I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l
E
ng
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
4
3
,
No
.
2
,
A
u
g
u
s
t
2
0
2
6
,
p
p
.
5
0
7
~
52
1
I
SS
N:
2
5
0
2
-
4
7
5
2
,
DOI
: 1
0
.
1
1
5
9
1
/ijeecs.v
4
3
.i
2
.
pp
507
-
5
2
1
507
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs.ia
esco
r
e.
co
m
Ada
ptive f
ea
ture
selection
for credi
t
sco
ring
mo
dels
M
o
s
t
a
f
a
M
o
ha
m
ed
S
eif
E
lna
s
r,
Ya
s
s
er
O
m
a
r,
Sa
leh Mes
ba
h
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
o
n
S
y
st
e
ms,
C
o
l
l
e
g
e
o
f
C
o
m
p
u
t
i
n
g
a
n
d
I
n
f
o
r
ma
t
i
o
n
T
e
c
h
n
o
l
o
g
y
,
A
r
a
b
A
c
a
d
e
m
y
f
o
r
S
c
i
e
n
c
e
,
T
e
c
h
n
o
l
o
g
y
a
n
d
M
a
r
i
t
i
m
e
Tr
a
n
s
p
o
r
t
(
A
A
S
T
M
T)
,
C
a
i
r
o
,
Eg
y
p
t
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
an
3
1
,
2
0
2
6
R
ev
is
ed
J
u
n
1
2
,
2
0
2
6
Acc
ep
ted
J
u
n
2
7
,
2
0
2
6
Ac
c
u
ra
te
c
re
d
it
sc
o
rin
g
a
n
d
d
e
li
n
q
u
e
n
c
y
p
re
d
icti
o
n
a
re
c
rit
ica
l
fo
r
fin
a
n
c
ial
in
stit
u
ti
o
n
s,
p
a
rti
c
u
larl
y
w
h
e
n
e
v
a
lu
a
ti
n
g
b
o
t
h
b
a
n
k
e
d
a
n
d
u
n
b
a
n
k
e
d
c
li
e
n
ts.
Trad
it
io
n
a
l
fe
a
tu
re
se
lec
ti
o
n
a
p
p
r
o
a
c
h
e
s
su
c
h
a
s
wra
p
p
e
r,
fil
ter,
a
n
d
e
m
b
e
d
d
e
d
m
e
th
o
d
s
o
ften
re
q
u
ire
re
p
e
a
ted
re
train
in
g
a
n
d
m
a
y
n
o
t
e
fficie
n
tl
y
e
x
p
lo
re
o
p
ti
m
a
l
fe
a
t
u
re
su
b
se
ts
in
m
u
lt
i
-
c
las
s
c
re
d
it
risk
se
tt
in
g
s.
Th
is
stu
d
y
p
ro
p
o
se
s
a
re
in
fo
rc
e
m
e
n
t
lea
rn
in
g
(RL)
–
b
a
se
d
a
d
a
p
ti
v
e
fe
a
tu
re
se
lec
ti
o
n
fra
m
e
wo
rk
u
sin
g
tab
u
lar
Q
-
lea
rn
in
g
to
d
y
n
a
m
ica
ll
y
i
d
e
n
ti
f
y
in
fo
rm
a
ti
v
e
fe
a
tu
re
su
b
se
ts
fo
r
c
re
d
it
b
u
c
k
e
t
c
las
sifica
ti
o
n
a
n
d
d
e
li
n
q
u
e
n
c
y
p
r
e
d
ictio
n
.
Th
e
fra
m
e
wo
rk
wa
s
e
v
a
lu
a
ted
o
n
two
d
a
tas
e
ts,
in
c
lu
d
i
n
g
p
r
o
p
r
ieta
ry
c
re
d
it
d
a
tas
e
ts
(2
4
,
5
3
3
re
c
o
rd
s
wit
h
1
8
–
1
9
fe
a
tu
re
s).
Us
in
g
stra
ti
fie
d
1
0
-
fo
l
d
c
ro
ss
-
v
a
li
d
a
ti
o
n
,
th
e
p
ro
p
o
se
d
a
p
p
ro
a
c
h
a
c
h
iev
e
d
6
2
.
8
4
%
a
c
c
u
ra
c
y
(F
1
=
0
.
6
3
)
fo
r
8
-
c
las
s
c
re
d
it
b
u
c
k
e
t
p
re
d
ictio
n
,
o
u
tp
e
rfo
rm
i
n
g
t
ra
d
it
io
n
a
l
wra
p
p
e
r
-
b
a
se
d
m
e
th
o
d
s
b
y
u
p
t
o
4
.
2
%
wh
il
e
re
d
u
c
in
g
c
o
m
p
u
ta
ti
o
n
a
l
c
o
st
c
o
m
p
a
re
d
to
e
x
h
a
u
stiv
e
s
u
b
se
t
e
v
a
lu
a
ti
o
n
.
F
o
r
d
e
li
n
q
u
e
n
c
y
p
re
d
i
c
ti
o
n
,
th
e
m
o
d
e
l
a
c
h
iev
e
d
a
p
p
ro
x
ima
tely
7
0
%
F
1
-
sc
o
re
,
d
e
m
o
n
stra
ti
n
g
imp
ro
v
e
d
m
in
o
r
it
y
-
c
las
s
se
n
siti
v
it
y
.
C
o
m
p
a
re
d
to
fil
ter
a
n
d
e
m
b
e
d
d
e
d
m
e
th
o
d
s,
t
h
e
RL
-
b
a
se
d
fra
m
e
wo
rk
p
ro
d
u
c
e
d
m
o
re
c
o
m
p
a
c
t
fe
a
tu
re
su
b
se
ts
wh
il
e
m
a
in
tain
in
g
c
o
m
p
e
ti
ti
v
e
c
o
m
p
u
tati
o
n
a
l
e
fficie
n
c
y
.
T
h
e
se
fin
d
i
n
g
s
d
e
m
o
n
stra
te
th
a
t
a
d
a
p
ti
v
e
RL
d
r
iv
e
n
fe
a
tu
re
se
lec
ti
o
n
p
r
o
v
i
d
e
s
a
p
ra
c
ti
c
a
l
a
n
d
sc
a
lab
le
m
e
c
h
a
n
ism
fo
r
e
n
h
a
n
c
in
g
p
re
d
icti
v
e
p
e
rfo
rm
a
n
c
e
in
c
re
d
it
risk
m
o
d
e
ll
in
g
,
su
p
p
o
r
ti
n
g
a
u
to
m
a
ted
a
n
d
c
o
n
siste
n
t
d
e
c
isio
n
m
a
k
i
n
g
in
fin
a
n
c
ia
l
in
stit
u
ti
o
n
s.
K
ey
w
o
r
d
s
:
C
r
ed
it sco
r
in
g
Delin
q
u
en
cy
p
r
ed
ictio
n
Featu
r
e
s
elec
tio
n
Fin
an
cial
r
is
k
m
o
d
ellin
g
Q
-
lear
n
in
g
R
ein
f
o
r
ce
m
en
t le
ar
n
i
n
g
W
r
ap
p
er
m
eth
o
d
s
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
:
Mo
s
taf
a
Mo
h
am
ed
Seif
E
ln
asr
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
Sy
s
tem
s
,
C
o
lleg
e
o
f
C
o
m
p
u
tin
g
an
d
I
n
f
o
r
m
atio
n
T
ec
h
n
o
lo
g
y
Ar
ab
Aca
d
em
y
f
o
r
Scien
ce
,
T
ec
h
n
o
lo
g
y
an
d
Ma
r
itime
T
r
a
n
s
p
o
r
t (
AAST
MT
)
C
air
o
,
E
g
y
p
t
E
m
ail:
m
m
s
eif
8
7
@
y
ah
o
o
.
c
o
m
1.
I
NT
RO
D
UCT
I
O
N
C
r
ed
it
s
co
r
in
g
r
ef
er
s
to
th
e
e
v
alu
atio
n
o
f
an
in
d
i
v
id
u
al
clien
t
’
s
cr
ed
itwo
r
th
in
ess
with
in
a
f
in
an
cial
in
s
titu
tio
n
.
T
h
e
r
ap
id
d
ev
elo
p
m
en
t
o
f
f
in
a
n
cial
s
er
v
ices
an
d
th
e
em
er
g
en
ce
o
f
f
in
tech
h
av
e
m
ad
e
co
m
p
a
n
ies
th
at
lev
er
ag
e
in
f
o
r
m
ati
o
n
tec
h
n
o
lo
g
y
to
p
r
o
v
id
e
f
i
n
an
cial
s
o
lu
tio
n
s
,
an
d
m
an
y
f
in
tech
s
tar
tu
p
s
ar
e
n
o
w
o
f
f
er
in
g
d
iv
e
r
s
e
s
er
v
ices,
s
u
ch
as
co
n
s
u
m
e
r
f
in
an
c
e,
m
icr
o
f
in
an
ce
,
an
d
tr
ea
s
u
r
y
an
d
tr
ad
e
f
in
an
ce
ap
p
licatio
n
s
.
B
ef
o
r
e
p
r
o
v
id
in
g
th
ese
f
in
an
cial
s
er
v
ices,
co
m
p
an
ies
m
u
s
t
ass
es
s
th
e
r
is
k
o
f
th
eir
clien
ts
as
m
an
d
ated
b
y
ce
n
tr
al
b
an
k
r
eg
u
latio
n
s
to
e
n
s
u
r
e
b
o
r
r
o
wer
s
a
r
e
ca
p
ab
le
o
f
r
ep
a
y
in
g
th
eir
li
ab
ilit
ies.
T
h
e
ad
v
e
n
t
o
f
b
ig
d
ata
h
as
s
ig
n
if
ican
tly
im
p
ac
ted
th
e
f
in
an
cial
s
ec
to
r
,
esp
ec
ially
in
th
e
ar
ea
s
o
f
cr
ed
it
s
co
r
in
g
an
d
r
is
k
ass
es
s
m
en
t.
C
r
ed
it
s
co
r
es,
wh
ich
r
esu
lt
f
r
o
m
co
m
p
r
eh
en
s
iv
e
r
is
k
ev
alu
atio
n
s
,
p
la
y
a
cr
u
cial
r
o
le
in
d
if
f
er
en
tiatin
g
clien
ts
an
d
m
a
n
ag
in
g
r
is
k
.
T
r
ad
itio
n
ally
,
b
a
n
k
s
h
av
e
r
elied
o
n
cr
ed
it
s
co
r
in
g
s
y
s
tem
s
th
at
u
s
e
s
tatis
t
ical
tech
n
iq
u
es
to
an
aly
s
e
clien
t
d
ata,
wh
ich
in
clu
d
es
h
is
to
r
ical
cr
ed
it
in
f
o
r
m
atio
n
.
Ho
wev
er
,
s
u
c
h
co
n
v
en
tio
n
al
ap
p
r
o
ac
h
es
ty
p
i
ca
lly
r
eq
u
i
r
e
s
u
b
s
tan
tial
m
a
n
u
al
in
ter
v
e
n
tio
n
an
d
in
cu
r
s
ig
n
if
ican
t
p
r
o
ce
s
s
in
g
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.
4
3
,
No
.
2
,
Au
g
u
s
t
20
2
6
:
507
-
5
2
1
508
tim
e.
Data
s
cien
ce
h
as
r
ev
o
lu
tio
n
ized
th
is
s
p
ac
e
b
y
in
tr
o
d
u
cin
g
ad
v
an
ce
d
p
r
ed
ictiv
e
t
o
o
ls
an
d
tech
n
iq
u
es.
T
h
ese
m
o
d
er
n
ap
p
r
o
ac
h
es
au
t
o
m
ate
th
e
r
is
k
ass
ess
m
en
t
p
r
o
ce
s
s
b
y
ev
alu
atin
g
clien
t
d
ata
to
g
en
er
ate
c
r
ed
it
s
co
r
es
u
s
in
g
b
o
th
h
is
to
r
ical
cr
ed
it
r
ec
o
r
d
s
an
d
d
e
m
o
g
r
a
p
h
ic
attr
ib
u
tes.
Ho
wev
er
,
a
m
ajo
r
ch
allen
g
e
r
em
ain
s
m
illi
o
n
s
o
f
in
d
iv
id
u
als
in
lo
w
-
in
co
m
e
ec
o
n
o
m
ies,
r
ef
er
r
e
d
to
as
“
u
n
b
an
k
e
d
”
clien
ts
,
lack
s
u
f
f
icien
t
cr
ed
it
h
is
to
r
y
t
o
g
e
n
er
ate
a
cr
e
d
it
s
co
r
e
o
r
ac
ce
s
s
f
i
n
an
cial
s
er
v
ice
s
.
T
o
a
d
d
r
ess
th
is
is
s
u
e,
d
ata
s
cien
ce
m
o
d
els
m
u
s
t
in
co
r
p
o
r
ate
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
es
th
at
ca
n
p
r
ed
ict
cr
ed
it
s
co
r
es
f
o
r
b
o
th
“
b
an
k
e
d
”
an
d
“
u
n
b
an
k
ed
”
clien
ts
.
A
cr
u
cial
c
o
m
p
o
n
en
t
o
f
th
is
p
r
o
ce
s
s
is
f
ea
tu
r
e
s
elec
tio
n
,
wh
ich
f
o
r
m
s
th
e
f
o
u
n
d
ati
o
n
o
f
th
ese
m
o
d
els.
E
f
f
ec
tiv
e
f
ea
t
u
r
e
s
elec
tio
n
tech
n
iq
u
es
ar
e
ess
en
tial
f
o
r
m
an
ag
in
g
c
r
ed
it
s
co
r
in
g
d
atasets
th
at
co
n
tain
a
lar
g
e
n
u
m
b
er
o
f
v
ar
iab
les.
T
r
ad
itio
n
al
ap
p
r
o
ac
h
es
to
cr
ed
it
s
co
r
in
g
:
th
ese
tech
n
iq
u
es
in
v
o
lv
e
ex
ten
s
iv
e
m
an
u
al
wo
r
k
an
d
ar
e
tim
e
-
co
n
s
u
m
in
g
.
Data
s
cien
ce
ad
d
r
ess
es
th
es
e
ch
allen
g
es
b
y
in
tr
o
d
u
cin
g
p
r
ed
ictiv
e
to
o
ls
an
d
tech
n
iq
u
es
th
at
an
aly
s
e
clien
t d
ata
to
p
er
f
o
r
m
r
is
k
ass
ess
m
en
ts
an
d
p
r
o
v
id
e
clien
ts
with
ad
ju
s
ted
cr
ed
it
s
co
r
es
b
ased
o
n
th
ei
r
c
r
ed
it h
is
to
r
y
a
n
d
o
th
e
r
d
em
o
g
r
a
p
h
ic
in
f
o
r
m
a
tio
n
.
T
h
e
n
ee
d
f
o
r
m
ac
h
in
e
lear
n
in
g
(
ML
)
in
c
r
ed
it
s
co
r
in
g
:
wit
h
th
e
in
cr
ea
s
in
g
n
u
m
b
er
o
f
i
n
d
iv
id
u
als
with
o
u
t
cr
ed
it
h
is
to
r
ies,
tr
ad
itio
n
al
cr
ed
it
s
co
r
i
n
g
m
o
d
els
ar
e
in
ad
eq
u
ate
in
e
v
alu
atin
g
t
h
e
r
is
k
f
o
r
u
n
b
an
k
e
d
o
r
n
o
n
-
b
an
k
in
g
clien
ts
.
T
h
ese
clien
ts
o
f
ten
lac
k
th
e
f
o
r
m
al
f
in
an
cial
b
ac
k
g
r
o
u
n
d
n
ec
ess
ar
y
f
o
r
co
n
v
en
tio
n
al
ass
es
s
m
en
ts
.
Ma
ch
in
e
lear
n
in
g
tech
n
iq
u
es
ad
d
r
ess
th
is
ch
allen
g
e
b
y
i
d
en
tify
in
g
co
m
p
lex
r
elatio
n
s
h
ip
s
with
in
lar
g
e
an
d
h
eter
o
g
en
eo
u
s
d
a
tasets
to
p
r
o
d
u
ce
d
ata
-
d
r
i
v
e
n
p
r
ed
ictio
n
s
.
