I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
,
p
p
.
788
~
797
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
1
.
p
p
7
8
8
-
7
9
7
788
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
A
no
v
el
BE
RT
-
lo
ng
sho
rt
-
te
rm me
mo
ry
hy
brid mo
d
el f
o
r
eff
ec
tive cr
edit
card fraud
det
ectio
n
O
us
s
a
m
a
Nda
m
a
1
,
Sa
f
a
e
Nd
a
m
a
1
,
I
s
ma
il B
ens
a
s
s
i
1,
2
,
E
l
M
o
k
hta
r
E
n
-
Na
im
i
1
1
D
S
A
I
2
S
R
e
sea
r
c
h
Te
a
m,
C
3
S
L
a
b
o
r
a
t
o
r
y
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
s a
n
d
T
e
c
h
n
i
q
u
e
s
o
f
T
a
n
g
i
e
r
,
A
b
d
e
l
m
a
l
e
k
Ess
a
â
d
i
U
n
i
v
e
r
si
t
y
,
Te
t
o
u
a
n
,
M
o
r
o
c
c
o
2
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
s
o
f
R
a
b
a
t
,
M
o
h
a
m
me
d
V
U
n
i
v
e
r
s
i
t
y
,
R
a
b
a
t
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
1
,
2
0
2
4
R
ev
is
ed
De
c
2
1
,
2
0
2
5
Acc
ep
ted
J
an
1
0
,
2
0
2
6
In
th
e
ra
p
i
d
ly
e
v
o
lv
in
g
lan
d
sc
a
p
e
o
f
fin
a
n
c
ial
tran
sa
c
ti
o
n
s,
th
e
d
e
tec
ti
o
n
o
f
fra
u
d
u
le
n
t
a
c
ti
v
it
ies
re
m
a
in
s
a
c
rit
ica
l
c
h
a
ll
e
n
g
e
fo
r
fi
n
a
n
c
ial
i
n
stit
u
t
io
n
s
wo
rld
wi
d
e
.
Th
is
st
u
d
y
in
tro
d
u
c
e
s
a
n
o
v
e
l
b
i
d
irec
ti
o
n
a
l
e
n
c
o
d
e
r
re
p
re
se
n
tatio
n
fr
o
m
tran
sf
o
rm
e
rs
(BERT
)
–
l
o
n
g
sh
o
rt
-
term
m
e
m
o
ry
(LS
TM
)
h
y
b
rid
m
o
d
e
l
th
a
t
in
teg
ra
tes
b
o
th
te
x
tu
a
l
a
n
d
n
u
m
e
rica
l
d
a
ta
t
o
e
n
h
a
n
c
e
c
re
d
it
c
a
rd
fra
u
d
d
e
tec
ti
o
n
.
Lev
e
ra
g
in
g
BERT
f
o
r
d
e
e
p
c
o
n
tex
t
u
a
l
e
m
b
e
d
d
in
g
s
a
n
d
LS
T
M
f
o
r
se
q
u
e
n
c
e
a
n
a
ly
sis,
th
e
m
o
d
e
l
p
r
o
v
i
d
e
s
a
c
o
m
p
re
h
e
n
siv
e
a
p
p
r
o
a
c
h
t
h
a
t
s
u
rp
a
ss
e
s
trad
it
io
n
a
l
fra
u
d
d
e
tec
ti
o
n
sy
ste
m
s
p
rima
ril
y
b
a
se
d
o
n
n
u
m
e
rica
l
a
n
a
ly
sis.
O
n
th
e
v
a
li
d
a
ti
o
n
se
t,
t
h
e
m
o
d
e
l
a
c
h
iev
e
d
a
re
c
a
ll
o
f
1
0
0
%
a
n
d
a
n
a
c
c
u
ra
c
y
o
f
9
9
.
1
1
%
,
h
ig
h
li
g
h
ti
n
g
stro
n
g
e
ffe
c
ti
v
e
n
e
ss
in
i
d
e
n
ti
f
y
in
g
fra
u
d
u
l
e
n
t
tran
sa
c
ti
o
n
s
u
n
d
e
r
c
las
s
i
m
b
a
lan
c
e
.
Th
ro
u
g
h
rig
o
ro
u
s
e
v
a
lu
a
ti
o
n
,
th
e
m
o
d
e
l
d
e
m
o
n
stra
ted
e
x
c
e
p
ti
o
n
a
l
a
c
c
u
ra
c
y
a
n
d
re
li
a
b
i
li
ty
,
p
r
o
m
isin
g
imp
r
o
v
e
m
e
n
ts
in
fra
u
d
d
e
tec
ti
o
n
a
n
d
m
it
ig
a
ti
o
n
.
Th
is
p
a
p
e
r
d
e
tails
t
h
e
d
e
v
e
l
o
p
m
e
n
t
a
n
d
v
a
li
d
a
ti
o
n
o
f
t
h
e
h
y
b
rid
m
o
d
e
l,
e
m
p
h
a
siz
in
g
it
s
u
se
o
f
m
ix
e
d
d
a
ta
ty
p
e
s
to
c
a
p
tu
re
c
o
m
p
lex
p
a
tt
e
rn
s
in
tran
sa
c
ti
o
n
d
a
ta.
Th
e
re
su
lt
s
in
d
ica
te
a
n
e
w
fr
o
n
t
ier
i
n
fra
u
d
d
e
t
e
c
ti
o
n
b
y
c
o
m
b
in
i
n
g
n
a
t
u
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
(NLP
)
a
n
d
se
q
u
e
n
ti
a
l
d
a
t
a
a
n
a
ly
sis
to
c
re
a
te
a
ro
b
u
st
so
l
u
ti
o
n
fo
r
r
eal
-
wo
rld
a
p
p
l
ica
ti
o
n
s,
s
u
p
p
o
rti
n
g
t
h
e
se
c
u
rit
y
a
n
d
in
teg
rit
y
o
f
fin
a
n
c
ial
sy
ste
m
s g
lo
b
a
ll
y
.
K
ey
w
o
r
d
s
:
B
E
R
T
-
L
STM
h
y
b
r
id
C
r
ed
it c
ar
d
f
r
au
d
d
etec
tio
n
Fin
an
cial
s
ec
u
r
ity
Hy
b
r
id
m
o
d
els
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
Seq
u
en
ce
an
aly
s
is
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
:
Ou
s
s
am
a
Nd
am
a
DSAI
2
S R
esear
ch
T
ea
m
,
C
3
S
L
ab
o
r
ato
r
y
,
Facu
lty
o
f
Scien
c
es a
n
d
T
ec
h
n
i
q
u
es o
f
T
a
n
g
ier
Ab
d
elm
alek
E
s
s
aâ
d
i U
n
iv
er
s
ity
T
eto
u
an
,
M
o
r
o
cc
o
E
m
ail: o
u
s
s
am
a.
n
d
am
a@
etu
.
u
ae
.
ac
.
m
a
1.
I
NT
RO
D
UCT
I
O
N
As
th
e
d
ig
ital
ag
e
p
r
o
g
r
ess
es,
cr
ed
it
ca
r
d
t
r
an
s
ac
tio
n
s
h
av
e
in
cr
ea
s
in
g
ly
b
ec
o
m
e
ce
n
tr
a
l
to
d
aily
co
m
m
er
ce
,
p
r
o
m
o
tin
g
co
n
v
e
n
ien
ce
an
d
ef
f
icien
cy
ac
r
o
s
s
g
lo
b
al
m
ar
k
ets.
W
ith
m
illi
o
n
s
o
f
tr
an
s
ac
tio
n
s
o
cc
u
r
r
in
g
ev
er
y
m
in
u
te,
th
e
f
in
an
cial
ec
o
s
y
s
tem
h
ea
v
ily
d
ep
en
d
s
o
n
th
e
tr
u
s
t
an
d
s
ec
u
r
ity
ass
o
ciate
d
wi
th
th
ese
in
ter
ac
tio
n
s
[
1
]
.
Ho
we
v
er
,
th
is
s
u
r
g
e
in
o
n
lin
e
tr
a
n
s
ac
tio
n
s
h
as
also
led
t
o
a
m
ar
k
ed
in
cr
ea
s
e
in
f
r
au
d
u
le
n
t
ac
tiv
ities
.
C
r
ed
it
ca
r
d
f
r
au
d
n
o
t
o
n
ly
in
f
licts
s
u
b
s
tan
tial
f
in
a
n
cial
lo
s
s
es
o
n
co
n
s
u
m
er
s
an
d
f
in
an
cial
in
s
titu
tio
n
s
b
u
t
also
er
o
d
es
tr
u
s
t
in
th
e
p
ay
m
en
t
s
y
s
tem
,
p
r
esen
tin
g
a
f
o
r
m
id
ab
le
th
r
ea
t
to
th
e
g
lo
b
al
ec
o
n
o
m
y
[
2
]
.
Dete
ctin
g
f
r
au
d
u
len
t
tr
an
s
ac
tio
n
s
is
a
co
m
p
l
ex
an
d
ch
allen
g
in
g
task
d
u
e
t
o
th
e
d
y
n
am
ic
a
n
d
ev
o
lv
in
g
n
atu
r
e
o
f
f
r
au
d
s
tr
ateg
ies
[
3
]
.
Fra
u
d
s
ter
s
co
n
tin
u
o
u
s
ly
ad
ap
t
an
d
em
p
l
o
y
s
o
p
h
is
ti
ca
ted
tech
n
iq
u
es
to
b
y
p
ass
tr
ad
itio
n
al
s
ec
u
r
ity
m
ea
s
u
r
es
[
4
]
,
[
5
]
.
C
o
n
s
eq
u
e
n
tly
,
th
er
e
is
a
p
r
ess
in
g
n
ee
d
f
o
r
ad
v
a
n
ce
d
an
d
ad
ap
tiv
e
f
r
a
u
d
d
etec
tio
n
s
y
s
tem
s
th
at
ca
n
ef
f
ec
tiv
ely
id
en
tif
y
an
d
m
itig
ate
f
r
au
d
u
len
t a
ctiv
ities
in
r
ea
l
-
tim
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l B
E
R
T
-
lo
n
g
s
h
o
r
t
-
term me
mo
r
y
h
yb
r
id
mo
d
el
fo
r
effe
ctive
cred
it c
a
r
d
…
(
Ou
s
s
a
ma
N
d
a
ma
)
789
R
ec
e
n
t
a
d
v
a
n
ce
m
e
n
ts
i
n
m
a
ch
i
n
e
l
ea
r
n
in
g
(
M
L
)
a
n
d
ar
tifi
ci
al
i
n
t
ell
ig
e
n
ce
(
AI
)
h
a
v
e
o
p
en
e
d
n
ew
av
e
n
u
es
f
o
r
d
e
v
el
o
p
i
s
n
g
s
o
p
h
is
t
ica
te
d
m
o
d
e
ls
ca
p
a
b
le
o
f
i
d
e
n
ti
f
y
i
n
g
s
u
b
tl
e
p
att
er
n
s
a
n
d
a
n
o
m
a
lies
in
d
ic
ati
v
e
o
f
f
r
a
u
d
.
T
h
ese
m
o
d
els
l
e
v
e
r
a
g
e
v
as
t
am
o
u
n
ts
o
f
t
r
a
n
s
a
cti
o
n
al
d
a
ta
t
o
le
ar
n
a
n
d
p
r
e
d
i
ct
f
r
au
d
u
l
en
t
b
e
h
a
v
i
o
r
w
it
h
h
i
g
h
ac
c
u
r
ac
y
.
Ho
w
ev
er
,
th
e
a
v
ail
ab
ilit
y
a
n
d
q
u
ali
ty
o
f
r
e
al
tr
an
s
ac
ti
o
n
al
d
a
ta
f
o
r
tr
ai
n
i
n
g
t
h
ese
m
o
d
els
a
r
e
o
f
t
en
li
m
it
ed
d
u
e
to
p
r
i
v
a
cy
c
o
n
ce
r
n
s
,
r
eg
u
l
at
o
r
y
r
es
tr
ict
io
n
s
,
a
n
d
t
h
e
in
h
e
r
e
n
t
s
en
s
iti
v
it
y
o
f
f
i
n
a
n
c
ial
d
at
a
[
6
]
,
[
7
]
.
T
o
ad
d
r
ess
th
ese
lim
itatio
n
s
,
r
esear
ch
er
s
h
av
e
tu
r
n
e
d
to
s
im
u
lated
d
atasets
,
wh
ich
ca
n
m
im
ic
th
e
co
m
p
lex
ity
an
d
v
a
r
iab
ilit
y
o
f
r
ea
l
-
wo
r
ld
d
ata
with
o
u
t
co
m
p
r
o
m
is
in
g
in
d
iv
i
d
u
al
p
r
iv
ac
y
.
Simu
lated
d
ataset
s
p
r
o
v
id
e
a
co
n
tr
o
lled
e
n
v
ir
o
n
m
en
t
to
test
an
d
v
alid
ate
ML
m
o
d
els,
o
f
f
e
r
in
g
v
alu
ab
le
i
n
s
ig
h
ts
an
d
en
ab
lin
g
th
e
d
ev
elo
p
m
e
n
t
o
f
m
o
r
e
r
o
b
u
s
t
f
r
au
d
d
etec
tio
n
s
y
s
tem
s
.
I
n
th
is
s
tu
d
y
,
a
s
im
u
lated
cr
e
d
it
-
ca
r
d
tr
an
s
ac
tio
n
d
ataset
p
r
o
d
u
ce
d
b
y
th
e
Sp
ar
k
o
v
d
ata
g
en
e
r
atio
n
to
o
l
is
em
p
lo
y
ed
[
8
]
.
A
h
y
b
r
i
d
f
r
au
d
-
d
etec
tio
n
m
o
d
el
is
p
r
o
p
o
s
ed
th
at
co
u
p
les
b
id
ir
ec
tio
n
al
en
c
o
d
er
r
ep
r
esen
tatio
n
f
r
o
m
tr
an
s
f
o
r
m
er
s
(
B
E
R
T
)
f
o
r
n
atu
r
al
la
n
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
with
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
n
etwo
r
k
s
f
o
r
s
eq
u
e
n
ce
m
o
d
elin
g
.
B
E
R
T
is
u
s
ed
to
en
co
d
e
tex
tu
al
f
ield
s
,
s
p
ec
if
ically
m
e
r
ch
an
t
n
am
es
a
n
d
tr
a
n
s
ac
tio
n
c
ateg
o
r
ies,
ca
p
tu
r
i
n
g
r
ich
co
n
te
x
tu
al
r
elatio
n
s
h
ip
s
[
9
]
,
[
1
0
]
,
wh
ile
L
STM
ca
p
tu
r
es
tem
p
o
r
al
d
ep
e
n
d
en
cies
p
r
esen
t
in
tr
a
n
s
ac
tio
n
s
eq
u
e
n
ce
s
[
1
1
]
,
[
1
2
]
.
B
y
in
teg
r
atin
g
th
ese
c
o
m
p
o
n
en
ts
,
th
e
m
o
d
el
lev
er
a
g
es
co
m
p
le
m
en
tar
y
tex
t
u
al
an
d
s
eq
u
en
ti
al
s
ig
n
als
to
b
etter
d
etec
t c
o
m
p
lex
f
r
au
d
p
atter
n
s
an
d
cu
r
b
f
alse p
o
s
itiv
es.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
p
ap
er
is
to
p
r
esen
t
th
e
d
ev
elo
p
m
en
t,
im
p
lem
en
tatio
n
,
a
n
d
v
alid
atio
n
o
f
o
u
r
h
y
b
r
id
B
E
R
T
-
L
STM
m
o
d
el
f
o
r
f
r
a
u
d
d
etec
tio
n
.
T
h
e
m
o
d
e
l's
ef
f
ec
tiv
en
ess
will
b
e
d
em
o
n
s
tr
ated
u
s
in
g
th
e
s
im
u
lated
d
ataset,
s
h
o
wca
s
in
g
its
p
o
ten
tial
to
s
ig
n
if
ican
tly
im
p
r
o
v
e
f
r
au
d
d
etec
tio
n
r
ate
s
.
Ad
d
itio
n
ally
,
th
e
im
p
licatio
n
s
o
f
o
u
r
f
in
d
in
g
s
,
t
h
e
s
tr
en
g
th
s
a
n
d
lim
itatio
n
s
o
f
o
u
r
ap
p
r
o
ac
h
,
a
n
d
p
o
ten
tial
av
en
u
es
f
o
r
f
u
tu
r
e
r
esear
ch
in
th
is
cr
itical
d
o
m
ai
n
will b
e
d
is
cu
s
s
ed
.
2.
RE
L
AT
E
D
WO
RK
S
2
.
1
.
