I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
1
6
,
No
.
2
,
A
p
r
il
20
2
6
,
p
p
.
991
~
1
0
0
1
I
SS
N:
2088
-
8
7
0
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijece.
v
1
6
i
2
.
pp
9
9
1
-
1
0
0
1
991
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
M
ultimo
da
l
ma
ch
ine learning
fram
ewo
rk f
o
r
fake r
e
v
iew
detec
tion
Ra
s
hm
i R
.
,
Sh
o
bh
a
T
.
,
Dha
n
us
hree
C
.
S
.
,
G
a
y
a
t
ri
S
.
Sa
nt
i,
J
ee
v
it
a
S
.
De
v
a
dig
,
H
a
rsh
it
ha
L
.
V
.
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
o
n
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
B
.
M
.
S
.
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
,
B
e
n
g
a
l
u
r
u
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
1
,
2
0
2
4
R
ev
is
ed
J
an
1
,
2
0
2
6
Acc
ep
ted
J
an
1
6
,
2
0
2
6
On
li
n
e
re
v
iew
s
sig
n
ifi
c
a
n
tl
y
i
n
fl
u
e
n
c
e
c
o
n
s
u
m
e
r
d
e
c
isio
n
-
m
a
k
in
g
,
y
e
t
th
e
ir
c
re
d
ib
il
it
y
is
i
n
c
re
a
sin
g
l
y
u
n
d
e
r
m
in
e
d
b
y
t
h
e
rise
o
f
fa
k
e
a
n
d
m
a
n
ip
u
late
d
c
o
n
ten
t.
T
h
is
stu
d
y
a
d
d
re
ss
e
s
th
e
g
ro
wi
n
g
c
h
a
ll
e
n
g
e
o
f
d
e
tec
ti
n
g
d
e
c
e
p
ti
v
e
o
n
li
n
e
re
v
iew
s
b
y
d
e
v
e
lo
p
in
g
a
h
ig
h
ly
a
c
c
u
ra
te,
ro
b
u
st,
a
n
d
e
x
p
lai
n
a
b
le
m
a
c
h
in
e
lea
rn
in
g
fra
m
e
wo
rk
th
a
t
su
p
p
o
rts
tru
st
a
n
d
re
li
a
b
il
it
y
in
d
i
g
it
a
l
m
a
rk
e
tp
lac
e
s.
Th
e
p
ro
p
o
se
d
m
u
lt
imo
d
a
l
fra
m
e
wo
rk
i
n
teg
ra
t
e
s
tex
tu
a
l,
b
e
h
a
v
i
o
u
ra
l,
tem
p
o
ra
l,
a
n
d
n
e
t
wo
rk
-
b
a
se
d
fe
a
t
u
re
s
to
e
n
h
a
n
c
e
d
e
tec
ti
o
n
pe
rfo
rm
a
n
c
e
.
Tex
t
u
a
l
c
h
a
ra
c
teristics
a
re
e
x
trac
ted
u
si
n
g
term
f
re
q
u
e
n
c
y
-
in
v
e
rse
d
o
c
u
m
e
n
t
fre
q
u
e
n
c
y
(
TF
-
IDF)
a
n
d
se
n
ti
m
e
n
t
a
n
a
l
y
sis,
wh
il
e
b
e
h
a
v
i
o
u
ra
l
a
n
d
tem
p
o
ra
l
a
t
tri
b
u
tes
m
o
d
e
l
re
v
iew
e
r
a
c
ti
v
it
y
p
a
tt
e
rn
s.
Ne
two
rk
-
o
rien
ted
fe
a
tu
re
s
c
a
p
t
u
re
su
sp
ici
o
u
s
re
v
iew
e
r
in
tera
c
ti
o
n
s.
To
m
it
ig
a
te
c
las
s
imb
a
lan
c
e
,
sy
n
t
h
e
ti
c
sa
m
p
les
a
re
g
e
n
e
ra
ted
u
sin
g
t
h
e
sy
n
t
h
e
ti
c
m
in
o
rit
y
o
v
e
r
-
sa
m
p
li
n
g
tec
h
n
i
q
u
e
(S
M
OTE
).
S
e
v
e
ra
l
m
a
c
h
in
e
lea
rn
in
g
m
o
d
e
ls
—
in
c
lu
d
i
n
g
lo
g
isti
c
re
g
r
e
ss
io
n
,
d
e
c
isio
n
tree
s,
XG
Bo
o
st,
a
n
d
a
sta
c
k
in
g
e
n
se
m
b
le
—
a
re
train
e
d
a
n
d
e
v
a
lu
a
ted
.
E
x
p
e
rime
n
tal
f
in
d
i
n
g
s
s
h
o
w
th
a
t
XG
Bo
o
st
a
n
d
t
h
e
sta
c
k
in
g
e
n
se
m
b
le
d
e
li
v
e
r
stro
n
g
b
a
lan
c
e
d
p
e
rfo
rm
a
n
c
e
,
a
c
h
iev
i
n
g
a
n
F
1
-
sc
o
re
o
f
a
p
p
r
o
x
ima
tely
0
.
8
7
a
n
d
a
n
a
c
c
u
ra
c
y
o
f
0
.
9
4
.
De
c
isio
n
Tree
s
e
x
h
i
b
it
h
ig
h
p
re
c
isio
n
(0
.
9
8
),
a
l
b
e
it
wit
h
c
o
m
p
a
ra
ti
v
e
ly
lo
we
r
re
c
a
ll
.
T
o
e
n
su
re
tran
sp
a
re
n
c
y
a
n
d
in
ter
p
re
tab
il
it
y
,
S
h
a
p
ley
a
d
d
it
iv
e
e
x
p
lan
a
ti
o
n
s
(S
HA
P
)
a
re
u
se
d
to
a
n
a
ly
se
m
o
d
e
l
p
re
d
ictio
n
s.
Re
su
lt
s
i
n
d
ica
te
th
a
t
re
v
iew
e
r
c
o
n
n
e
c
ti
v
it
y
,
c
o
-
re
v
iew
e
r
c
o
u
n
ts,
a
n
d
se
n
ti
m
e
n
t
–
ra
ti
n
g
in
c
o
n
siste
n
c
ies
a
re
a
m
o
n
g
th
e
m
o
st i
n
fl
u
e
n
ti
a
l
fe
a
tu
re
s.
Ov
e
ra
ll
,
th
e
p
ro
p
o
se
d
fra
m
e
wo
rk
e
n
h
a
n
c
e
s
d
e
tec
ti
o
n
a
c
c
u
ra
c
y
a
n
d
p
ro
v
i
d
e
s
m
e
a
n
in
g
fu
l
,
e
x
p
lai
n
a
b
le
in
si
g
h
ts,
m
a
k
in
g
it
we
ll
-
su
it
e
d
f
o
r
d
e
p
l
o
y
m
e
n
t
in
re
a
l
-
wo
rld
d
i
g
it
a
l
m
a
rk
e
tp
lac
e
s.
F
u
tu
re
wo
rk
will
f
o
c
u
s
o
n
e
x
te
n
d
i
n
g
t
h
e
fra
m
e
wo
rk
to
m
u
lt
il
i
n
g
u
a
l
d
a
tas
e
ts
a
n
d
i
n
c
o
rp
o
ra
ti
n
g
a
d
a
p
ti
v
e
lea
rn
in
g
m
e
c
h
a
n
ism
s to
a
d
d
re
ss
e
v
o
lv
i
n
g
d
e
c
e
p
ti
v
e
b
e
h
a
v
io
u
r.
K
ey
w
o
r
d
s
:
E
n
s
em
b
le
m
o
d
els
E
x
p
lain
ab
ilit
y
Fak
e
r
ev
iew
d
etec
tio
n
Ma
ch
in
e
lear
n
in
g
Mu
ltimo
d
al
f
ea
tu
r
es
Sh
ap
ley
ad
d
itiv
e
e
x
p
lan
atio
n
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
:
R
ash
m
i R
.
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
Scien
ce
an
d
E
n
g
in
ee
r
i
n
g
,
B
.
M.
S
.
C
o
lleg
e
o
f
E
n
g
in
ee
r
in
g
B
u
ll T
em
p
le
R
o
ad
,
B
en
g
alu
r
u
-
5
6
0
0
1
9
,
KA,
I
n
d
ia
E
m
ail: r
ash
m
ir
.
is
e@
b
m
s
ce
.
ac
.
in
1.
I
NT
RO
D
UCT
I
O
N
E
v
er
y
tim
e
a
co
n
s
u
m
er
o
r
b
u
s
in
ess
lo
o
k
s
o
n
lin
e,
th
ey
s
e
e
th
at
o
n
lin
e
r
ev
iews
h
av
e
b
ec
o
m
e
an
ess
en
tial
p
ar
t
o
f
th
ei
r
r
esear
ch
p
r
o
ce
s
s
wh
en
ch
o
o
s
in
g
wh
at
p
r
o
d
u
cts
to
b
u
y
.