T
h
ese
m
o
d
el
s
ca
n
in
co
r
p
o
r
ate
alter
n
ativ
e
in
f
o
r
m
atio
n
s
o
u
r
ce
s
,
in
clu
d
in
g
tr
an
s
ac
tio
n
r
ec
o
r
d
s
,
b
eh
av
i
o
u
r
al
in
d
icato
r
s
,
an
d
o
th
er
n
o
n
-
tr
ad
itio
n
al
attr
ib
u
tes,
to
ev
alu
ate
clien
t
cr
ed
itwo
r
th
in
ess
.
T
h
e
f
lex
ib
ilit
y
an
d
ad
ap
tab
ilit
y
o
f
ML
m
o
d
els
ar
e
p
ar
ticu
lar
ly
u
s
ef
u
l
in
ex
p
a
n
d
i
n
g
cr
ed
it
ac
ce
s
s
to
a
wid
er
p
o
p
u
latio
n
,
in
clu
d
in
g
th
o
s
e
wh
o
d
o
n
o
t
h
av
e
f
o
r
m
al
cr
ed
it h
is
to
r
ies
[
1
]
.
T
h
e
r
o
le
o
f
f
ea
t
u
r
e
s
elec
tio
n
in
m
ac
h
in
e
lear
n
in
g
:
f
ea
tu
r
e
s
elec
tio
n
en
h
an
ce
s
m
o
d
el
g
en
e
r
aliza
tio
n
b
y
r
e
m
o
v
in
g
ir
r
ele
v
an
t,
n
o
is
y
,
o
r
r
e
d
u
n
d
an
t
attr
ib
u
tes
wh
ile
r
etain
in
g
ea
c
h
f
ea
t
u
r
e
’
s
o
r
ig
in
al
in
f
o
r
m
atio
n
u
n
lik
e
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
f
ea
tu
r
e
s
p
ac
e
is
r
ed
u
ce
d
ac
c
o
r
d
in
g
to
a
d
e
f
in
ed
o
p
tim
izatio
n
o
b
jectiv
e
.
Mo
s
t
r
esear
ch
in
th
is
ar
ea
f
o
cu
s
es
o
n
cr
e
d
it
s
co
r
in
g
u
s
in
g
m
a
ch
in
e
lear
n
in
g
o
r
d
ee
p
lear
n
in
g
alg
o
r
ith
m
s
[
2
]
.
Featu
r
e
s
elec
tio
n
m
et
h
o
d
s
f
all
in
to
two
m
ai
n
ca
teg
o
r
ies:
i
)
Statis
tical
m
eth
o
d
s
,
wh
ich
m
ea
s
u
r
e
th
e
co
r
r
elatio
n
b
etwe
en
in
d
e
p
en
d
en
t
v
ar
iab
le
s
an
d
class
lab
els,
ty
p
ically
a
p
p
ly
in
g
least
-
s
q
u
ar
es
esti
m
atio
n
a
lo
n
g
s
id
e
s
u
b
s
et
s
elec
tio
n
,
s
h
r
in
k
ag
e
,
an
d
d
i
m
en
s
io
n
ality
r
e
d
u
ctio
n
to
i
m
p
r
o
v
e
p
r
ed
ictiv
e
p
er
f
o
r
m
a
n
ce
[
3
]
.
ii
)
W
r
ap
p
er
m
eth
o
d
s
,
wh
ich
ev
alu
ate
f
ea
tu
r
e
s
u
b
s
ets
b
y
iter
ativ
ely
ad
d
in
g
o
r
r
em
o
v
i
n
g
p
r
e
d
icto
r
s
u
s
in
g
a
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
to
s
co
r
e
p
er
f
o
r
m
a
n
ce
,
with
co
m
m
o
n
te
ch
n
iq
u
es
in
clu
d
in
g
f
o
r
war
d
s
elec
tio
n
,
b
ac
k
war
d
elim
in
atio
n
,
an
d
s
tep
wis
e
s
elec
tio
n
[
4
]
a
k
e
y
d
r
awb
ac
k
o
f
tr
ad
itio
n
al
s
tatis
tical
tech
n
iq
u
es
is
th
eir
in
ab
ilit
y
to
ca
p
tu
r
e
d
e
p
en
d
e
n
cies
am
o
n
g
f
ea
tu
r
es.
W
r
ap
p
e
r
-
b
ased
m
eth
o
d
s
ar
e
m
o
r
e
ex
p
r
ess
i
v
e
b
u
t
in
c
u
r
h
i
g
h
co
m
p
u
tatio
n
al
co
s
ts
d
u
e
to
r
ep
ea
ted
m
o
d
el
tr
ain
in
g
ac
r
o
s
s
f
ea
tu
r
e
s
u
b
s
ets
an
d
ar
e
p
r
o
n
e
to
o
v
er
f
itti
n
g
,
as
s
elec
ted
f
ea
tu
r
es m
ay
n
o
t g
en
e
r
alize
b
ey
o
n
d
th
e
tr
ai
n
in
g
d
ata.
R
ein
f
o
r
ce
m
en
t
lear
n
in
g
(
RL
)
f
o
r
f
ea
tu
r
e
s
elec
tio
n
:
RL
in
t
r
o
d
u
ce
s
a
m
o
r
e
d
y
n
am
ic
an
d
ad
ap
tiv
e
ap
p
r
o
ac
h
to
f
ea
tu
r
e
s
elec
tio
n
co
m
p
ar
ed
t
o
tr
ad
itio
n
al
m
et
h
o
d
s
.
I
n
RL
,
an
ag
en
t
im
p
r
o
v
es
it
s
b
eh
av
io
r
th
r
o
u
g
h
co
n
tin
u
o
u
s
in
ter
ac
tio
n
with
t
h
e
en
v
ir
o
n
m
e
n
t,
u
s
in
g
r
ewa
r
d
-
b
ased
f
ee
d
b
ac
k
to
r
e
f
in
e
i
ts
ac
tio
n
s
to
war
d
m
ax
im
izin
g
l
o
n
g
-
te
r
m
cu
m
u
l
ativ
e
r
etu
r
n
s
.
W
h
en
ap
p
lied
t
o
f
ea
tu
r
e
s
elec
tio
n
,
a
n
R
L
ag
en
t
ca
n
i
n
tellig
en
tly
n
av
ig
ate
th
r
o
u
g
h
th
e
f
ea
tu
r
e
s
p
ac
e,
s
elec
tin
g
th
e
m
o
s
t
in
f
o
r
m
ativ
e
f
ea
tu
r
es
b
ased
o
n
th
eir
co
n
tr
ib
u
tio
n
t
o
th
e
m
o
d
el
’
s
p
r
ed
ictiv
e
ac
c
u
r
ac
y
.
Un
lik
e
wr
ap
p
er
m
eth
o
d
s
,
wh
ich
r
ely
o
n
p
r
ed
ef
i
n
ed
r
u
les
an
d
ex
h
au
s
tiv
e
s
ea
r
ch
es,
R
L
ag
en
ts
ca
n
au
t
o
n
o
m
o
u
s
ly
a
d
ap
t
to
d
if
f
er
e
n
t
d
at
asets
an
d
co
n
ti
n
u
o
u
s
ly
im
p
r
o
v
e
f
ea
tu
r
e
s
elec
tio
n
.
T
h
is
f
lex
ib
ilit
y
m
ak
es R
L
p
ar
ticu
lar
ly
p
o
wer
f
u
l w
h
en
d
ea
lin
g
with
lar
g
e
-
s
ca
le
d
atasets
o
r
n
o
n
-
b
an
k
in
g
clien
ts
with
lim
ited
h
is
to
r
ical
d
ata
[
5
]
.
T
h
is
p
ap
er
will
f
o
c
u
s
o
n
h
o
w
to
o
p
tim
ize
th
e
cr
ed
it
s
co
r
in
g
r
esu
lts
b
y
u
s
in
g
RL
to
o
p
tim
ize
th
e
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
to
e
n
h
an
ce
th
e
r
esu
lt
o
f
cr
ed
it
s
c
o
r
e
p
r
e
d
ictio
n
;
also
,
we
will
g
o
to
u
s
e
th
e
cr
e
d
it
s
co
r
in
g
p
r
ed
ictio
n
r
esu
lt
t
o
p
r
ed
ict
th
e
d
elin
q
u
en
cy
b
e
h
av
io
u
r
f
o
r
th
is
clien
t
af
ter
u
s
in
g
th
is
cr
e
d
it
lim
it.
T
h
e
co
n
tr
ib
u
tio
n
will
h
elp
f
in
an
cial
o
r
g
an
izatio
n
s
to
m
ak
e
a
g
o
o
d
r
is
k
ass
ess
m
en
t
an
d
ad
d
v
alu
e
to
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
.
Statem
en
t
o
f
p
r
o
b
lem
,
t
h
e
r
a
p
id
g
r
o
wth
o
f
Fin
T
ec
h
an
d
t
h
e
in
cr
ea
s
in
g
av
ailab
ilit
y
o
f
l
ar
g
e
-
s
ca
le
f
in
an
cial
d
ata
h
a
v
e
s
ig
n
if
ica
n
tly
elev
ated
th
e
im
p
o
r
ta
n
c
e
o
f
cr
e
d
it
s
co
r
in
g
i
n
b
an
k
in
g
an
d
co
n
s
u
m
e
r
f
in
an
cin
g
.
T
r
a
d
itio
n
al
cr
ed
it
s
co
r
in
g
ap
p
r
o
ac
h
es,
wh
ich
r
ely
h
ea
v
ily
o
n
h
is
to
r
ical
cr
ed
it
r
ec
o
r
d
s
an
d
class
ical
s
tatis
t
ical
an
aly
s
is
,
ar
e
in
cr
ea
s
in
g
ly
in
ad
eq
u
ate
f
o
r
m
o
d
er
n
r
is
k
ass
es
s
m
en
t.
C
o
n
s
eq
u
en
tly
,
d
ata
-
d
r
i
v
en
an
d
m
ac
h
in
e
lear
n
in
g
b
ased
m
o
d
e
ls
h
av
e
b
ec
o
m
e
ess
en
tial
f
o
r
b
u
ild
in
g
r
o
b
u
s
t
d
ec
is
io
n
s
u
p
p
o
r
t
s
y
s
tem
s
th
at
ca
n
ac
cu
r
ately
ass
ess
cr
ed
itwo
r
th
in
ess
an
d
m
a
n
ag
e
f
in
an
cial
r
is
k
.
C
u
r
r
e
n
t
lim
itatio
n
s
in
c
r
ed
i
t
s
co
r
in
g
m
eth
o
d
s
ca
n
b
e
b
r
o
ad
ly
ca
teg
o
r
ize
d
in
to
b
u
s
in
ess
r
elate
d
an
d
tech
n
i
ca
l
ch
allen
g
es.
Fro
m
a
b
u
s
in
e
s
s
p
er
s
p
ec
tiv
e,
o
n
e
m
ajo
r
is
s
u
e
is
th
e
tr
ea
tm
en
t
o
f
u
n
b
a
n
k
ed
clie
n
ts
.
Mo
s
t
ex
is
tin
g
cr
ed
it
s
co
r
in
g
m
o
d
els
ar
e
tr
ain
ed
ex
clu
s
iv
ely
o
n
h
is
to
r
ical
c
r
ed
it
d
ata
f
r
o
m
b
an
k
e
d
cu
s
to
m
e
r
s
,
r
en
d
er
in
g
th
em
u
n
s
u
itab
le
f
o
r
clien
ts
with
o
u
t
p
r
io
r
cr
e
d
it
h
is
to
r
ies.
As
a
r
esu
lt,
cr
ed
it
lim
it
d
eter
m
in
atio
n
f
o
r
u
n
b
an
k
ed
in
d
iv
id
u
als
o
f
ten
d
e
p
en
d
s
o
n
m
an
u
al,
tim
e
-
co
n
s
u
m
in
g
r
is
k
ass
ess
m
en
t
p
r
o
ce
s
s
es.
I
n
co
r
p
o
r
atin
g
a
lter
n
ativ
e
d
ata
s
o
u
r
ce
s
s
u
ch
a
s
d
em
o
g
r
ap
h
ic
an
d
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
a
p
tive
fea
tu
r
e
s
elec
tio
n
fo
r
cred
it sco
r
in
g
mo
d
els
(
Mo
s
ta
fa
Mo
h
a
med
S
eifE
ln
a
s
r
)
509
s
o
cio
-
ec
o
n
o
m
ic
attr
ib
u
tes
o
f
f
er
a
p
r
ac
tical
s
o
lu
tio
n
f
o
r
au
to
m
atin
g
cr
ed
it
lim
it
ass
ig
n
m
en
t
an
d
im
p
r
o
v
in
g
d
ec
is
io
n
co
n
s
is
ten
cy
.
An
o
th
e
r
b
u
s
in
ess
-
r
elate
d
ch
allen
g
e
co
n
ce
r
n
s
d
elin
q
u
en
cy
b
e
h
av
i
o
u
r
.
W
h
ile
m
an
y
s
tu
d
ies
f
o
cu
s
o
n
p
r
ed
ictin
g
c
r
ed
it
s
co
r
es
o
r
elig
ib
ilit
y
,
th
ey
o
f
ten
n
eg
lect
th
e
lik
elih
o
o
d
o
f
d
elin
q
u
en
c
y
af
ter
cr
ed
it
is
g
r
an
ted
.
Delin
q
u
e
n
cy
p
r
e
d
ictio
n
i
s
c
r
itical,
as
it
d
ir
ec
tly
in
f
lu
e
n
ce
s
r
is
k
e
x
p
o
s
u
r
e
an
d
m
ay
alter
th
e
ass
ig
n
ed
cr
ed
it
b
u
ck
et
o
r
len
d
in
g
d
ec
is
io
n
.
Fro
m
a
tech
n
i
ca
l
p
er
s
p
ec
tiv
e,
f
ea
tu
r
e
s
elec
tio
n
r
em
ain
s
a
k
ey
ch
allen
g
e.
C
o
n
v
e
n
tio
n
al
f
ea
t
u
r
e
s
elec
tio
n
tech
n
iq
u
es
in
clu
d
in
g
f
ilter
,
wr
ap
p
er
,
a
n
d
em
b
e
d
d
ed
m
eth
o
d
s
ar
e
ty
p
ically
s
tatic
an
d
u
n
ab
le
to
ex
p
lo
r
e
th
e
f
u
ll
co
m
b
in
ato
r
ial
f
ea
tu
r
e
s
p
ac
e
ef
f
icien
tly
.
Dee
p
lear
n
in
g
-
b
ased
s
elec
tio
n
ap
p
r
o
ac
h
es,
wh
ile
p
o
wer
f
u
l,
r
e
q
u
ir
e
lar
g
e
d
ata
s
ets,
in
cu
r
h
ig
h
co
m
p
u
tatio
n
al
co
s
t,
an
d
lack
ad
ap
tab
ilit
y
to
ev
o
lv
in
g
d
ata
d
is
tr
ib
u
tio
n
s
o
r
n
ewly
in
tr
o
d
u
ce
d
f
ea
tu
r
es.
Mo
r
e
o
v
er
,
th
ese
m
eth
o
d
s
g
en
e
r
ally
f
ail
to
in
co
r
p
o
r
ate
o
n
lin
e
f
e
ed
b
ac
k
m
ec
h
an
is
m
s
,
lim
itin
g
th
eir
ef
f
ec
tiv
en
ess
in
d
y
n
a
m
ic
cr
ed
it
s
co
r
in
g
en
v
ir
o
n
m
en
ts
[
6
]
.
B
ac
k
g
r
o
u
n
d
,
C
r
ed
it
lo
a
n
s
an
d
co
n
s
u
m
e
r
f
in
a
n
cin
g
co
n
s
titu
te
a
co
r
e
co
m
p
o
n
e
n
t
o
f
th
e
b
an
k
i
n
g
in
d
u
s
tr
y
,
a
n
d
th
e
ef
f
ec
tiv
e
n
ess
o
f
cr
ed
it
r
is
k
ass
ess
m
en
t
d
ir
ec
tly
im
p
ac
ts
a
b
an
k
’
s
p
r
o
f
i
ta
b
ilit
y
an
d
f
in
an
cial
s
tab
ilit
y
.