T
ra
ditio
na
l
m
a
chine le
a
rning
a
nd
im
ba
la
nce
ha
nd
li
ng
Ad
d
r
ess
in
g
im
b
ala
n
ce
r
e
m
ain
s
ce
n
tr
al
to
f
r
a
u
d
d
etec
tio
n
.
B
r
esk
u
v
ien
ė
an
d
Dze
m
y
d
a
[
1
3
]
p
r
o
p
o
s
e
f
ea
tu
r
e
im
p
o
r
tan
ce
–
d
r
iv
e
n
(
F
I
D)
s
elf
-
o
r
g
an
izin
g
m
ap
s
(
S
OM
)
,
a
SOM
-
b
ased
f
ea
t
u
r
e
s
elec
tio
n
tailo
r
ed
to
s
k
ewe
d
d
ata,
alig
n
in
g
well
wi
th
m
o
d
er
n
tr
ee
/b
o
o
s
tin
g
p
ip
eli
n
es.
C
h
u
n
g
an
d
L
ee
[
1
4
]
em
p
h
asize
r
ec
all
v
ia
a
lig
h
tweig
h
t
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
,
lin
ea
r
d
is
cr
im
in
a
n
t
an
aly
s
is
(
L
DA)
-
lin
ea
r
r
e
g
r
es
s
io
n
en
s
em
b
le
with
s
im
p
le
r
u
le
lo
g
ic
to
f
av
o
r
m
i
n
o
r
ity
d
etec
tio
n
.
Af
r
iy
ie
et
a
l
.
[
1
5
]
r
ea
f
f
ir
m
th
e
p
r
ac
ticality
o
f
class
ical
b
aselin
es
u
n
d
er
u
n
d
e
r
s
am
p
lin
g
,
n
o
tab
l
y
r
an
d
o
m
f
o
r
est.
Ou
r
wo
r
k
co
m
p
lem
en
ts
th
ese
d
ir
ec
tio
n
s
b
y
in
tr
o
d
u
cin
g
a
u
n
if
ied
tex
t
–
n
u
m
e
r
ic
f
u
s
io
n
(
B
E
R
T
em
b
ed
d
in
g
s
p
lu
s
am
o
u
n
t)
with
in
a
s
in
g
le
en
d
-
to
-
e
n
d
m
o
d
el.
2
.
2
.
Dee
p sequ
ence
m
o
dels
S
e
q
u
e
n
c
e
-
a
w
a
r
e
m
e
t
h
o
d
s
c
a
p
t
u
r
e
t
e
m
p
o
r
a
l
r
e
g
u
l
a
r
i
t
i
e
s
i
n
t
r
a
n
s
a
c
t
i
o
n
s
t
r
e
a
m
s
.
F
o
r
o
u
g
h
a
n
d
M
o
m
t
a
z
i
[
1
6
]
ca
s
t
f
r
au
d
d
etec
tio
n
as
s
eq
u
e
n
ce
lab
elin
g
v
ia
L
STM
co
n
d
i
tio
n
al
r
an
d
o
m
f
ield
(
CRF
)
s
tack
,
o
u
tp
e
r
f
o
r
m
in
g
L
STM
,
g
ated
r
ec
u
r
r
en
t
u
n
it
(
GR
U)
,
an
d
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
ANN)
b
aselin
es
an
d
in
tr
o
d
u
cin
g
a
s
eq
u
en
ce
-
awa
r
e
u
n
d
er
s
am
p
li
n
g
m
eth
o
d
(
Seq
-
US)
th
at
p
r
eser
v
es
p
r
e
-
f
r
a
u
d
co
n
tex
t.
C
o
m
p
lem
en
tar
ily
,
B
o
u
lier
is
et
a
l.
[
1
7
]
in
teg
r
at
e
ex
p
lain
ab
le
A
I
with
L
ST
M
ar
ch
itectu
r
es
to
e
n
h
an
ce
tr
an
s
p
ar
en
cy
,
wh
ile
Mie
n
y
e
an
d
J
er
e
[
1
8
]
s
u
r
v
e
y
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN
)
,
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
(
R
NN)
,
L
STM
,
an
d
GR
U
ad
v
an
ce
s
t
h
at
m
o
d
el
co
m
p
le
x
s
eq
u
e
n
tial
p
atter
n
s
.
E
x
ten
d
in
g
b
ey
o
n
d
p
u
r
el
y
s
eq
u
en
tial
v
iews,
C
h
er
if
et
a
l.
[
1
9
]
em
p
lo
y
a
n
en
co
d
er
–
d
ec
o
d
er
g
r
a
p
h
n
e
u
r
al
n
etwo
r
k
o
n
lar
g
e
Sp
ar
k
o
v
d
ata
to
ex
p
lo
it
cu
s
to
m
er
–
m
er
c
h
an
t
r
elatio
n
s
,
u
s
in
g
a
g
r
ap
h
co
n
v
er
ter
an
d
b
atch
n
o
r
m
aliza
tio
n
to
s
ta
b
ilize
tr
ain
in
g
an
d
r
ep
o
r
tin
g
g
ain
s
in
p
r
ec
is
io
n
,
r
ec
all,
F1
-
s
co
r
e
,
an
d
r
ec
eiv
er
o
p
er
atin
g
ch
ar
ac
ter
is
tic
(
R
OC
)
;
th
ey
f
u
r
th
er
u
n
d
er
s
co
r
e
g
eo
s
p
atial
m
er
c
h
a
n
t
–
cu
s
to
m
er
d
is
tan
ce
as a
n
in
f
o
r
m
ativ
e
s
ig
n
al
f
o
r
f
r
au
d
.
2
.
3
.
T
ra
ns
f
o
rm
er
-
ba
s
ed
a
nd
hy
brid m
o
dels
R
ec
en
t
wo
r
k
co
m
b
in
es
tr
an
s
f
o
r
m
er
r
ep
r
esen
tatio
n
s
with
o
t
h
er
lear
n
e
r
s
.
I
leb
er
i
an
d
Su
n
[
2
0
]
p
r
esen
t
a
s
tack
in
g
en
s
em
b
le
(
C
NN,
L
STM
,
T
r
an
s
f
o
r
m
er
)
with
an
e
x
tr
em
e
g
r
ad
ie
n
t b
o
o
s
tin
g
(
XG
B
o
o
s
t
)
m
eta
-
lear
n
er
,
ac
h
iev
in
g
h
i
g
h
s
en
s
itiv
ity
,
s
p
ec
if
icity
,
an
d
ar
ea
u
n
d
er
t
h
e
cu
r
v
e
(
AUC
)
o
n
E
u
r
o
p
ea
n
an
d
T
aiwa
n
d
atasets
.
Hew
ap
ath
ir
an
a
et
a
l.
[
2
1
]
i
n
v
esti
g
ate
T
ab
B
E
R
T
f
o
r
tr
a
n
s
ac
tio
n
al
d
ep
e
n
d
en
cies,
a
n
d
N
L
P
-
ce
n
tr
ic
lin
es
b
y
[
2
2
]
,
[
2
3
]
lev
er
a
g
e
lan
g
u
ag
e
t
ec
h
n
o
lo
g
ies (
in
clu
d
in
g
ch
atb
o
ts
)
to
d
etec
t o
r
m
itig
ate
f
r
au
d
.
Mo
s
t
h
y
b
r
id
s
y
s
tem
s
ar
e
c
o
m
p
u
te
-
h
ea
v
y
an
d
eith
er
c
o
n
v
er
t
tex
t
f
ield
s
in
to
p
u
r
el
y
n
u
m
er
ic
s
u
r
r
o
g
ates
o
r
f
u
s
e
m
o
d
alities
o
n
ly
at
t
h
e
s
co
r
e
lev
el.
Ou
r
n
o
v
elty
is
a
s
in
g
le
-
b
r
an
c
h
,
r
ep
r
esen
tatio
n
-
lev
el
f
u
s
io
n
th
at
p
r
eser
v
es
th
e
NL
P
s
ig
n
al
b
y
u
s
in
g
co
n
tex
tu
al
B
E
R
T
em
b
ed
d
in
g
s
o
f
m
e
r
ch
an
t
an
d
ca
teg
o
r
y
,
an
d
in
teg
r
ates
th
is
with
th
e
n
u
m
er
ic
am
o
u
n
t
in
o
n
e
e
n
d
-
to
-
en
d
(
B
E
R
T
→f
u
s
io
n
→L
STM
)
ar
ch
itectu
r
e
r
ath
er
th
a
n
f
latten
in
g
ev
er
y
t
h
in
g
in
to
n
u
m
er
ic
.
T
h
is
p
r
eser
v
es
m
o
d
alit
y
-
s
p
ec
if
ic
in
f
o
r
m
atio
n
an
d
a
v
o
id
s
h
ea
v
y
s
tack
in
g
,
wh
ich
d
is
tin
g
u
is
h
es o
u
r
a
p
p
r
o
ac
h
f
r
o
m
p
r
io
r
wo
r
k
th
at
h
o
m
o
g
en
izes a
ll in
p
u
ts
in
to
n
u
m
er
ic
v
ec
to
r
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
788
-
7
9
7
790
3.
M
E
T
H
O
D
T
h
is
s
ec
tio
n
d
etails
th
e
d
ata
s
o
u
r
ce
s
,
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
,
an
d
th
e
m
eth
o
d
o
l
o
g
ies
em
p
lo
y
ed
in
d
ev
elo
p
in
g
th
e
cr
ed
it
ca
r
d
f
r
au
d
d
etec
tio
n
m
o
d
el.
T
h
e
d
ataset
is
f
ir
s
t
d
escr
ib
e
d
,
f
o
llo
wed
b
y
th
e
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
a
p
p
l
ied
to
p
r
e
p
ar
e
th
e
d
ata
f
o
r
m
o
d
elin
g
.
Su
b
s
eq
u
en
tly
,
t
h
e
im
p
lem
en
tatio
n
d
etails
ar
e
o
u
tlin
ed
,
in
clu
d
in
g
th
e
al
g
o
r
ith
m
s
u
s
ed
a
n
d
th
e
a
r
ch
it
ec
tu
r
e
o
f
p
r
o
p
o
s
ed
B
E
R
T
-
L
STM
h
y
b
r
i
d
m
o
d
el.
Fig
u
r
e
1
illu
s
tr
ates
th
e
h
y
b
r
id
m
o
d
el
ar
ch
itectu
r
e
d
iag
r
am
,
wh
ich
v
is
u
ally
r
ep
r
esen
ts
all
th
e
s
tep
s
f
o
llo
wed
to
b
u
ild
th
e
m
o
d
el
an
d
p
r
o
v
id
es
a
f
r
am
ewo
r
k
f
o
r
th
e
d
is
cu
s
s
io
n
s
in
th
is
s
ec
tio
n
.
Fig
u
r
e
1
.
T
h
e
ar
ch
itectu
r
e
d
ia
g
r
am
o
f
th
e
B
E
R
T
-
L
STM
h
y
b
r
id
m
o
d
el
f
o
r
c
r
ed
it c
ar
d
f
r
au
d
d
etec
tio
n
3
.
1
.
Da
t
a
s
et
T
h
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
is
a
s
im
u
lated
cr
ed
it
ca
r
d
tr
an
s
ac
tio
n
d
ataset
g
en
er
ated
u
s
in
g
th
e
Sp
ar
k
o
v
Data
Gen
er
ati
o
n
t
o
o
l
.
I
t
c
o
n
tain
s
tr
a
n
s
ac
tio
n
s
f
r
o
m
1
,
0
0
0
cu
s
to
m
er
s
an
d
8
0
0
m
er
ch
an
ts
,
s
p
a
n
n
in
g
th
e
p
er
io
d
f
r
o
m
J
an
u
ar
y
1
,
2
0
1
9
to
Dec
em
b
er
3
1
,
2
0
2
0
with
a
to
tal
o
f
1
,
2
9
6
,
6
7
5
r
ec
o
r
d
s
.
T
h
e
d
ataset
in
clu
d
es
b
o
th
leg
itima
te
an
d
f
r
au
d
u
len
t
tr
an
s
ac
tio
n
s
,
with
f
ea
tu
r
e
s
s
u
ch
as
tr
an
s
ac
tio
n
am
o
u
n
t,
m
er
ch
an
t
n
am
e,
ca
teg
o
r
y
,
a
n
d
a
b
in
a
r
y
f
r
a
u
d
la
b
el
in
d
icatin
g
wh
eth
er
a
tr
a
n
s
ac
tio
n
is
f
r
au
d
u
len
t,
t
h
e
f
ea
tu
r
es in
th
e
d
ataset
ar
e
as p
r
esen
ted
in
T
ab
le
1
.
T
h
is
d
iv
er
s
e
s
et
o
f
f
ea
tu
r
es
p
r
o
v
id
es
a
r
ich
co
n
te
x
t
f
o
r
ea
ch
tr
an
s
ac
tio
n
,
ca
p
tu
r
in
g
n
o
t
o
n
l
y
tr
an
s
ac
tio
n
al
d
etails
b
u
t
also
d
em
o
g
r
ap
h
ic
an
d
s
p
atial
in
f
o
r
m
atio
n
ab
o
u
t
b
o
th
ca
r
d
h
o
ld
e
r
s
an
d
m
er
c
h
an
ts
.
B
y
co
m
b
in
in
g
tex
tu
al,
ca
teg
o
r
ica
l,
an
d
n
u
m
er
ic
attr
ib
u
tes,
th
e
d
ataset
en
ab
les
in
ter
ac
tio
n
s
th
at
r
ev
ea
l
s
u
b
tle
p
atter
n
s
ass
o
ciate
d
with
f
r
au
d
.
T
h
is
r
ich
n
ess
allo
ws
f
o
r
co
m
p
r
eh
en
s
iv
e
a
n
aly
s
is
an
d
m
o
r
e
r
o
b
u
s
t
m
o
d
elin
g
o
f
f
r
au
d
u
le
n
t b
eh
a
v
io
r
.
T
ab
le
1
.
Data
s
et
f
ea
tu
r
es
F
e
a
t
u
r
e
D
e
scri
p
t
i
o
n
Ty
p
e
t
r
a
n
s
_
d
a
t
e
_
t
r
a
n
s
_
t
i
me
Th
e
d
a
t
e
a
n
d
t
i
m
e
o
f
t
h
e
t
r
a
n
sa
c
t
i
o
n
o
b
j
e
c
t
c
c
_
n
u
m
Th
e
c
r
e
d
i
t
c
a
r
d
n
u
m
b
e
r
u
s
e
d
i
n
t
h
e
t
r
a
n
sa
c
t
i
o
n
i
n
t
6
4
merch
a
n
t
Th
e
n
a
me
o
f
t
h
e
m
e
r
c
h
a
n
t
w
h
e
r
e
t
h
e
t
r
a
n
sa
c
t
i
o
n
o
c
c
u
r
r
e
d
o
b
j
e
c
t
c
a
t
e
g
o
r
y
Th
e
c
a
t
e
g
o
r
y
o
f
t
h
e
m
e
r
c
h
a
n
t
o
b
j
e
c
t
a
mt
Th
e
a
m
o
u
n
t
o
f
t
h
e
t
r
a
n
s
a
c
t
i
o
n
f
l
o
a
t
6
4
f
i
r
st
Th
e
f
i
r
st
n
a
me
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
l
a
st
Th
e
l
a
st
n
a
me
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
g
e
n
d
e
r
Th
e
g
e
n
d
e
r
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
st
r
e
e
t
Th
e
st
r
e
e
t
a
d
d
r
e
ss
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
c
i
t
y
Th
e
c
i
t
y
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
st
a
t
e
Th
e
st
a
t
e
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
z
i
p
Th
e
ZI
P
c
o
d
e
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
i
n
t
6
4
l
a
t
Th
e
l
a
t
i
t
u
d
e
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
's
l
o
c
a
t
i
o
n
f
l
o
a
t
6
4
l
o
n
g
Th
e
l
o
n
g
i
t
u
d
e
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
's
l
o
c
a
t
i
o
n
f
l
o
a
t
6
4
c
i
t
y
_
p
o
p
Th
e
p
o
p
u
l
a
t
i
o
n
o
f
t
h
e
c
i
t
y
w
h
e
r
e
t
h
e
c
a
r
d
h
o
l
d
e
r
r
e
s
i
d
e
s
i
n
t
6
4
j
o
b
Th
e
o
c
c
u
p
a
t
i
o
n
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
d
o
b
Th
e
d
a
t
e
o
f
b
i
r
t
h
o
f
t
h
e
c
a
r
d
h
o
l
d
e
r
o
b
j
e
c
t
t
r
a
n
s
_
n
u
m
Th
e
t
r
a
n
sa
c
t
i
o
n
n
u
m
b
e
r
o
b
j
e
c
t
u
n
i
x
_
t
i
m
e
Th
e
t
r
a
n
sa
c
t
i
o
n
t
i
me
i
n
U
n
i
x
t
i
me
f
o
r
mat
i
n
t
6
4
merch
_
l
a
t
Th
e
l
a
t
i
t
u
d
e
o
f
t
h
e
merc
h
a
n
t
's
l
o
c
a
t
i
o
n
f
l
o
a
t
6
4
merch
_
l
o
n
g
Th
e
l
o
n
g
i
t
u
d
e
o
f
t
h
e
mer
c
h
a
n
t
's
l
o
c
a
t
i
o
n
f
l
o
a
t
6
4
i
s_
f
r
a
u
d
A
b
i
n
a
r
y
l
a
b
e
l
i
n
d
i
c
a
t
i
n
g
w
h
e
t
h
e
r
t
h
e
t
r
a
n
s
a
c
t
i
o
n
i
s
f
r
a
u
d
u
l
e
n
t
i
n
t
6
4
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l B
E
R
T
-
lo
n
g
s
h
o
r
t
-
term me
mo
r
y
h
yb
r
id
mo
d
el
fo
r
effe
ctive
cred
it c
a
r
d
…
(
Ou
s
s
a
ma
N
d
a
ma
)
791
3
.