W
h
en
it
c
o
m
es
to
o
n
lin
e
r
etail
an
d
h
o
s
p
itality
s
er
v
ices,
ab
o
u
t
9
0
%
o
f
to
d
ay
’
s
co
n
s
u
m
er
s
lo
o
k
at
r
ev
iews
b
ef
o
r
e
d
ec
id
i
n
g
o
n
a
p
ar
ticu
lar
p
r
o
d
u
ct
o
r
s
er
v
ice
[
1
]
.
R
ec
en
t
s
tu
d
ies
h
av
e
s
h
o
wn
th
at
9
3
%
o
f
co
n
s
u
m
er
s
r
ely
o
n
o
n
lin
e
r
ev
iews
as
a
b
asi
s
f
o
r
d
ec
id
in
g
wh
eth
e
r
to
p
u
r
ch
ase
a
p
r
o
d
u
ct
an
d
th
at
th
ey
tr
u
s
t
th
ese
r
ev
iews
ju
s
t
as
m
u
ch
as
th
ey
wo
u
ld
t
r
u
s
t
p
er
s
o
n
al
r
ec
o
m
m
en
d
atio
n
s
f
r
o
m
f
r
ien
d
s
o
r
f
am
ily
[
2
]
.
I
n
t
h
e
h
o
s
p
itality
in
d
u
s
tr
y
,
o
n
lin
e
r
e
v
iews
p
lay
an
e
v
en
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
9
1
-
1
0
0
1
992
m
o
r
e
p
iv
o
tal
r
o
le
in
in
f
lu
e
n
c
in
g
co
n
s
u
m
er
b
eh
av
i
o
u
r
s
,
g
i
v
en
th
at
m
o
s
t
o
f
th
e
p
r
o
d
u
cts
(
e.
g
.
,
h
o
tels
an
d
r
estau
r
an
ts
)
ar
e
in
tan
g
ib
le
an
d
ca
n
n
o
t
b
e
ex
p
er
ie
n
ce
d
u
n
til
a
f
ter
th
e
p
u
r
ch
ase
h
as
b
ee
n
m
a
d
e.
Un
f
o
r
tu
n
ately
,
m
an
y
co
n
s
u
m
e
r
s
ar
e
h
av
in
g
th
eir
v
iews
o
f
h
o
tels
an
d
r
estau
r
an
ts
d
is
to
r
ted
b
y
th
e
p
r
o
life
r
atio
n
o
f
f
a
k
e
r
ev
iews.
Un
f
o
r
t
u
n
ately
,
th
e
la
ck
o
f
s
tr
ict
p
u
r
ch
asin
g
v
er
if
ic
atio
n
p
r
o
ce
s
s
es
am
o
n
g
m
an
y
p
latf
o
r
m
s
th
at
allo
w
f
o
r
co
n
s
u
m
er
p
r
o
d
u
ct
r
atin
g
(
i.e
.
,
Yelp
an
d
T
r
ip
a
d
v
is
o
r
)
h
as
ad
d
ed
an
o
th
e
r
lay
er
o
f
v
u
l
n
er
ab
ilit
y
t
o
in
d
ep
en
d
en
t
h
o
tels
an
d
r
estau
r
an
ts
th
at
r
ely
h
ea
v
ily
o
n
p
o
s
itiv
e
o
n
lin
e
r
ev
iews
to
estab
lis
h
th
em
s
elv
es
with
in
th
e
m
ar
k
etp
lace
.
Sev
er
al
s
tu
d
ies
h
av
e
d
em
o
n
s
tr
ated
th
e
ef
f
ec
tiv
e
n
ess
o
f
s
u
p
er
v
is
ed
m
ac
h
in
e
lear
n
in
g
al
g
o
r
ith
m
s
,
s
u
ch
as
d
ec
is
io
n
tr
ee
s
,
r
an
d
o
m
f
o
r
ests
,
an
d
s
u
p
p
o
r
t v
ec
to
r
m
ac
h
in
es,
in
id
en
tify
in
g
f
ak
e
r
e
v
iews
u
s
in
g
tex
tu
al
an
d
r
e
v
iewe
r
-
b
ased
f
ea
tu
r
es
[
3
]
.
R
ec
en
t
wo
r
k
h
as
also
s
h
o
wn
th
at
co
m
b
i
n
in
g
m
u
ltip
le
class
ical
m
ac
h
in
e
lear
n
in
g
m
o
d
els
ca
n
im
p
r
o
v
e
f
ak
e
r
ev
iew
d
etec
tio
n
ac
cu
r
ac
y
,
ev
en
wh
en
la
b
elled
d
ata
is
lim
ited
[
4
]
.
E
ar
ly
ap
p
r
o
ac
h
es
to
f
ak
e
r
ev
iew
d
etec
tio
n
p
r
im
ar
ily
r
elied
o
n
t
ex
tu
al
an
aly
s
is
,
f
o
cu
s
in
g
o
n
lin
g
u
is
tic
p
atter
n
s
,
s
en
tim
en
t
p
o
lar
ity
,
an
d
wr
itin
g
s
ty
les.
Alth
o
u
g
h
th
ese
m
et
h
o
d
s
ac
h
ie
v
ed
r
ea
s
o
n
ab
le
p
er
f
o
r
m
an
ce
,
th
e
y
ar
e
lim
ited
b
y
th
eir
in
ab
ilit
y
t
o
ca
p
tu
r
e
n
o
n
-
te
x
tu
al
cu
es,
s
u
c
h
a
s
r
ev
iewe
r
b
eh
a
v
io
u
r
,
tem
p
o
r
al
p
o
s
tin
g
p
atter
n
s
,
an
d
co
o
r
d
in
ated
r
e
v
iew
ac
tiv
ities
.
Mo
r
e
r
ec
en
t
s
tu
d
ies
h
av
e
ex
p
lo
r
ed
b
eh
av
i
o
u
r
al
an
d
n
et
wo
r
k
-
b
ased
s
ig
n
als;
h
o
wev
er
,
m
an
y
o
f
th
ese
ap
p
r
o
ac
h
es
s
till
r
ely
o
n
a
s
in
g
le
f
ea
tu
r
e
m
o
d
ality
o
r
lack
in
ter
p
r
etab
ilit
y
,
m
ak
in
g
th
eir
ad
o
p
tio
n
in
r
ea
l
-
wo
r
ld
p
latf
o
r
m
s
ch
allen
g
in
g
.
W
ith
ad
v
an
ce
s
in
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
,
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
s
u
ch
as
DeBERTa
an
d
lar
g
e
lan
g
u
ag
e
m
o
d
els
h
a
v
e
d
em
o
n
s
tr
ated
s
tr
o
n
g
p
er
f
o
r
m
an
ce
b
y
ca
p
tu
r
i
n
g
d
ee
p
er
s
em
an
tic
a
n
d
im
p
licat
io
n
al
ch
a
r
ac
ter
is
tics
o
f
d
ec
e
p
tiv
e
r
ev
iews
[
5
]
.
Ho
wev
er
,
d
esp
ite
h
ig
h
ac
c
u
r
a
cy
,
s
u
ch
m
o
d
els
o
f
ten
f
ac
e
ch
allen
g
es
r
elate
d
to
in
ter
p
r
etab
ilit
y
an
d
r
ea
l
-
wo
r
l
d
d
ep
lo
y
m
e
n
t
[
6
]
.
T
h
e
o
b
jectiv
e
o
f
th
is
s
tu
d
y
is
to
d
esig
n
a
r
elia
b
le,
ac
cu
r
ate,
an
d
in
ter
p
r
etab
le
f
r
am
e
wo
r
k
f
o
r
d
etec
tin
g
f
ak
e
r
e
v
iews
b
y
o
v
e
r
co
m
in
g
th
e
lim
itatio
n
s
o
f
tr
a
d
itio
n
al
tex
t
-
b
ased
an
d
s
in
g
le
-
m
o
d
ality
m
eth
o
d
s
.
T
o
ac
h
iev
e
t
h
is
,
th
e
s
tu
d
y
in
te
g
r
ates
m
u
ltip
le
co
m
p
lem
en
tar
y
f
ea
tu
r
e
ty
p
es
wh
ile
m
ain
tain
in
g
p
r
ac
ticality
f
o
r
r
ea
l
-
wo
r
ld
d
e
p
lo
y
m
e
n
t
o
n
o
n
l
in
e
p
latf
o
r
m
s
.