Acc
u
r
ate
ev
alu
atio
n
o
f
a
cu
s
to
m
er
’
s
f
in
an
cial
b
ac
k
g
r
o
u
n
d
an
d
c
r
ed
itwo
r
th
in
ess
is
th
er
ef
o
r
e
ess
en
tial
p
r
io
r
to
an
y
len
d
in
g
d
ec
is
io
n
an
d
p
lay
s
a
cr
u
cial
r
o
le
in
m
i
tig
atin
g
cr
ed
it
r
is
k
[
7
]
.
I
n
ef
f
e
ctiv
e
cr
ed
it
s
co
r
in
g
p
r
ac
tice
s
co
n
tr
ib
u
te
to
a
n
in
cr
ea
s
e
in
n
o
n
-
p
er
f
o
r
m
in
g
l
o
a
n
s
(
NPLs)
an
d
d
ef
au
lt
r
ates,
p
o
s
in
g
s
ig
n
if
ican
t
ch
allen
g
es
to
b
an
k
in
g
in
s
titu
t
io
n
s
.
As
illu
s
tr
ated
in
Fig
u
r
e
1
,
q
u
a
r
ter
ly
NPL
s
tatis
tic
s
f
r
o
m
2
0
22
to
2
0
2
4
r
ef
lect
p
er
s
is
ten
t
lev
els
o
f
cr
e
d
it
d
ef
au
lt.
Alth
o
u
g
h
cr
ed
it
d
e
cisi
o
n
s
ar
e
g
en
er
ally
b
ased
o
n
av
ailab
le
f
in
a
n
cial
in
f
o
r
m
atio
n
,
ex
ter
n
al
f
ac
to
r
s
s
u
ch
as
ec
o
n
o
m
ic
v
o
latilit
y
,
in
ter
est
r
ate
f
lu
ctu
atio
n
s
,
an
d
r
eg
u
lato
r
y
o
r
tax
ch
an
g
es
ca
n
ad
v
e
r
s
ely
af
f
ec
t
r
ep
ay
m
en
t
b
eh
a
v
io
u
r
.
C
o
n
s
eq
u
en
tly
,
cr
e
d
it
r
is
k
r
e
m
ain
s
t
h
e
p
r
im
ar
y
ca
u
s
e
o
f
b
an
k
f
ailu
r
e
an
d
r
ep
r
esen
ts
o
n
e
o
f
th
e
m
o
s
t c
r
itical
r
is
k
s
f
ac
ed
b
y
b
an
k
in
g
m
a
n
ag
em
en
t.
On
th
e
o
th
er
h
an
d
,
th
e
n
ee
d
f
o
r
m
icr
o
f
i
n
an
cin
g
a
n
d
s
m
a
ll
lo
an
s
in
cr
ea
s
ed
th
ese
d
ay
s
with
h
ig
h
in
f
latio
n
s
an
d
b
ad
ec
o
n
o
m
y
s
tatu
s
an
d
m
o
s
t
p
eo
p
le
g
o
in
g
t
o
th
e
c
o
n
ce
p
t
o
f
b
u
y
n
o
w
p
ay
later
“
B
NPL
”
an
d
th
e
n
u
m
b
er
o
f
co
n
s
u
m
er
f
in
a
n
cin
g
Fin
tech
’
s
in
c
r
ea
s
ed
to
f
u
lf
il
th
ese
d
em
an
d
s
f
r
o
m
th
e
clien
ts
b
ec
au
s
e
all
clien
ts
g
o
in
g
to
in
s
talm
en
ts
p
l
an
s
in
s
tead
o
f
ca
s
h
p
ay
m
en
t.
Acc
o
r
d
in
g
to
s
tatis
tics
p
u
b
lis
h
ed
b
y
th
e
Fin
an
ce
an
d
L
ea
s
in
g
Ass
o
ciatio
n
(
FL
A)
[
8
]
,
c
o
n
s
u
m
er
f
in
a
n
ce
ac
ti
v
ity
in
cr
ea
s
ed
b
y
1
4
%
in
Au
g
u
s
t
2
0
2
2
r
elativ
e
to
Au
g
u
s
t
2
0
2
1
,
with
cu
m
u
lativ
e
g
r
o
wth
r
ea
c
h
in
g
2
1
%
o
v
er
th
e
f
ir
s
t
eig
h
t
m
o
n
th
s
o
f
2
0
2
2
co
m
p
ar
e
d
to
th
e
co
r
r
esp
o
n
d
in
g
p
er
io
d
in
th
e
p
r
ev
io
u
s
y
ea
r
(
Fig
u
r
e
2
)
.
Fig
u
r
e
1
.
NPL
r
atio
q
u
ar
ter
l
y
:
E
g
y
p
t
Fig
u
r
e
2
.
C
o
n
s
u
m
e
r
f
in
a
n
cin
g
g
r
o
wth
an
al
y
s
is
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.
4
3
,
No
.
2
,
Au
g
u
s
t
20
2
6
:
507
-
5
2
1
510
B
an
k
s
an
d
Fin
tech
’
s
n
ee
d
a
q
u
ick
way
to
ev
alu
ate
th
e
clien
t
f
r
o
m
cr
e
d
it
p
er
s
p
ec
tiv
e
to
g
i
v
e
th
em
a
p
ee
d
lo
an
with
o
u
t
tim
e
co
n
s
u
m
in
g
.
o
n
th
e
s
am
e
tim
e
th
ese
f
in
an
cial
in
s
titu
tio
n
s
n
ee
d
to
g
u
ar
an
tee
cr
ed
it
r
is
k
o
f
th
e
clien
ts
.
C
r
ed
it
s
co
r
in
g
tech
n
iq
u
es
h
av
e
b
ee
n
d
ev
elo
p
ed
to
ad
d
r
es
s
cr
ed
it
r
is
k
a
s
s
es
s
m
en
t
b
y
l
ev
er
ag
in
g
d
ata
-
d
r
iv
e
n
m
o
d
els
to
esti
m
ate
ap
p
r
o
p
r
iate
cr
e
d
it
s
co
r
es
b
a
s
ed
o
n
d
e
m
o
g
r
ap
h
ic
a
n
d
f
i
n
a
n
cial
attr
ib
u
tes
[
9
]
.
E
ar
ly
f
o
u
n
d
atio
n
al
wo
r
k
[
1
0
]
in
tr
o
d
u
ce
d
s
tatis
tical
m
ea
s
u
r
e
m
en
ts
to
d
is
tin
g
u
is
h
b
etwe
en
g
o
o
d
a
n
d
b
ad
l
o
an
s
th
r
o
u
g
h
an
al
y
s
is
o
f
ap
p
lican
t
ch
ar
ac
ter
is
tics
,
f
o
r
m
in
g
th
e
b
a
s
is
o
f
m
o
d
er
n
cr
ed
it
s
co
r
in
g
r
esear
ch
.
Sin
ce
th
en
,
cr
ed
it
s
co
r
in
g
m
eth
o
d
o
lo
g
ies
h
av
e
e
v
o
lv
ed
in
to
t
h
r
ee
p
r
in
cip
al
ca
teg
o
r
ies:
ex
p
e
r
t
-
b
ased
m
o
d
els,
s
tatis
tical
m
o
d
els,
an
d
a
r
tific
ial
in
tellig
en
ce
(
AI
)
-
b
ased
ap
p
r
o
ac
h
es.
Statis
t
ical
tech
n
iq
u
es
s
u
ch
as
lin
ea
r
d
is
cr
im
in
an
t
an
aly
s
is
(
L
DA)
an
d
lo
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
h
av
e
b
ee
n
wid
ely
ad
o
p
ted
,
with
L
DA
em
p
lo
y
in
g
lin
ea
r
d
is
cr
im
in
an
t
f
u
n
ctio
n
s
to
s
ep
ar
at
e
class
es
an
d
L
R
esti
m
atin
g
d
ef
au
lt
p
r
o
b
ab
ilit
ie
s
wh
ile
id
en
tify
in
g
b
e
h
av
i
o
u
r
-
r
elate
d
v
ar
iab
les.
Mo
r
e
r
ec
en
tly
,
AI
-
b
ased
m
eth
o
d
s
in
clu
d
in
g
ar
tific
ial
n
eu
r
al
n
etwo
r
k
s
(
ANN)
,
k
-
n
ea
r
est
n
eig
h
b
o
u
r
(
KNN)
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM)
,
an
d
g
en
etic
alg
o
r
ith
m
s
(
GA)
h
av
e
d
em
o
n
s
tr
ated
s
u
p
er
io
r
p
r
e
d
ictiv
e
p
er
f
o
r
m
a
n
ce
in
co
m
p
lex
cr
e
d
it
s
co
r
in
g
task
s
[
1
1
]
,
h
ig
h
lig
h
tin
g
th
e
g
r
o
win
g
r
o
le
o
f
ad
v
a
n
ce
d
m
ac
h
in
e
lear
n
in
g
i
n
f
in
a
n
cial
r
is
k
ass
es
s
m
en
t.
R
elate
d
wo
r
k
an
d
liter
atu
r
e
s
u
r
v
ey
,
m
a
n
y
s
tu
d
ies
h
av
e
in
v
esti
g
ated
th
e
ap
p
licat
io
n
o
f
d
ata
m
in
in
g
an
d
m
ac
h
in
e
lear
n
i
n
g
tech
n
iq
u
es
in
clu
d
in
g
s
tatis
tical
m
o
d
els
an
d
n
eu
r
al
n
etwo
r
k
s
in
b
u
s
in
ess
an
d
f
in
an
cial
d
ec
is
io
n
-
m
ak
in
g
.
Mo
s
t
o
f
th
is
wo
r
k
f
o
cu
s
es
o
n
p
r
ed
ictin
g
cr
ed
it
s
co
r
es
o
r
cr
ed
it
elig
ib
ilit
y
u
s
in
g
h
is
to
r
ical
d
ata
f
r
o
m
clien
ts
wi
th
estab
lis
h
ed
cr
ed
it
h
is
to
r
ies.
Featu
r
e
s
elec
tio
n
h
as
b
ee
n
ad
d
r
ess
ed
u
s
in
g
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
o
r
d
ee
p
lea
r
n
in
g
ap
p
r
o
ac
h
es
to
im
p
r
o
v
e
p
r
ed
ictiv
e
ac
cu
r
ac
y
;
h
o
wev
e
r
,
alter
n
ativ
e
f
ea
tu
r
e
s
elec
tio
n
s
tr
ateg
ies
b
ey
o
n
d
s
tep
wis
e
o
r
d
ee
p
lear
n
i
n
g
-
b
ase
d
m
e
th
o
d
s
r
em
ain
u
n
d
er
e
x
p
l
o
r
ed
.
Fu
r
th
er
m
o
r
e,
th
e
m
ajo
r
ity
o
f
ex
is
tin
g
r
esear
ch
o
v
er
lo
o
k
s
u
n
b
an
k
ed
clien
ts
wh
o
lack
p
r
io
r
cr
ed
it
r
ec
o
r
d
s
,
with
o
n
ly
lim
ited
s
tu
d
ies
in
co
r
p
o
r
atin
g
alter
n
ativ
e
d
ata
s
o
u
r
ce
s
.
On
e
n
o
tab
le
e
x
ce
p
tio
n
le
v
er
ag
es
m
o
b
ile
an
d
d
em
o
g
r
ap
h
ic
d
ata
to
en
h
an
ce
cr
ed
it sco
r
in
g
p
er
f
o
r
m
an
ce
f
o
r
clien
ts
with
o
u
t c
r
ed
it h
is
to
r
y
.
R
ec
en
t
s
tu
d
ies
h
av
e
also
ex
am
in
ed
RL
f
o
r
im
p
r
o
v
in
g
cr
ed
i
t
s
co
r
in
g
d
ec
is
io
n
s
.
Her
asy
m
o
v
y
ch
et
a
l.
[
1
2
]
p
r
o
p
o
s
ed
a
RL
f
r
a
m
ewo
r
k
to
d
y
n
am
ically
o
p
tim
ize
a
cc
ep
tan
ce
th
r
esh
o
l
d
s
in
co
n
s
u
m
er
cr
ed
it
s
co
r
in
g
,
ad
d
r
ess
in
g
ch
allen
g
es
s
u
ch
a
s
s
elec
tio
n
b
ias,
p
o
p
u
latio
n
d
r
if
t,
an
d
m
is
alig
n
ed
b
u
s
in
ess
o
b
jectiv
es.
Usi
n
g
h
is
to
r
ical
d
ata
f
r
o
m
an
in
ter
n
a
tio
n
al
cr
ed
it
co
m
p
an
y
,
th
e
au
t
h
o
r
s
co
m
b
in
ed
RL
with
M
o
n
t
e
C
ar
lo
s
im
u
latio
n
s
an
d
g
au
s
s
ian
r
a
d
ial
b
asis
f
u
n
ctio
n
s
f
o
r
v
al
u
e
a
p
p
r
o
x
im
atio
n
.
T
h
eir
r
esu
lts
d
em
o
n
s
tr
ated
th
at
s
tatic,
co
s
t
-
s
en
s
itiv
e
th
r
esh
o
ld
o
p
tim
izati
o
n
ap
p
r
o
ac
h
es
ar
e
in
s
u
f
f
icien
t
an
d
m
ay
lea
d
to
s
ig
n
i
f
ican
t
f
in
an
cial
lo
s
s
es,
w
h
er
ea
s
th
e
p
r
o
p
o
s
ed
ad
ap
tiv
e
s
y
s
tem
ac
h
iev
e
d
s
u
b
s
tan
tially
h
i
g
h
er
p
r
o
f
itab
ilit
y
in
b
o
th
s
im
u
lated
an
d
r
ea
l
-
wo
r
ld
s
ettin
g
s
.
T
h
e
s
tu
d
y
was
lim
ited
to
clien
ts
with
e
x
is
tin
g
cr
ed
it
h
is
to
r
ies
an
d
f
o
cu
s
ed
s
o
lely
o
n
t
h
r
esh
o
ld
o
p
tim
izatio
n
,
with
o
u
t a
d
d
r
ess
i
n
g
f
ea
t
u
r
e
s
elec
tio
n
f
o
r
im
p
r
o
v
in
g
th
e
u
n
d
e
r
ly
in
g
c
r
ed
it sco
r
in
g
m
o
d
el.
A
s
ig
n
if
ican
t
s
tu
d
y
o
n
clien
t
ev
alu
atio
n
in
c
r
ed
it
s
co
r
in
g
is
p
r
esen
ted
in
[
1
3
]
,
wh
e
r
e
t
h
e
au
th
o
r
s
in
v
esti
g
ate
o
p
tim
al
co
m
b
in
atio
n
s
o
f
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
an
d
class
if
icatio
n
alg
o
r
ith
m
s
f
o
r
p
r
ed
ictin
g
cr
ed
it
s
co
r
es.
T
h
e
s
tu
d
y
ev
alu
ates
f
ilter
,
wr
ap
p
er
,
an
d
em
b
ed
d
ed
f
ea
tu
r
e
s
elec
tio
n
tech
n
iq
u
e
s
p
ec
if
ically
f
o
r
war
d
s
elec
tio
n
an
d
b
ac
k
wa
r
d
elim
in
atio
n
to
g
eth
er
with
m
u
ltip
le
class
if
ier
s
,
in
clu
d
in
g
d
ec
is
io
n
tr
ee
s
(
DT
)
,
r
an
d
o
m
f
o
r
est
(
R
F)
,
Naïv
e
B
ay
es,
LR
,
an
d
KNN
s
.
E
x
p
er
i
m
en
ts
co
n
d
u
cted
o
n
a
d
ataset
o
f
9
1
,
7
5
9
r
ec
o
r
d
s
with
2
7
2
f
ea
tu
r
es
(
7
0
/3
0
tr
ain
–
test
s
p
lit)
s
h
o
wed
th
at
co
r
r
elatio
n
-
b
ased
f
ilter
s
elec
tio
n
ac
h
iev
ed
t
h
e
b
est
f
ea
tu
r
e
s
elec
tio
n
p
er
f
o
r
m
an
ce
,
wh
ile
RF
y
ield
ed
th
e
h
ig
h
est
p
r
ed
ictiv
e
ac
cu
r
ac
y
.