2
.
P
re
pro
ce
s
s
ing
Data
s
et
p
r
ep
r
o
ce
s
s
in
g
is
a
cr
u
cial
s
tep
in
p
r
ep
ar
in
g
t
h
e
d
a
ta
f
o
r
m
o
d
el
tr
ain
i
n
g
.
T
h
e
p
r
e
p
r
o
ce
s
s
in
g
s
tep
s
f
o
r
th
is
s
tu
d
y
in
clu
d
e:
i)
Data
s
am
p
lin
g
:
t
h
e
o
r
ig
in
al
d
a
taset
co
n
tain
s
1
,
2
9
6
,
6
7
5
r
ec
o
r
d
s
,
wh
ich
ca
n
b
e
co
m
p
u
tatio
n
ally
in
ten
s
iv
e
to
p
r
o
ce
s
s
.
T
o
b
alan
ce
c
o
m
p
u
tatio
n
al
ef
f
icien
cy
with
m
o
d
el
p
er
f
o
r
m
an
ce
,
a
r
a
n
d
o
m
s
am
p
le
o
f
1
0
0
,
0
0
0
r
ec
o
r
d
s
was
s
elec
ted
f
o
r
a
n
a
ly
s
is
.
T
h
is
s
am
p
lin
g
en
s
u
r
es
th
at
th
e
s
u
b
s
et
o
f
d
ata
r
etai
n
s
th
e
o
v
er
all
ch
ar
ac
ter
is
tics
an
d
d
is
tr
ib
u
tio
n
o
f
th
e
en
tire
d
ataset,
f
ac
ilit
atin
g
ef
f
ec
tiv
e
m
o
d
el
tr
ain
i
n
g
a
n
d
e
v
alu
atio
n
.
Fro
m
th
is
s
am
p
le
d
d
ata,
o
n
l
y
th
e
r
elev
a
n
t
co
l
u
m
n
s
'
m
er
ch
an
t'
,
'
ca
teg
o
r
y
'
,
'
am
t
'
,
an
d
'
is
_
f
r
au
d
'
wer
e
s
elec
ted
.
T
h
is
s
tep
f
o
cu
s
ed
th
e
an
aly
s
is
o
n
th
e
f
ea
tu
r
es
d
ir
ec
tly
p
er
tin
e
n
t
to
d
etec
ti
n
g
f
r
a
u
d
u
len
t
tr
an
s
ac
tio
n
s
,
s
im
p
lify
in
g
th
e
d
ataset
,
an
d
en
h
an
ci
n
g
co
m
p
u
ta
tio
n
al
ef
f
icien
cy
.
ii)
Han
d
lin
g
m
is
s
in
g
v
alu
es
:
a
n
y
m
is
s
in
g
v
alu
es
in
th
e
d
atase
t
wer
e
h
an
d
led
u
s
in
g
ap
p
r
o
p
r
i
ate
im
p
u
tatio
n
tech
n
iq
u
es
to
m
ain
tain
d
ata
in
teg
r
ity
[
2
4
]
.
T
h
is
s
tep
en
s
u
r
es
th
at
th
e
d
ataset
is
co
m
p
lete
an
d
s
u
itab
le
f
o
r
m
o
d
el
tr
ain
in
g
with
o
u
t in
t
r
o
d
u
cin
g
b
ias o
r
in
ac
cu
r
ac
ies.
iii)
Featu
r
e
en
g
i
n
ee
r
in
g
:
t
o
en
h
a
n
ce
th
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
,
a
d
d
itio
n
al
f
ea
tu
r
es
wer
e
en
g
in
e
er
ed
f
r
o
m
th
e
ex
is
tin
g
d
ata.
T
h
e
g
o
al
o
f
f
e
atu
r
e
en
g
in
ee
r
in
g
is
to
cr
ea
te
n
ew
in
p
u
t
f
ea
tu
r
es
th
at
ca
n
im
p
r
o
v
e
th
e
p
r
ed
ictiv
e
p
o
wer
o
f
th
e
m
o
d
el
[
2
5
]
,
[
2
6
]
.
I
n
th
is
s
tu
d
y
,
te
x
tu
al
f
ea
tu
r
es
s
u
ch
as
'
m
er
ch
an
t'
an
d
'
ca
teg
o
r
y
'
wer
e
co
m
b
in
ed
to
cr
ea
te
a
s
in
g
le,
en
r
ich
ed
tex
tu
al
in
p
u
t.
T
h
is
co
m
b
in
ed
f
ea
tu
r
e
en
ca
p
s
u
lates
th
e
id
en
tity
o
f
th
e
m
er
ch
a
n
t
an
d
th
e
ty
p
e
o
f
tr
an
s
ac
tio
n
,
p
r
o
v
id
in
g
a
r
ich
e
r
co
n
tex
t
f
o
r
t
h
e
m
o
d
el.
B
y
m
er
g
in
g
th
ese
two
co
lu
m
n
s
,
we
aim
to
ca
p
tu
r
e
th
e
r
elatio
n
s
h
ip
b
etwe
en
wh
er
e
th
e
tr
a
n
s
ac
tio
n
to
o
k
p
lace
an
d
wh
at
k
in
d
o
f
tr
an
s
ac
tio
n
it
was,
wh
ich
ca
n
b
e
cr
u
cial
f
o
r
id
en
tify
in
g
f
r
au
d
p
atter
n
s
.
T
h
is
n
ew
tex
tu
al
f
ea
tu
r
e
was th
en
p
r
ep
a
r
ed
f
o
r
f
u
r
th
er
p
r
o
ce
s
s
in
g
an
d
em
b
ed
d
in
g
.
iv
)
T
o
k
en
izatio
n
an
d
em
b
e
d
d
in
g
:
t
h
e
co
m
b
in
ed
tex
tu
al
f
ea
tu
r
e
s
wer
e
to
k
en
ize
d
an
d
em
b
ed
d
ed
u
s
in
g
th
e
B
E
R
T
to
k
en
izer
an
d
m
o
d
el
[
2
7
]
.
T
o
k
en
izatio
n
is
th
e
p
r
o
ce
s
s
o
f
co
n
v
er
tin
g
tex
t
in
t
o
s
m
aller
u
n
its
ca
lle
d
to
k
en
s
[
2
8
]
.
T
h
e
B
E
R
T
to
k
en
izer
b
r
ea
k
s
d
o
wn
th
e
co
m
b
in
ed
tex
tu
al
in
p
u
t
in
to
to
k
e
n
s
a
n
d
m
a
p
s
ea
ch
to
k
en
t
o
a
u
n
iq
u
e
n
u
m
er
ical
i
d
en
tifie
r
,
c
r
ea
tin
g
to
k
en
I
Ds.
T
h
ese
to
k
en
I
Ds
ar
e
th
en
f
e
d
in
to
th
e
B
E
R
T
m
o
d
el,
wh
ich
g
en
er
ates
n
u
m
e
r
ical
em
b
ed
d
in
g
s
f
o
r
ea
ch
to
k
en
.
B
E
R
T
em
b
ed
d
in
g
s
ar
e
co
n
tex
tu
ally
r
ic
h
v
ec
to
r
r
ep
r
esen
tatio
n
s
th
at
ca
p
tu
r
e
th
e
s
em
an
tic
m
ea
n
in
g
a
n
d
r
elatio
n
s
h
ip
s
with
in
th
e
te
x
t
[
2
9
]
.
T
h
is
p
r
o
ce
s
s
tr
an
s
f
o
r
m
s
th
e
tex
tu
al
d
ata
in
to
a
f
o
r
m
at
th
at
th
e
m
o
d
el
ca
n
ef
f
ec
tiv
ely
u
s
e,
allo
win
g
it
to
u
n
d
er
s
tan
d
an
d
le
v
er
ag
e
th
e
c
o
n
tex
tu
al
n
u
an
ce
s
o
f
th
e
in
p
u
t
tex
t.
v)
C
o
m
b
in
in
g
f
ea
tu
r
es
:
th
e
g
e
n
er
ated
te
x
t
em
b
ed
d
in
g
s
we
r
e
co
m
b
in
ed
with
th
e
n
u
m
er
ical
f
ea
tu
r
e,
tr
an
s
ac
tio
n
am
o
u
n
t
(
a
m
t)
.
T
h
is
s
tep
in
teg
r
ates
th
e
c
o
n
te
x
tu
ally
r
ich
tex
tu
al
in
f
o
r
m
at
io
n
with
th
e
q
u
an
titativ
e
d
ata,
f
o
r
m
in
g
a
c
o
m
p
r
eh
e
n
s
iv
e
f
ea
tu
r
e
s
et
f
o
r
ea
ch
tr
an
s
ac
tio
n
.
T
h
e
co
m
b
in
atio
n
o
f
tex
tu
a
l
an
d
n
u
m
er
ical
d
ata
allo
ws
th
e
m
o
d
el
t
o
co
n
s
id
er
b
o
th
th
e
s
em
an
tic
co
n
tex
t
o
f
th
e
tr
a
n
s
ac
tio
n
an
d
its
m
o
n
etar
y
v
alu
e.
B
y
m
e
r
g
in
g
th
ese
d
iv
er
s
e
f
ea
tu
r
es,
th
e
m
o
d
el
g
ain
s
a
h
o
lis
tic
v
iew
o
f
ea
c
h
tr
an
s
ac
tio
n
,
en
h
an
cin
g
its
ab
ilit
y
to
id
en
tif
y
f
r
au
d
u
len
t
ac
tiv
ities
b
ased
o
n
b
o
th
th
e
n
atu
r
e
o
f
th
e
tr
an
s
ac
tio
n
an
d
its
f
in
an
cial
attr
ib
u
tes.
v
i)
C
las
s
im
b
alan
ce
h
an
d
li
n
g
:
th
e
d
ataset
ex
h
ib
ited
class
im
b
alan
ce
,
with
f
r
a
u
d
u
len
t
tr
an
s
ac
tio
n
s
b
ein
g
s
ig
n
if
ican
tly
f
ewe
r
th
an
leg
itima
te
o
n
es.
Fra
u
d
u
len
t
r
ec
o
r
d
s
r
ep
r
esen
t
ap
p
r
o
x
im
ately
0
.
6
0
%
o
f
th
e
en
tir
e
d
ataset.
C
lass
im
b
alan
ce
ca
n
n
eg
ativ
ely
im
p
ac
t
th
e
m
o
d
e
l's
p
er
f
o
r
m
a
n
ce
,
as
it
m
ay
b
ec
o
m
e
b
iased
to
war
d
s
th
e
m
aj
o
r
ity
class
(
le
g
itima
te
tr
an
s
ac
tio
n
s
)
an
d
f
ail
to
d
etec
t
f
r
au
d
u
len
t
o
n
es.
T
o
ad
d
r
ess
th
is
is
s
u
e,
th
e
s
y
n
th
etic
m
in
o
r
ity
o
v
er
-
s
am
p
lin
g
tec
h
n
iq
u
e
(
SMOT
E
)
was
ap
p
lied
.
SMO
T
E
g
en
er
ate
s
s
y
n
th
etic
s
am
p
les
f
o
r
th
e
m
in
o
r
ity
class
(
f
r
au
d
u
len
t
tr
a
n
s
ac
tio
n
s
)
b
y
in
ter
p
o
latin
g
b
et
wee
n
ex
is
tin
g
m
in
o
r
ity
class
s
am
p
les
[
3
0
]
.
T
h
is
tech
n
iq
u
e
e
n
s
u
r
es
a
b
al
an
ce
d
r
ep
r
esen
tatio
n
o
f
b
o
t
h
class
es
d
u
r
in
g
m
o
d
el
tr
ain
in
g
,
en
ab
lin
g
t
h
e
m
o
d
el
to
lear
n
an
d
d
etec
t
f
r
au
d
u
len
t
tr
an
s
ac
tio
n
s
m
o
r
e
ef
f
ec
tiv
ely
[
3
1
]
.
B
y
ad
d
r
ess
in
g
class
im
b
alan
ce
,
th
e
m
o
d
el'
s
s
en
s
i
tiv
ity
to
f
r
au
d
u
le
n
t
ac
tiv
ities
is
im
p
r
o
v
ed
,
an
d
th
e
lik
elih
o
o
d
o
f
f
alse n
eg
ativ
es is
r
ed
u
ce
d
.
T
h
ese
p
r
ep
r
o
ce
s
s
in
g
s
tep
s
p
r
e
p
ar
e
th
e
d
ataset
f
o
r
ef
f
e
ctiv
e
m
o
d
el
tr
ain
i
n
g
.
B
y
co
n
s
o
lid
ati
n
g
r
ele
v
an
t
f
ea
tu
r
es,
g
en
e
r
atin
g
B
E
R
T
-
r
e
ad
y
tex
t
em
b
ed
d
i
n
g
s
,
an
d
c
o
r
r
ec
tin
g
class
im
b
alan
ce
,
th
ey
im
p
r
o
v
e
s
ig
n
al
q
u
ality
an
d
s
tab
ilit
y
.
C
o
n
s
eq
u
en
tly
,
th
e
m
o
d
el
lear
n
s
m
o
r
e
r
o
b
u
s
tly
an
d
d
etec
ts
f
r
a
u
d
u
le
n
t
tr
an
s
ac
tio
n
s
with
h
ig
h
er
ac
c
u
r
ac
y
.
3
.
3
.
Det
a
ils
o
f
im
plem
ent
a
t
i
o
n
T
h
e
im
p
lem
en
tatio
n
o
f
o
u
r
B
E
R
T
-
L
STM
h
y
b
r
id
m
o
d
el
f
o
r
cr
ed
it
ca
r
d
f
r
au
d
d
etec
tio
n
was
ca
r
r
ied
o
u
t
o
n
Go
o
g
le
C
o
lab
,
u
tili
zin
g
th
e
p
o
wer
f
u
l
NVI
DI
A
A1
0
0
g
r
ap
h
ics
p
r
o
ce
s
s
in
g
u
n
it
(
GPU
)
to
m
ee
t
th
e
co
m
p
u
tatio
n
al
n
ee
d
s
o
f
d
ee
p
l
ea
r
n
in
g
.
Py
t
h
o
n
s
er
v
ed
as
th
e
p
r
im
ar
y
la
n
g
u
a
g
e,
s
u
p
p
o
r
ted
b
y
lib
r
ar
ies
s
u
ch
a
s
Pan
d
as
an
d
Nu
m
Py
f
o
r
d
ata
p
r
o
ce
s
s
in
g
,
Py
T
o
r
ch
f
o
r
m
o
d
el
d
ev
elo
p
m
en
t,
an
d
Hu
g
g
in
g
F
ac
e’
s
tr
an
s
f
o
r
m
e
r
s
lib
r
ar
y
f
o
r
e
f
f
icien
t
h
a
n
d
lin
g
o
f
tex
t
u
al
d
ata
with
B
E
R
T
.
Scik
it
-
lear
n
was
u
s
ed
f
o
r
d
ata
s
p
litt
in
g
an
d
ev
alu
atio
n
,
w
h
ile
im
b
alan
ce
d
-
lear
n
an
d
SMOT
E
ad
d
r
ess
ed
class
im
b
alan
ce
.
T
h
is
s
etu
p
en
s
u
r
e
d
ef
f
icien
t,
ef
f
ec
tiv
e
m
o
d
el
d
ev
elo
p
m
en
t,
lev
er
ag
in
g
s
tate
-
of
-
th
e
-
ar
t
to
o
l
s
f
o
r
lar
g
e
-
s
ca
le
d
ata
a
n
d
d
ee
p
lear
n
in
g
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
788
-
7
9
7
792
3
.
4
.