T
h
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
f
o
cu
s
es
o
n
i
d
en
tify
in
g
f
ak
e
r
ev
iews
in
d
o
m
ain
s
s
u
ch
as
e
-
co
m
m
e
r
ce
an
d
h
o
s
p
itality
u
s
in
g
a
s
tack
in
g
m
ac
h
in
e
lear
n
i
n
g
a
p
p
r
o
ac
h
.
I
t
a
g
g
r
eg
ates
th
e
p
r
ed
ictio
n
s
o
f
s
ev
er
al
b
ase
m
o
d
els
an
d
em
p
lo
y
s
XGBo
o
s
t
as
th
e
f
in
al
class
if
ier
to
en
h
an
ce
d
etec
tio
n
p
er
f
o
r
m
an
ce
.
T
h
e
ap
p
r
o
ac
h
in
co
r
p
o
r
ates
tex
tu
al
f
ea
tu
r
es
(
d
er
iv
ed
th
r
o
u
g
h
n
atu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
(
NL
P)
an
d
ter
m
f
r
eq
u
en
cy
-
in
v
er
s
e
d
o
cu
m
en
t
f
r
e
q
u
en
c
y
(
TF
-
I
DF
)
)
,
b
eh
av
io
u
r
al
attr
ib
u
tes
(
b
ased
o
n
r
ev
iewe
r
ac
tiv
ity
p
atter
n
s
,
in
clu
d
in
g
f
r
e
q
u
en
c
y
an
d
co
n
s
is
ten
cy
)
,
a
n
d
tim
e
-
s
e
r
ies
ch
ar
ac
ter
is
tics
(
f
r
o
m
d
ata
s
ets
s
u
ch
as
Yelp
)
.
T
h
is
m
u
ltimo
d
al
f
ea
tu
r
e
in
teg
r
atio
n
en
ab
les
h
ig
h
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
an
d
r
ec
all
ac
r
o
s
s
b
o
th
s
m
all
an
d
lar
g
e
d
atasets
[
7
]
.
Fu
r
th
er
m
o
r
e,
th
e
in
clu
s
io
n
o
f
Sh
ap
ley
a
d
d
itiv
e
ex
p
lan
atio
n
s
(
SHAP)
with
in
th
e
m
eth
o
d
o
lo
g
y
en
h
an
ce
s
m
o
d
el
in
te
r
p
r
etab
ilit
y
,
o
f
f
e
r
in
g
d
ee
p
er
in
s
ig
h
t
in
to
f
ea
tu
r
e
co
n
tr
i
b
u
tio
n
s
an
d
p
o
t
en
tial
b
ias.
Ov
er
all,
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
n
o
t
o
n
ly
im
p
r
o
v
es
d
etec
tio
n
ac
c
u
r
ac
y
b
u
t
also
p
r
o
v
id
es
m
ea
n
i
n
g
f
u
l
ex
p
lain
ab
ilit
y
,
m
ak
in
g
it we
ll
-
s
u
ited
f
o
r
im
p
l
em
en
tatio
n
in
r
e
al
-
wo
r
ld
d
ig
it
al
m
ar
k
etp
lace
s
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
Fak
e
r
ev
iew
d
etec
tio
n
is
b
ec
o
m
in
g
a
p
o
p
u
lar
f
ield
o
f
s
tu
d
y
with
r
esp
ec
t
to
th
e
u
s
e
o
f
tec
h
n
o
lo
g
y
in
o
n
lin
e
m
a
r
k
etp
lace
s
a
n
d
th
e
p
o
ten
tial
im
p
ac
t
o
n
co
n
s
u
m
e
r
co
n
f
id
en
ce
with
in
th
o
s
e
m
a
r
k
ets.
His
to
r
ically
,
r
esear
ch
er
s
h
av
e
f
o
c
u
s
ed
th
e
ir
ef
f
o
r
ts
o
n
id
en
tify
i
n
g
d
ec
ep
tio
n
s
b
ased
o
n
t
h
e
u
s
e
o
f
lan
g
u
a
g
e
an
d
tex
t
an
aly
s
is
tech
n
iq
u
es.
Fo
r
e
x
a
m
p
le,
Ott
et
a
l.
[
8
]
wer
e
ab
le
to
s
h
o
w
th
at
th
e
r
e
wer
e
m
a
n
y
d
if
f
e
r
en
t
ty
p
es
o
f
lin
g
u
is
tic
f
ea
tu
r
es
ass
o
ciate
d
with
d
ec
ep
tiv
e
r
ev
iews,
in
clu
d
in
g
s
en
tim
en
t
p
o
lar
ity
an
d
b
ag
-
of
-
wo
r
d
s
ap
p
r
o
ac
h
es.
Me
an
wh
ile,
J
in
d
a
l
an
d
L
iu
[
9
]
d
ev
elo
p
e
d
m
o
d
e
ls
to
an
aly
s
e
o
p
in
io
n
s
p
am
b
a
s
ed
o
n
r
u
le
-
b
ased
s
y
s
tem
s
an
d
r
ev
iewe
r
b
eh
av
io
u
r
.
W
h
ile
th
ese
ea
r
ly
r
esear
ch
s
tu
d
ies
d
id
p
r
o
v
id
e
s
o
m
e
v
alu
ab
le
in
s
ig
h
ts
in
t
o
id
en
tify
in
g
d
ec
e
p
tiv
e
r
ev
iew
s
,
th
ey
wer
e
lim
ited
in
th
ei
r
ab
ilit
y
to
d
etec
t
co
o
r
d
in
at
ed
b
eh
av
io
u
r
an
d
b
eh
av
io
u
r
-
d
r
iv
en
f
r
a
u
d
b
ec
a
u
s
e
th
ey
r
elied
h
ea
v
ily
o
n
tex
t
u
al
-
b
ased
f
ea
tu
r
es.
As
r
esear
ch
er
s
lear
n
ed
m
o
r
e
ab
o
u
t
h
o
w
co
n
s
u
m
er
s
in
ter
ac
ted
with
an
d
wr
o
te
r
ev
iews,
th
ey
b
e
g
an
u
s
in
g
alter
n
ativ
e
m
eth
o
d
s
to
b
etter
u
n
d
er
s
tan
d
t
h
e
b
eh
a
v
io
u
r
s
a
n
d
r
elatio
n
s
h
ip
s
b
etwe
en
c
o
n
s
u
m
er
s
.
Fo
r
ex
am
p
le,
Mu
k
h
er
jee
et
a
l.
[
1
0
]
co
n
d
u
cte
d
a
s
tu
d
y
o
n
h
o
w
co
n
s
u
m
er
s
b
eh
a
v
e
an
d
d
ev
el
o
p
r
elatio
n
s
h
ip
s
o
n
r
ev
iew
s
ites
lik
e
y
elp
to
b
etter
u
n
d
er
s
tan
d
h
o
w
r
ev
iew
s
ites
f
ilter
o
u
t
f
r
au
d
u
len
t
r
ev
iews.
T
h
ey
f
o
u
n
d
th
at
a
b
etter
u
n
d
er
s
tan
d
in
g
o
f
th
o
s
e
b
eh
av
io
u
r
s
an
d
r
elatio
n
s
h
ip
s
c
o
u
ld
im
p
r
o
v
e
th
e
ac
c
u
r
ac
y
o
f
d
etec
tin
g
f
r
a
u
d
u
le
n
t
r
ev
iews.
T
o
d
ay
,
r
esear
ch
er
s
ar
e
in
cr
ea
s
in
g
l
y
u
s
in
g
n
etwo
r
k
-
b
ased
a
p
p
r
o
ac
h
es
to
id
e
n
tif
y
f
r
a
u
d
u
le
n
t
ac
tiv
ity
.
Fo
r
ex
a
m
p
le,
He
et
a
l.
[
1
1
]
d
ev
elo
p
e
d
a
m
o
d
el
f
o
r
id
e
n
tif
y
in
g
p
r
o
d
u
cts
p
u
r
ch
ased
with
f
r
au
d
u
len
t
r
ev
iews
o
n
Am
az
o
n
u
s
in
g
n
etwo
r
k
-
lev
el
s
tr
u
ctu
r
al
attr
ib
u
tes s
u
ch
as d
eg
r
ee
ce
n
tr
ality
,
Pag
eRan
k
,
an
d
clu
s
ter
in
g
co
ef
f
icien
ts
.
T
h
is
m
eth
o
d
wo
r
k
s
v
er
y
well
b
u
t
is
lim
ited
to
id
e
n
tify
in
g
th
e
f
r
au
d
u
len
t
ac
ti
v
ities
o
f
p
r
o
d
u
cts
o
n
Am
az
o
n
s
in
ce
it
r
eq
u
ir
es
d
ata
f
r
o
m
th
at
p
latf
o
r
m
.
