M
.
S
o
u
i
,
“
B
a
n
k
r
u
p
t
c
y
p
r
e
d
i
c
t
i
o
n
u
si
n
g
d
e
e
p
l
e
a
r
n
i
n
g
a
p
p
r
o
a
c
h
b
a
se
d
o
n
b
o
r
d
e
r
l
i
n
e
S
M
O
TE,
”
I
n
f
o
rm
a
t
i
o
n
S
y
st
e
m
s
Fro
n
t
i
e
rs
,
v
o
l
.
2
2
,
n
o
.
5
,
p
p
.
1
0
6
7
–
1
0
8
3
,
A
u
g
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
7
9
6
-
0
2
0
-
1
0
0
3
1
-
6.
[
4
]
S
.
H
.
S
y
e
d
N
o
r
,
S
.
I
smai
l
,
a
n
d
B
.
W
.
Y
a
p
,
“
P
e
r
so
n
a
l
b
a
n
k
r
u
p
t
c
y
p
r
e
d
i
c
t
i
o
n
u
s
i
n
g
d
e
c
i
s
i
o
n
t
r
e
e
mo
d
e
l
,
”
J
o
u
rn
a
l
o
f
Ec
o
n
o
m
i
c
s
,
Fi
n
a
n
c
e
a
n
d
A
d
m
i
n
i
st
r
a
t
i
v
e
S
c
i
e
n
c
e
,
v
o
l
.
2
4
,
n
o
.
4
7
,
p
p
.
1
5
7
–
1
7
0
,
M
a
r
.
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
8
/
j
e
f
a
s
-
08
-
2
0
1
8
-
0
0
7
6
.
[
5
]
Q
.
F
u
,
K
.
L
i
,
J.
C
h
e
n
,
J.
W
a
n
g
,
Y
.
L
u
,
a
n
d
Y
.
W
a
n
g
,
“
B
u
i
l
d
i
n
g
e
n
e
r
g
y
c
o
n
su
mp
t
i
o
n
p
r
e
d
i
c
t
i
o
n
u
si
n
g
a
d
e
e
p
-
f
o
r
e
st
-
b
a
se
d
D
Q
N
met
h
o
d
,
”
B
u
i
l
d
i
n
g
s
,
v
o
l
.
1
2
,
n
o
.
2
,
p
.
1
3
1
,
Ja
n
.
2
0
2
2
,
d
o
i
:
1
0
.
3
3
9
0
/
b
u
i
l
d
i
n
g
s1
2
0
2
0
1
3
1
.
[
6
]
T.
Zh
o
u
,
X
.
S
u
n
,
X
.
X
i
a
,
B
.
L
i
,
a
n
d
X
.
C
h
e
n
,
“
I
mp
r
o
v
i
n
g
d
e
f
e
c
t
p
r
e
d
i
c
t
i
o
n
w
i
t
h
d
e
e
p
f
o
r
e
s
t
,
”
I
n
f
o
rm
a
t
i
o
n
a
n
d
S
o
f
t
w
a
re
T
e
c
h
n
o
l
o
g
y
,
v
o
l
.
1
1
4
,
p
p
.
2
0
4
–
2
1
6
,
O
c
t
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
f
s
o
f
.
2
0
1
9
.
0
7
.
0
0
3
.
[
7
]
Y
.
Zh
o
n
g
a
n
d
H
.
W
a
n
g
,
“
I
n
t
e
r
n
e
t
f
i
n
a
n
c
i
a
l
c
r
e
d
i
t
s
c
o
r
i
n
g
m
o
d
e
l
s
b
a
se
d
o
n
d
e
e
p
f
o
r
e
st
a
n
d
r
e
s
a
m
p
l
i
n
g
me
t
h
o
d
s,”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
8
6
8
9
–
8
7
0
0
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
a
c
c
e
ss.
2
0
2
3
.
3
2
3
9
8
8
9
.
[
8
]
S
.
P
o
p
o
v
,
S
.
M
o
r
o
z
o
v
,
a
n
d
A
.