Alg
o
rit
hm
s
T
h
e
d
e
v
elo
p
m
e
n
t
o
f
t
h
e
cr
e
d
i
t
ca
r
d
f
r
au
d
d
etec
tio
n
m
o
d
el
em
p
lo
y
ed
two
m
ain
alg
o
r
ith
m
s
:
B
E
R
T
an
d
L
STM
.
E
ac
h
o
f
th
ese
al
g
o
r
ith
m
s
p
la
y
s
a
cr
itical
r
o
le
in
h
a
n
d
lin
g
d
if
f
e
r
en
t
asp
ec
ts
o
f
t
h
e
d
ataset
an
d
m
o
d
elin
g
c
h
allen
g
es.
B
E
R
T
en
co
d
es
co
n
te
x
tu
al
s
em
an
tics
f
r
o
m
m
er
c
h
an
t
a
n
d
ca
teg
o
r
y
te
x
t,
wh
er
ea
s
L
STM
ca
p
tu
r
es
tem
p
o
r
al
d
ep
en
d
en
ci
es
ac
r
o
s
s
tr
an
s
ac
tio
n
s
,
allo
win
g
th
e
h
y
b
r
id
to
ad
d
r
ess
co
m
p
l
em
en
tar
y
f
ac
ets
o
f
f
r
au
d
d
etec
tio
n
.
3
.
4
.
1
.
B
idi
re
ct
io
na
l e
nco
der
re
presenta
t
io
ns
f
ro
m
t
ra
ns
f
o
rm
er
s
T
h
e
B
E
R
T
,
d
ev
el
o
p
ed
b
y
G
o
o
g
le,
is
a
g
r
o
u
n
d
b
r
ea
k
in
g
m
o
d
el
in
th
e
f
ield
o
f
NL
P.
I
t
u
s
es
a
m
ec
h
an
is
m
k
n
o
wn
as
t
r
an
s
f
o
r
m
er
s
to
u
n
d
er
s
tan
d
th
e
co
n
te
x
t o
f
a
wo
r
d
with
in
a
tex
t,
r
ath
er
th
an
ju
s
t th
e
wo
r
d
in
is
o
latio
n
[
3
2
]
.
T
h
is
ab
ilit
y
m
ak
es
B
E
R
T
ex
tr
em
ely
ef
f
ec
tiv
e
f
o
r
task
s
th
at
r
ely
o
n
th
e
co
n
tex
tu
al
u
s
e
o
f
wo
r
d
s
,
s
u
ch
as
s
en
tim
en
t
an
aly
s
is
,
n
am
ed
en
tity
r
ec
o
g
n
itio
n
,
an
d
in
o
u
r
ca
s
e,
p
r
o
ce
s
s
in
g
an
d
u
n
d
er
s
tan
d
in
g
tex
tu
al
d
ata
r
elate
d
to
tr
a
n
s
ac
tio
n
s
[
3
3
]
.
I
n
th
is
p
r
o
ject,
B
E
R
T
was
u
tili
ze
d
to
p
r
o
ce
s
s
tex
t
u
al
f
ea
tu
r
es
s
u
ch
as
m
er
ch
an
t
n
a
m
es
an
d
tr
an
s
ac
t
io
n
ca
teg
o
r
ies.
B
y
tr
an
s
f
o
r
m
i
n
g
th
ese
tex
tu
al
in
p
u
ts
in
to
em
b
ed
d
e
d
v
ec
to
r
s
,
B
E
R
T
p
r
o
v
id
ed
a
n
u
a
n
ce
d
r
ep
r
esen
tatio
n
o
f
th
e
tex
t,
ca
p
tu
r
in
g
s
u
b
tle
m
ea
n
i
n
g
s
th
at
co
u
l
d
in
d
icate
f
r
a
u
d
u
le
n
t
ac
tiv
ity
.
T
h
e
em
b
e
d
d
in
g
s
g
en
er
ated
b
y
B
E
R
T
s
er
v
e
as
a
s
o
p
h
is
ticated
in
p
u
t
to
th
e
s
u
b
s
eq
u
en
t
s
tag
es
o
f
th
e
m
o
d
el,
en
h
an
cin
g
its
ab
ilit
y
to
d
is
ce
r
n
p
atter
n
s
in
d
icativ
e
o
f
f
r
au
d
.
3
.
4
.
2
.
L
o
ng
s
ho
rt
-
t
er
m m
e
mo
ry
T
h
e
L
STM
n
etwo
r
k
s
ar
e
a
ty
p
e
o
f
R
N
N
s
p
ec
if
ically
d
esig
n
ed
to
h
an
d
le
s
eq
u
en
ce
p
r
ed
ictio
n
p
r
o
b
lem
s
[
3
4
]
.
L
STM
s
ar
e
ca
p
ab
le
o
f
lear
n
in
g
lo
n
g
-
te
r
m
d
e
p
en
d
en
cies
in
s
eq
u
en
ce
d
ata,
wh
ich
is
cr
u
cial
f
o
r
ap
p
licatio
n
s
lik
e
tim
e
-
s
er
ies
an
aly
s
is
,
s
p
ee
ch
r
ec
o
g
n
itio
n
,
an
d
im
p
o
r
tan
tly
,
tr
a
n
s
ac
tio
n
s
eq
u
en
ce
an
aly
s
is
[
3
5
]
.
I
n
th
e
co
n
tex
t
o
f
cr
e
d
it
ca
r
d
f
r
au
d
d
etec
tio
n
,
L
STM
n
etwo
r
k
s
wer
e
em
p
lo
y
ed
to
an
aly
ze
th
e
s
eq
u
en
ce
s
o
f
tr
an
s
ac
tio
n
s
,
co
n
s
id
er
in
g
th
e
tem
p
o
r
al
r
elatio
n
s
h
ip
s
an
d
p
atter
n
s
th
at
em
er
g
e
o
v
er
ti
m
e.
B
y
in
teg
r
atin
g
L
STM
with
B
E
R
T
em
b
ed
d
in
g
s
,
th
e
m
o
d
el
c
o
u
ld
e
f
f
ec
tiv
ely
lev
er
ag
e
b
o
th
th
e
co
n
tex
t
u
al
an
d
s
eq
u
e
n
tial
in
f
o
r
m
atio
n
in
th
e
d
ataset.
T
h
i
s
in
teg
r
atio
n
allo
ws
th
e
L
ST
M
to
in
ter
p
r
et
th
e
em
b
ed
d
ed
tex
t
in
th
e
co
n
te
x
t
o
f
tr
an
s
ac
tio
n
s
eq
u
en
ce
s
,
en
h
a
n
c
in
g
its
ab
ilit
y
to
p
r
ed
ict
f
r
au
d
u
len
t
tr
a
n
s
ac
tio
n
s
b
ased
o
n
b
eh
av
io
r
al
p
atter
n
s
th
at
u
n
f
o
ld
o
v
er
tim
e.
T
o
g
eth
e
r
,
B
E
R
T
an
d
L
STM
f
o
r
m
a
p
o
wer
f
u
l c
o
m
b
in
atio
n
f
o
r
tack
lin
g
th
e
co
m
p
lex
ities
o
f
f
r
a
u
d
d
etec
tio
n
in
tr
a
n
s
ac
tio
n
d
ata.
B
E
R
T
’
s
d
ee
p
u
n
d
er
s
t
an
d
in
g
o
f
tex
tu
al
c
o
n
tex
t,
c
o
u
p
led
with
L
STM
’
s
p
r
o
f
icien
c
y
in
s
eq
u
en
ce
m
o
d
elin
g
,
p
r
o
v
id
es
a
co
m
p
r
eh
e
n
s
iv
e
ap
p
r
o
ac
h
to
i
d
en
tify
in
g
f
r
au
d
u
len
t
ac
tiv
ities
with
h
ig
h
er
ac
c
u
r
ac
y
a
n
d
ef
f
ic
ien
cy
.
4.
T
H
E
P
RO
P
O
SE
D
B
E
RT
-
L
ST
M
H
YB
R
I
D
M
O
D
E
L
I
n
th
is
s
tu
d
y
,
a
h
y
b
r
id
m
o
d
el
was
d
ev
elo
p
ed
th
at
co
m
b
in
e
s
th
e
s
tr
en
g
th
s
o
f
B
E
R
T
an
d
L
STM
to
en
h
an
ce
c
r
ed
it
ca
r
d
f
r
a
u
d
d
etec
tio
n
ca
p
ab
ilit
ies.
T
h
is
s
ec
tio
n
d
etails
th
e
m
o
d
el
ar
c
h
itectu
r
e
an
d
th
e
tr
ai
n
in
g
an
d
v
alid
atio
n
m
eth
o
d
o
lo
g
ies
u
s
ed
to
co
n
s
tr
u
ct
a
n
d
d
e
p
lo
y
th
e
s
y
s
tem
ef
f
ec
tiv
ely
.
I
t
also
h
ig
h
lig
h
ts
th
e
d
esig
n
ch
o
ices
th
at
d
is
tin
g
u
is
h
o
u
r
a
p
p
r
o
ac
h
,
p
r
eser
v
in
g
t
o
k
en
-
lev
el
B
E
R
T
r
ep
r
esen
tatio
n
s
f
o
r
L
STM
-
b
ased
s
eq
u
en
ce
m
o
d
elin
g
an
d
m
itig
a
tin
g
class
im
b
alan
ce
.
4
.
1
.
M
o
del
a
rc
hite
ct
ure
T
h
e
B
E
R
T
–
L
STM
h
y
b
r
id
m
o
d
el
lev
er
ag
es
B
E
R
T
to
p
r
o
d
u
ce
d
ee
p
co
n
te
x
tu
al
em
b
ed
d
in
g
s
f
r
o
m
tex
tu
al
in
p
u
ts
,
wh
ile
L
STM
c
ap
tu
r
es
tem
p
o
r
al
d
ep
e
n
d
en
cie
s
to
an
aly
ze
tr
an
s
ac
tio
n
p
atter
n
s
o
v
er
tim
e.
T
h
ese
f
u
s
io
n
p
r
eser
v
es
to
k
e
n
-
lev
el
s
em
an
tics
an
d
s
eq
u
en
tial
d
y
n
am
ics,
en
ab
lin
g
th
e
d
etec
to
r
to
ex
p
lo
it
co
m
p
lem
en
tar
y
cu
es
f
o
r
f
r
au
d
id
en
tific
atio
n
,
esp
ec
ially
u
n
d
er
class
im
b
alan
ce
.
At
a
h
ig
h
lev
el,
th
e
p
ip
elin
e
p
r
o
ce
ed
s
th
r
o
u
g
h
tex
t f
ea
t
u
r
e
p
r
o
ce
s
s
in
g
,
s
eq
u
en
ce
m
o
d
elin
g
,
an
d
f
in
al
in
teg
r
atio
n
an
d
cla
s
s
if
icatio
n
.
4
.
1
.
1
.
T
ex
t
f
ea
t
ure
pro
ce
s
s
in
g
Fo
r
ea
ch
tr
an
s
ac
tio
n
,
th
e
two
tex
tu
al
f
ield
s
,
m
er
ch
an
t
an
d
ca
teg
o
r
y
,
ar
e
c
o
n
ca
ten
ated
i
n
to
a
s
in
g
le
in
p
u
t
s
tr
in
g
.
T
h
is
tex
t
is
to
k
en
ized
with
th
e
s
tan
d
ar
d
B
E
R
T
t
o
k
en
izer
an
d
p
ass
ed
th
r
o
u
g
h
th
e
B
E
R
T
en
co
d
er
.
T
h
e
[
C
L
S]
to
k
en
em
b
ed
d
i
n
g
is
u
s
ed
as
a
f
ix
ed
-
len
g
th
r
ep
r
esen
tatio
n
o
f
th
e
tr
an
s
ac
ti
o
n
tex
t,
ca
p
tu
r
in
g
co
n
tex
tu
al
in
f
o
r
m
atio
n
ab
o
u
t t
h
e
m
er
ch
a
n
t a
n
d
t
h
e
p
u
r
ch
ase
ty
p
e
in
a
7
6
8
-
d
im
en
s
io
n
al
v
ec
to
r
.
4
.
1
.
2
.
Sequ
ence
m
o
delin
g
T
h
e
[
C
L
S]
em
b
ed
d
i
n
g
is
co
n
ca
ten
ate
d
with
t
h
e
tr
an
s
ac
tio
n
am
o
u
n
t
(
am
t)
,
y
ield
in
g
a
769
-
d
im
en
s
io
n
al
f
ea
tu
r
e
v
ec
t
o
r
.
T
h
is
v
ec
to
r
is
p
r
esen
ted
to
an
L
STM
lay
er
as
a
s
i
n
g
le
-
s
tep
s
eq
u
en
ce
(
s
eq
u
en
ce
len
g
t
h
=1
)
,
w
h
ich
f
u
n
ctio
n
s
as
a
g
ated
p
r
o
jectio
n
th
at
ca
n
m
o
d
el
n
o
n
lin
ea
r
i
n
ter
ac
tio
n
s
b
etwe
en
tex
tu
al
co
n
tex
t
a
n
d
a
m
o
u
n
t.
Alth
o
u
g
h
n
o
tem
p
o
r
al
s
eq
u
e
n
ce
ac
r
o
s
s
to
k
en
s
o
r
tr
an
s
ac
tio
n
s
is
u
s
ed
in
t
h
e
cu
r
r
en
t
im
p
lem
en
tatio
n
,
t
h
e
L
STM
’
s
g
atin
g
s
till
p
r
o
v
id
e
s
a
lear
n
ab
le
tr
an
s
f
o
r
m
atio
n
th
at
ca
n
im
p
r
o
v
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l B
E
R
T
-
lo
n
g
s
h
o
r
t
-
term me
mo
r
y
h
yb
r
id
mo
d
el
fo
r
effe
ctive
cred
it c
a
r
d
…
(
Ou
s
s
a
ma
N
d
a
ma
)
793
s
ep
ar
ab
ilit
y
r
elativ
e
to
a
p
u
r
ely
lin
ea
r
h
ea
d
.
T
h
is
s
in
g
le
-
s
tep
d
esig
n
is
ju
s
t
if
ied
b
y
o
u
r
p
er
-
tr
a
n
s
ac
tio
n
s
cr
ee
n
in
g
o
b
jectiv
e
an
d
th
e
a
b
s
en
ce
o
f
r
eliab
le,
s
ess
io
n
-
lev
el
o
r
d
er
i
n
g
ac
r
o
s
s
tr
an
s
ac
tio
n
s
,
wh
er
e
f
a
b
r
icatin
g
win
d
o
ws
co
u
ld
i
n
tr
o
d
u
ce
s
p
u
r
io
u
s
d
y
n
am
ics.
I
t
also
k
ee
p
s
laten
cy
an
d
p
ar
am
eter
co
u
n
t
lo
w,
wh
ich
h
elp
s
cu
r
b
o
v
er
f
itti
n
g
u
n
d
er
class
im
b
alan
ce
,
wh
ile
leav
in
g
a
clea
r
p
ath
to
m
u
lti
-
s
tep
s
eq
u
en
ce
s
wh
en
co
n
s
is
ten
t
in
ter
-
tr
an
s
ac
tio
n
o
r
d
er
in
g
b
ec
o
m
es a
v
ailab
le.
4
.
1
.
3
.
I
nte
g
ra
t
io
n a
nd
o
utput
T
h
e
L
STM
o
u
tp
u
ts
a
co
m
p
a
ct
r
ep
r
esen
tatio
n
th
at
in
te
g
r
a
tes
co
n
tex
tu
al
m
ea
n
in
g
with
tem
p
o
r
al
s
tr
u
ctu
r
e,
wh
ich
is
s
u
b
s
eq
u
en
t
ly
f
ed
to
a
f
u
lly
co
n
n
ec
ted
cl
ass
if
icatio
n
h
ea
d
.
A
s
ig
m
o
id
ac
tiv
atio
n
m
ap
s
th
e
o
u
tp
u
t
to
a
p
r
o
b
ab
ilit
y
o
f
f
r
au
d
,
en
a
b
lin
g
th
r
esh
o
ld
-
b
ase
d
d
ec
is
io
n
m
ak
in
g
alig
n
e
d
w
ith
o
p
er
atio
n
al
r
is
k
p
r
ef
er
en
ce
s
.
T
h
is
en
d
-
to
-
en
d
d
esig
n
allo
ws
th
e
m
o
d
el
to
co
m
b
in
e
n
u
an
ce
d
tex
t
u
n
d
er
s
tan
d
i
n
g
with
s
eq
u
en
tial
p
atter
n
r
ec
o
g
n
itio
n
i
n
a
s
in
g
le
tr
ain
ab
le
p
ip
elin
e.
4
.
2
.