Natu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
h
as
b
ee
n
m
ak
in
g
g
r
ea
t
s
tr
id
es
d
u
e
to
r
ap
id
ad
v
an
ce
m
e
n
ts
in
d
ee
p
lear
n
in
g
an
d
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
.
T
r
an
s
f
o
r
m
er
-
b
ased
ar
ch
i
tectu
r
es
h
av
e
b
ee
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Mu
ltimo
d
a
l m
a
ch
in
e
lea
r
n
i
n
g
fr
a
mewo
r
k
fo
r
fa
ke
r
ev
ie
w
d
etec
tio
n
(
R
a
s
h
mi
R
.
)
993
u
s
ed
b
y
Salm
in
en
et
a
l.
[
1
2
]
t
o
class
if
y
d
ec
ep
tiv
e
r
ev
iews.
Similar
ly
,
C
h
en
et
a
l.
[
1
3
]
u
s
e
d
b
id
ir
ec
tio
n
al
-
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
B
iLST
M
)
m
o
d
els
with
atten
tio
n
m
ec
h
an
is
m
s
to
ca
p
tu
r
e
s
em
an
tic
d
is
p
er
s
io
n
b
etwe
en
d
if
f
er
en
t
r
ev
iew
tex
ts
.
Sev
er
al
s
tu
d
ies
h
av
e
v
alid
ated
t
h
e
u
s
e
o
f
DeBERTa
[
1
4
]
an
d
R
o
B
E
R
T
a
[
1
5
]
as
ef
f
ec
tiv
e
m
o
d
els
f
o
r
d
etec
tin
g
f
ak
e
r
ev
iews.
Fu
r
th
er
,
Su
et
a
l.
[
1
6
]
wo
r
k
ed
o
n
m
eth
o
d
s
to
id
en
tify
AI
-
g
en
er
ated
(
u
s
es
p
r
etr
ain
ed
lan
g
u
ag
e
m
o
d
els
s
u
ch
as
B
E
R
T
an
d
DeBERTa)
.
Alth
o
u
g
h
th
e
s
e
ty
p
es
o
f
m
o
d
els
h
av
e
b
ee
n
f
o
u
n
d
to
ac
h
iev
e
h
i
g
h
lev
els
o
f
ac
cu
r
ac
y
,
th
ey
r
e
m
ain
tex
t
-
b
ased
,
h
ea
v
ily
co
m
p
u
tatio
n
al,
an
d
lac
k
in
ter
p
r
etab
ilit
y
as th
ey
g
en
er
al
ly
d
o
n
o
t c
ap
tu
r
e
co
o
r
d
in
ated
h
u
m
an
f
r
au
d
an
d
r
ev
iewe
r
c
o
llu
s
io
n
.
T
h
e
u
s
e
o
f
en
s
em
b
le
an
d
m
u
lti
-
f
ea
tu
r
e
lear
n
in
g
ap
p
r
o
ac
h
es
h
as
em
er
g
ed
as
a
way
to
im
p
r
o
v
e
r
o
b
u
s
tn
ess
.
Fo
r
ex
am
p
le,
C
ao
et
a
l.
[
1
7
]
p
r
o
p
o
s
ed
a
d
ec
ep
tiv
e
r
ev
iew
d
etec
tio
n
s
y
s
tem
wh
ich
s
ep
ar
ated
m
u
lti
-
f
ea
tu
r
e
lear
n
in
g
an
d
c
lass
if
icatio
n
p
h
ases
to
cr
ea
t
e
an
ef
f
ec
tiv
e
d
ec
ep
tio
n
d
e
tectio
n
m
o
d
el
b
y
im
p
r
o
v
in
g
g
en
e
r
aliza
tio
n
p
e
r
f
o
r
m
a
n
ce
.
Mo
h
awe
s
h
et
a
l
.
[
1
8
]
p
r
esen
ted
an
ex
p
lain
ab
le
en
s
em
b
le
o
f
m
u
ltiv
iew
DL
m
o
d
els
wh
ich
co
m
b
in
ed
tex
tu
al
an
d
b
eh
a
v
io
u
r
al
f
ea
tu
r
es
o
f
u
s
er
s
to
en
h
an
ce
tr
an
s
p
ar
en
c
y
wh
ile
p
r
o
v
id
in
g
a
m
eth
o
d
to
im
p
r
o
v
e
e
x
p
lain
ab
ilit
y
.
T
h
e
u
s
e
o
f
s
tack
in
g
-
b
ased
e
n
s
em
b
le
f
r
am
ewo
r
k
s
[
1
]
h
av
e
s
h
o
wn
im
p
r
o
v
em
en
ts
in
m
o
d
el
p
er
f
o
r
m
an
ce
th
r
o
u
g
h
th
e
co
m
b
in
atio
n
o
f
m
u
ltip
le
class
if
ier
s
;
h
o
wev
er
,
th
ese
ap
p
r
o
ac
h
es
o
f
ten
em
p
h
a
s
ized
th
e
ar
ch
itectu
r
e
o
f
m
o
d
els
r
ath
er
th
an
p
r
o
v
i
d
in
g
th
o
r
o
u
g
h
in
te
g
r
atio
n
o
f
f
ea
tu
r
es.
I
n
ad
d
itio
n
to
tex
t
-
b
ased
an
d
en
s
em
b
le
-
b
ased
ap
p
r
o
ac
h
es,
a
lter
n
ativ
e
tech
n
iq
u
es
ar
e
b
ein
g
ex
p
lo
r
e
d
f
o
r
id
en
tify
i
n
g
m
is
lead
in
g
o
n
lin
e
p
r
o
d
u
ct
r
ev
iews.
Fo
r
ex
am
p
le,
Sh
ah
ar
iar
et
a
l.
[
1
9
]
h
av
e
cr
ea
ted
a
b
en
ch
m
ar
k
d
ataset
o
f
f
ictitio
u
s
B
en
g
ali
r
ev
iews
th
at
ca
n
b
e
u
s
ed
to
ev
alu
ate
lan
g
u
ag
e
-
s
p
ec
if
ic
d
etec
tio
n
s
y
s
tem
s
d
u
e
to
th
e
ab
s
en
ce
o
f
m
u
ltil
in
g
u
al
d
atasets
.
B
ir
im
et
a
l.
[
2
0
]
u
s
ed
to
p
ic
m
o
d
els
to
id
en
tify
th
e
u
n
iq
u
e
q
u
ality
o
f
d
ec
ep
ti
v
e
r
ev
iew
p
atter
n
s
b
y
id
en
tify
in
g
u
n
u
s
u
al
o
r
u
n
ex
p
ec
ted
d
is
tr
ib
u
tio
n
s
a
m
o
n
g
th
e
to
p
ics
o
f
th
e
r
ev
iews,
wh
ile
Yao
et
a
l.
[
2
1
]
d
e
v
elo
p
e
d
a
g
r
ap
h
-
b
ase
d
ap
p
r
o
ac
h
to
p
er
f
o
r
m
i
n
g
co
n
tr
asti
v
e
lear
n
in
g
to
m
o
d
el
h
ie
r
ar
ch
ical
c
o
n
tr
ac
tio
n
s
o
f
r
e
v
iewe
r
b
e
h
av
io
u
r
.
O
n
a
b
r
o
a
d
er
s
ca
le,
W
alth
er
et
a
l.
[
2
2
]
ex
p
lo
r
ed
h
o
w
p
eo
p
le
u
s
e
d
if
f
er
en
t
f
ac
to
r
s
to
d
eter
m
in
e
wh
eth
er
o
r
n
o
t
r
ev
iews
ar
e
a
u
th
en
tic,
wh
ile
J
ab
eu
r
et
a
l.
[
2
3
]
s
u
m
m
ar
is
ed
th
e
m
ajo
r
r
esear
ch
tr
en
d
s
in
th
is
ar
ea
an
d
o
u
t
lin
ed
p
o
s
s
ib
le
f
u
tu
r
e
d
ir
ec
tio
n
s
f
o
r
in
v
esti
g
atio
n
th
r
o
u
g
h
b
ib
lio
m
etr
ic
an
aly
s
is
.
W
h
ile
m
an
y
s
tu
d
ies
h
av
e
p
r
o
d
u
ce
d
h
ig
h
le
v
els
o
f
p
e
r
f
o
r
m
an
ce
m
o
d
ellin
g
f
ak
e
r
ev
iews
with
tex
t
-
b
ased
m
eth
o
d
s
,
d
ee
p
lear
n
i
n
g
tech
n
iq
u
es,
a
n
d
g
r
ap
h
/co
m
m
u
n
ity
-
b
ased
a
p
p
r
o
ac
h
es,
th
er
e
ar
e
s
till
s
ev
er
al
ch
allen
g
es
ass
o
ciate
d
with
th
ese
ap
p
r
o
ac
h
es,
in
clu
d
in
g
th
eir
ab
ilit
y
to
b
e
in
ter
p
r
etab
le
(
i.e
.