B
a
b
e
n
k
o
,
“
N
e
u
r
a
l
o
b
l
i
v
i
o
u
s
d
e
c
i
si
o
n
e
n
se
m
b
l
e
s
f
o
r
d
e
e
p
l
e
a
r
n
i
n
g
o
n
t
a
b
u
l
a
r
d
a
t
a
,
”
a
rXi
v
p
re
p
ri
n
t
a
rXi
v
:
1
9
0
9
.
0
6
3
1
2
,
2
0
1
9
.
[
9
]
K
.
D
.
H
u
m
b
i
r
d
,
J
.
L.
P
e
t
e
r
so
n
,
a
n
d
R
.
G
.
M
c
c
l
a
r
r
e
n
,
“
D
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
i
n
i
t
i
a
l
i
z
a
t
i
o
n
w
i
t
h
d
e
c
i
si
o
n
t
r
e
e
s,
”
I
E
EE
T
ra
n
s
a
c
t
i
o
n
s
o
n
N
e
u
r
a
l
N
e
t
w
o
r
k
s
a
n
d
L
e
a
rn
i
n
g
S
y
s
t
e
m
s
,
v
o
l
.
3
0
,
n
o
.
5
,
p
p
.
1
2
8
6
–
1
2
9
5
,
M
a
y
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
t
n
n
l
s
.
2
0
1
8
.
2
8
6
9
6
9
4
.
[
1
0
]
F
.
M
.
Ta
l
a
a
t
,
A
.
A
l
j
a
d
a
n
i
,
M
.
B
a
d
a
w
y
,
a
n
d
M
.
El
h
o
sse
i
n
i
,
“
T
o
w
a
r
d
i
n
t
e
r
p
r
e
t
a
b
l
e
c
r
e
d
i
t
s
c
o
r
i
n
g
:
i
n
t
e
g
r
a
t
i
n
g
e
x
p
l
a
i
n
a
b
l
e
a
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
w
i
t
h
d
e
e
p
l
e
a
r
n
i
n
g
f
o
r
c
r
e
d
i
t
c
a
r
d
d
e
f
a
u
l
t
p
r
e
d
i
c
t
i
o
n
,
”
N
e
u
r
a
l
C
o
m
p
u
t
i
n
g
a
n
d
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
3
6
,
n
o
.
9
,
p
p
.
4
8
4
7
–
4
8
6
5
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
0
0
5
2
1
-
0
2
3
-
0
9
2
3
2
-
2.
[
1
1
]
Q
.
Z
h
a
n
g
,
L.
S
u
n
,
G
.
Y
a
n
g
,
B
.
Lu
,
X
.
N
i
n
g
,
a
n
d
W
.
Li
,
“
TB
N
N
:
t
o
t
a
l
l
y
-
b
i
n
a
r
y
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
i
m
a
g
e
c
l
a
s
si
f
i
c
a
t
i
o
n
,
”
O
p
t
o
e
l
e
c
t
r
o
n
i
c
s L
e
t
t
e
rs
,
v
o
l
.
1
9
,
n
o
.
2
,
p
p
.
1
1
7
–
1
2
2
,
F
e
b
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
1
8
0
1
-
0
2
3
-
2
1
1
3
-
2.
[
1
2
]
M
.
S
t
e
v
e
n
s
o
n
,
C
.
M
u
e
s
,
a
n
d
C
.
B
r
a
v
o
,
“
T
h
e
v
a
l
u
e
o
f
t
e
x
t
f
o
r
s
mal
l
b
u
s
i
n
e
ss
d
e
f
a
u
l
t
p
r
e
d
i
c
t
i
o
n
:
A
D
e
e
p
Le
a
r
n
i
n
g
a
p
p
r
o
a
c
h
,
”
Eu
r
o
p
e
a
n
J
o
u
r
n
a
l
o
f
O
p
e
ra
t
i
o
n
a
l
R
e
s
e
a
r
c
h
,
v
o
l
.
2
9
5
,
n
o
.
2
,
p
p
.
7
5
8
–
7
7
1
,
D
e
c
.
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
j
o
r
.
2
0
2
1
.
0
3
.
0
0
8
.
Evaluation Warning : The document was created with Spire.PDF for Python.