M
o
del
co
ns
t
ruct
io
n a
nd
t
ra
ini
ng
m
et
ho
do
lo
g
ies
C
o
n
s
tr
u
ctin
g
an
d
tr
ai
n
in
g
th
e
B
E
R
T
–
L
STM
h
y
b
r
id
m
o
d
el
i
n
v
o
lv
e
d
p
r
ag
m
atic
ch
o
ices
to
m
ax
im
ize
p
er
f
o
r
m
an
ce
u
n
d
er
class
im
b
alan
ce
wh
ile
p
r
eser
v
in
g
g
e
n
e
r
aliza
tio
n
.
A
p
r
e
-
tr
ain
e
d
B
E
R
T
-
b
ase
-
u
n
ca
s
ed
is
f
in
e
-
tu
n
e
d
to
o
b
tain
d
o
m
ain
-
s
p
ec
if
ic
tex
t
em
b
ed
d
in
g
s
an
d
m
o
d
el
t
o
k
en
-
le
v
el
s
eq
u
e
n
ce
s
with
an
L
STM
.
An
im
b
alan
ce
-
awa
r
e
o
b
jectiv
e
an
d
a
h
eld
-
o
u
t v
alid
atio
n
s
p
lit g
u
id
e
iter
ativ
e
r
ef
in
em
e
n
ts
an
d
e
ar
ly
s
to
p
p
in
g
.
4
.
2
.
1
.
P
re
-
t
r
a
ini
ng
,
f
ine
-
t
un
ing
a
nd
s
eque
ntia
l da
t
a
ha
nd
lin
g
B
E
R
T
is
in
itia
lized
with
p
u
b
licly
av
ailab
le
p
r
e
-
tr
ain
ed
weig
h
ts
to
en
co
d
e
s
tr
o
n
g
le
x
ical
an
d
s
em
an
tic
p
r
io
r
s
.
I
t
is
th
en
f
in
e
-
tu
n
ed
o
n
tr
an
s
ac
tio
n
te
x
t
(
m
er
ch
an
t+c
ateg
o
r
y
)
s
o
t
h
e
e
m
b
ed
d
in
g
s
ad
ap
t
to
d
o
m
ain
-
s
p
ec
if
ic,
f
r
au
d
-
r
elev
a
n
t
r
eg
u
lar
ities
.
T
h
is
two
-
s
tag
e
tr
an
s
f
er
ac
ce
ler
ates
co
n
v
er
g
e
n
ce
,
ad
d
s
m
in
im
al
ex
tr
a
p
ar
am
eter
s
,
a
n
d
y
ield
s
m
o
r
e
d
is
cr
im
in
ativ
e
r
e
p
r
esen
tatio
n
s
f
o
r
f
r
au
d
cu
es.
T
h
e
L
STM
lay
er
is
tr
ain
ed
o
n
th
e
to
k
e
n
-
lev
el
em
b
ed
d
in
g
s
to
m
o
d
el
tem
p
o
r
al
d
e
p
en
d
e
n
cies
with
in
ea
ch
tex
tu
al
s
eq
u
en
ce
.
B
y
le
ar
n
in
g
h
o
w
m
ea
n
in
g
u
n
f
o
ld
s
ac
r
o
s
s
to
k
en
s
,
th
e
n
etwo
r
k
e
x
p
o
s
es
b
eh
av
io
r
ally
r
elev
an
t
s
ig
n
als
th
at
s
tatic
p
o
o
lin
g
m
i
g
h
t
o
v
er
lo
o
k
.
T
h
is
s
eq
u
en
tial
tr
ea
tm
e
n
t
co
m
p
lem
en
ts
th
e
co
n
tex
tu
al
p
o
wer
o
f
B
E
R
T
an
d
s
u
p
p
o
r
ts
h
ig
h
er
r
ec
all
o
n
m
i
n
o
r
ity
f
r
au
d
ca
s
es.
4
.
2
.
2
.
H
y
perpa
ra
m
et
er
s
a
nd
t
ra
ini
ng
s
et
t
ing
s
Fo
r
th
e
h
y
b
r
id
m
o
d
el,
b
er
t
-
b
ase
-
u
n
ca
s
ed
is
u
s
ed
as
a
f
r
o
ze
n
tex
t
en
c
o
d
er
,
with
to
k
en
izatio
n
co
n
f
ig
u
r
ed
f
o
r
p
ad
d
in
g
an
d
tr
u
n
ca
tio
n
at
t
h
e
d
ef
a
u
lt
m
ax
im
u
m
len
g
th
.
T
h
e
f
u
s
ed
r
ep
r
esen
tatio
n
co
n
ca
ten
ates
th
e
B
E
R
T
[
C
L
S]
v
ec
to
r
(
7
6
8
d
im
en
s
io
n
s
)
with
th
e
n
u
m
er
i
c
am
o
u
n
t
to
f
o
r
m
a
7
6
9
-
d
im
e
n
s
io
n
al
in
p
u
t.
T
h
e
class
if
ier
h
ea
d
is
a
s
in
g
le
-
lay
er
L
STM
with
h
id
d
en
s
ize
2
5
6
,
b
atch
_
f
ir
s
t=T
r
u
e,
u
n
id
ir
ec
tio
n
al,
an
d
d
r
o
p
o
u
t
s
et
to
0
,
f
o
llo
wed
b
y
a
s
ig
m
o
id
o
u
tp
u
t.
Op
tim
izatio
n
u
s
es
Ad
am
with
a
lear
n
in
g
r
ate
o
f
0
.
0
0
1
an
d
b
in
a
r
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
.
T
r
ain
in
g
u
s
es
f
u
ll
-
b
atch
u
p
d
ates
(
b
atch
s
ize
eq
u
al
to
t
h
e
n
u
m
b
er
o
f
tr
ai
n
in
g
ex
a
m
p
les)
f
o
r
5
0
0
e
p
o
ch
s
,
a
n
d
e
v
alu
atio
n
is
p
er
f
o
r
m
e
d
o
n
an
8
0
/2
0
r
an
d
o
m
t
r
ain
-
v
alid
atio
n
s
p
lit.
T
o
ad
d
r
ess
class
im
b
alan
ce
,
SMOT
E
is
ap
p
lie
d
to
th
e
co
m
b
in
ed
f
ea
t
u
r
e
m
atr
ix
b
ef
o
r
e
th
e
s
p
lit.
Un
less
o
th
er
wis
e
n
o
ted
,
m
etr
ics ar
e
co
m
p
u
te
d
at
a
d
ec
i
s
io
n
th
r
esh
o
ld
o
f
0
.
5
.
4
.
2
.
3
.
Va
lid
a
t
io
n a
nd
it
er
a
t
iv
e
im
pro
v
e
m
ent
Mo
d
el
d
e
v
elo
p
m
e
n
t
f
o
llo
ws
a
h
eld
-
o
u
t
v
alid
atio
n
s
ch
em
e
to
ass
ess
g
en
er
aliza
tio
n
t
o
u
n
s
ee
n
d
ata
an
d
g
u
id
e
ea
r
ly
s
to
p
p
i
n
g
.
Sta
n
d
ar
d
m
et
r
ics
s
u
ch
as
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
F1
-
s
co
r
e
ar
e
m
o
n
ito
r
ed
,
an
d
h
y
p
er
p
ar
am
eter
s
an
d
ar
ch
itectu
r
al
d
etails
ar
e
ad
j
u
s
ted
in
r
esp
o
n
s
e
to
v
alid
atio
n
tr
en
d
s
.
I
ter
ativ
e
r
ef
in
em
en
ts
f
o
cu
s
o
n
s
tab
iliz
in
g
tr
ain
in
g
,
i
m
p
r
o
v
in
g
m
in
o
r
ity
-
class
s
en
s
itiv
ity
,
an
d
e
n
s
u
r
in
g
th
e
p
ip
elin
e
r
em
ain
s
r
o
b
u
s
t u
n
d
er
class
im
b
alan
ce
.
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
B
E
R
T
-
L
STM
h
y
b
r
id
m
o
d
el
s
h
o
wca
s
ed
o
u
ts
tan
d
i
n
g
p
er
f
o
r
m
a
n
ce
o
n
th
e
v
alid
atio
n
s
et,
ac
h
iev
in
g
an
ac
c
u
r
ac
y
o
f
9
9
.
1
1
%.
T
h
is
h
ig
h
lev
el
o
f
ac
cu
r
a
cy
h
ig
h
lig
h
ts
th
e
m
o
d
el'
s
r
o
b
u
s
t
ab
ilit
y
to
class
if
y
tr
an
s
ac
tio
n
s
ef
f
ec
tiv
ely
.
Pre
ci
s
io
n
was
n
o
tab
l
y
h
i
g
h
at
9
8
.
2
7
%,
wh
ile
th
e
m
o
d
el
ac
h
ie
v
e
d
a
p
er
f
ec
t
r
ec
all
o
f
1
0
0
%,
in
d
icatin
g
its
s
u
cc
ess
in
id
en
tify
in
g
all
f
r
a
u
d
u
len
t
tr
an
s
ac
tio
n
s
with
in
th
e
d
ataset.
T
h
e
F1
-
s
co
r
e,
b
alan
cin
g
p
r
ec
is
io
n
an
d
r
ec
all
,
s
to
o
d
im
p
r
ess
iv
ely
at
9
9
.
1
3
%.
T
h
e
f
o
llo
win
g
is
a
co
n
cise
s
u
m
m
ar
y
o
f
th
e
k
ey
p
er
f
o
r
m
an
ce
m
etr
ics,
as illu
s
tr
ated
in
T
ab
le
2
.
T
h
e
m
o
d
el
also
m
ain
tain
ed
a
l
o
w
v
alid
atio
n
lo
s
s
o
f
0
.
0
3
7
5
,
f
u
r
th
er
v
alid
atin
g
its
ef
f
icien
cy
in
f
r
au
d
d
etec
tio
n
.
T
h
is
b
len
d
o
f
h
i
g
h
p
r
ec
is
io
n
,
r
ec
all,
an
d
ac
cu
r
ac
y
u
n
d
er
s
co
r
es
th
e
m
o
d
el’
s
ca
p
ab
ilit
ies
in
ef
f
ec
tiv
ely
d
etec
tin
g
f
r
au
d
,
p
o
s
itio
n
in
g
it
as
a
p
o
ten
t
to
o
l
in
f
in
an
cial
s
ec
u
r
ity
s
y
s
tem
s
.
T
h
e
class
if
icatio
n
r
ep
o
r
t
f
u
r
th
e
r
d
etails
th
ese
r
esu
lts
in
Fig
u
r
e
2
,
p
r
o
v
i
d
in
g
a
c
o
m
p
r
e
h
en
s
iv
e
b
r
ea
k
d
o
wn
o
f
th
e
m
o
d
el’
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
788
-
7
9
7
794
p
er
f
o
r
m
an
ce
ac
r
o
s
s
d
if
f
er
e
n
t
class
es.
T
h
i
s
v
is
u
aliza
tio
n
en
h
an
ce
s
u
n
d
er
s
tan
d
in
g
o
f
th
e
m
o
d
el'
s
p
r
ec
is
io
n
an
d
r
ec
all
b
y
class
,
illu
s
tr
atin
g
its
b
alan
ce
d
ef
f
ec
tiv
en
ess
in
f
r
au
d
d
etec
tio
n
.
T
ab
le
2
.
Per
f
o
r
m
an
ce
m
etr
ics o
f
th
e
B
E
R
T
-
L
STM
h
y
b
r
i
d
m
o
d
el
M
o
d
e
l
A
c
c
u
r
a
c
y
A
c
c
u
r
a
c
y
0
.
9
9
1
1
P
r
e
c
i
s
i
o
n
0
.
9
8
2
7
R
e
c
a
l
l
F
1
sc
o
r
e
1
.
0
0
0
0
0
.
9
9
1
3
Fig
u
r
e
2
.
C
lass
if
icatio
n
r
ep
o
r
t
o
f
th
e
B
E
R
T
-
L
STM
h
y
b
r
i
d
m
o
d
el
T
h
e
B
E
R
T
-
L
STM
h
y
b
r
id
m
o
d
el
ef
f
ec
tiv
ely
co
m
b
in
es
B
E
R
T
'
s
co
n
tex
tu
al
em
b
ed
d
in
g
s
wi
th
L
STM
's
s
eq
u
en
ce
an
aly
s
is
ca
p
ab
ilit
ies,
o
f
f
er
in
g
a
r
o
b
u
s
t
ap
p
r
o
ac
h
to
f
r
au
d
d
etec
tio
n
.
T
h
i
s
m
o
d
el
n
o
t
o
n
ly
d
em
o
n
s
tr
ates
s
tr
o
n
g
s
tatis
tica
l
p
er
f
o
r
m
a
n
ce
b
u
t
also
ex
h
i
b
its
a
d
ee
p
u
n
d
er
s
tan
d
in
g
o
f
c
o
m
p
lex
tr
an
s
ac
tio
n
p
atter
n
s
ess
en
tial
f
o
r
id
en
tify
in
g
f
r
au
d
u
len
t
ac
tiv
ities
.
I
ts
h
ig
h
r
ec
all
r
ate
is
c
r
u
cial
in
a
f
r
au
d
d
etec
tio
n
co
n
tex
t,
en
s
u
r
i
n
g
n
o
f
r
a
u
d
u
len
t
tr
an
s
ac
tio
n
is
m
is
s
ed
,
wh
ich
co
u
ld
o
th
er
wis
e
h
av
e
s
ev
er
e
f
in
an
cial
im
p
licatio
n
s
.
Ad
d
itio
n
ally
,
th
e
m
o
d
el’
s
im
p
r
ess
iv
e
p
r
ec
is
io
n
m
in
im
izes
f
alse
p
o
s
itiv
es,
th
er
eb
y
p
r
eser
v
i
n
g
cu
s
to
m
er
tr
u
s
t a
n
d
o
p
er
atio
n
al
ef
f
icien
cy
.
Fo
r
ex
ter
n
al
co
n
tex
t,
T
ab
le
3
co
n
tr
asts
o
u
r
r
esu
lts
with
r
ep
r
esen
tativ
e
ex
is
tin
g
m
et
h
o
d
s
e
v
alu
ated
o
n
th
e
s
am
e
d
ataset
f
am
ily
.
R
elat
iv
e
to
r
an
d
o
m
f
o
r
est
b
aselin
es
[
1
3
]
,
[
1
5
]
,
a
lig
h
tweig
h
t
class
ical
en
s
em
b
le
[
1
4
]
,
an
d
an
en
co
d
er
-
d
ec
o
d
er
GNN
[
1
9
]
,
th
e
p
r
o
p
o
s
ed
B
E
R
T
-
L
STM
f
u
s
io
n
attain
s
th
e
h
ig
h
est
F1
-
s
co
r
e
an
d
r
ec
all
wh
ile
p
r
eser
v
in
g
s
tr
o
n
g
p
r
ec
is
io
n
.
T
h
ese
co
m
p
ar
is
o
n
s
s
u
g
g
e
s
t th
at
u
n
if
y
in
g
tex
tu
al
m
er
ch
a
n
t/categ
o
r
y
s
ig
n
als
with
n
u
m
er
ic
a
m
o
u
n
t
in
s
id
e
o
n
e
m
o
d
el
ca
n
b
e
ad
v
an
tag
eo
u
s
.
Pro
to
co
ls
an
d
r
esam
p
lin
g
s
t
r
ateg
ies
v
ar
y
ac
r
o
s
s
s
tu
d
ies,
s
o
th
e
n
u
m
b
e
r
s
ar
e
in
d
icativ
e
r
ath
er
th
a
n
s
tr
ictly
co
m
p
ar
ab
le.
Ho
wev
er
,
th
e
m
o
d
el
f
ac
es
ch
allen
g
es,
p
ar
ticu
lar
ly
with
th
e
ad
ap
tab
ilit
y
to
s
o
p
h
is
ticated
o
r
p
r
ev
io
u
s
ly
u
n
s
ee
n
f
r
au
d
tactics
as
f
r
au
d
s
ter
s
co
n
tin
u
ally
ev
o
lv
e
th
eir
s
tr
ateg
ies.
T
o
en
h
a
n
ce
its
ad
ap
tab
ilit
y
,
f
u
tu
r
e
en
h
an
ce
m
e
n
ts
co
u
ld
i
n
clu
d
e
d
y
n
am
ic
lear
n
i
n
g
an
d
u
p
d
atin
g
m
ec
h
an
is
m
s
to
b
et
ter
r
esp
o
n
d
t
o
n
ew
f
r
au
d
p
atter
n
s
.
Fu
r
th
e
r
r
esear
ch
m
ig
h
t
also
ex
p
lo
r
e
in
te
g
r
atin
g
ad
d
itio
n
al
d
ata
ty
p
es
s
u
ch
as
cu
s
to
m
er
b
eh
av
io
r
o
r
m
ac
r
o
ec
o
n
o
m
ic
in
d
icato
r
s
to
b
o
o
s
t
th
e
m
o
d
el's
p
r
ed
ictiv
e
p
o
wer
.