,
ea
s
ily
u
n
d
er
s
to
o
d
)
,
th
eir
ca
p
ac
ity
f
o
r
s
ca
lin
g
(
i.e
.
,
g
r
o
win
g
in
s
ize
an
d
n
u
m
b
e
r
)
,
an
d
th
eir
v
u
ln
er
ab
ilit
y
to
b
ein
g
ex
p
lo
ited
f
o
r
co
o
r
d
in
ated
f
r
a
u
d
u
len
t
ac
tiv
ities
.
T
o
s
o
lv
e
th
e
ch
allen
g
es
s
tated
ab
o
v
e,
t
h
is
wo
r
k
p
r
o
p
o
s
es
a
u
n
if
ied
s
o
lu
tio
n
th
at
in
teg
r
ates
f
o
u
r
d
if
f
er
e
n
t
ty
p
es
o
f
in
f
o
r
m
atio
n
th
at
m
ig
h
t
s
ig
n
al
th
e
p
r
esen
ce
o
f
a
f
ak
e
r
ev
iew
(
e.
g
.
,
tex
tu
al,
b
eh
a
v
io
u
r
al
,
tem
p
o
r
al,
an
d
n
etwo
r
k
-
b
ased
i
n
d
icato
r
s
)
u
s
in
g
a
s
tack
in
g
en
s
em
b
l
e
m
o
d
ellin
g
ar
ch
itectu
r
e.
Ad
d
i
tio
n
ally
,
b
y
a
p
p
ly
in
g
Sh
ap
le
y
ad
d
itiv
e
ex
p
lan
atio
n
s
(
SH
AP)
,
we
p
r
o
v
id
e
a
m
ea
n
s
o
f
im
p
r
o
v
in
g
m
o
d
el
in
ter
p
r
etab
ilit
y
f
o
r
d
ec
is
io
n
-
m
ak
er
s
.
B
y
u
s
in
g
a
co
m
b
in
ed
m
u
ltimo
d
al
an
d
ex
p
lain
ab
le
f
r
am
ewo
r
k
to
d
et
ec
t
f
ak
e
r
ev
iews,
o
u
r
s
o
lu
tio
n
ca
n
b
e
d
ep
lo
y
e
d
in
d
ep
en
d
en
tly
ac
r
o
s
s
m
u
ltip
le
e
-
co
m
m
er
ce
p
latf
o
r
m
s
to
p
r
o
t
ec
t c
o
n
s
u
m
er
s
f
r
o
m
b
ein
g
m
is
led
b
y
f
r
au
d
u
len
t o
r
d
ec
ep
tiv
e
r
ev
iews.
3.
M
E
T
H
O
DO
L
O
G
Y
3
.
1
.
P
re
-
pro
ce
s
s
ing
T
h
e
r
aw
r
ev
iew
d
ataset
u
n
d
er
g
o
es
s
tan
d
ar
d
p
r
e
p
r
o
ce
s
s
in
g
s
tep
s
to
en
s
u
r
e
d
ata
q
u
ality
an
d
co
n
s
is
ten
cy
.
T
ex
tu
al
co
n
ten
t
is
clea
n
ed
b
y
r
em
o
v
in
g
p
u
n
ctu
atio
n
,
s
to
p
wo
r
d
s
,
an
d
s
p
ec
ial
ch
ar
ac
ter
s
,
f
o
llo
wed
b
y
to
k
e
n
is
atio
n
an
d
n
o
r
m
alis
atio
n
.
Miss
in
g
v
alu
es
in
n
o
n
-
tex
tu
al
attr
ib
u
tes
ar
e
h
an
d
led
ap
p
r
o
p
r
iately
,
an
d
ca
teg
o
r
ical
v
ar
iab
les ar
e
en
co
d
ed
wh
e
r
e
r
eq
u
ir
ed
.
3
.
1
.
1
.
Da
t
a
ba
la
ncing
Fak
e
r
ev
iew
d
atasets
ar
e
n
at
u
r
ally
im
b
alan
ce
d
b
ec
a
u
s
e
g
en
u
in
e
r
ev
iews
s
ig
n
if
ican
tly
o
u
tn
u
m
b
er
f
r
au
d
u
le
n
t
o
n
es.
T
o
ad
d
r
ess
th
is
is
s
u
e,
we
u
s
ed
th
e
s
y
n
th
etic
m
in
o
r
ity
o
v
er
-
s
am
p
lin
g
te
ch
n
iq
u
e
(
SMOT
E
)
,
wh
ich
g
en
e
r
ates
n
ew,
s
y
n
th
e
tic
ex
am
p
les
f
o
r
th
e
m
in
o
r
it
y
class
r
ath
er
th
an
s
im
p
ly
d
u
p
licatin
g
ex
is
tin
g
s
am
p
les.
SMOT
E
wo
r
k
s
b
y
i
d
en
tify
in
g
th
e
n
ea
r
est
n
eig
h
b
o
u
r
s
o
f
m
in
o
r
ity
-
class
s
am
p
les
an
d
cr
ea
tin
g
n
ew
p
o
in
ts
alo
n
g
th
e
lin
e
th
at
co
n
n
ec
ts
th
em
.
T
h
is
h
elp
s
ex
p
an
d
th
e
d
ec
is
io
n
b
o
u
n
d
ar
y
f
o
r
t
h
e
m
in
o
r
ity
(
f
ak
e)
class
an
d
r
ed
u
ce
s
m
is
class
if
icatio
n
.
As
a
r
esu
lt,
th
e
clas
s
if
ier
b
ec
o
m
es
m
o
r
e
s
en
s
itiv
e
an
d
ac
cu
r
ate
in
id
en
tify
in
g
f
ak
e
r
e
v
iews.
3
.
1
.
2
.
F
ea
t
ure
eng
ineering
Ou
r
m
eth
o
d
o
l
o
g
y
in
teg
r
ates
a
b
r
o
ad
s
et
o
f
f
ea
tu
r
es
an
d
m
u
lti
p
le
m
ac
h
in
e
lear
n
in
g
m
o
d
els
to
ca
p
tu
r
e
th
e
v
ar
io
u
s
p
atter
n
s
p
r
esen
t
i
n
b
o
th
g
en
u
i
n
e
an
d
f
ak
e
r
ev
iew
s
.
T
h
e
ap
p
r
o
ac
h
co
m
b
i
n
es
tem
p
o
r
al,
b
eh
av
io
u
r
al,
n
etwo
r
k
,
tex
tu
al,
an
d
s
y
n
th
eti
c
s
ig
n
als,
all
e
x
tr
ac
ted
o
r
en
g
i
n
ee
r
ed
f
r
o
m
th
e
d
ataset
f
ield
s
:
r
ev
iew_
id
,
u
s
er
_
id
,
b
u
s
in
ess
_
id
,
r
atin
g
,
r
ev
iew_
te
x
t,
d
ate,
an
d
f
lag
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
9
1
-
1
0
0
1
994
a.
T
em
p
o
r
al
f
ea
t
u
r
es
Usi
n
g
th
e
r
ev
iew
tim
estam
p
,
we
g
en
er
ated
tim
e
-
b
ased
b
e
h
a
v
io
u
r
al
p
atter
n
s
to
u
n
d
er
s
tan
d
wh
en
th
e
r
ev
iew
was
p
o
s
ted
.
T
h
e
f
o
llo
win
g
f
ea
tu
r
es
wer
e
ex
tr
ac
ted
:
Yea
r
,
Mo
n
th
,
Day
,
W
ee
k
d
a
y
,
Ho
u
r
o
f
r
ev
iew
p
o
s
tin
g
,
I
s
_
wee
k
en
d
(
Satu
r
d
ay
/Su
n
d
ay
)
,
I
s
_
n
ig
h
t
(
p
o
s
ted
b
etwe
en
1
0
p
m
an
d
6
am
)
,
I
s
_
b
u
s
in
ess
_
h
o
u
r
s
(
p
o
s
ted
d
u
r
in
g
9
am
–
5
p
m
)
T
h
ese
f
ea
tu
r
es
ar
e
u
s
ef
u
l
b
e
ca
u
s
e
f
r
au
d
u
len
t
ac
tiv
ity
o
f
t
en
clu
s
ter
s
ar
o
u
n
d
u
n
u
s
u
al
h
o
u
r
s
o
r
s
p
ec
if
ic
d
ay
s
,
in
d
icatin
g
a
b
n
o
r
m
al
p
o
s
tin
g
b
eh
av
io
u
r
.
b.