Alth
o
u
g
h
th
e
s
tu
d
y
p
r
o
v
i
d
es
a
co
m
p
r
e
h
en
s
iv
e
co
m
p
ar
is
o
n
o
f
f
ea
tu
r
e
s
elec
tio
n
an
d
cla
s
s
if
icatio
n
m
eth
o
d
s
,
it
is
lim
ited
to
clien
ts
with
ex
is
tin
g
cr
ed
it
h
is
to
r
ies,
d
o
e
s
n
o
t
ex
p
lo
r
e
ex
h
a
u
s
tiv
e
f
e
atu
r
e
co
m
b
i
n
atio
n
s
,
an
d
lac
k
s
m
ec
h
an
is
m
s
f
o
r
in
co
r
p
o
r
atin
g
r
ea
l
-
tim
e
f
ee
d
b
ac
k
o
r
ad
a
p
tin
g
to
n
ewly
in
tr
o
d
u
ce
d
f
ea
tu
r
es.
Ku
m
ar
et
a
l
.
[
1
4
]
,
th
e
au
th
o
r
s
p
r
o
p
o
s
e
a
h
y
b
r
id
c
r
ed
it
s
co
r
i
n
g
f
r
am
ewo
r
k
c
o
m
b
in
in
g
d
e
ep
lear
n
i
n
g
a
n
d
K
-
m
ea
n
s
cl
u
s
ter
in
g
.
T
h
e
m
o
d
el
f
o
llo
ws a
th
r
ee
-
s
tag
e
p
ip
eli
n
e
co
n
s
is
tin
g
o
f
d
ata
p
r
ep
r
o
ce
s
s
in
g
,
d
im
e
n
s
io
n
ality
r
ed
u
ctio
n
v
ia
f
ea
tu
r
e
s
elec
tio
n
,
an
d
d
ee
p
n
e
u
r
al
n
etwo
r
k
–
b
a
s
ed
p
r
ed
ictio
n
.
C
r
ed
it
s
co
r
in
g
is
ca
teg
o
r
ized
in
to
b
eh
av
i
o
u
r
al
s
co
r
in
g
an
d
co
llectio
n
s
co
r
in
g
,
with
K
-
m
e
an
s
u
s
ed
to
clu
s
ter
cu
s
to
m
er
s
b
ased
o
n
b
e
h
av
io
u
r
al
p
atter
n
s
.
Usi
n
g
th
e
Kag
g
le
Ho
m
e
C
r
ed
it
Def
au
lt
R
i
s
k
d
ataset,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
a
ch
iev
ed
ap
p
r
o
x
im
ately
8
7
%
ac
cu
r
ac
y
.
Desp
ite
its
s
tr
o
n
g
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
an
d
e
x
ten
s
iv
e
u
s
e
o
f
e
v
alu
atio
n
m
et
r
ics
in
clu
d
in
g
ac
cu
r
ac
y
,
R
MSE
,
p
r
o
b
a
b
ilit
y
o
f
d
e
f
au
lt,
a
n
d
AUC
th
e
f
r
am
ewo
r
k
r
em
ain
s
c
o
n
s
tr
ain
ed
t
o
b
a
n
k
ed
clien
ts
,
d
o
es
n
o
t
ex
p
lo
r
e
all
p
o
s
s
ib
le
f
ea
tu
r
e
s
u
b
s
ets,
an
d
l
ac
k
s
ad
ap
tab
ilit
y
to
r
ea
l
-
tim
e
d
ata
ch
an
g
es.A
co
m
p
ar
ativ
e
d
ee
p
lear
n
in
g
s
tu
d
y
is
p
r
esen
ted
in
[
1
5
]
,
w
h
er
e
d
ee
p
s
eq
u
e
n
tial
n
eu
r
al
n
etwo
r
k
s
(
DSNN)
an
d
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NN)
ar
e
ev
alu
ated
a
g
ain
s
t
tr
ad
itio
n
al
class
if
ier
s
s
u
ch
as
K
NN,
class
if
icat
io
n
an
d
r
eg
r
e
s
s
io
n
tr
ee
(
C
AR
T
)
,
Naïv
e
B
ay
es,
an
d
SVM
.
E
x
p
er
im
en
ts
wer
e
co
n
d
u
cted
u
s
in
g
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
o
n
th
r
ee
d
atasets
:
Ger
m
an
C
r
ed
it
Data
,
Au
s
tr
a
lian
C
r
ed
it
Ap
p
r
o
v
al,
an
d
Kag
g
le
C
r
ed
it
Data
.
T
h
e
d
e
ep
lear
n
in
g
m
o
d
els
co
n
s
is
ten
tly
o
u
tp
er
f
o
r
m
ed
tr
a
d
itio
n
al
class
if
ier
s
,
ac
h
iev
in
g
ac
cu
r
ac
ies
o
f
u
p
to
9
3
.
6
3
%
o
n
th
e
Kag
g
le
d
ataset.
W
h
ile
th
e
s
tu
d
y
d
em
o
n
s
tr
ates
th
e
ef
f
ec
tiv
e
n
ess
o
f
d
ee
p
lea
r
n
in
g
i
n
cr
e
d
it
s
co
r
in
g
ac
r
o
s
s
m
u
ltip
le
co
u
n
tr
ies,
its
r
elian
ce
o
n
lar
g
e
d
atasets
li
m
its
g
en
er
aliza
tio
n
to
s
m
aller
d
atasets
,
an
d
it
d
o
es
n
o
t
ad
d
r
e
s
s
clas
s
im
b
alan
ce
or
in
teg
r
ate
f
ea
tu
r
e
s
elec
tio
n
with
tr
ad
itio
n
al
class
if
ier
s
.
An
alter
n
ativ
e
d
ir
ec
tio
n
f
o
r
cr
ed
it
s
co
r
in
g
in
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
a
p
tive
fea
tu
r
e
s
elec
tio
n
fo
r
cred
it sco
r
in
g
mo
d
els
(
Mo
s
ta
fa
Mo
h
a
med
S
eifE
ln
a
s
r
)
511
u
n
b
an
k
ed
p
o
p
u
latio
n
s
is
in
tr
o
d
u
ce
d
in
Mo
b
iSco
r
e
[
1
6
]
,
wh
i
ch
lev
er
ag
es
m
o
b
ile
p
h
o
n
e
d
at
a
as
a
s
u
b
s
titu
te
f
o
r
co
n
v
en
tio
n
al
f
in
an
cial
r
ec
o
r
d
s
.
T
h
e
s
tu
d
y
ex
tr
ac
ts
co
n
s
u
m
p
tio
n
,
m
o
b
ilit
y
,
a
n
d
s
o
cial
n
et
wo
r
k
f
ea
tu
r
es
f
r
o
m
ca
ll
d
etail
r
ec
o
r
d
s
an
d
ev
alu
a
tes
m
u
ltip
le
clas
s
if
ier
s
,
in
clu
d
in
g
LR
,
SVM,
an
d
g
r
ad
ien
t
b
o
o
s
ted
tr
ee
s
,
u
s
in
g
f
iv
e
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
.
R
esu
lts
s
h
o
w
th
at
in
co
r
p
o
r
atin
g
m
o
b
ile
d
ata
im
p
r
o
v
es
p
r
ed
i
ctio
n
ac
cu
r
ac
y
to
7
1
.
5
%,
co
m
p
ar
ed
with
5
7
.
7
%
wh
en
u
s
in
g
tr
ad
itio
n
al
cr
ed
it
b
u
r
ea
u
d
ata
alo
n
e.
Alth
o
u
g
h
th
is
ap
p
r
o
ac
h
s
ig
n
if
ican
tly
ad
v
an
ce
s
cr
ed
it
s
co
r
in
g
f
o
r
u
n
b
an
k
ed
clien
ts
,
it
d
o
es
n
o
t
em
p
lo
y
f
ea
tu
r
e
s
elec
tio
n
tech
n
iq
u
es
an
d
lack
s
a
clea
r
c
o
m
p
ar
ati
v
e
r
esu
lts
tab
le,
lim
itin
g
in
ter
p
r
et
ab
ilit
y
an
d
o
p
tim
izatio
n
.
Gr
o
u
p
-
b
ase
d
f
ea
tu
r
e
s
e
le
cti
o
n
f
o
r
c
r
e
d
it
s
co
r
i
n
g
is
ex
p
l
o
r
e
d
i
n
[
1
7
]
,
wh
er
e
Gr
o
u
p
L
ass
o
r
e
g
r
ess
i
o
n
is
p
r
o
p
o
s
e
d
as
a
n
a
lte
r
n
a
ti
v
e
t
o
s
t
ep
w
is
e
s
e
lec
ti
o
n
m
e
th
o
d
s
.
Usi
n
g
a
UC
I
cr
e
d
it
d
at
ase
t w
it
h
1
,
0
0
0
r
e
co
r
d
s
a
n
d
2
1
f
e
at
u
r
es,
th
e
a
u
t
h
o
r
s
c
o
m
p
ar
e
Gr
o
u
p
L
ass
o
w
it
h
LR
m
o
d
els
o
p
ti
m
i
ze
d
v
ia
AI
C
,
B
I
C
,
cr
o
s
s
-
v
ali
d
ati
o
n
e
r
r
o
r
,
an
d
b
a
ck
wa
r
d
el
im
i
n
ati
o
n
.
T
h
e
r
esu
lts
d
e
m
o
n
s
tr
ate
c
o
m
p
a
r
a
b
le
ac
c
u
r
ac
i
es
a
cr
o
s
s
m
e
th
o
d
s
,
wit
h
Gr
o
u
p
L
ass
o
o
f
f
er
i
n
g
i
m
p
r
o
v
e
d
in
te
r
p
r
et
a
b
i
lit
y
th
r
o
u
g
h
g
r
o
u
p
e
d
v
a
r
i
a
b
le
s
ele
cti
o
n
.
H
o
w
ev
er
,
t
h
e
a
n
a
l
y
s
is
is
l
im
i
te
d
b
y
t
h
e
r
el
ati
v
e
ly
s
m
all
d
at
aset
,
th
e
u
s
e
o
f
a
r
estr
ict
e
d
s
e
t
o
f
c
lass
i
f
i
er
s
,
a
n
d
in
s
u
f
f
ici
e
n
t
p
r
ese
n
t
ati
o
n
o
f
e
x
p
e
r
i
m
e
n
ta
l
r
es
u
lts
.
R
aso
u
l
et
a
l.
[1
5
]
,
t
h
e
au
t
h
o
r
s
f
u
r
t
h
er
e
x
p
l
o
r
e
RL
b
as
ed
f
ea
t
u
r
e
s
el
ec
t
io
n
b
y
m
o
d
el
li
n
g
t
h
e
p
r
o
b
le
m
as
a
m
a
r
k
o
v
d
ec
is
i
o
n
p
r
o
ce
s
s
(
M
DP)
a
n
d
e
m
p
lo
y
i
n
g
a
t
em
p
o
r
al
d
i
f
f
e
r
e
n
c
e
(
T
D
)
le
ar
n
i
n
g
s
t
r
at
e
g
y
.
T
h
e
s
ele
ct
ed
f
e
at
u
r
e
s
u
b
s
ets
we
r
e
e
v
al
u
ate
d
u
s
i
n
g
SVM
a
cr
o
s
s
t
h
r
ee
d
ata
s
ets:
A
u
s
tr
ali
an
C
r
ed
it,
B
r
ea
s
t
C
an
ce
r
W
is
co
n
s
i
n
(
Pr
o
g
n
o
s
tic
)
,
a
n
d
C
o
n
n
e
cti
o
n
i
s
t
B
e
n
c
h
.
T
h
e
r
ep
o
r
te
d
p
er
f
o
r
m
a
n
c
e
r
es
u
lts
d
em
o
n
s
t
r
at
e
t
h
e
ef
f
e
cti
v
en
ess
o
f
RL
in
a
d
a
p
t
iv
el
y
id
e
n
ti
f
y
i
n
g
i
n
f
o
r
m
ati
v
e
f
e
at
u
r
e
s
u
b
s
e
ts
N
e
v
e
r
th
e
less
,
th
e
s
t
u
d
y
la
c
k
s
b
e
n
ch
m
a
r
k
co
m
p
a
r
is
o
n
s
wit
h
tr
a
d
iti
o
n
a
l
f
ea
t
u
r
e
s
e
le
ct
io
n
m
et
h
o
d
s
,
p
r
o
v
i
d
es
l
im
i
te
d
d
at
aset
an
d
f
e
at
u
r
e
d
esc
r
i
p
t
io
n
s
,
an
d
ev
al
u
at
es
p
e
r
f
o
r
m
a
n
c
e
u
s
i
n
g
a
s
i
n
g
l
e
c
l
ass
if
ie
r
,
r
est
r
ict
in
g
th
e
g
e
n
e
r
al
iza
b
il
it
y
o
f
t
h
e
f
i
n
d
i
n
g
s
.
T
a
b
l
e
1
S
u
m
m
ar
izes
t
h
e
r
e
v
i
ewe
d
s
tu
d
i
es
ac
r
o
s
s
th
ei
r
c
o
r
e
a
p
p
r
o
a
ch
,
d
atas
et
,
t
ar
g
et
p
o
p
u
lati
o
n
,
f
e
at
u
r
e
s
ele
cti
o
n
s
tr
ate
g
y
,
c
o
n
t
r
i
b
u
ti
o
n
s
,
an
d
li
m
it
ati
o
n
s
.
T
ab
le
1
.
L
iter
atu
r
e
s
u
r
v
e
y
s
u
m
m
ar
y
Ref
C
o
r
e
a
p
p
r
o
a
c
h
D
a
t
a
s
e
t
t
y
p
e
Ta
r
g
e
t
p
o
p
u
l
a
t
i
o
n
F
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
st
r
a
t
e
g
y
K
e
y
c
o
n
t
r
i
b
u
t
i
o
n
I
d
e
n
t
i
f
i
e
d
l
i
m
i
t
a
t
i
o
n
H
e
r
a
sy
m
o
v
y
c
h
e
t
a
l
.
[
1
2
]
R
L
f
o
r
d
y
n
a
m
i
c
a
c
c
e
p
t
a
n
c
e
t
h
r
e
s
h
o
l
d
o
p
t
i
m
i
z
a
t
i
o
n
H
i
st
o
r
i
c
a
l
c
o
n
su
mer
c
r
e
d
i
t
d
a
t
a
B
a
n
k
e
d
c
l
i
e
n
t
s
N
o
t
f
e
a
t
u
r
e
-
f
o
c
u
se
d
A
d
a
p
t
i
v
e
t
h
r
e
s
h
o
l
d
s;
h
i
g
h
e
r
p
r
o
f
i
t
a
b
i
l
i
t
y
N
o
a
d
a
p
t
i
v
e
sel
e
c
t
i
o
n
;
l
i
mi
t
e
d
t
o
b
a
n
k
e
d
c
l
i
e
n
t
s
Zi
e
m
b
a
e
t
a
l
.
[
1
3
]
R
L
f
o
r
d
y
n
a
m
i
c
a
c
c
e
p
t
a
n
c
e
t
h
r
e
s
h
o
l
d
o
p
t
i
m
i
z
a
t
i
o
n
9
1
,
7
5
9
r
e
c
o
r
d
s,
2
7
2
f
e
a
t
u
r
e
s
B
a
n
k
e
d
c
l
i
e
n
t
s
F
o
r
w
a
r
d
,
B
a
c
k
w
a
r
d
,
C
o
r
r
e
l
a
t
i
o
n
f
i
l
t
e
r
C
o
m
p
r
e
h
e
n
si
v
e
F
S
a
n
d
c
l
a
ss
i
f
i
e
r
c
o
m
p
a
r
i
so
n
N
o
a
d
a
p
t
i
v
e
sel
e
c
t
i
o
n
;
n
o
r
e
a
l
-
t
i
m
e
f
e
e
d
b
a
c
k
;
n
o
u
n
b
a
n
k
e
d
f
o
c
u
s
K
u
mar
e
t
a
l
.