Dep
lo
y
in
g
th
is
m
o
d
el
in
r
ea
l
-
wo
r
ld
f
in
a
n
cial
s
y
s
tem
s
wo
u
ld
n
ec
ess
itate
r
o
b
u
s
t
in
f
r
astru
ctu
r
e
f
o
r
r
ea
l
-
tim
e
a
n
a
ly
s
is
an
d
s
ea
m
les
s
in
teg
r
atio
n
with
ex
is
tin
g
m
o
n
i
to
r
in
g
f
r
am
ewo
r
k
s
,
alo
n
g
s
id
e
en
s
u
r
in
g
c
o
m
p
lian
ce
with
s
tr
i
n
g
en
t
d
ata
p
r
iv
ac
y
r
eg
u
latio
n
s
an
d
m
ain
tain
in
g
h
i
g
h
d
ata
s
ec
u
r
ity
s
tan
d
ar
d
s
.
T
ab
le
3
.
C
o
m
p
a
r
is
o
n
with
ex
i
s
tin
g
m
eth
o
d
s
(
Sp
ar
k
o
v
d
ataset
f
am
ily
)
M
o
d
e
l
A
c
c
u
r
a
c
y
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F
1
sc
o
r
e
O
u
r
mo
d
e
l
(
B
ER
T
-
LST
M
)
0
.
9
9
1
1
0
.
9
8
2
7
1
.
0
0
0
0
0
.
9
9
1
3
R
a
n
d
o
m F
o
r
e
s
t
[
1
3
]
-
-
-
0
.
8
3
5
0
K
N
N
,
LD
A
,
L
R
[
1
4
]
R
a
n
d
o
m
f
o
r
e
s
t
[
1
5
]
-
0
.
9
6
0
0
-
0
.
0
9
0
0
0
.
9
7
0
1
0
.
9
7
0
0
-
0
.
1
7
0
0
En
c
o
d
e
r
-
d
e
c
o
d
e
r
G
N
N
[
1
9
]
0
.
9
7
0
0
0
.
8
2
0
0
0
.
9
2
0
0
0
.
8
6
0
0
6.
CO
NCLU
SI
O
N
T
h
e
B
E
R
T
-
L
STM
h
y
b
r
id
m
o
d
el
d
ev
elo
p
ed
in
th
is
s
tu
d
y
r
ep
r
esen
ts
a
s
ig
n
if
ican
t
s
tep
f
o
r
war
d
in
cr
ed
it
ca
r
d
f
r
au
d
d
etec
tio
n
b
y
in
teg
r
atin
g
b
o
th
te
x
tu
al
an
d
n
u
m
e
r
ical
d
ata.
B
y
co
m
b
i
n
in
g
B
E
R
T
’
s
d
ee
p
co
n
tex
tu
al
em
b
e
d
d
in
g
s
with
L
STM
’
s
s
eq
u
en
ce
an
aly
s
is
,
t
h
e
ap
p
r
o
ac
h
o
u
t
p
er
f
o
r
m
s
tr
ad
itio
n
al
m
o
d
els
th
at
o
n
ly
u
s
e
n
u
m
er
ical
f
ea
t
u
r
es,
s
ettin
g
a
n
ew
b
en
c
h
m
ar
k
f
o
r
h
an
d
lin
g
h
eter
o
g
en
e
o
u
s
d
ata
in
f
r
a
u
d
d
etec
tio
n
.
Ho
wev
er
,
th
e
s
tu
d
y
is
lim
ited
b
y
its
r
elian
ce
o
n
s
y
n
th
etic
d
ata,
ab
s
en
ce
o
f
tr
u
e
tr
an
s
ac
tio
n
s
eq
u
en
ce
s
,
an
d
a
n
ar
r
o
w
f
ea
tu
r
e
s
et
(
m
er
ch
an
t,
ca
teg
o
r
y
,
am
o
u
n
t)
.
T
o
ad
v
a
n
ce
f
u
r
th
er
,
f
u
tu
r
e
wo
r
k
s
h
o
u
l
d
test
th
e
m
o
d
el
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l B
E
R
T
-
lo
n
g
s
h
o
r
t
-
term me
mo
r
y
h
yb
r
id
mo
d
el
fo
r
effe
ctive
cred
it c
a
r
d
…
(
Ou
s
s
a
ma
N
d
a
ma
)
795
r
ea
l
tr
an
s
ac
tio
n
d
ata
with
p
r
iv
ac
y
s
af
e
g
u
ar
d
s
,
ap
p
ly
e
x
p
lain
ab
le
AI
m
eth
o
d
s
lik
e
S
h
ap
ley
a
d
d
itiv
e
ex
p
lan
atio
n
s
(
SHAP
)
an
d
lo
c
al
in
ter
p
r
etab
le
m
o
d
el
-
ag
n
o
s
tic
ex
p
lan
atio
n
s
(
LIME
)
f
o
r
d
e
cisi
o
n
au
d
itin
g
,
an
d
ex
p
er
im
en
t
with
tr
an
s
f
o
r
m
er
-
b
ased
tab
u
lar
m
o
d
els
(
e.
g
.
,
T
ab
T
r
an
s
f
o
r
m
er
a
n
d
FT
-
T
r
an
s
f
o
r
m
er
)
f
o
r
r
ich
e
r
in
p
u
t
h
an
d
lin
g
.
Ad
d
itio
n
ally
,
d
ep
lo
y
in
g
a
r
ea
l
-
tim
e
s
tr
ea
m
in
g
p
ip
elin
e
with
m
o
d
el
-
d
r
if
t
m
o
n
ito
r
in
g
will
b
e
cr
u
cial
f
o
r
m
ain
tain
i
n
g
r
o
b
u
s
t
an
d
co
m
p
lia
n
t
p
er
f
o
r
m
an
ce
as
f
r
au
d
tactics
ev
o
lv
e.
T
h
ese
d
ir
ec
tio
n
s
ar
e
k
ey
f
o
r
tr
an
s
latin
g
t
h
e
ap
p
r
o
ac
h
i
n
to
a
r
eliab
le
f
r
a
u
d
d
etec
tio
n
s
o
lu
tio
n
in
r
ea
l
-
w
o
r
ld
f
i
n
an
cial
s
y
s
tem
s
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
n
o
f
u
n
d
in
g
in
v
o
lv
ed
.
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
Ou
s
s
am
a
Nd
am
a
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
Saf
ae
Nd
am
a
✓
✓
✓
✓
I
s
m
ail
B
en
s
as
s
i
✓
✓
E
l M
o
k
h
tar
E
n
-
Naim
i
✓
✓
✓
✓
✓
✓
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
T
h
e
d
ataset
u
s
ed
in
th
is
s
tu
d
y
is
p
u
b
licly
av
ailab
le
an
d
o
b
tain
ed
f
r
o
m
a
well
-
estab
lis
h
ed
r
ep
o
s
ito
r
y
.
T
h
e
d
ata
s
u
p
p
o
r
tin
g
th
e
cr
ed
it
ca
r
d
f
r
au
d
d
etec
tio
n
(
C
C
FD)
ex
p
er
im
en
ts
ar
e
o
p
en
ly
av
ail
ab
le
o
n
GitHu
b
at:
h
ttp
s
://g
ith
u
b
.
co
m
/n
a
m
eb
r
an
d
o
n
/Sp
ar
k
o
v
_
Data
_
Gen
er
ati
o
n
.
RE
F
E
R
E
NC
E
S
[
1
]
H
.
M
.
A
l
z
o
u
b
i
,
M
.
T.
A
l
s
h
u
r
i
d
e
h
,
B
.
A
.
K
u
r
d
i
,
K
.
M
.
K
.
A
l
h
y
a
sa
t
,
a
n
d
T
.
M
.
G
h
a
z
a
l
,
“
Th
e
e
f
f
e
c
t
o
f
e
-
p
a
y
me
n
t
a
n
d
o
n
l
i
n
e
sh
o
p
p
i
n
g
o
n
sal
e
s
g
r
o
w
t
h
:
e
v
i
d
e
n
c
e
f
r
o
m
b
a
n
k
i
n
g
i
n
d
u
st
r
y
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
D
a
t
a
a
n
d
N
e
t
w
o
r
k
S
c
i
e
n
c
,
v
o
l
.
6
,
n
o
.
4
,
p
p
.
1
3
6
9
–
1
3
8
0
,
2
0
2
2
,
d
o
i
:
1
0
.
5
2
6
7
/
j
.
i
j
d
n
s.
2
0
2
2
.
5
.
0
1
4
.
[
2
]
M
.
H
.
U
.
S
h
a
r
i
f
a
n
d
M
.
A
.
M
o
h
a
m
med
,
“
A
l
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
o
f
f
i
n
a
n
c
i
a
l
l
o
ss
e
s
st
a
t
i
s
t
i
c
s
f
o
r
c
y
b
e
r
se
c
u
r
i
t
y
a
n
d
f
u
t
u
r
e
t
r
e
n
d
,
”
Wo
r
l
d
J
o
u
r
n
a
l
o
f
A
d
v
a
n
c
e
d
Re
s
e
a
r
c
h
a
n
d
Re
v
i
e
w
s
,
v
o
l
.
1
5
,
n
o
.
1
,
p
p
.
1
3
8
–
1
5
6
,
J
u
l
.
2
0
2
2
,
d
o
i
:
1
0
.
3
0
5
7
4
/
w
j
a
r
r
.
2
0
2
2
.
1
5
.
1
.
0
5
7
3
.
[
3
]
O
.
A
.
B
e
l
l
o
a
n
d
K
.
O
l
u
f
e
m
i
,
“
A
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
i
n
f
r
a
u
d
p
r
e
v
e
n
t
i
o
n
:
e
x
p
l
o
r
i
n
g
t
e
c
h
n
i
q
u
e
s
a
n
d
a
p
p
l
i
c
a
t
i
o
n
s
c
h
a
l
l
e
n
g
e
s
a
n
d
o
p
p
o
r
t
u
n
i
t
i
e
s,
”
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
I
n
f
o
rm
a
t
i
o
n
T
e
c
h
n
o
l
o
g
y
Re
s
e
a
rc
h
J
o
u
r
n
a
l
,
v
o
l
.
5
,
n
o
.
6
,
p
p
.
1
5
0
5
–
1
5
2
0
,
Ju
n
.
2
0
2
4
,
d
o
i
:
1
0
.
5
1
5
9
4
/
c
si
t
r
j
.
v
5
i
6
.
1
2
5
2
.
[
4
]
S
.
C
h
e
r
n
i
a
v
s
k
y
i
,
V
.
B
a
b
a
n
i
n
a
,
I
.
V
a
r
t
y
l
e
t
s
k
a
,
a
n
d
O
.
M
y
k
y
t
c
h
y
k
,
“
P
e
c
u
l
i
a
r
i
t
i
e
s o
f
t
h
e
e
c
o
n
o
m
i
c
c
r
i
mes
c
o
mm
i
t
t
e
d
w
i
t
h
t
h
e
u
s
e
o
f
i
n
f
o
r
mat
i
o
n
t
e
c
h
n
o
l
o
g
i
e
s,
”
Eu
r
o
p
e
a
n
J
o
u
r
n
a
l
o
f
S
u
st
a
i
n
a
b
l
e
D
e
v
e
l
o
p
m
e
n
t
,
v
o
l
.
1
0
,
n
o
.
1
,
F
e
b
.
2
0
2
1
,
d
o
i
:
1
0
.
1
4
2
0
7
/
e
j
s
d
.
2
0
2
1
.
v
1
0
n
1
p
4
2
0
.
[
5
]
J.
R
o
b
i
n
s
o
n
a
n
d
M
.
E
d
w
a
r
d
s
,
“
F
r
a
u
d
st
e
r
s
t
a
r
g
e
t
t
h
e
e
l
d
e
r
l
y
:
b
e
h
a
v
i
o
u
r
a
l
e
v
i
d
e
n
c
e
f
r
o
m
r
a
n
d
o
mi
s
e
d
c
o
n
t
r
o
l
l
e
d
sca
m
-
b
a
i
t
i
n
g
e
x
p
e
r
i
me
n
t
s,
”
S
e
c
u
r
i
t
y
J
o
u
r
n
a
l
,
v
o
l
.
3
7
,
n
o
.
4
,
p
p
.
1
1
7
3
–
1
1
9
6
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
5
7
/
s4
1
2
8
4
-
0
2
3
-
0
0
4
1
0
-
4.
[
6
]
S
.
W
a
n
g
,
M
.
A
s
i
f
,
M
.
F
.
S
h
a
h
z
a
d
,
a
n
d
M
.
A
sh
f
a
q
,
“
D
a
t
a
p
r
i
v
a
c
y
a
n
d
c
y
b
e
r
sec
u
r
i
t
y
c
h
a
l
l
e
n
g
e
s
i
n
t
h
e
d
i
g
i
t
a
l
t
r
a
n
sf
o
r
mat
i
o
n
o
f
t
h
e
b
a
n
k
i
n
g
sec
t
o
r
,
”
C
o
m
p
u
t
e
rs &
S
e
c
u
ri
t
y
,
v
o
l
.
1
4
7
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
o
s
e
.
2
0
2
4
.
1
0
4
0
5
1
.
[
7
]
M
.
C
a
n
a
y
a
z
,
I
.
K
a
n
t
o
r
o
v
i
t
c
h
,
a
n
d
R
.
M
i
h
e
t
,
“
C
o
n
s
u
mer
p
r
i
v
a
c
y
a
n
d
v
a
l
u
e
o
f
c
o
n
s
u
m
e
r
d
a
t
a
,
”
S
S
R
N
El
e
c
t
r
o
n
i
c
J
o
u
rn
a
l
,
2
0
2
1
,
d
o
i
:
1
0
.
2
1
3
9
/
s
sr
n
.
3
9
8
6
5
6
2
.
[
8
]
B
.
H
a
r
r
i
s
,
“
G
i
t
H
u
b
-
n
a
me
b
r
a
n
d
o
n
/
sp
a
r
k
o
v
_
d
a
t
a
_
g
e
n
e
r
a
t
i
o
n
:
sy
n
t
h
e
t
i
c
c
r
e
d
i
t
c
a
r
d
t
r
a
n
sa
c
t
i
o
n
g
e
n
e
r
a
t
o
r
u
s
e
d
i
n
t
h
e
s
p
a
r
k
o
v
p
r
o
g
r
a
m
,
”
G
i
t
H
u
b
.
A
c
c
e
sse
d
:
J
u
n
.
2
6
,
2
0
2
4
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
g
i
t
h
u
b
.
c
o
m/
n
a
m
e
b
r
a
n
d
o
n
/
S
p
a
r
k
o
v
_
D
a
t
a
_
G
e
n
e
r
a
t
i
o
n
[
9
]
S
.
S
i
n
g
l
a
,
P
r
i
y
a
n
sh
u
,
A
.
Th
a
k
u
r
,
A
.
S
w
a
mi
,
U
.
S
a
w
a
r
n
,
a
n
d
P
.
S
i
n
g
l
a
,
“
A
d
v
a
n
c
e
me
n
t
s
i
n
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
:
B
E
R
T
a
n
d
t
r
a
n
sf
o
r
mer
-
b
a
se
d
m
o
d
e
l
s
f
o
r
t
e
x
t
u
n
d
e
r
s
t
a
n
d
i
n
g
,
”
i
n
2
0
2
4
S
e
c
o
n
d
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
Ad
v
a
n
c
e
d
C
o
m
p
u
t
i
n
g
&
C
o
m
m
u
n
i
c
a
t
i
o
n
T
e
c
h
n
o
l
o
g
i
e
s (I
C
AC
C
T
e
c
h
)
,
I
EEE,
N
o
v
.
2
0
2
4
,
p
p
.
3
7
2
–
379
,
d
o
i
:
1
0
.
1
1
0
9
/
I
C
A
C
C
T
e
c
h
6
5
0
8
4
.
2
0
2
4
.
0
0
0
6
8
.
[
1
0
]
A
.
Tu
r
c
h
i
n
,
S
.
M
a
s
h
a
r
sk
y
,
a
n
d
M
.
Z
i
t
n
i
k
,
“
C
o
mp
a
r
i
s
o
n
o
f
B
E
R
T
i
mp
l
e
m
e
n
t
a
t
i
o
n
s
f
o
r
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
s
si
n
g
o
f
n
a
r
r
a
t
i
v
e
med
i
c
a
l
d
o
c
u
me
n
t
s
,
”
I
n
f
o
rm
a
t
i
c
s
i
n
Me
d
i
c
i
n
e
U
n
l
o
c
k
e
d
,
v
o
l
.
3
6
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
mu
.
2
0
2
2
.
1
0
1
1
3
9
.