B
eh
av
io
u
r
al
f
ea
tu
r
es
T
o
c
ap
tu
r
e
r
e
v
i
ewe
r
b
eh
a
v
i
o
u
r
,
we
c
o
m
p
u
t
ed
s
e
v
e
r
al
u
s
e
r
-
le
v
e
l
s
t
atis
tics
:
u
s
e
r
_
t
o
t
al_
r
ev
iews
,
u
s
e
r
_
a
v
g
_
r
a
ti
n
g
,
u
s
er
_
r
at
in
g
_
s
td
,
u
s
e
r
_
m
in
_
r
at
in
g
,
u
s
er
_
m
a
x
_
r
ati
n
g
,
r
e
v
i
ews
_
p
er
_
d
a
y
,
an
d
r
at
in
g
_
c
o
n
s
is
t
en
c
y
(
h
o
w
o
f
te
n
a
u
s
e
r
g
i
v
es
th
e
s
am
e
r
a
ti
n
g
)
.
T
h
ese
f
ea
t
u
r
es
s
u
g
g
est
wh
eth
er
th
e
u
s
er
b
e
h
av
es
lik
e
a
ty
p
ical
r
ev
iewe
r
o
r
s
h
o
ws s
ig
n
s
o
f
au
t
o
m
ated
,
r
e
p
etitiv
e,
o
r
s
u
s
p
icio
u
s
r
atin
g
b
e
h
av
io
u
r
.
c.
Netwo
r
k
f
ea
tu
r
es
W
e
co
n
s
tr
u
cted
a
u
s
er
–
b
u
s
in
ess
in
ter
ac
tio
n
g
r
ap
h
u
s
in
g
Netwo
r
k
X
to
u
n
d
er
s
tan
d
r
elatio
n
s
h
ip
s
an
d
d
etec
t p
atter
n
s
o
f
c
o
llu
s
io
n
.
Fr
o
m
th
is
g
r
ap
h
,
we
ex
tr
ac
te
d
:
−
u
s
er
_
n
etwo
r
k
_
s
ize
—
n
u
m
b
e
r
o
f
co
n
n
ec
ted
u
s
er
s
−
u
s
er
_
n
etwo
r
k
_
d
en
s
ity
—
co
n
n
ec
tio
n
s
p
er
r
ev
iew
−
co
-
r
ev
iewe
r
c
o
u
n
t
—
u
s
er
s
r
e
v
iewin
g
th
e
s
am
e
b
u
s
in
ess
es
−
d
eg
r
ee
ce
n
t
r
ality
—
h
o
w
in
f
l
u
en
tial th
e
u
s
er
is
in
th
e
n
etwo
r
k
T
h
ese
n
etwo
r
k
s
ig
n
als
h
elp
r
ev
ea
l
s
u
s
p
icio
u
s
clu
s
ter
s
o
f
r
ev
iewe
r
s
p
o
s
tin
g
to
g
eth
er
,
wh
ich
is
co
m
m
o
n
in
c
o
o
r
d
i
n
ated
f
ak
e
r
ev
iew
ca
m
p
aig
n
s
.
Fig
u
r
e
1
illu
s
tr
ates
th
e
r
ev
iewe
r
–
b
u
s
in
ess
in
ter
ac
tio
n
n
etwo
r
k
,
h
ig
h
lig
h
tin
g
in
f
lu
e
n
tial
r
ev
iewe
r
s
an
d
d
e
n
s
ely
co
n
n
ec
ted
clu
s
ter
s
th
at
m
ay
in
d
ic
ate
co
o
r
d
in
ated
f
ak
e
r
ev
iew
b
eh
av
i
o
u
r
.
Fig
u
r
e
1
.
Netwo
r
k
an
aly
s
is
o
f
f
ea
tu
r
es
d.
T
ex
tu
al
f
ea
tu
r
es a
n
d
tex
t e
m
b
ed
d
in
g
T
o
cr
ea
te
r
e
p
r
esen
tatio
n
s
o
f
r
e
v
iew
d
ata,
we
an
aly
s
ed
r
ev
ie
w
len
g
th
,
wo
r
d
co
u
n
t,
a
n
d
s
en
tim
en
t
(
v
ia
a
B
E
R
T
b
ased
s
en
tim
en
t
s
co
r
in
g
m
o
d
el
)
,
as
well
as
T
F
-
I
DF
em
b
ed
d
in
g
v
alu
es
(
lim
ite
d
to
a
m
ax
im
u
m
o
f
5
0
0
0
f
ea
tu
r
es).
T
h
e
T
F
-
I
DF
m
eth
o
d
allo
ws
f
o
r
th
e
id
e
n
tific
atio
n
o
f
k
ey
ter
m
s
in
d
o
c
u
m
en
ts
b
y
g
iv
in
g
lo
w
weig
h
ts
to
co
m
m
o
n
ly
u
s
ed
w
o
r
d
s
,
wh
ile
th
e
s
en
tim
en
t
s
co
r
e,
wh
ich
in
d
icate
s
th
e
em
o
tio
n
al
v
alen
ce
/q
u
ality
o
f
th
e
r
e
v
iew'
s
co
n
ten
t,
h
ig
h
l
ig
h
ted
in
c
o
n
s
is
ten
cies
with
in
th
e
d
ata
if
t
h
er
e
a
r
e
v
er
y
n
eg
ativ
e
r
ev
iews
with
5
-
s
tar
r
atin
g
s
,
t
h
er
eb
y
allo
win
g
p
o
te
n
tial
id
en
tific
atio
n
o
f
d
e
ce
p
tiv
e
r
ev
iews.
T
F
-
I
DF
was
ap
p
lied
ag
ai
n
s
t
th
e
tex
t
o
f
all
r
ev
iews,
p
r
o
d
u
cin
g
n
u
m
er
ic
v
alu
es
th
at
wer
e
u
s
ed
in
co
m
b
in
atio
n
with
b
eh
a
v
io
u
r
al
an
d
n
etwo
r
k
f
ea
tu
r
es
to
en
ab
le
m
o
d
el
in
p
u
t
o
f
b
o
th
u
n
s
tr
u
ctu
r
ed
(
r
e
v
iew
tex
t)
an
d
s
tr
u
ctu
r
ed
(
n
u
m
e
r
ically
r
ep
r
esen
ted
)
d
ata.
Fig
u
r
e
2
s
h
o
ws
th
e
f
ea
tu
r
e
im
p
o
r
tan
ce
d
er
i
v
ed
f
r
o
m
th
e
Dec
is
io
n
T
r
ee
m
o
d
el,
in
d
i
ca
tin
g
th
at
n
etwo
r
k
ce
n
tr
ality
an
d
r
ev
iewe
r
ac
tiv
it
y
m
etr
ics p
lay
a
d
o
m
in
a
n
t r
o
le
in
f
ak
e
r
ev
iew
d
etec
tio
n
.
e.
Sy
n
th
etic
b
eh
av
i
o
u
r
al
f
ea
t
u
r
es a
n
d
d
ata
p
r
o
ce
s
s
in
g
T
h
e
d
ataset
was
en
h
an
ce
d
with
s
y
n
th
etic
b
eh
a
v
io
u
r
al
p
r
o
x
y
f
ea
tu
r
es
d
er
iv
ed
ex
clu
s
iv
ely
f
r
o
m
p
u
b
licly
av
ailab
le
r
e
v
iew
m
etad
ata.
T
h
ese
p
r
o
x
ies
ap
p
r
o
x
i
m
ate
r
ea
l
-
wo
r
ld
r
ev
iewe
r
ac
ti
v
ity
p
atter
n
s
—
s
u
ch
as
in
f
er
r
ed
ac
tiv
ity
r
eg
u
lar
it
y
,
in
ter
ac
tio
n
co
n
s
is
ten
cy
,
r
e
v
iewin
g
s
p
ee
d
,
an
d
ac
co
u
n
t
lo
n
g
ev
ity
—
with
o
u
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Mu
ltimo
d
a
l m
a
ch
in
e
lea
r
n
i
n
g
fr
a
mewo
r
k
fo
r
fa
ke
r
ev
ie
w
d
etec
tio
n
(
R
a
s
h
mi
R
.
)
995
r
ely
in
g
o
n
an
y
p
latf
o
r
m
-
in
ter
n
al
lo
g
s
o
r
p
r
i
v
ate
u
s
er
in
f
o
r
m
atio
n
.
T
h
e
o
r
ig
in
al
f
lag
attr
i
b
u
te
was
co
n
v
er
ted
in
to
a
b
in
ar
y
lab
el
(
0
:
g
e
n
u
in
e
r
ev
iew,
1
:
f
ak
e
r
ev
iew)
.