[
1
4
]
H
y
b
r
i
d
D
e
e
p
Le
a
r
n
i
n
g
+
K
-
M
e
a
n
s
c
l
u
st
e
r
i
n
g
K
a
g
g
l
e
H
o
me
C
r
e
d
i
t
d
a
t
a
se
t
B
a
n
k
e
d
c
l
i
e
n
t
s
D
i
me
n
si
o
n
a
l
i
t
y
r
e
d
u
c
t
i
o
n
+
D
L
H
i
g
h
p
r
e
d
i
c
t
i
v
e
a
c
c
u
r
a
c
y
(
~
8
7
%)
N
o
e
x
h
a
u
st
i
v
e
F
S
;
n
o
a
d
a
p
t
i
v
e
mec
h
a
n
i
sm
;
b
a
n
k
e
d
o
n
l
y
R
a
s
o
u
l
e
t
a
l
.
[
1
5
]
H
y
b
r
i
d
D
e
e
p
Le
a
r
n
i
n
g
+
K
-
M
e
a
n
s
c
l
u
st
e
r
i
n
g
G
e
r
man
,
A
u
st
r
a
l
i
a
n
,
K
a
g
g
l
e
d
a
t
a
se
t
s
B
a
n
k
e
d
c
l
i
e
n
t
s
N
o
e
x
p
l
i
c
i
t
F
S
i
n
t
e
g
r
a
t
i
o
n
S
t
r
o
n
g
d
e
e
p
l
e
a
r
n
i
n
g
p
e
r
f
o
r
m
a
n
c
e
(
u
p
t
o
9
3
%)
R
e
q
u
i
r
e
s
l
a
r
g
e
d
a
t
a
se
t
s;
n
o
i
mb
a
l
a
n
c
e
h
a
n
d
l
i
n
g
;
n
o
F
S
i
n
t
e
g
r
a
t
i
o
n
P
e
d
r
o
e
t
a
l
.
[
1
6
]
D
e
e
p
S
e
q
u
e
n
t
i
a
l
NN
a
n
d
C
N
N
v
s
t
r
a
d
i
t
i
o
n
a
l
M
L
M
o
b
i
l
e
+
f
i
n
a
n
c
i
a
l
r
e
c
o
r
d
s
U
n
B
a
n
k
e
d
c
l
i
e
n
t
s
N
o
e
x
p
l
i
c
i
t
F
S
i
n
t
e
g
r
a
t
i
o
n
A
l
t
e
r
n
a
t
i
v
e
d
a
t
a
f
o
r
u
n
b
a
n
k
e
d
sco
r
i
n
g
N
o
f
e
a
t
u
r
e
o
p
t
i
m
i
z
a
t
i
o
n
;
l
i
m
i
t
e
d
i
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
C
h
e
n
a
n
X
i
a
n
g
[
1
7
]
M
o
b
i
S
c
o
r
e
(
mo
b
i
l
e
p
h
o
n
e
d
a
t
a
-
b
a
se
d
c
r
e
d
i
t
sco
r
i
n
g
)
U
C
I
d
a
t
a
set
(
1
,
0
0
0
r
e
c
o
r
d
s)
B
a
n
k
e
d
c
l
i
e
n
t
s
Emb
e
d
d
e
d
(
G
r
o
u
p
La
sso
)
I
mp
r
o
v
e
d
g
r
o
u
p
e
d
f
e
a
t
u
r
e
i
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
S
mal
l
d
a
t
a
se
t
;
l
i
m
i
t
e
d
c
l
a
ssi
f
i
e
r
s;
n
o
a
d
a
p
t
i
v
e
e
x
p
l
o
r
a
t
i
o
n
R
a
s
o
u
l
e
t
a
l
.
[
15
]
RL
-
b
a
se
d
f
e
a
t
u
r
e
sel
e
c
t
i
o
n
(
TD
l
e
a
r
n
i
n
g
)
M
u
l
t
i
p
l
e
st
r
u
c
t
u
r
e
d
d
a
t
a
se
t
s
G
e
n
e
r
a
l
RL
-
b
a
se
d
I
mp
r
o
v
e
d
g
r
o
u
p
e
d
f
e
a
t
u
r
e
i
n
t
e
r
p
r
e
t
a
b
i
l
i
t
y
N
o
b
e
n
c
h
mar
k
c
o
m
p
a
r
i
so
n
;
si
n
g
l
e
c
l
a
ss
i
f
i
e
r
;
l
i
m
i
t
e
d
d
a
t
a
se
t
d
e
t
a
i
l
s
R
esear
ch
g
ap
,
d
esp
ite
t
h
e
g
r
o
win
g
ad
o
p
tio
n
o
f
m
a
ch
in
e
le
ar
n
in
g
a
n
d
d
ee
p
lear
n
in
g
tec
h
n
iq
u
es
in
cr
ed
it
s
co
r
in
g
,
s
ev
er
al
lim
ita
tio
n
s
r
em
ain
u
n
r
eso
lv
ed
.
Fir
s
t,
m
o
s
t
ex
is
tin
g
s
tu
d
ies
f
o
cu
s
o
n
o
p
tim
izin
g
p
r
ed
ictiv
e
m
o
d
els
u
s
in
g
s
tatic
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
,
s
u
ch
as
f
ilter
,
wr
ap
p
er
,
o
r
e
m
b
ed
d
ed
ap
p
r
o
ac
h
es,
wh
ich
r
ely
o
n
p
r
e
d
ef
in
ed
ev
al
u
atio
n
cr
iter
ia
an
d
lack
ad
a
p
ta
b
ilit
y
to
ev
o
lv
in
g
d
ata
d
is
tr
ib
u
tio
n
s
.
Seco
n
d
,
p
r
io
r
RL
ap
p
licatio
n
s
in
cr
e
d
it
r
is
k
m
o
d
ellin
g
h
av
e
p
r
im
ar
ily
c
o
n
ce
n
tr
ated
o
n
th
r
esh
o
l
d
o
p
tim
izatio
n
o
r
p
o
licy
d
ec
is
io
n
s
,
r
ath
er
th
an
ad
ap
tiv
e
f
ea
tu
r
e
s
u
b
s
et
d
is
co
v
er
y
.
T
h
ir
d
,
ex
is
tin
g
R
L
-
b
ased
f
ea
tu
r
e
s
elec
tio
n
s
tu
d
ies ar
e
g
en
er
ally
ev
alu
ate
d
o
n
lim
ited
b
in
ar
y
d
atasets
an
d
d
o
n
o
t
ad
d
r
ess
m
u
lti
-
class
cr
ed
it
b
u
ck
et
p
r
ed
ictio
n
co
m
b
in
ed
with
d
o
wn
s
tr
ea
m
d
elin
q
u
en
c
y
m
o
d
ell
in
g
.
Fu
r
t
h
er
m
o
r
e
,
th
e
s
ca
lab
ilit
y
an
d
s
tab
ilit
y
o
f
RL
f
o
r
s
tr
u
ctu
r
ed
f
in
a
n
cial
d
atasets
r
em
ain
in
s
u
f
f
icien
tly
e
x
am
in
ed
.
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.
4
3
,
No
.
2
,
Au
g
u
s
t
20
2
6
:
507
-
5
2
1
512
C
o
n
s
eq
u
en
tly
,
th
er
e
is
a
cle
ar
n
ee
d
f
o
r
a
r
ig
o
r
o
u
s
ly
v
al
id
ated
f
r
am
ewo
r
k
th
at
ap
p
lies
RL
f
o
r
ad
ap
tiv
e
f
ea
tu
r
e
s
elec
tio
n
in
m
u
lti
-
class
c
r
ed
it
s
co
r
in
g
wh
i
le
ex
p
licitly
e
v
alu
atin
g
its
im
p
ac
t
o
n
d
elin
q
u
en
c
y
p
r
ed
ictio
n
,
c
o
m
p
u
tatio
n
al
ef
f
i
cien
cy
,
an
d
s
u
b
s
et
s
tab
ilit
y
u
n
d
er
co
n
t
r
o
lled
ex
p
er
im
en
tal
s
e
ttin
g
s
.
C
o
n
tr
ib
u
tio
n
o
f
th
is
Stu
d
y
,
i
n
lig
h
t
o
f
t
h
e
lim
itatio
n
s
id
e
n
tifie
d
in
p
r
io
r
r
esear
c
h
,
t
h
i
s
s
tu
d
y
m
a
k
es
th
e
f
o
llo
win
g
c
o
n
tr
ib
u
tio
n
s
:
−
Un
if
ied
R
L
-
b
ased
f
r
am
ewo
r
k
f
o
r
cr
e
d
it
s
co
r
in
g
an
d
d
elin
q
u
en
cy
p
r
ed
ictio
n
:
t
h
is
s
tu
d
y
is
am
o
n
g
th
e
f
ir
s
t
to
u
s
e
RL
–
b
ased
ad
ap
tiv
e
f
e
atu
r
e
s
elec
tio
n
in
a
s
in
g
le
f
r
am
ewo
r
k
th
at
c
o
v
er
s
b
o
th
cr
ed
it
s
co
r
in
g
an
d
d
elin
q
u
e
n
c
y
p
r
e
d
ictio
n
,
m
ak
in
g
it m
o
r
e
r
elev
an
t f
o
r
p
r
ac
tical
f
in
an
cial
r
is
k
ass
ess
m
en
t.
−
C
o
m
p
ar
is
o
n
with
c
o
n
v
e
n
tio
n
al
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
:
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
is
c
o
m
p
ar
ed
with
wr
ap
p
er
,
f
ilter
,
an
d
em
b
ed
d
e
d
f
ea
tu
r
e
s
elec
tio
n
tech
n
iq
u
es
u
n
d
er
th
e
s
am
e
p
r
ep
r
o
ce
s
s
in
g
an
d
ev
alu
atio
n
s
ettin
g
s
,
p
r
o
v
id
in
g
a
f
air
ass
ess
m
en
t o
f
its
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
an
d
co
m
p
u
tatio
n
al
co
s
t
.
−
Ad
ap
tiv
e
f
ea
tu
r
e
s
elec
tio
n
in
s
tead
o
f
s
tatic
s
elec
tio
n
:
t
h
e
p
r
o
p
o
s
ed
m
eth
o
d
tr
ea
ts
f
ea
tu
r
e
s
elec
tio
n
as
a
s
eq
u
en
tial
d
ec
is
io
n
p
r
o
ce
s
s
,
allo
win
g
th
e
m
o
d
el
to
ex
p
l
o
r
e
f
ea
tu
r
e
s
u
b
s
ets
d
y
n
am
ic
ally
r
ath
er
th
an
d
ep
en
d
i
n
g
o
n
ly
o
n
s
tatic
f
ilter
,
wr
ap
p
er
,
o
r
em
b
ed
d
e
d
m
eth
o
d
s
.
−
E
v
alu
atio
n
o
n
m
u
lti
-
class
cr
ed
it
p
r
o
b
lem
s
:
t
h
e
f
r
am
ewo
r
k
is
test
ed
o
n
cr
e
d
it
b
u
c
k
et
p
r
ed
ictio
n
an
d
d
elin
q
u
en
c
y
p
r
e
d
ictio
n
task
s
,
wh
ich
m
ak
es
it
s
u
itab
le
f
o
r
m
u
lti
-
class
cr
ed
it
r
is
k
m
o
d
ellin
g
an
d
m
o
r
e
r
ea
lis
tic
f
in
an
cial
d
atasets
.
−
Stab
ilit
y
an
d
r
o
b
u
s
tn
ess
an
aly
s
is
:
t
h
e
co
n
s
is
ten
cy
o
f
th
e
s
elec
ted
f
ea
tu
r
es
is
ex
am
in
e
d
th
r
o
u
g
h
r
e
p
ea
ted
r
u
n
s
an
d
s
im
ilar
ity
a
n
aly
s
is
,
p
r
o
v
id
in
g
ad
d
itio
n
al
ev
id
en
ce
o
f
r
o
b
u
s
tn
ess
an
d
in
ter
p
r
eta
b
ilit
y
.
−
Fair
an
d
r
ep
r
o
d
u
cib
le
ev
al
u
atio
n
:
t
h
e
s
tu
d
y
ap
p
lies
s
tr
atif
ied
cr
o
s
s
-
v
alid
atio
n
,
c
lass
im
b
alan
ce
h
an
d
lin
g
,
an
d
th
e
s
am
e
p
r
e
p
r
o
ce
s
s
in
g
s
tep
s
ac
r
o
s
s
all
b
a
s
elin
e
m
eth
o
d
s
to
en
s
u
r
e
a
f
air
an
d
r
eliab
le
c
o
m
p
ar
is
o
n
.
T
h
e
r
em
ain
d
er
o
f
th
is
p
ap
er
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws.
I
n
s
e
ctio
n
2
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
RL
-
b
ased
f
ea
tu
r
e
s
elec
tio
n
f
r
am
ewo
r
k
.
I
n
s
ec
tio
n
3
d
escr
ib
es
t
h
e
Me
th
o
d
in
clu
d
in
g
d
ataset
d
escr
ip
tio
n
,
d
ata
p
r
ep
r
o
ce
s
s
in
g
,
ev
alu
atio
n
s
et
u
p
,
a
n
d
ev
alu
atio
n
m
atr
i
x
w
h
ile
s
ec
tio
n
4
r
e
p
o
r
ts
th
e
r
es
u
lts
an
d
d
is
cu
s
s
io
n
.
I
n
s
ec
tio
n
5
c
o
n
clu
d
es
an
d
o
u
t
lin
es f
u
tu
r
e
r
esear
ch
d
ir
ec
tio
n
s
.
2.
P
RO
P
O
SE
D
RE
I
NF
O
RC
E
M
E
N
T
L
E
AR
NING
F
RA
M
E
WO
RK
RL
tech
n
iq
u
es
a
r
e
g
en
er
ally
class
if
ied
in
to
v
alu
e
-
b
ased
an
d
p
o
licy
-
b
ased
ca
teg
o
r
ies.
Va
lu
e
-
b
ased
ap
p
r
o
ac
h
es
ap
p
r
o
x
im
ate
a
v
al
u
e
f
u
n
ctio
n
th
at
r
e
f
lects
th
e
e
x
p
ec
ted
l
o
n
g
-
te
r
m
r
ewa
r
d
a
n
d
in
f
er
th
e
o
p
tim
al
p
o
licy
b
y
ch
o
o
s
in
g
ac
tio
n
s
th
at
m
ax
im
ize
th
is
v
al
u
e,
wi
th
Q
-
lear
n
in
g
s
er
v
in
g
as
a
ty
p
ical
ex
am
p
le.
Po
licy
-
b
ased
m
eth
o
d
s
,
b
y
c
o
n
tr
ast,
f
o
cu
s
o
n
d
ir
ec
tly
l
ea
r
n
in
g
a
p
ar
am
ete
r
ized
p
o
l
icy
th
at
o
p
tim
izes
cu
m
u
lativ
e
r
ewa
r
d
.
T
h
is
wo
r
k
ad
o
p
ts
Q
-
lear
n
in
g
,
a
m
o
d
e
l
-
f
r
ee
,
v
alu
e
-
b
ased
R
L
alg
o
r
ith
m
th
at
d
o
es
n
o
t
r
eq
u
ir
e
p
r
io
r
k
n
o
wled
g
e
o
f
th
e
en
v
ir
o
n
m
e
n
t
d
y
n
am
ics.
Q
-
le
ar
n
in
g
d
er
i
v
es
an
o
p
tim
al
p
o
licy
th
r
o
u
g
h
iter
ativ
e
r
ef
in
em
en
t
o
f
a
s
tate
–
ac
tio
n
v
alu
e
f
u
n
ctio
n
,
r
ef
e
r
r
ed
to
a
s
th
e
Q
-
f
u
n
ctio
n
,
wh
ich
ass
o
ciate
s
ea
ch
s
tate
–
ac
tio
n
p
air
with
its
ex
p
ec
ted
l
o
n
g
-
ter
m
r
etu
r
n
.