[
1
1
]
S
.
N
o
s
o
u
h
i
a
n
,
F
.
N
o
s
o
u
h
i
a
n
,
a
n
d
A
.
K
.
K
h
o
s
h
o
u
e
i
,
“
A
r
e
v
i
e
w
o
f
r
e
c
u
r
r
e
n
t
n
e
u
r
a
l
n
e
t
w
o
r
k
a
r
c
h
i
t
e
c
t
u
r
e
f
o
r
s
e
q
u
e
n
c
e
l
e
a
r
n
i
n
g
:
c
o
m
p
a
r
i
so
n
b
e
t
w
e
e
n
LST
M
a
n
d
G
R
U
,
”
Pr
e
p
r
i
n
t
s
,
2
0
2
1
,
d
o
i
:
1
0
.
2
0
9
4
4
/
p
r
e
p
r
i
n
t
s
2
0
2
1
0
7
.
0
2
5
2
.
v
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
15
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
:
788
-
7
9
7
796
[
1
2
]
Y
.
C
h
e
n
a
n
d
M
.
D
u
,
“
F
i
n
a
n
c
i
a
l
f
r
a
u
d
t
r
a
n
sa
c
t
i
o
n
p
r
e
d
i
c
t
i
o
n
a
p
p
r
o
a
c
h
b
a
se
d
o
n
g
l
o
b
a
l
e
n
h
a
n
c
e
d
G
C
N
a
n
d
b
i
d
i
r
e
c
t
i
o
n
a
l
LST
M
,
”
C
o
m
p
u
t
a
t
i
o
n
a
l
Ec
o
n
o
m
i
c
s
,
v
o
l
.
6
6
,
n
o
.
2
,
p
p
.
1
7
4
7
–
1
7
6
6
,
A
u
g
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
6
1
4
-
0
2
4
-
1
0
7
9
1
-
2.
[
1
3
]
D
.
B
r
e
s
k
u
v
i
e
n
ė
a
n
d
G
.
D
z
e
my
d
a
,
“
En
h
a
n
c
i
n
g
c
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
:
h
i
g
h
l
y
i
m
b
a
l
a
n
c
e
d
d
a
t
a
c
a
s
e
,
”
J
o
u
rn
a
l
o
f
B
i
g
D
a
t
a
,
v
o
l
.
1
1
,
n
o
.
1
,
D
e
c
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
8
6
/
s
4
0
5
3
7
-
0
2
4
-
0
1
0
5
9
-
5.
[
1
4
]
J.
C
h
u
n
g
a
n
d
K
.
Le
e
,
“
C
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
:
a
n
i
mp
r
o
v
e
d
st
r
a
t
e
g
y
f
o
r
h
i
g
h
r
e
c
a
l
l
u
s
i
n
g
K
N
N
,
LD
A
,
a
n
d
l
i
n
e
a
r
r
e
g
r
e
ssi
o
n
,
”
S
e
n
so
rs
,
v
o
l
.
2
3
,
n
o
.
1
8
,
S
e
p
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
s2
3
1
8
7
7
8
8
.
[
1
5
]
J.
K
.
A
f
r
i
y
i
e
e
t
a
l
.
,
“
A
su
p
e
r
v
i
s
e
d
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
f
o
r
d
e
t
e
c
t
i
n
g
a
n
d
p
r
e
d
i
c
t
i
n
g
f
r
a
u
d
i
n
c
r
e
d
i
t
c
a
r
d
t
r
a
n
s
a
c
t
i
o
n
s,”
D
e
c
i
s
i
o
n
An
a
l
y
t
i
c
s
J
o
u
r
n
a
l
,
v
o
l
.
6
,
M
a
r
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
d
a
j
o
u
r
.
2
0
2
3
.
1
0
0
1
6
3
.
[
1
6
]
J.
F
o
r
o
u
g
h
a
n
d
S
.
M
o
mt
a
z
i
,
“
S
e
q
u
e
n
t
i
a
l
c
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
:
a
j
o
i
n
t
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
a
n
d
p
r
o
b
a
b
i
l
i
st
i
c
g
r
a
p
h
i
c
a
l
m
o
d
e
l
a
p
p
r
o
a
c
h
,
”
Ex
p
e
rt
S
y
s
t
e
m
s
,
v
o
l
.
3
9
,
n
o
.
1
,
Ja
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
1
1
1
/
e
x
s
y
.
1
2
7
9
5
.
[
1
7
]
P
.
B
o
u
l
i
e
r
i
s,
J.
P
a
v
l
o
p
o
u
l
o
s,
A
.
X
e
n
o
s,
a
n
d
V
.
V
a
ssa
l
o
s
,
“
F
r
a
u
d
d
e
t
e
c
t
i
o
n
w
i
t
h
n
a
t
u
r
a
l
l
a
n
g
u
a
g
e
p
r
o
c
e
ssi
n
g
,
”
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
,
v
o
l
.
1
1
3
,
n
o
.
8
,
p
p
.
5
0
8
7
–
5
1
0
8
,
A
u
g
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
0
9
9
4
-
0
2
3
-
0
6
3
5
4
-
5.
[
1
8
]
I
.
D
.
M
i
e
n
y
e
a
n
d
N
.
J
e
r
e
,
“
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
f
r
a
u
d
d
e
t
e
c
t
i
o
n
:
a
r
e
v
i
e
w
o
f
a
l
g
o
r
i
t
h
ms,
c
h
a
l
l
e
n
g
e
s,
a
n
d
so
l
u
t
i
o
n
s
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
9
6
8
9
3
–
9
6
9
1
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
4
2
6
9
5
5
.
[
1
9
]
A
.
C
h
e
r
i
f
,
H
.
A
mm
a
r
,
M
.
K
a
l
k
a
t
a
w
i
,
S
.
A
l
s
h
e
h
r
i
,
a
n
d
A
.
I
mi
n
e
,
“
E
n
c
o
d
e
r
–
d
e
c
o
d
e
r
g
r
a
p
h
n
e
u
r
a
l
n
e
t
w
o
r
k
f
o
r
c
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
,
”
J
o
u
rn
a
l
o
f
K
i
n
g
S
a
u
d
U
n
i
v
e
rs
i
t
y
–
C
o
m
p
u
t
e
r
a
n
d
I
n
f
o
r
m
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
3
6
,
n
o
.
3
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
k
s
u
c
i
.
2
0
2
4
.
1
0
2
0
0
3
.
[
2
0
]
E.
I
l
e
b
e
r
i
a
n
d
Y
.
S
u
n
,
“
A
h
y
b
r
i
d
d
e
e
p
l
e
a
r
n
i
n
g
e
n
se
mb
l
e
mo
d
e
l
f
o
r
c
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
,
”
I
EEE
Ac
c
e
ss
,
v
o
l
.
1
2
,
p
p
.
1
7
5
8
2
9
–
1
7
5
8
3
8
,
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
4
.
3
5
0
2
5
4
2
.
[
2
1
]
I
.
H
e
w
a
p
a
t
h
i
r
a
n
a
,
N
.
K
e
k
a
y
a
n
,
a
n
d
D
.
D
i
y
a
se
n
a
,
“
A
s
y
s
t
e
m
a
t
i
c
i
n
v
e
st
i
g
a
t
i
o
n
o
n
t
h
e
e
f
f
e
c
t
i
v
e
n
e
ss
o
f
t
h
e
t
a
b
b
e
r
t
m
o
d
e
l
f
o
r
c
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
,
”
i
n
2
0
2
2
I
n
t
e
r
n
a
t
i
o
n
a
l
Re
s
e
a
r
c
h
C
o
n
f
e
re
n
c
e
o
n
S
m
a
rt
C
o
m
p
u
t
i
n
g
a
n
d
S
y
st
e
m
s
E
n
g
i
n
e
e
r
i
n
g
(
S
C
S
E)
,
I
EEE,
S
e
p
.
2
0
2
2
,
p
p
.
9
6
–
1
0
1
,
d
o
i
:
1
0
.
1
1
0
9
/
S
C
S
E5
6
5
2
9
.
2
0
2
2
.
9
9
0
5
2
0
8
.
[
2
2
]
J.
R
a
v
a
l
e
t
a
l
.
,
“
R
a
K
S
h
A
:
a
t
r
u
s
t
e
d
e
x
p
l
a
i
n
a
b
l
e
LST
M
m
o
d
e
l
t
o
c
l
a
ssi
f
y
f
r
a
u
d
p
a
t
t
e
r
n
s
o
n
c
r
e
d
i
t
c
a
r
d
t
r
a
n
sa
c
t
i
o
n
s,”
M
a
t
h
e
m
a
t
i
c
s
,
v
o
l
.
1
1
,
n
o
.
8
,
A
p
r
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
m
a
t
h
1
1
0
8
1
9
0
1
.
[
2
3
]
J.
-
W
.
C
h
a
n
g
,
N
.
Y
e
n
,
a
n
d
J.
C
.
H
u
n
g
,
“
D
e
si
g
n
o
f
a
N
LP
-
e
m
p
o
w
e
r
e
d
f
i
n
a
n
c
e
f
r
a
u
d
a
w
a
r
e
n
e
ss
m
o
d
e
l
:
t
h
e
a
n
t
i
-
f
r
a
u
d
c
h
a
t
b
o
t
f
o
r
f
r
a
u
d
d
e
t
e
c
t
i
o
n
a
n
d
f
r
a
u
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
a
s
a
n
i
n
st
a
n
c
e
,
”
J
o
u
rn
a
l
o
f
Am
b
i
e
n
t
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
H
u
m
a
n
i
ze
d
C
o
m
p
u
t
i
n
g
,
v
o
l
.
1
3
,
n
o
.
1
0
,
p
p
.
4
6
6
3
–
4
6
7
9
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
0
7
/
s1
2
6
5
2
-
021
-
0
3
5
1
2
-
2.
[
2
4
]
Y
.
H
a
n
y
f
a
n
d
H
.
S
i
l
k
a
n
,
“
A
me
t
h
o
d
f
o
r
mi
ssi
n
g
v
a
l
u
e
s
i
m
p
u
t
a
t
i
o
n
o
f
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
d
a
t
a
se
t
s,
”
I
AE
S
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
Art
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
3
,
n
o
.
1
,
pp
.
8
8
8
-
898
,
M
a
r
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
a
i
.
v
1
3
.
i
1
.
p
p
8
8
8
-
8
9
8
.
[
2
5
]
T.
V
e
r
d
o
n
c
k
,
B
.
B
a
e
s
e
n
s
,
M
.
Ó
sk
a
r
sd
ó
t
t
i
r
,
a
n
d
S
.
V
.
B
r
o
u
c
k
e
,
“
S
p
e
c
i
a
l
i
ss
u
e
o
n
f
e
a
t
u
r
e
e
n
g
i
n
e
e
r
i
n
g
e
d
i
t
o
r
i
a
l
,
”
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
v
o
l
.
1
1
3
,
n
o
.
7
,
p
p
.
3
9
1
7
–
3
9
2
8
,
Ju
l
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
0
7
/
s
1
0
9
9
4
-
0
2
1
-
0
6
0
4
2
-
2.
[
2
6
]
E.
K
a
t
y
a
,
“
E
x
p
l
o
r
i
n
g
f
e
a
t
u
r
e
e
n
g
i
n
e
e
r
i
n
g
s
t
r
a
t
e
g
i
e
s
f
o
r
i
m
p
r
o
v
i
n
g
p
r
e
d
i
c
t
i
v
e
m
o
d
e
l
s
i
n
d
a
t
a
sc
i
e
n
c
e
,
”
R
e
se
a
rc
h
J
o
u
r
n
a
l
o
f
C
o
m
p
u
t
e
r
S
y
st
e
m
s E
n
g
i
n
e
e
ri
n
g
,
v
o
l
.
4
,
n
o
.
2
,
p
p
.
2
0
1
–
2
1
5
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
5
2
7
1
0
/
r
j
c
se.
8
8
.
[
2
7
]
F
.
C
.
S
o
u
z
a
,
R
.
F
.
N
o
g
u
e
i
r
a
,
a
n
d
R
.
A
.
Lo
t
u
f
o
,
“
B
ER
T
mo
d
e
l
s
f
o
r
b
r
a
z
i
l
i
a
n
p
o
r
t
u
g
u
e
se
:
p
r
e
t
r
a
i
n
i
n
g
,
e
v
a
l
u
a
t
i
o
n
a
n
d
t
o
k
e
n
i
z
a
t
i
o
n
a
n
a
l
y
si
s
,
”
Ap
p
l
i
e
d
S
o
f
t
C
o
m
p
u
t
i
n
g
,
v
o
l
.
1
4
9
,
D
e
c
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
a
s
o
c
.
2
0
2
3
.
1
1
0
9
0
1
.
[
2
8
]
A
.
N
a
y
a
k
,
H
.
T
i
mm
a
p
a
t
h
i
n
i
,
K
.
P
o
n
n
a
l
a
g
u
,
a
n
d
V
.
G
.
V
e
n
k
o
p
a
r
a
o
,
“
D
o
m
a
i
n
a
d
a
p
t
a
t
i
o
n
c
h
a
l
l
e
n
g
e
s
o
f
B
E
R
T
i
n
t
o
k
e
n
i
z
a
t
i
o
n
a
n
d
su
b
-
w
o
r
d
r
e
p
r
e
se
n
t
a
t
i
o
n
s
o
f
O
u
t
-
of
-
V
o
c
a
b
u
l
a
r
y
w
o
r
d
s,”
i
n
F
i
rst
Wo
r
k
sh
o
p
o
n
I
n
si
g
h
t
s
f
r
o
m
N
e
g
a
t
i
v
e
Re
su
l
t
s
i
n
N
L
P
,
S
t
r
o
u
d
s
b
u
r
g
,
P
A
,
U
S
A
:
A
sso
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s,
2
0
2
0
,
p
p
.
1
–
5
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
0
.
i
n
s
i
g
h
t
s
-
1
.
1
.
[
2
9
]
G
.
P
u
c
c
e
t
t
i
,
A
.
M
i
a
s
c
h
i
,
a
n
d
F
.
D
e
l
l
’
O
r
l
e
t
t
a
,
“
H
o
w
d
o
B
E
R
T
e
m
b
e
d
d
i
n
g
s
o
r
g
a
n
i
z
e
l
i
n
g
u
i
s
t
i
c
k
n
o
w
l
e
d
g
e
?
,
”
i
n
P
r
o
c
e
e
d
i
n
g
s
o
f
D
e
e
p
L
e
a
r
n
i
n
g
I
n
s
i
d
e
O
u
t
(
D
e
e
L
I
O
)
:
T
h
e
2
n
d
W
o
r
k
s
h
o
p
o
n
K
n
o
w
l
e
d
g
e
E
x
t
r
a
c
t
i
o
n
a
n
d
I
n
t
e
g
r
a
t
i
o
n
f
o
r
D
e
e
p
L
e
a
r
n
i
n
g
A
r
c
h
i
t
e
c
t
u
r
e
s
,
S
t
r
o
u
d
s
b
u
r
g
,
P
A
,
U
S
A
:
A
s
s
o
c
i
a
t
i
o
n
f
o
r
C
o
m
p
u
t
a
t
i
o
n
a
l
L
i
n
g
u
i
s
t
i
c
s
,
2
0
2
1
,
p
p
.
4
8
–
57
,
d
o
i
:
1
0
.
1
8
6
5
3
/
v
1
/
2
0
2
1
.
d
e
e
l
i
o
-
1
.
6
.
[
3
0
]
F
.
A
.
G
h
a
l
e
b
,
F
.
S
a
e
e
d
,
M
.
A
l
-
S
a
r
e
m,
S
.
N
.
Q
a
sem
,
a
n
d
T.
A
l
-
H
a
d
h
r
a
m
i
,
“
E
n
sem
b
l
e
s
y
n
t
h
e
s
i
z
e
d
m
i
n
o
r
i
t
y
o
v
e
r
sam
p
l
i
n
g
-
b
a
s
e
d
g
e
n
e
r
a
t
i
v
e
a
d
v
e
r
sari
a
l
n
e
t
w
o
r
k
s
a
n
d
r
a
n
d
o
m
f
o
r
e
s
t
a
l
g
o
r
i
t
h
m
f
o
r
c
r
e
d
i
t
c
a
r
d
f
r
a
u
d
d
e
t
e
c
t
i
o
n
,
”
I
EE
E
Ac
c
e
ss
,
v
o
l
.
1
1
,
p
p
.
8
9
6
9
4
–
8
9
7
1
0
,
2
0
2
3
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
3
.
3
3
0
6
6
2
1
.
[
3
1
]
H
.
H
a
i
r
a
n
i
,
T.
W
i
d
i
y
a
n
i
n
g
t
y
a
s,
a
n
d
D
.
D
.