I
n
a
d
d
itio
n
,
a
co
m
p
o
s
ite
s
u
s
p
icio
u
s
n
ess
s
co
r
e,
ter
m
ed
f
ak
e_
s
co
r
e,
was
co
m
p
u
ted
t
o
ca
p
tu
r
e
a
b
n
o
r
m
al
b
e
h
av
io
u
r
al
p
atter
n
s
in
clu
d
in
g
u
n
u
s
u
ally
h
ig
h
r
ev
iew
f
r
eq
u
e
n
cy
,
r
atin
g
h
o
m
o
g
en
eit
y
,
r
ap
i
d
r
e
v
iew
p
o
s
tin
g
,
an
d
p
o
s
tin
g
d
u
r
in
g
aty
p
ical
tim
e
p
er
io
d
s
.
Miss
in
g
v
alu
es
wer
e
h
an
d
led
u
s
in
g
li
n
ea
r
in
ter
p
o
latio
n
,
n
u
m
e
r
ical
f
ea
tu
r
es
wer
e
s
tan
d
ar
d
ized
u
s
in
g
Stan
d
ar
d
Scaler
,
an
d
th
e
d
ataset
was
s
p
lit
i
n
to
8
0
%
tr
ain
in
g
a
n
d
2
0
%
test
in
g
s
ets.
A
ll
f
ea
tu
r
e
en
g
in
ee
r
in
g
s
tep
s
ar
e
d
eter
m
in
is
tic,
r
ep
r
o
d
u
cib
le,
a
n
d
d
o
cu
m
en
ted
i
n
th
e
p
u
b
lic
Gi
tHu
b
r
ep
o
s
ito
r
y
.
Fig
u
r
e
2
.
Featu
r
e
im
p
o
r
ta
n
ce
f
o
r
th
e
d
ec
is
io
n
f.
Fak
e
r
ev
iew
d
etec
tio
n
m
o
d
els
A
f
r
am
ewo
r
k
th
at
u
s
es
v
ar
io
u
s
m
eth
o
d
s
f
o
r
c
r
ea
tin
g
a
m
ac
h
in
e
lear
n
in
g
en
v
ir
o
n
m
e
n
t
to
id
en
tify
p
o
s
s
ib
le
f
r
au
d
u
len
t
o
r
m
is
lead
in
g
r
e
v
iews
is
p
r
esen
ted
in
t
h
is
p
ap
er
.
W
ith
lo
g
is
tic
r
eg
r
e
s
s
io
n
,
d
ec
is
io
n
tr
ee
,
XGBo
o
s
t,
an
d
s
tack
in
g
en
s
em
b
les,
th
ese
m
o
d
els
ar
e
tr
ain
ed
u
s
in
g
s
tr
atif
ied
s
am
p
lin
g
to
m
ain
tain
th
e
b
alan
ce
o
f
class
es.
L
o
g
is
tic
r
eg
r
ess
io
n
ac
ts
as
a
b
asel
in
e
m
o
d
el
to
s
ee
if
lin
ea
r
s
ep
ar
ab
ilit
y
ex
is
ts
with
th
e
tr
ain
ed
d
ata
an
d
to
b
e
a
b
le
to
in
ter
p
r
et
th
e
ef
f
ec
t
o
f
f
ea
t
u
r
es
o
n
th
e
tr
a
in
ed
d
ata.
XGBo
o
s
t
u
tili
ze
s
c
o
m
p
lex
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
with
in
th
e
s
tr
u
c
tu
r
ed
f
ea
t
u
r
es
o
f
th
is
d
ata
s
et
,
wh
ich
allo
ws
it
to
p
r
o
v
id
e
s
u
p
er
io
r
p
r
ed
ictiv
e
ca
p
ab
ilit
ies
o
v
er
all
o
th
er
m
o
d
els.
Stack
in
g
en
s
em
b
les
co
m
b
in
e
lo
g
is
tic
r
eg
r
ess
io
n
an
d
X
GB
o
o
s
t
b
y
way
o
f
lo
g
is
tic
r
eg
r
ess
io
n
b
ein
g
th
e
m
eta
-
class
if
ier
,
allo
win
g
f
o
r
im
p
r
o
v
e
d
r
o
b
u
s
tn
ess
an
d
d
ec
r
ea
s
ed
class
if
icatio
n
er
r
o
r
.
T
h
e
d
ec
is
io
n
tr
ee
m
o
d
el
ad
d
s
to
th
e
o
v
e
r
all
in
ter
p
r
eta
b
ilit
y
o
f
th
is
m
o
d
el
b
y
allo
win
g
u
s
er
s
to
id
en
tify
th
e
m
o
s
t
im
p
o
r
tan
t
f
ea
tu
r
es
in
f
lu
en
cin
g
th
e
class
if
icatio
n
o
f
a
f
r
a
u
d
u
le
n
t
r
ev
iew.
Fig
u
r
e
3
p
r
esen
ts
th
e
Dec
is
io
n
T
r
ee
s
tr
u
ct
u
r
e,
d
em
o
n
s
tr
atin
g
h
o
w
k
ey
f
ea
tu
r
es
s
u
ch
as
r
e
v
iewe
r
c
o
u
n
t,
d
e
g
r
ee
ce
n
tr
ality
,
an
d
ac
co
u
n
t a
g
e
in
f
lu
en
c
e
class
if
icatio
n
d
ec
is
io
n
s
.
3
.
2
.
O
v
er
v
iew
o
f
t
he
pro
po
s
ed
f
ra
m
ewo
r
k
A
m
ac
h
in
e
lear
n
in
g
b
ased
ap
p
r
o
ac
h
was
d
e
v
elo
p
ed
a
n
d
te
s
ted
to
d
etec
t
f
r
au
d
/
f
alse
r
ev
iews
o
n
th
e
in
ter
n
et
.
T
h
e
g
o
al
o
f
th
is
s
y
s
tem
was
to
m
ak
e
co
n
s
u
m
er
s
m
o
r
e
co
m
f
o
r
tab
le
wh
en
u
tili
s
in
g
o
n
lin
e
r
ev
iew
s
ites
.
T
h
e
m
eth
o
d
o
l
o
g
y
u
s
ed
t
o
d
ev
el
o
p
th
e
m
ac
h
in
e
lear
n
i
n
g
b
ased
m
o
d
el
in
v
o
l
v
ed
e
x
ten
s
iv
e
an
d
c
o
m
p
lex
d
ata
p
r
ep
ar
atio
n
.
Dif
f
er
e
n
t
ty
p
es
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els
wer
e
cr
ea
ted
to
an
aly
s
e
o
n
lin
e
r
ev
iew
d
ata
;
th
ese
m
o
d
els
in
clu
d
e
d
lo
g
is
ti
c
r
eg
r
ess
io
n
,
XGBo
o
s
t,
an
d
s
tack
in
g
en
s
em
b
lin
g
.
T
h
e
co
m
b
in
atio
n
o
f
th
ese
m
u
ltip
le
m
o
d
els
h
as
y
ield
ed
s
tr
o
n
g
r
esu
lts
as
th
ey
ca
p
t
u
r
e
m
u
ltip
le
d
im
en
s
io
n
s
o
f
f
r
au
d
u
le
n
t
wr
itin
g
b
eh
av
io
u
r
s
an
d
tr
en
d
s
.
Ad
d
it
io
n
ally
,
p
er
f
o
r
m
an
ce
m
ea
s
u
r
es,
s
tr
en
g
th
s
an
d
lim
itatio
n
s
o
f
ea
ch
m
o
d
el
ar
e
d
is
cu
s
s
ed
in
th
e
f
o
llo
win
g
s
u
b
s
ec
tio
n
s
;
it
i
s
s
tr
ess
ed
th
at
r
o
b
u
s
t
ap
p
r
o
ac
h
es
ar
e
cr
itical
to
p
r
ev
en
tin
g
th
e
m
an
ip
u
latio
n
o
f
r
ev
iews,
as
w
ell
as
to
b
etter
u
n
d
e
r
s
tan
d
th
e
f
u
n
d
am
e
n
tal
asp
ec
ts
o
f
o
n
lin
e
d
ec
ep
tio
n
t
o
cr
ea
te
m
o
r
e
r
eliab
le
e
-
co
m
m
er
c
e
en
v
ir
o
n
m
e
n
ts
.
Fig
u
r
e
4
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
en
d
-
to
-
en
d
f
r
am
ewo
r
k
f
o
r
f
a
k
e
r
ev
iew
d
etec
tio
n
,
illu
s
tr
atin
g
th
e
p
r
e
p
r
o
ce
s
s
in
g
p
ip
elin
e,
f
e
atu
r
e
ex
tr
ac
tio
n
s
tag
es,
b
ase
class
if
ier
s
,
s
tack
in
g
en
s
em
b
le,
an
d
e
v
alu
atio
n
m
etr
ics
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
9
1
-
1
0
0
1
996
Fig
u
r
e
3
.