T
h
e
a
l
g
o
r
ith
m
o
p
er
ates
o
f
f
-
p
o
lic
y
,
u
p
d
atin
g
Q
-
v
alu
es
b
ased
o
n
o
b
s
er
v
ed
tr
an
s
itio
n
s
wh
ile
b
alan
cin
g
ex
p
lo
r
atio
n
an
d
ex
p
lo
itatio
n
t
h
r
o
u
g
h
an
ex
p
lo
r
atio
n
s
tr
ateg
y
.
At
in
itializatio
n
,
Q
-
v
alu
es
a
r
e
u
n
if
o
r
m
ly
ass
ig
n
ed
ze
r
o
v
alu
es,
in
d
icatin
g
th
at
th
e
a
g
en
t
h
as
n
o
p
r
io
r
in
f
o
r
m
atio
n
ab
o
u
t th
e
en
v
ir
o
n
m
en
t.
F
o
r
f
e
a
t
u
r
e
s
e
le
c
t
i
o
n
,
t
h
e
R
L
en
v
i
r
o
n
m
e
n
t
i
s
d
e
f
i
n
e
d
s
u
c
h
t
h
at
e
a
c
h
s
t
at
e
r
e
p
r
e
s
e
n
t
s
a
s
u
b
s
e
t
o
f
s
e
l
e
c
t
e
d
f
e
a
t
u
r
e
s
,
e
a
c
h
ac
t
i
o
n
c
o
r
r
es
p
o
n
d
s
t
o
a
d
d
i
n
g
a
n
e
w
f
e
a
t
u
r
e
to
t
h
e
s
u
b
s
e
t
,
a
n
d
t
h
e
r
e
w
a
r
d
is
d
e
t
e
r
m
i
n
e
d
b
y
t
h
e
pr
e
d
i
c
t
i
v
e
p
e
r
f
o
r
m
a
n
c
e
o
f
a
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
o
d
e
l
t
r
ai
n
e
d
u
s
i
n
g
t
h
e
s
el
e
c
te
d
f
e
a
t
u
r
es
.
A
Q
-
t
a
b
l
e
is
m
a
i
n
t
a
i
n
e
d
t
o
s
t
o
r
e
Q
-
v
a
l
u
es
f
o
r
a
l
l
s
ta
t
e
-
a
c
t
i
o
n
p
a
i
r
s
.
A
t
e
a
c
h
s
t
e
p
,
t
h
e
a
g
e
n
t
s
el
e
c
ts
a
n
a
ct
i
o
n
,
u
p
d
a
t
es
t
h
e
f
e
at
u
r
e
s
u
b
s
e
t
,
e
v
a
l
u
a
t
es
m
o
d
e
l
p
e
r
f
o
r
m
a
n
c
e
,
a
n
d
u
p
d
a
t
e
s
t
h
e
c
o
r
r
es
p
o
n
d
i
n
g
Q
-
v
a
l
u
e
u
s
i
n
g
t
h
e
s
t
a
n
d
a
r
d
Q
-
le
a
r
n
i
n
g
u
p
d
a
t
e
r
u
l
e
.
I
n
t
h
i
s
s
t
u
d
y
,
a
RF
cl
as
s
i
f
i
e
r
is
e
m
p
l
o
y
e
d
t
o
c
o
m
p
u
t
e
t
h
e
r
ew
ar
d
m
e
t
r
i
c
,
w
h
i
c
h
s
e
r
v
es
as
f
e
e
d
b
a
c
k
f
o
r
t
h
e
a
g
e
n
t
.
A
s
a
v
a
l
u
e
-
b
a
s
e
d
R
L
m
e
t
h
o
d
,
Q
-
l
e
a
r
n
i
n
g
f
o
l
l
o
w
s
a
t
h
r
e
e
-
s
t
a
g
e
w
o
r
k
f
l
o
w
:
es
t
i
m
a
t
i
n
g
th
e
v
a
l
u
e
f
u
n
c
t
i
o
n
,
d
e
r
i
v
i
n
g
t
h
e
o
p
t
i
m
a
l
p
o
li
c
y
b
y
m
a
x
i
m
i
z
i
n
g
t
h
e
es
t
i
m
a
t
e
d
v
a
l
u
e
s
,
a
n
d
e
x
e
c
u
ti
n
g
t
h
is
p
o
l
i
c
y
t
o
m
a
x
i
m
i
ze
c
u
m
u
l
a
t
i
v
e
r
ew
a
r
d
s
.
T
w
o
f
o
r
m
s
o
f
v
al
u
e
f
u
n
c
ti
o
n
s
a
r
e
r
e
lev
a
n
t
i
n
t
h
i
s
c
o
n
t
e
x
t:
t
h
e
s
t
a
t
e
v
a
l
u
e
f
u
n
c
t
i
o
n
V
(
s
)
,
w
h
i
c
h
q
u
a
n
t
i
f
i
e
s
t
h
e
e
x
p
e
c
te
d
r
e
w
a
r
d
o
f
b
e
i
n
g
i
n
a
g
i
v
e
n
s
t
a
te
,
a
n
d
t
h
e
s
t
a
t
e
-
a
ct
i
o
n
v
a
lu
e
f
u
n
c
t
i
o
n
Q
(
s
,
a
)
,
w
h
i
c
h
e
s
t
i
m
a
te
s
t
h
e
e
x
p
e
c
t
e
d
r
e
w
a
r
d
o
f
t
a
k
i
n
g
a
s
p
e
ci
f
i
c
a
ct
i
o
n
i
n
a
g
i
v
e
n
s
t
at
e
[1
8
]
.
I
n
d
is
c
r
et
e
e
n
v
i
r
o
n
m
e
n
ts
,
Q
-
v
alu
es
a
r
e
ty
p
i
ca
ll
y
s
t
o
r
ed
i
n
a
Q
-
t
ab
le
.
T
h
e
Q
-
ta
b
le
is
s
t
r
u
c
tu
r
e
d
as
a
two
-
d
i
m
e
n
s
i
o
n
al
m
a
tr
ix
i
n
w
h
ic
h
r
o
ws
c
o
r
r
es
p
o
n
d
t
o
en
v
i
r
o
n
m
e
n
t
s
tat
es,
c
o
l
u
m
n
s
d
e
n
o
te
a
v
ail
ab
le
ac
ti
o
n
s
,
an
d
ea
ch
e
n
t
r
y
Q(
s
,
a
)
s
to
r
es
t
h
e
esti
m
a
te
d
v
al
u
e
o
f
e
x
ec
u
ti
n
g
ac
t
io
n
(
a)
i
n
s
tat
e
(
s
)
.
Ov
e
r
t
im
e,
as
t
h
e
Q
-
l
ea
r
n
in
g
al
g
o
r
it
h
m
u
p
d
a
tes
th
e
Q
-
v
al
u
es
u
s
in
g
t
h
e
te
m
p
o
r
al
-
d
i
f
f
er
e
n
ce
r
u
le
,
t
h
e
Q
-
ta
b
l
e
c
o
n
v
e
r
g
es
t
o
w
ar
d
th
e
o
p
tim
al
ac
ti
o
n
-
v
a
lu
e
f
u
n
cti
o
n
.
Op
ti
m
al
a
cti
o
n
s
e
le
cti
o
n
is
ac
h
i
ev
e
d
b
y
c
h
o
o
s
i
n
g
t
h
e
a
cti
o
n
ass
o
c
iat
e
d
w
it
h
th
e
m
a
x
i
m
u
m
Q
-
v
al
u
e
f
o
r
th
e
cu
r
r
e
n
t
s
tat
e.
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
a
p
tive
fea
tu
r
e
s
elec
tio
n
fo
r
cred
it sco
r
in
g
mo
d
els
(
Mo
s
ta
fa
Mo
h
a
med
S
eifE
ln
a
s
r
)
513
2
.
1
.
M
DP
f
o
r
m
ula
t
io
n
T
h
e
f
ea
tu
r
e
s
elec
tio
n
task
is
f
o
r
m
u
lated
as a
MD
P
,
d
ef
i
n
ed
b
y
th
e
tu
p
le
(
,
,
,
,
)
:
-
State
(
)
:
a
s
u
b
s
et
o
f
f
ea
tu
r
es selecte
d
f
r
o
m
th
e
f
u
ll
f
ea
tu
r
e
s
et
Ψ
.
-
Actio
n
(
)
:
s
elec
tin
g
a
f
ea
tu
r
e
f
∈
Ψ
th
at
is
n
o
t y
et
in
clu
d
e
d
in
t
h
e
cu
r
r
e
n
t su
b
s
et.
-
T
r
an
s
itio
n
(
→
′
)
:
d
eter
m
in
is
tic
u
p
d
ate
o
f
th
e
s
u
b
s
et
b
y
ad
d
i
n
g
th
e
s
elec
ted
f
ea
tu
r
e.
-
R
ewa
r
d
(
)
:
a
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
m
etr
ic
(
e.
g
.
,
ac
cu
r
ac
y
,
F1
,
o
r
AUC)
o
b
tain
e
d
b
y
tr
ain
in
g
a
class
if
ier
o
n
th
e
r
esu
ltin
g
f
ea
t
u
r
e
s
u
b
s
et.
-
Po
licy
(
)
:
t
h
e
f
ea
tu
r
e
s
elec
tio
n
s
tr
ateg
y
th
at
m
ax
i
m
izes th
e
ex
p
ec
ted
lo
n
g
-
te
r
m
p
r
e
d
ictiv
e
p
er
f
o
r
m
an
ce
.
-
Dis
co
u
n
t f
ac
to
r
(
)
:
c
o
n
tr
o
ls
th
e
weig
h
tin
g
o
f
f
u
tu
r
e
r
ewa
r
d
s
r
e
lativ
e
to
im
m
ed
iate
r
ewa
r
d
s
(
0
≤
≤
1
)
.
W
ith
in
th
is
f
r
am
ewo
r
k
,
Q
-
lear
n
in
g
is
ap
p
lied
to
ap
p
r
o
x
im
ate
th
e
o
p
tim
al
s
tate
–
ac
tio
n
v
alu
e
f
u
n
ctio
n
∗
(
,
)
,
wh
ich
r
e
p
r
esen
ts
th
e
ex
p
ec
ted
cu
m
u
lativ
e
r
ewa
r
d
o
f
ta
k
in
g
ac
tio
n
“
a
”
in
s
tate
“
s
”
an
d
ac
tin
g
o
p
tim
ally
th
er
ea
f
ter
.
T
h
e
Q
-
ta
b
le
is
in
itialized
to
ze
r
o
f
o
r
all
s
tate
–
ac
tio
n
p
air
s
.
At
ea
ch
s
tep
,
th
e
a
g
en
t
s
elec
ts
an
ac
tio
n
u
s
in
g
a
n
ε
-
g
r
ee
d
y
p
o
licy
,
ch
o
o
s
in
g
a
r
an
d
o
m
ac
ti
o
n
with
p
r
o
b
ab
ilit
y
ε
to
e
n
co
u
r
ag
e
ex
p
lo
r
atio
n
,
an
d
th
e
ac
tio
n
with
t
h
e
h
i
g
h
est cu
r
r
en
t Q
-
v
alu
e
with
p
r
o
b
a
b
ilit
y
(
1
−
)
.
Af
ter
tak
in
g
ac
tio
n
“
a
”
in
s
t
ate
“
s
”
an
d
o
b
s
er
v
in
g
r
ewa
r
d
“
r
”
a
n
d
n
ex
t
s
tate
“
s′
”
,
th
e
ag
en
t
co
m
p
u
tes th
e
T
D
tar
g
et,
d
er
i
v
ed
f
r
o
m
th
e
B
ellm
an
o
p
tim
alit
y
eq
u
atio
n
,
as sh
o
wn
in
(
1
)
:
=
+
′
ma
x
(
′
,
′
)
(
1
)
-
r
:
t
h
e
im
m
ed
iate
r
ewa
r
d
o
b
tain
ed
af
ter
tr
a
n
s
itio
n
in
g
to
s
′.
-
γ
:
t
h
e
d
is
co
u
n
t
f
ac
to
r
.
-
a
’
m
ax
Q(
s
′,
a′)
:
t
h
e
m
ax
im
u
m
esti
m
ated
v
alu
e
o
v
er
all
p
o
s
s
ib
le
ac
tio
n
s
in
th
e
n
ex
t
s
tate
s
′,
r
etr
iev
ed
d
ir
ec
tly
f
r
o
m
th
e
Q
-
tab
le.
T
h
e
T
D
er
r
o
r
m
ea
s
u
r
es
th
e
g
a
p
b
etwe
en
th
is
n
ewly
co
m
p
u
t
ed
tar
g
et
an
d
th
e
cu
r
r
en
t
Q
-
v
a
lu
e
s
to
r
ed
in
th
e
tab
le,
as sh
o
wn
i
n
(
2
)
.
=
−
(
,
)
(
2
)
Fin
ally
,
th
e
Q
-
v
alu
e
is
u
p
d
ate
d
b
y
m
o
v
in
g
th
e
c
u
r
r
en
t
esti
m
ate
to
war
d
th
e
T
D
tar
g
et
b
y
a
f
r
ac
tio
n
α
(
th
e
lear
n
in
g
r
ate)
,
as sh
o
w
n
i
n
(
3
)
.
(
,
)
←
(
,
)
+
(
3
)
-
α
(
lear
n
in
g
r
ate)
:
c
o
n
tr
o
ls
h
o
w
s
tr
o
n
g
ly
n
ew
i
n
f
o
r
m
atio
n
in
f
lu
en
ce
s
th
e
ex
is
tin
g
Q
-
v
al
u
e
(
0
<
α
≤
1
)
.
Su
b
s
titu
tin
g
(
1
)
-
(
3
)
y
ield
s
th
e
s
tan
d
ar
d
Q
-
lear
n
i
n
g
u
p
d
ate
in
(
4
)
:
(
,
)
←
(
,
)
+
[
+
′
(
′
,
′
)
−
(
,
)
]
(
4
)
T
h
is
p
r
o
ce
s
s
is
r
ep
ea
ted
ac
r
o
s
s
m
an
y
ep
is
o
d
es
u
n
til
th
e
Q
-
v
alu
es
co
n
v
er
g
e,
at
wh
ich
p
o
in
t
t
h
e
lear
n
ed
p
o
licy
co
r
r
esp
o
n
d
s
to
s
elec
tin
g
th
e
ac
tio
n
with
th
e
h
ig
h
est Q
-
v
alu
e
in
ea
c
h
s
tate.
2
.
2
.
Wo
r
k
f
lo
w
T
h
e
p
r
o
p
o
s
ed
f
r
a
m
ewo
r
k
a
p
p
l
ies
Q
-
lear
n
in
g
–
b
ased
RL
to
a
d
ap
tiv
ely
s
elec
t
in
f
o
r
m
ativ
e
f
ea
tu
r
es
f
o
r
cr
ed
it
s
co
r
in
g
an
d
d
elin
q
u
e
n
c
y
p
r
ed
ictio
n
,
im
p
r
o
v
in
g
o
v
er
tr
ad
itio
n
al
g
r
ee
d
y
m
eth
o
d
s
Fig
u
r
e
3
.
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
is
im
p
lem
en
ted
th
r
o
u
g
h
t
h
e
f
o
llo
win
g
s
eq
u
en
tial
s
tep
s
:
−
Step
1
–
RL
-
b
ased
f
ea
tu
r
e
s
el
ec
tio
n
:
a
RL
ag
en
t
ex
p
lo
r
es
th
e
f
ea
tu
r
e
s
p
ac
e
wh
er
e
ea
ch
s
tate
r
ep
r
esen
ts
a
s
u
b
s
et
o
f
f
ea
t
u
r
es.
T
h
e
ag
e
n
t
ev
alu
ates
ea
ch
s
u
b
s
et
u
s
in
g
a
class
if
icatio
n
m
o
d
el
a
n
d
r
e
ce
iv
es
a
r
ewa
r
d
b
ased
o
n
p
r
ed
ictiv
e
p
er
f
o
r
m
a
n
ce
.