P
r
a
s
e
t
y
a
,
“
A
d
d
r
e
ssi
n
g
c
l
a
ss
i
m
b
a
l
a
n
c
e
o
f
h
e
a
l
t
h
d
a
t
a
:
a
sy
s
t
e
m
a
t
i
c
l
i
t
e
r
a
t
u
r
e
r
e
v
i
e
w
o
n
mo
d
i
f
i
e
d
s
y
n
t
h
e
t
i
c
m
i
n
o
r
i
t
y
o
v
e
r
sa
mp
l
i
n
g
t
e
c
h
n
i
q
u
e
(
S
M
O
TE)
st
r
a
t
e
g
i
e
s,”
J
O
I
V
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
rn
a
l
o
f
I
n
f
o
rm
a
t
i
c
s
a
n
d
Vi
su
a
l
i
z
a
t
i
o
n
,
v
o
l
.
8
,
n
o
.
3
,
S
e
p
.
2
0
2
4
,
d
o
i
:
1
0
.
6
2
5
2
7
/
j
o
i
v
.
8
.
3
.
2
2
8
3
.
[
3
2
]
B
.
G
h
o
j
o
g
h
a
n
d
A
.
G
h
o
d
s
i
,
“
A
t
t
e
n
t
i
o
n
me
c
h
a
n
i
sm,
t
r
a
n
sf
o
r
mers
,
B
ER
T,
a
n
d
G
P
T
:
t
u
t
o
r
i
a
l
a
n
d
s
u
r
v
e
y
,
”
O
S
F
Pre
p
r
i
n
t
s
,
D
e
c
.
2
0
2
0
,
d
o
i
:
1
0
.
3
1
2
1
9
/
o
sf
.
i
o
/
m
6
g
c
n
.
[
3
3
]
L.
Z
h
a
o
,
L
.
Li
,
X
.
Z
h
e
n
g
,
a
n
d
J.
Z
h
a
n
g
,
“
A
B
E
R
T
b
a
se
d
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
a
n
d
k
e
y
e
n
t
i
t
y
d
e
t
e
c
t
i
o
n
a
p
p
r
o
a
c
h
f
o
r
o
n
l
i
n
e
f
i
n
a
n
c
i
a
l
t
e
x
t
s,
”
i
n
2
0
2
1
I
EE
E
2
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
e
n
c
e
o
n
C
o
m
p
u
t
e
r
S
u
p
p
o
rt
e
d
C
o
o
p
e
r
a
t
i
v
e
W
o
r
k
i
n
D
e
si
g
n
(
C
S
C
WD)
,
I
EEE,
M
a
y
2
0
2
1
,
p
p
.
1
2
3
3
–
1
2
3
8
,
d
o
i
:
1
0
.
1
1
0
9
/
C
S
C
W
D
4
9
2
6
2
.
2
0
2
1
.
9
4
3
7
6
1
6
.
[
3
4
]
R
.
D
i
P
i
e
t
r
o
a
n
d
G
.
D
.
H
a
g
e
r
,
“
D
e
e
p
l
e
a
r
n
i
n
g
:
R
N
N
s
a
n
d
LS
TM
,
”
i
n
H
a
n
d
b
o
o
k
o
f
M
e
d
i
c
a
l
I
m
a
g
e
C
o
m
p
u
t
i
n
g
a
n
d
C
o
m
p
u
t
e
r
Assi
st
e
d
I
n
t
e
r
v
e
n
t
i
o
n
,
El
s
e
v
i
e
r
,
2
0
2
0
,
p
p
.
5
0
3
–
519
,
d
o
i
:
1
0
.
1
0
1
6
/
B
9
7
8
-
0
-
12
-
8
1
6
1
7
6
-
0
.
0
0
0
2
6
-
0.
[
3
5
]
C
.
U
b
a
l
,
G
.
D
.
-
G
i
o
r
g
i
,
J.
E.
C
.
-
R
e
y
e
s,
a
n
d
R
.
S
a
l
a
s
,
“
P
r
e
d
i
c
t
i
n
g
t
h
e
l
o
n
g
-
t
e
r
m
d
e
p
e
n
d
e
n
c
i
e
s
i
n
t
i
me
ser
i
e
s
u
si
n
g
r
e
c
u
r
r
e
n
t
a
r
t
i
f
i
c
i
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
s,
”
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
a
n
d
K
n
o
w
l
e
d
g
e
Ex
t
ra
c
t
i
o
n
,
v
o
l
.
5
,
n
o
.
4
,
p
p
.
1
3
4
0
–
1
3
5
8
,
O
c
t
.
2
0
2
3
,
d
o
i
:
1
0
.
3
3
9
0
/
ma
k
e
5
0
4
0
0
6
8
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
O
u
ss
a
m
a
Nd
a
m
a
is
a
d
o
c
to
r
P
h
.
D
.
in
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
Artifi
c
ial
In
telli
g
e
n
c
e
with
th
e
DSAI
2
S
(
Da
ta
S
c
ien
c
e
,
Artifi
c
ial
In
tell
ig
e
n
c
e
a
n
d
S
m
a
rt
S
y
ste
m
s)
Re
se
a
rc
h
Tea
m
,
C3
S
Lab
o
ra
to
r
y
,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s
a
n
d
Tec
h
n
o
lo
g
ies
,
Tan
g
ier,
M
o
r
o
c
c
o
.
He
is
a
lso
a
b
u
si
n
e
ss
in
tell
ig
e
n
c
e
e
n
g
i
n
e
e
r
with
m
o
re
t
h
a
n
6
y
e
a
rs
o
f
e
x
p
e
rien
c
e
in
m
u
lt
in
a
ti
o
n
a
l
c
o
m
p
a
n
ies
.
H
is
re
s
e
a
rc
h
in
tere
sts
in
c
lu
d
e
sm
a
rt
sy
ste
m
s,
m
a
c
h
in
e
lea
rn
in
g
,
d
e
e
p
lea
rn
in
g
,
n
a
tu
ra
l
lan
g
u
a
g
e
p
r
o
c
e
ss
in
g
(NLP
),
a
rti
ficia
l
n
e
u
ra
l
n
e
two
rk
s
(AN
N),
se
n
ti
m
e
n
t
a
n
a
ly
sis,
a
n
d
sm
a
rt
c
it
ies
.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
o
u
ss
a
m
a
.
n
d
a
m
a
@e
tu
.
u
a
e
.
a
c
.
m
a
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
A
n
o
ve
l B
E
R
T
-
lo
n
g
s
h
o
r
t
-
term me
mo
r
y
h
yb
r
id
mo
d
el
fo
r
effe
ctive
cred
it c
a
r
d
…
(
Ou
s
s
a
ma
N
d
a
ma
)
797
S
a
fa
e
N
d
a
m
a
is
a
P
h
.
D
.
st
u
d
e
n
t
i
n
DSAI
2
S
(Da
ta
S
c
ie
n
c
e
,
Arti
ficia
l
In
telli
g
e
n
c
e
a
n
d
S
m
a
rt
S
y
ste
m
s
Re
se
a
rc
h
Tea
m
),
C3
S
Lab
o
ra
to
ry
,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s
a
n
d
Tec
h
n
o
l
o
g
ies
(F
S
T),
Tan
g
ier,
M
o
r
o
c
c
o
.
S
h
e
e
a
r
n
e
d
h
e
r
m
a
ste
r'
s
in
C
o
m
p
u
ter
S
c
i
e
n
c
e
a
n
d
B
ig
Da
ta
fro
m
th
e
F
S
T
o
f
Tan
g
ier.
Wi
th
two
y
e
a
rs
o
f
e
x
p
e
rien
c
e
a
s
a
d
a
ta
sc
ien
ti
st,
h
e
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
sm
a
rt
sy
ste
m
s,
d
a
ta
sc
ien
c
e
,
a
rti
ficia
l
in
tell
ig
e
n
c
e
,
se
n
ti
m
e
n
t
a
n
a
ly
sis,
a
n
d
sm
a
rt
c
it
ies
.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
sa
fa
e
.
n
d
a
m
a
@e
tu
.
u
a
e
.
a
c
.
m
a
.
Is
m
a
il
Be
n
sa
ss
i
is
a
d
o
c
to
r
in
DSAI2
S
(Da
ta
S
c
ien
c
e
,
Artif
icia
l
In
telli
g
e
n
c
e
a
n
d
S
m
a
rt
S
y
ste
m
s
Re
se
a
rc
h
Tea
m
)
,
C3
S
La
b
o
ra
t
o
ry
,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s
a
n
d
Tec
h
n
o
lo
g
ies
,
Tan
g
ier,
M
o
r
o
c
c
o
.
He
is
a
n
e
n
g
i
n
e
e
r
in
Co
m
p
u
ter
S
c
ien
c
e
,
lau
re
a
te
o
f
F
S
T
o
f
Tan
g
ier.
His
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
sm
a
rt
c
o
n
n
e
c
ti
o
n
o
f
u
se
r
p
ro
fi
les
in
a
b
i
g
d
a
ta
c
o
n
te
x
t,
m
u
lt
i
-
a
g
e
n
t
sy
ste
m
s
(M
AS),
c
a
se
-
b
a
se
d
re
a
s
o
n
i
n
g
(CBR),
o
n
to
l
o
g
y
,
m
a
c
h
in
e
lea
rn
in
g
,
sm
a
rt
c
it
ies
,
a
n
d
e
Lea
rn
in
g
/M
OO
C/
S
P
OC.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
b
e
n
sa
ss
i.
is
m
a
il
@g
m
a
il
.
c
o
m
.
Dr
.
El
Mo
k
h
t
a
r
En
-
Na
i
m
i
is
a
fu
ll
p
ro
fe
ss
o
r
in
t
h
e
Un
iv
e
rsi
ty
o
f
A
b
d
e
lma
lek
Essa
â
d
i
(UA
E),
F
a
c
u
lt
y
o
f
S
c
ie
n
c
e
s
a
n
d
Tec
h
n
o
l
o
g
ies
(F
S
T)
o
f
Tan
g
ier,
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
s.
He
wa
s
tem
p
o
ra
ry
p
r
o
fe
ss
o
r
fr
o
m
1
9
9
9
to
2
0
0
3
a
n
d
p
e
rm
a
n
e
n
t
p
ro
fe
ss
o
r
sin
c
e
2
0
0
3
/2
0
0
4
u
n
ti
l
n
o
w.
He
is
a
F
u
ll
P
r
o
fe
ss
o
r
i
n
UA
E,
F
S
T
o
f
Tan
g
ier.
He
wa
s
a
h
e
a
d
i
n
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
s,
sin
c
e
Oc
to
b
e
r
2
0
1
6
u
n
ti
l
t
h
e
e
n
d
o
f
De
c
e
m
b
e
r
2
0
2
0
.
He
wa
s
re
sp
o
n
sib
le
fo
r
a
Li
c
e
n
se
o
f
S
c
ien
c
e
a
n
d
Tec
h
n
o
lo
g
y
,
L
S
T
Co
m
p
u
ter
En
g
in
e
e
rin
g
(“
Li
c
e
n
c
e
LS
T
-
G
I
”
),
fro
m
Ja
n
u
a
ry
2
0
1
2
to
Oc
to
b
e
r
2
0
1
6
.
He
is
a
c
h
ief
o
f
Da
ta
S
c
i
e
n
c
e
,
Artifi
c
ial
In
telli
g
e
n
c
e
a
n
d
S
m
a
rt
S
y
ste
m
s
(DSAI2
S
)
Re
se
a
rc
h
Tea
m
sin
c
e
th
e
a
c
a
d
e
m
ic
y
e
a
r
2
0
2
2
/
2
0
2
3
.
He
is also
a
fo
u
n
d
in
g
m
e
m
b
e
r
o
f
th
e
b
o
t
h
lab
o
ra
to
ries
:
Lab
o
ra
to
ire
d
'
In
fo
rm
a
ti
q
u
e
,
S
y
stè
m
e
s
e
t
Télé
c
o
m
m
u
n
ica
ti
o
n
s
(LI
S
T)
Lab
o
ra
to
r
y
(fro
m
2
0
0
8
to
2
0
2
2
)
a
n
d
C
o
m
p
u
ter
S
c
ien
c
e
a
n
d
S
m
a
rt
S
y
ste
m
s
(C3
S
)
Lab
o
ra
to
r
y
sin
c
e
th
e
a
c
a
d
e
m
ic
y
e
a
r
2
0
2
2
/
2
0
2
3
u
n
t
il
n
o
w,
th
e
Un
i
v
e
rsity
o
f
Ab
d
e
lma
lek
Essa
â
d
i,
F
S
T
o
f
Tan
g
ier,
M
o
ro
c
c
o
.
He
is
a
lso
a
n
e
x
p
e
r
t
e
v
a
lu
a
to
r
with
t
h
e
AN
EAQ,
sin
c
e
th
e
a
c
a
d
e
m
ic
y
e
a
r
2
0
1
6
/
2
0
1
7
u
n
ti
l
n
o
w
,
t
h
a
t
a
n
e
x
p
e
rt
o
f
th
e
p
ri
v
a
te
e
sta
b
li
sh
m
e
n
ts
b
e
l
o
n
g
in
g
to
th
e
territ
o
ry
o
f
t
h
e
UA
E
a
n
d
a
lso
a
n
e
x
p
e
rt
o
f
t
h
e
in
it
ial
o
r
f
u
n
d
a
m
e
n
tal
fo
rm
a
ti
o
n
s
a
n
d
fo
rm
a
ti
o
n
s
c
o
n
t
in
u
o
u
s
a
t
th
e
M
in
istr
y
o
f
Hig
h
e
r
Ed
u
c
a
ti
o
n
,
S
c
ien
ti
f
ic
Re
se
a
rc
h
a
n
d
E
x
e
c
u
ti
v
e
Train
in
g
a
n
d
a
ls
o
a
t
t
h
e
UA
E
U
n
iv
e
rsit
y
a
n
d
th
e
F
S
T
Tan
g
ier
sin
c
e
2
0
1
2
/2
0
1
3
u
n
ti
l
n
o
w.
He
is
a
n
a
u
t
h
o
r/co
-
a
u
th
o
rs
o
f
se
v
e
ra
l
a
rti
c
les
,
p
u
b
li
sh
e
d
i
n
T
h
e
In
tern
a
ti
o
n
a
l
J
o
u
r
n
a
ls
in
Co
m
p
u
ter
S
c
ien
c
e
s,
i
n
p
a
rti
c
u
lar,
i
n
m
u
lt
i
-
a
g
e
n
t
sy
ste
m
s
(M
AS),
c
a
se
s
-
b
a
se
d
re
a
so
n
in
g
(CBR)
,
a
rti
ficia
l
in
telli
g
e
n
t
(AI),
m
a
c
h
in
e
lea
rn
in
g
(M
L),
d
e
e
p
lea
rn
in
g
(DL),
e
Lea
r
n
in
g
,
M
OO
C/S
P
OC,
b
i
g
d
a
ta,
d
a
ta
-
m
in
in
g
,
wire
les
s
se
n
so
r
n
e
two
rk
,
VA
Ne
t,
M
AN
e
t,
a
n
d
s
m
a
rt
c
it
y
.
He
wa
s/is
a
lso
d
irec
to
r
o
f
se
v
e
ra
l
d
o
c
to
ra
l
th
e
se
s
in
c
o
m
p
u
ter
sc
ien
c
e
s.
He
h
a
s
t
o
o
s
e
rv
e
d
a
s
a
g
e
n
e
ra
l
c
h
a
ir,
tec
h
n
ic
a
l
p
ro
g
ra
m
c
h
a
ir,
tec
h
n
ica
l
p
ro
g
ra
m
c
o
m
m
it
tee
m
e
m
b
e
r,
o
rg
a
n
izin
g
c
o
m
m
it
tee
m
e
m
b
e
r,
se
ss
io
n
c
h
a
ir
,
a
n
d
re
v
iew
e
r
fo
r
m
a
n
y
i
n
tern
a
ti
o
n
a
l
c
o
n
fe
re
n
c
e
s
a
n
d
wo
rk
s
h
o
p
s.
I
n
a
d
d
i
ti
o
n
,
h
e
is
a
n
a
ss
o
c
iate
m
e
m
b
e
r
o
f
th
e
IS
CN
-
In
stit
u
te
o
f
Co
m
p
lex
S
y
ste
m
s
in
N
o
rm
a
n
d
y
,
th
e
U
n
iv
e
rsit
y
o
f
t
h
e
Ha
v
re
,
F
ra
n
c
e
,
si
n
c
e
2
0
0
9
u
n
t
il
n
o
w.
He
c
a
n
b
e
c
o
n
tac
ted
a
t
e
m
a
il
:
e
n
-
n
a
imi@u
a
e
.
a
c
.
m
a
.
Evaluation Warning : The document was created with Spire.PDF for Python.