T
h
r
o
u
g
h
iter
ativ
e
ex
p
lo
r
atio
n
,
th
e
s
u
b
s
et
with
th
e
h
ig
h
e
s
t
cu
m
u
lativ
e
r
ewa
r
d
is
s
elec
ted
as th
e
o
p
ti
m
al
f
ea
tu
r
e
s
et
f
o
r
cr
ed
it sco
r
i
n
g
.
−
Step
2
–
C
r
ed
it
s
co
r
in
g
class
if
icatio
n
:
th
e
s
elec
ted
f
ea
t
u
r
e
s
u
b
s
et
is
u
s
ed
to
tr
ain
a
class
if
icatio
n
m
o
d
el
th
at
p
r
ed
icts
cr
ed
it sco
r
in
g
b
u
ck
ets.
−
Step
3
–
Mo
d
el
v
alid
atio
n
:
m
u
l
tip
le
s
u
p
er
v
is
ed
lear
n
in
g
al
g
o
r
ith
m
s
ar
e
ap
p
lied
to
e
v
alu
ate
th
e
s
elec
ted
f
ea
tu
r
e
s
u
b
s
et
an
d
id
en
tify
t
h
e
class
if
ier
th
at
ac
h
iev
es th
e
b
e
s
t p
er
f
o
r
m
a
n
ce
.
−
Step
4
–
Delin
q
u
en
cy
m
o
d
e
llin
g
:
th
e
p
r
ed
icted
cr
ed
it
s
co
r
in
g
r
esu
lts
ar
e
co
m
b
in
ed
with
ad
d
itio
n
al
f
ea
tu
r
es to
co
n
s
tr
u
ct
a
d
ataset
f
o
r
d
elin
q
u
en
c
y
p
r
e
d
ictio
n
.
−
Step
5
–
Delin
q
u
en
cy
f
ea
tu
r
e
s
elec
tio
n
:
th
e
RL
ag
en
t
is
ap
p
lied
ag
ain
to
id
e
n
tify
th
e
m
o
s
t
r
elev
an
t
f
ea
tu
r
es f
o
r
d
elin
q
u
e
n
cy
class
if
icatio
n
an
d
t
o
b
u
ild
a
p
r
e
d
ictiv
e
m
o
d
el
f
o
r
clien
t
r
is
k
b
eh
a
v
io
u
r
.
−
St
ep
6
–
Mo
d
el
ass
ess
m
en
t:
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
is
co
m
p
ar
ed
with
tr
a
d
itio
n
al
f
e
atu
r
e
s
elec
tio
n
tech
n
iq
u
es,
in
clu
d
i
n
g
wr
ap
p
er
m
eth
o
d
s
(
f
o
r
war
d
an
d
b
ac
k
war
d
s
elec
tio
n
)
an
d
em
b
ed
d
ed
m
eth
o
d
s
(
tr
ee
-
b
ased
im
p
o
r
tan
ce
)
,
u
s
in
g
s
tan
d
ar
d
ev
al
u
atio
n
m
etr
ics.
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.
4
3
,
No
.
2
,
Au
g
u
s
t
20
2
6
:
507
-
5
2
1
514
Fig
u
r
e
1
.
Featu
r
e
s
elec
tio
n
ag
en
t w
o
r
k
f
lo
w
B
en
ef
its
o
f
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
:
th
e
u
s
e
o
f
RL
i
n
f
e
atu
r
e
s
elec
tio
n
all
o
ws
f
o
r
d
y
n
am
ic
an
d
ad
ap
tiv
e
o
p
tim
izatio
n
o
f
th
e
f
ea
tu
r
e
s
et,
p
o
ten
tially
lead
in
g
to
h
ig
h
er
ac
cu
r
ac
y
an
d
m
o
r
e
r
o
b
u
s
t
cr
ed
it
s
co
r
in
g
m
o
d
els.
Fu
r
th
er
m
o
r
e,
t
h
is
ap
p
r
o
ac
h
ad
d
r
ess
es
th
e
ch
allen
g
e
o
f
h
an
d
lin
g
clien
ts
with
n
o
p
r
io
r
cr
ed
it
h
is
to
r
y
b
y
in
co
r
p
o
r
atin
g
alter
n
ati
v
e
d
ata
s
o
u
r
ce
s
an
d
p
r
ed
ictiv
e
tech
n
iq
u
es.
3.
M
E
T
H
O
D
3
.
1
.
Da
t
a
s
et
s
d
escript
io
n
Data
s
et
1
–
C
r
ed
it
b
u
c
k
et:
th
is
d
ataset
T
ab
le
2
is
em
p
lo
y
e
d
to
p
r
e
d
ict
th
e
cr
ed
it
b
u
ck
et
,
wh
ich
is
class
if
ied
in
to
eig
h
t
ca
teg
o
r
ie
s
r
ep
r
esen
tin
g
th
e
clien
t
’
s
o
v
er
all
cr
ed
itwo
r
th
in
ess
an
d
co
r
r
esp
o
n
d
i
n
g
cr
ed
it
lim
it e
lig
ib
ilit
y
.
T
h
e
ca
teg
o
r
ie
s
r
an
g
e
f
r
o
m
less
th
an
5
K
u
p
t
o
a
b
o
v
e
1
0
0
K
as in
T
ab
le
3
.
Data
s
et
2
–
Delin
q
u
en
cy
p
r
e
d
ictio
n
:
th
is
d
ataset
T
ab
le
2
is
em
p
lo
y
ed
to
p
r
ed
ict
th
e
d
elin
q
u
en
cy
b
eh
av
io
u
r
o
f
clien
ts
.
T
h
e
p
r
e
d
icted
elig
ib
le
cr
ed
it
b
u
ck
et
s
er
v
es
as
an
ad
d
itio
n
al
i
n
p
u
t
f
ea
tu
r
e,
en
ab
lin
g
th
e
m
o
d
el
to
ass
ess
th
e
lik
elih
o
o
d
o
f
d
elin
q
u
en
c
y
if
th
e
b
an
k
g
r
an
ts
a
cr
ed
it
f
ac
ilit
y
.
T
h
e
d
eli
n
q
u
en
c
y
b
e
h
av
io
u
r
is
class
if
ied
in
to
two
ca
teg
o
r
ie
s
as in
T
ab
le
3
.
T
h
e
p
r
e
d
ictio
n
task
s
ar
e
b
ased
p
r
im
ar
il
y
o
n
d
em
o
g
r
a
p
h
i
c
attr
ib
u
tes.
An
ex
te
n
d
ed
f
o
r
m
u
latio
n
in
co
r
p
o
r
ates
b
o
th
d
em
o
g
r
ap
h
ic
in
f
o
r
m
atio
n
an
d
th
e
ass
ig
n
ed
cr
ed
it
b
u
ck
et
to
en
h
an
ce
d
elin
q
u
en
c
y
p
r
ed
ictio
n
.
T
h
e
d
ataset
f
u
r
th
er
in
clu
d
es
f
ea
tu
r
es
ca
p
tu
r
in
g
d
em
o
g
r
a
p
h
ic,
f
in
a
n
cial,
an
d
s
o
cio
-
ec
o
n
o
m
ic
ch
ar
ac
ter
is
tics
o
f
b
o
th
b
an
k
e
d
an
d
u
n
b
an
k
ed
clien
ts
.
T
ab
le
2
.
Data
s
et
1
,
2
f
ea
tu
r
es
D
a
t
a
s
e
t
F
e
a
t
u
r
e
s
D
a
t
a
s
e
t
1
G
e
n
d
e
r
,
A
g
e
,
M
a
r
i
t
a
l
s
t
a
t
u
s
,
N
o
.
o
f
d
e
p
e
n
d
e
n
t
s,
R
e
si
d
e
n
t
i
a
l
s
t
a
t
u
s
,
T
i
me
a
t
c
u
r
r
e
n
t
a
d
d
r
e
ss
,
P
r
o
f
e
ssi
o
n
,
E
mp
l
o
y
e
r
t
y
p
e
,
I
n
d
u
s
t
r
y
,
I
n
c
o
me
b
r
a
c
k
e
t
,
N
e
t
m
o
n
t
h
l
y
i
n
c
o
me
,
N
e
t
m
o
n
t
h
l
y
e
x
p
e
n
ses
,
R
e
g
i
o
n
,
G
o
v
e
r
n
a
t
e
,
A
c
t
u
a
l
sa
v
e
d
m
o
n
e
y
,
C
a
r
br
a
n
d
,
C
l
u
b
m
e
m
b
e
r
s
h
i
p
,
e
l
i
g
i
b
l
e
l
i
mi
t
b
u
c
k
e
t
D
a
t
a
s
e
t
2
G
e
n
d
e
r
,
A
g
e
,
M
a
r
i
t
a
l
s
t
a
t
u
s
,
N
o
.
o
f
de
p
e
n
d
e
n
t
s,
R
e
si
d
e
n
t
i
a
l
st
a
t
u
s
,
T
i
me
a
t
c
u
r
r
e
n
t
a
d
d
r
e
ss,
P
r
o
f
e
ssi
o
n
,
E
mp
l
o
y
e
r
t
y
p
e
,
I
n
d
u
s
t
r
y
,
I
n
c
o
me
b
r
a
c
k
e
t
,
N
e
t
m
o
n
t
h
l
y
i
n
c
o
me
,
N
e
t
m
o
n
t
h
l
y
e
x
p
e
n
ses
,
R
e
g
i
o
n
,
G
o
v
e
r
n
a
t
e
,
A
c
t
u
a
l
sa
v
e
d
m
o
n
e
y
,
C
a
r
b
r
a
n
d
,
C
l
u
b
m
e
m
b
e
r
s
h
i
p
,
e
l
i
g
i
b
l
e
l
i
mi
t
b
u
c
k
e
t
,
d
e
l
i
n
q
u
e
n
c
y
b
e
h
a
v
i
o
u
r
T
ab
le
3
.
Data
s
et
1
,
2
lab
els
D
a
t
a
s
e
t
La
b
e
l
V
a
l
u
e
s
D
a
t
a
s
e
t
1
El
i
g
i
b
l
e
l
i
m
i
t
b
u
c
k
e
t
Le
ss
t
h
a
n
5
K
,
5
K
–
1
0
K
,
1
0
K
–
2
0
K
,
2
0
K
–
3
0
K
,
3
0
K
–
5
0
K
,
5
0
K
–
7
0
K
,
7
0
K
–
1
0
0
K
,
A
b
o
v
e
1
0
0
K
(
8
c
l
a
ss
e
s)
D
a
t
a
s
e
t
2
D
e
l
i
n
q
u
e
n
c
y
b
e
h
a
v
i
o
u
r
G
o
o
d
,
B
a
d
(
2
c
l
a
sses)
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
a
p
tive
fea
tu
r
e
s
elec
tio
n
fo
r
cred
it sco
r
in
g
mo
d
els
(
Mo
s
ta
fa
Mo
h
a
med
S
eifE
ln
a
s
r
)
515
3
.
1
.
1
.
B
a
la
ncing
d
a
t
a
s
et
On
e
o
f
th
e
cr
itical
ch
allen
g
e
s
in
tr
ain
in
g
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els
o
n
f
i
n
an
cial
d
atasets
is
th
e
p
r
esen
ce
o
f
im
b
ala
n
ce
d
class
d
is
tr
ib
u
tio
n
s
.
I
n
th
e
elig
ib
le
cr
ed
it
b
u
ck
et
d
ataset,
th
e
tar
g
et
v
ar
iab
le
is
co
m
p
o
s
ed
o
f
2
4
,
5
3
3
r
ec
o
r
d
s
d
is
tr
ib
u
ted
ac
r
o
s
s
eig
h
t c
ateg
o
r
ies,
as sh
o
wn
in
T
ab
le
4
.
T
ab
le
4
.
C
lass
d
is
tr
ib
u
tio
n
o
f
cr
ed
it lim
it b
u
ck
et
C
r
e
d
i
t
b
u
c
k
e
t
c
a
t
e
g
o
r
y
N
u
mb
e
r
o
f
r
e
c
o
r
d
s
P
e
r
c
e
n
t
a
g
e
o
f
t
o
t
a
l
Le
ss
t
h
a
n
5
K
3
3
7
1
.
3
7
%
5K
–
1
0
K
1
,
4
7
0
6
.
0
0
%
1
0
K
–
2
0
K
7
,
0
8
8
2
8
.
8
9
%
2
0
K
–
3
0
K
7
,
6
4
7
3
1
.
1
7
%
3
0
K
–
5
0
K
4
,
0
2
4
1
6
.
4
0
%
5
0
K
–
7
0
K
1
,
7
1
3
6
.
9
8
%
7
0
K
–
1
0
0
K
1
,
0
9
4
4
.
4
6
%
A
b
o
v
e
1
0
0
K
1
,
1
3
0
4
.
6
1
%
To
t
a
l
2
4
,
5
3
3
1
0
0
%
As
s
h
o
wn
in
Fig
u
r
e
4
,
th
e
d
is
tr
ib
u
tio
n
is
h
ea
v
ily
s
k
ewe
d
to
war
d
th
e
m
id
d
le
r
a
n
g
es
(
1
0
K
–
2
0
K
a
n
d
20K
–
3
0
K
)
,
wh
ich
t
o
g
eth
er
ac
co
u
n
t
f
o
r
o
v
er
6
0
%
o
f
t
h
e
d
at
aset.
C
o
n
v
er
s
ely
,
ca
teg
o
r
ies
s
u
ch
as
l
ess
th
an
5
K
an
d
a
b
o
v
e
1
0
0
K
ar
e
s
ev
e
r
ely
u
n
d
er
r
ep
r
esen
ted
.
Fig
u
r
e
4
.
C
lass
d
is
tr
ib
u
tio
n
o
f
cr
ed
it lim
it b
u
ck
et
C
las
s
im
b
alan
ce
m
ay
ca
u
s
e
lear
n
in
g
alg
o
r
ith
m
s
to
f
a
v
o
r
m
ajo
r
ity
class
es,
lead
in
g
to
d
eg
r
ad
e
d
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
f
o
r
m
in
o
r
ity
class
es.
T
h
is
o
f
ten
r
esu
lts
in
m
is
lead
in
g
l
y
h
i
g
h
ac
cu
r
ac
y
v
alu
es
ac
co
m
p
an
ied
b
y
r
e
d
u
ce
d
R
ec
all
an
d
F1
-
s
co
r
e,
as
well
as
p
o
o
r
g
en
e
r
aliza
tio
n
f
o
r
u
n
d
e
r
r
ep
r
esen
ted
ca
teg
o
r
ies,
s
u
ch
as c
lien
ts
b
elo
n
g
in
g
to
lo
w
-
f
r
eq
u
e
n
cy
cr
e
d
it g
r
o
u
p
s
.
S
o
l
v
i
n
g
i
m
b
a
la
n
c
e
I
s
s
u
e
:
T
o
ad
d
r
e
s
s
t
h
e
is
s
u
e
o
f
c
l
ass
i
m
b
ala
n
c
e
,
t
h
e
s
y
n
t
h
e
t
ic
m
i
n
o
r
i
t
y
o
v
e
r
s
a
m
p
l
i
n
g
t
e
c
h
n
i
q
u
e
(
SM
O
T
E
)
w
a
s
a
p
p
l
i
e
d
[
19
]
.
T
h
is
m
e
t
h
o
d
s
y
n
t
h
e
s
i
z
es
n
ew
m
i
n
o
r
it
y
c
l
ass
s
a
m
p
l
es
t
h
r
o
u
g
h
i
n
t
e
r
p
o
l
a
t
i
o
n
i
n
t
h
e
f
e
a
t
u
r
e
s
p
a
c
e
,
t
h
e
r
e
b
y
e
n
h
a
n
c
i
n
g
d
a
t
a
d
iv
e
r
s
i
t
y
c
o
m
p
a
r
e
d
t
o
s
i
m
p
l
e
d
u
p
l
i
c
a
t
i
o
n
s
t
r
a
te
g
i
es
.
T
o
a
v
o
i
d
d
a
t
a
l
e
a
k
a
g
e
,
o
v
e
r
s
a
m
p
l
i
n
g
w
a
s
r
e
s
t
r
i
c
t
e
d
t
o
t
h
e
t
r
a
i
n
i
n
g
d
a
t
a
,
w
h
i
l
e
t
h
e
t
es
t
s
et
r
e
t
a
i
n
e
d
i
ts
o
r
i
g
i
n
a
l
c
l
a
s
s
d
is
t
r
i
b
u
t
i
o
n
[
2
0
]
.
A
f
t
e
r
SM
O
T
E
,
t
h
e
d
a
t
as
e
t
w
as
m
o
r
e
b
a
l
a
n
c
e
d
a
c
r
o
s
s
c
a
t
e
g
o
r
i
es
,
e
n
a
b
l
i
n
g
t
h
e
m
o
d
e
l
to
l
e
a
r
n
r
o
b
u
s
t
d
e
c
is
i
o
n
b
o
u
n
d
a
r
i
es
.
T
h
e
s
a
m
e
b
a
l
a
n
c
i
n
g
m
e
t
h
o
d
o
l
o
g
y
a
p
p
l
i
e
d
o
n
t
h
e
o
t
h
e
r
o
t
h
e
r
d
a
t
a
s
e
t
[
2
1
].
3.
2
.
Da
t
a
p
re
pro
ce
s
s
ing
Prio
r
to
RL
a
n
d
m
o
d
el
co
n
s
tr
u
ctio
n
,
m
u
ltip
le
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
p
r
o
ce
d
u
r
es
wer
e
ap
p
lied
to
im
p
r
o
v
e
p
r
ed
ictiv
e
p
e
r
f
o
r
m
an
ce
.
T
h
ese
tr
an
s
f
o
r
m
atio
n
s
en
s
u
r
e
th
at
ca
teg
o
r
ical,
n
u
m
er
ica
l,
an
d
m
is
s
in
g
d
ata
ar
e
co
n
s
is
ten
tly
r
ep
r
esen
ted
a
n
d
s
u
itab
le
f
o
r
Q
-
lear
n
in
g
–
b
as
ed
f
ea
tu
r
e
s
elec
tio
n
[
2
2
].
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.
4
3
,
No
.
2
,
Au
g
u
s
t
20
2
6
:
507
-
5
2
1
516
Featu
r
e
en
co
d
in
g
:
ca
teg
o
r
ical
v
ar
iab
les,
s
u
ch
as
g
en
d
er
,
p
r
o
f
ess
io
n
,
em
p
lo
y
er
t
y
p
e,
in
d
u
s
tr
y
,
r
eg
io
n
,
g
o
v
er
n
o
r
ate,
m
ar
ital
s
tatu
s
,
r
e
s
id
en
tial
s
tatu
s
,
ca
r
b
r
a
n
d
,
an
d
clu
b
m
em
b
er
s
h
ip
,
ar
e
tr
an
s
f
o
r
m
ed
in
to
b
in
a
r
y
in
d
icato
r
v
a
r
iab
les
u
s
in
g
o
n
e
-
h
o
t
en
co
d
in
g
.
T
h
is
r
ep
r
esen
tatio
n
av
o
id
s
o
r
d
in
al
ass
u
m
p
tio
n
s
an
d
en
a
b
les
th
e
R
L
ag
en
t to
ev
alu
ate
th
e
p
r
e
d
ictiv
e
co
n
tr
ib
u
tio
n
o
f
ea
ch
ca
te
g
o
r
y
i
n
d
ep
e
n
d
en
tly
[
2
3
].
B
in
n
in
g
o
f
c
o
n
tin
u
o
u
s
v
ar
ia
b
les:
co
n
tin
u
o
u
s
attr
ib
u
tes,
in
clu
d
in
g
ag
e,
tim
e
at
c
u
r
r
en
t
ad
d
r
ess
,
n
et
m
o
n
th
ly
i
n
co
m
e,
a
n
d
n
et
m
o
n
th
ly
e
x
p
en
s
es
,
ar
e
d
is
cr
et
ized
in
to
f
ix
ed
b
in
s
.
B
in
n
in
g
ca
p
tu
r
es
p
o
ten
tial
non
-
lin
ea
r
ef
f
ec
ts
,
im
p
r
o
v
es in
ter
p
r
etab
ilit
y
,
an
d
r
ed
u
ce
s
s
en
s
itiv
ity
to
o
u
tlier
s
.
E
q
u
al
-
wid
t
h
o
r
q
u
an
tile
-
b
ased
b
in
n
in
g
s
tr
ateg
ies ar
e
co
n
s
i
d
er
ed
,
d
ep
e
n
d
in
g
o
n
v
ar
iab
le
d
is
tr
ib
u
tio
n
s
.
Featu
r
e
s
ca
lin
g
:
n
u
m
er
ical
f
ea
tu
r
es
ar
e
s
tan
d
ar
d
ize
d
o
r
n
o
r
m
alize
d
t
o
e
n
s
u
r
e
c
o
m
p
ar
ab
ilit
y
ac
r
o
s
s
d
if
f
er
en
t
r
an
g
es.
Featu
r
e
s
ca
lin
g
en
s
u
r
es
th
at
v
ar
iab
les
with
lar
g
er
n
u
m
e
r
ical
r
an
g
es,
s
u
ch
a
s
in
co
m
e,
d
o
n
o
t
d
o
m
in
ate
th
e
tr
ain
in
g
m
o
d
el
an
d
im
p
r
o
v
es
th
e
s
tab
ilit
y
o
f
d
is
tan
ce
-
b
ased
alg
o
r
ith
m
s
wh
er
e
ap
p
licab
le.
Data
im
p
u
tatio
n
:
m
is
s
in
g
d
ata
is
h
an
d
led
s
y
s
tem
atica
lly
th
r
o
u
g
h
tailo
r
ed
im
p
u
tatio
n
s
tr
ate
g
ies
b
ased
o
n
th
e
f
ea
tu
r
e
t
y
p
e.
Fo
r
ca
te
g
o
r
ical
v
ar
ia
b
les
with
a
l
o
w
p
r
o
p
o
r
tio
n
o
f
m
is
s
in
g
v
alu
es,
m
is
s
in
g
en
tr
ies
ar
e
im
p
u
ted
u
s
in
g
th
e
m
o
s
t
f
r
eq
u
en
t
ca
teg
o
r
y
.
Fo
r
n
u
m
er
ical
v
ar
iab
les,
m
ea
n
o
r
m
ed
ian
im
p
u
tatio
n
is
ap
p
lied
as
an
in
itial a
p
p
r
o
a
ch
.
T
ar
g
et
v
ar
ia
b
le
tr
an
s
f
o
r
m
atio
n
:
th
e
d
e
p
en
d
e
n
t
v
ar
ia
b
le
in
t
h
is
s
tu
d
y
d
if
f
er
s
ac
r
o
s
s
th
e
tw
o
d
atasets
.
Fo
r
d
ataset
1
,
th
e
d
ep
en
d
en
t
v
ar
iab
le
is
th
e
elig
ib
le
lim
it
b
u
ck
et
,
wh
ich
r
ep
r
esen
ts
th
e
clien
t
’
s
elig
ib
le
cr
ed
it
lim
it.
T
h
is
v
ar
iab
le
is
d
is
cr
etize
d
in
to
o
r
d
in
al
ca
te
g
o
r
ies,
o
r
b
u
c
k
ets,
r
ef
lectin
g
d
if
f
er
en
t
le
v
els
o
f
cr
ed
itwo
r
th
in
ess
an
d
cr
ed
it
lim
it
elig
ib
ilit
y
.
Fo
r
d
ataset
2
,
t
h
e
d
ep
en
d
e
n
t
v
ar
iab
le
is
Delin
q
u
en
c
y
b
eh
av
i
o
u
r
,
wh
ich
in
d
icate
s
wh
eth
er
th
e
clien
t
is
lik
ely
to
d
em
o
n
s
tr
ate
g
o
o
d
o
r
b
ad
r
ep
a
y
m
en
t
b
eh
a
v
io
u
r
.
T
h
is
v
ar
iab
le
is
r
ep
r
esen
ted
as a
b
i
n
ar
y
class
if
icatio
n
o
u
tco
m
e
.
I
n
th
e
ca
s
e
o
f
th
e
d
ataset
1
,
ea
ch
s
tate
ca
n
b
e
in
ter
p
r
eted
as
a
s
u
b
s
et
o
f
s
e
lecte
d
f
ea
tu
r
es
(
e.
g
.
,
Ag
e
g
r
o
u
p
,
I
n
c
o
m
e
r
a
n
g
e,
a
n
d
Ma
r
ital
s
tatu
s
)
,
wh
ile
ea
ch
ac
tio
n
co
r
r
esp
o
n
d
s
to
th
e
d
ec
is
io
n
o
f
wh
eth
er
to
in
clu
d
e
o
r
ex
clu
d
e
a
s
p
ec
if
ic
f
ea
tu
r
e
in
th
e
f
ea
tu
r
e
s
u
b
s
et
d
u
r
in
g
th
e
s
elec
tio
n
p
r
o
ce
s
s
.
T
h
e
Q
-
t
ab
le
th
u
s
s
to
r
es
th
e
lear
n
ed
Q
-
v
alu
es
f
o
r
all
p
o
s
s
ib
le
s
tate
–
ac
tio
n
p
air
s
,
r
ef
lect
in
g
th
e
ex
p
ec
te
d
co
n
tr
i
b
u
tio
n
o
f
in
clu
d
in
g
a
g
iv
e
n
f
ea
tu
r
e
in
im
p
r
o
v
in
g
m
o
d
el
p
er
f
o
r
m
a
n
ce
.
T
ab
le
5
p
r
o
v
id
es
a
s
im
p
lifie
d
ex
am
p
le
with
th
r
ee
s
tate
s
(
S1
,
S2
,
S3
)
r
ep
r
esen
tin
g
d
if
f
er
en
t
clien
t
p
r
o
f
iles
an
d
th
r
ee
ac
tio
n
s
(
A1
,
A2
,
A3
)
r
ep
r
esen
tin
g
ca
n
d
id
ate
fe
atu
r
es (
e.
g
.
,
Ag
e,
I
n
co
m
e,
Ma
r
ital Statu
s
)
.
-
(
I
n
co
m
e
,
y
o
u
n
g
,
lo
w
-
in
c
o
m
e,
s
in
g
le
clien
t)
,
th
e
Q
-
lear
n
in
g
ag
en
t
ass
ig
n
s
th
e
h
ig
h
est
Q
-
v
alu
e
to
I
n
co
m
e
(
0
.
7
)
,
m
ea
n
i
n
g
th
at
in
clu
d
in
g
I
n
co
m
e
as
a
f
ea
tu
r
e
co
n
t
r
ib
u
t
es
m
o
s
t
to
im
p
r
o
v
in
g
p
r
e
d
ictio
n
ac
cu
r
ac
y
f
o
r
th
is
p
r
o
f
ile.
-
Fo
r
State
S2
I
n
c
o
m
e
(
0
.
9
)
is
ag
ain
t
h
e
m
o
s
t
v
alu
ab
le
f
ea
tu
r
e,
in
d
icatin
g
s
tr
o
n
g
p
r
ed
ic
tiv
e
p
o
wer
f
o
r
clien
ts
with
m
ed
iu
m
in
co
m
e
.
-
Fo
r
State
S3
(
s
en
io
r
,
h
ig
h
-
in
co
m
e,
m
ar
r
ied
clien
ts
)
,
Ma
r
i
tal
s
tatu
s
(
0
.
8
)
em
er
g
es
as
t
h
e
m
o
s
t
u
s
ef
u
l
f
ea
tu
r
e.
T
h
is
m
ap
p
in
g
illu
s
tr
ates
h
o
w
th
e
Q
-
tab
le
h
elp
s
th
e
R
L
ag
en
t
to
iter
ativ
ely
id
en
tify
an
d
p
r
i
o
r
itize
th
e
m
o
s
t
in
f
o
r
m
ativ
e
f
ea
tu
r
es
f
o
r
p
r
ed
ictin
g
th
e
clien
t
’
s
elig
ib
le
cr
ed
it
b
u
ck
et
.
B
y
s
elec
tin
g
th
e
f
ea
tu
r
e
with
th
e
h
ig
h
est
Q
-
v
alu
e
in
ea
ch
s
tate,
th
e
ag
en
t
p
r
o
g
r
ess
iv
ely
co
n
v
er
g
es
to
war
d
an
o
p
tim
al
f
ea
tu
r
e
s
u
b
s
et
th
at
m
ax
im
izes c
lass
if
icatio
n
p
er
f
o
r
m
an
ce
.
T
ab
le
5
.
Q
-
l
ea
r
n
in
g
tab
le
s
am
p
le
f
o
r
d
ataset
1
S
t
a
t
e
A
c
t
i
o
n
(
A
g
e
)
A
c
t
i
o
n
(
I
n
c
o
m
e
)
A
c
t
i
o
n
(
M
a
r
i
t
a
l
s
t
a
t
u
s)
S
1
(
Y
o
u
n
g
-
l
o
w
i
n
c
o
m
e
,
s
i
n
g
l
e
)
0
.
5
0
.
7
0
.
2
S
2
(
M
i
d
d
l
e
-
A
g
e
,
M
e
d
i
u
m
,
M
a
r
r
i
e
d
)
0
.
1
0
.
9
0
.
4
S
3
(
S
e
n
i
o
r
,
H
i
g
h
i
n
c
o
m
e
,
M
a
r
r
i
e
d
)
0
.
3
0
.
6
0
.
8
3
.
3
.
Cro
s
s
v
a
lid
a
t
i
o
n
T
o
en
s
u
r
e
a
f
air
an
d
r
o
b
u
s
t
ev
alu
atio
n
,
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
p
r
o
ce
d
u
r
e
was
em
p
lo
y
e
d
.
T
h
e
f
u
ll
d
ataset
was
d
iv
id
ed
in
to
ten
ap
p
r
o
x
im
ately
eq
u
al
f
o
ld
s
wh
ile
p
r
eser
v
in
g
th
e
class
d
is
tr
ib
u
tio
n
in
ea
c
h
f
o
l
d
.
I
n
ea
ch
iter
atio
n
,
n
in
e
f
o
ld
s
w
er
e
u
s
ed
f
o
r
tr
ai
n
in
g
an
d
o
n
e
-
f
o
ld
was
r
eser
v
ed
f
o
r
v
alid
atio
n
,
an
d
th
e
p
r
o
ce
s
s
was
r
ep
ea
ted
u
n
til
e
v
er
y
f
o
l
d
h
ad
b
ee
n
u
s
ed
o
n
ce
as
v
alid
at
io
n
d
ata
.
T
h
e
f
in
al
p
er
f
o
r
m
an
ce
was
r
ep
o
r
ted
as
th
e
m
ea
n
v
alu
e
ac
r
o
s
s
th
e
te
n
f
o
l
d
s
.
Fo
r
im
b
alan
ce
d
d
ata
s
ets,
SMOT
E
was
ap
p
lied
o
n
ly
t
o
th
e
tr
ai
n
in
g
p
o
r
tio
n
with
in
ea
ch
f
o
ld
to
a
v
o
id
d
ata
leak
ag
e.
T
h
e
v
alid
atio
n
f
o
ld
r
em
ain
e
d
u
n
to
u
ch
e
d
an
d
r
etain
ed
th
e
o
r
ig
in
al
class
d
is
tr
ib
u
tio
n
.
T
h
is
p
r
o
to
co
l
en
s
u
r
es
th
at
p
er
f
o
r
m
an
ce
esti
m
ates
ar
e
u
n
b
iased
an
d
m
o
r
e
r
ep
r
esen
tativ
e
o
f
r
ea
l
-
w
o
r
ld
g
en
er
ali
za
tio
n
.
I
n
a
d
d
itio
n
,
th
e
RL
p
r
o
ce
d
u
r
e
was
r
ep
ea
ted
a
cr
o
s
s
m
u
ltip
le
r
u
n
s
to
r
ed
u
ce
th
e
ef
f
ec
t
o
f
r
an
d
o
m
n
ess
in
ex
p
lo
r
atio
n
a
n
d
s
u
b
s
et
s
elec
tio
n
.
W
h
er
e
ap
p
licab
le,
r
esu
lts
ar
e
r
ep
o
r
ted
as m
ea
n
±
s
tan
d
ar
d
d
ev
iatio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.