Dec
is
io
n
tr
ee
Fig
u
r
e
4
.
Pro
p
o
s
ed
f
r
am
ewo
r
k
f
o
r
f
a
k
e
r
e
v
iew
d
etec
tio
n
4.
RE
SU
L
T
S
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
ex
p
er
im
en
tal
ev
alu
atio
n
o
f
th
e
p
r
o
p
o
s
ed
m
u
ltimo
d
al
f
a
k
e
r
ev
iew
d
etec
tio
n
f
r
am
ewo
r
k
.
T
h
e
m
o
d
els
ar
e
a
s
s
es
s
ed
u
s
in
g
ac
cu
r
ac
y
,
p
r
ec
i
s
io
n
,
r
ec
all,
a
n
d
F1
-
s
co
r
e
t
o
p
r
o
v
id
e
a
b
alan
ce
d
ev
alu
atio
n
u
n
d
er
class
im
b
ala
n
ce
co
n
d
itio
n
s
as p
r
o
v
i
d
ed
in
(
1
)
,
(
2
)
,
(
3
)
a
n
d
(
4
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Mu
ltimo
d
a
l m
a
ch
in
e
lea
r
n
i
n
g
fr
a
mewo
r
k
fo
r
fa
ke
r
ev
ie
w
d
etec
tio
n
(
R
a
s
h
mi
R
.
)
997
4
.
1
.
P
er
f
o
r
m
a
nce
ev
a
lua
t
io
n
T
o
ass
ess
th
e
ef
f
ec
tiv
en
ess
o
f
o
u
r
f
ak
e
r
ev
iew
d
etec
tio
n
s
y
s
tem
,
s
ev
er
al
s
tan
d
ar
d
cla
s
s
if
icatio
n
m
etr
ics
wer
e
u
s
ed
.
T
h
ese
in
c
lu
d
e
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
a
ll,
an
d
th
e
F1
-
s
co
r
e.
Per
f
o
r
m
an
ce
is
b
ased
o
n
a
b
in
ar
y
class
if
icatio
n
s
ettin
g
,
wh
er
e
r
ev
iews
wer
e
lab
elled
as
eith
er
“
T
r
u
e
”
o
r
“
Fak
e
”
.
T
ab
le
1
p
r
esen
ts
th
e
co
n
f
u
s
io
n
m
atr
ix
o
f
th
e
b
est
-
p
er
f
o
r
m
in
g
m
o
d
el
T
ab
le
1
.
C
o
n
f
u
s
io
n
m
atr
ix
o
f
t
h
e
b
est
-
p
er
f
o
r
m
in
g
m
o
d
el
A
c
t
u
a
l
/
P
r
e
d
i
c
t
e
d
Tr
u
t
h
f
u
l
F
a
k
e
Tr
u
t
h
f
u
l
Tr
u
e
P
o
s
i
t
i
v
e
(
TP)
F
a
l
se
N
e
g
a
t
i
v
e
(FN)
F
a
k
e
F
a
l
se
P
o
s
i
t
i
v
e
(
F
P
)
Tr
u
e
N
e
g
a
t
i
v
e
(
TN
)
T
h
e
co
n
f
u
s
io
n
m
atr
ix
in
d
icate
s
a
r
ed
u
ce
d
n
u
m
b
er
o
f
f
alse
n
eg
ativ
es,
wh
ich
is
cr
itical
in
f
ak
e
r
e
v
iew
d
etec
tio
n
s
ce
n
ar
io
s
wh
er
e
u
n
d
etec
ted
f
r
au
d
u
len
t
r
ev
iews
m
ay
d
ir
ec
tly
af
f
ec
t
co
n
s
u
m
er
tr
u
s
t.
T
h
e
b
alan
ce
d
d
is
tr
ib
u
tio
n
o
f
f
alse
p
o
s
itiv
es
an
d
f
alse
n
eg
ativ
es
d
em
o
n
s
tr
a
tes
th
at
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
ac
h
iev
es
r
eliab
le
p
er
f
o
r
m
an
ce
s
u
itab
le
f
o
r
r
ea
l
-
wo
r
ld
d
ep
l
o
y
m
en
t.
T
h
e
e
v
alu
a
tio
n
m
etr
ics ar
e
co
m
p
u
ted
as
(
1
)
,
(
2
)
,
(
3
)
,
(
4
)
:
=
+
+
+
+
(
1
)
=
+
(
2
)
=
+
(
3
)
F1
Score
=
2
x
Pr
ecis
i
o
n
x
Recal
l
Pr
ecis
i
o
n
+
Recal
l
(
4
)
T
h
ese
m
etr
ics
co
llectiv
ely
p
r
o
v
id
e
a
co
m
p
r
e
h
en
s
iv
e
ass
es
s
m
en
t
o
f
m
o
d
el
ef
f
ec
tiv
en
ess
,
p
ar
ticu
lar
ly
u
n
d
e
r
im
b
alan
ce
d
class
d
is
tr
ib
u
tio
n
s
.
4
.
2
.
M
o
del
-
wis
e
perf
o
rm
a
nc
e
co
m
pa
riso
n
T
ab
le
2
p
r
esen
ts
th
e
co
m
p
ar
a
tiv
e
p
er
f
o
r
m
a
n
ce
o
f
d
if
f
e
r
en
t
m
ac
h
in
e
lear
n
in
g
m
o
d
els
tr
ain
ed
u
s
in
g
m
u
ltimo
d
al
f
ea
tu
r
es.
L
o
g
is
tic
r
eg
r
ess
io
n
s
h
o
ws
m
o
d
er
ate
p
e
r
f
o
r
m
a
n
ce
d
u
e
to
its
lin
ea
r
d
e
cisi
o
n
b
o
u
n
d
ar
ies,
wh
ich
lim
it
its
ab
ilit
y
to
m
o
d
el
co
m
p
lex
in
ter
ac
tio
n
s
a
m
o
n
g
te
x
tu
al,
b
eh
av
i
o
u
r
al,
tem
p
o
r
al,
a
n
d
n
etwo
r
k
-
b
ased
f
ea
tu
r
es.
Dec
is
io
n
T
r
ee
s
ac
h
iev
e
th
e
h
ig
h
est
p
r
e
cisi
o
n
(
0
.
9
8
)
,
in
d
icatin
g
s
tr
o
n
g
c
o
n
f
id
e
n
ce
in
id
en
tify
in
g
f
ak
e
r
ev
iews;
h
o
wev
er
,
th
eir
l
o
wer
r
ec
all
s
u
g
g
ests
th
at
a
co
n
s
id
er
ab
le
n
u
m
b
er
o
f
f
r
a
u
d
u
le
n
t
r
ev
iews
r
em
ain
u
n
d
etec
ted
.
I
n
co
n
tr
ast,
XGBo
o
s
t
an
d
th
e
s
tack
in
g
en
s
em
b
le
ac
h
ie
v
e
t
h
e
m
o
s
t
b
alan
ce
d
p
er
f
o
r
m
an
ce
ac
r
o
s
s
all
m
etr
ics.
T
h
eir
ab
ilit
y
to
m
o
d
el
n
o
n
-
lin
ea
r
r
elatio
n
s
h
ip
s
an
d
lev
er
ag
e
f
ea
tu
r
e
in
ter
ac
tio
n
s
en
ab
les
im
p
r
o
v
ed
d
etec
tio
n
o
f
c
o
o
r
d
i
n
ated
a
n
d
s
u
b
tle
d
ec
e
p
tiv
e
b
e
h
av
io
u
r
s
.
T
h
e
en
s
em
b
le
m
o
d
el
f
u
r
th
er
e
n
h
an
ce
s
r
o
b
u
s
tn
ess
b
y
co
m
b
i
n
in
g
th
e
s
tr
en
g
th
s
o
f
in
d
iv
id
u
al
class
if
ier
s
.
T
ab
le
2
.
C
o
m
p
a
r
ativ
e
p
er
f
o
r
m
an
ce
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els u
s
in
g
m
u
ltimo
d
al
f
ea
tu
r
es
M
o
d
e
l
P
r
e
c
i
s
i
o
n
R
e
c
a
l
l
F1
-
sc
o
r
e
A
c
c
u
r
a
c
y
Lo
g
i
s
t
i
c
r
e
g
r
e
ssi
o
n
0
.
6
4
0
.
9
5
0
.
7
7
0
.
8
6
X
G
B
o
o
st
0
.
8
3
0
.
9
3
0
.
8
7
0
.
9
4
S
t
a
c
k
i
n
g
e
n
s
e
mb
l
e
0
.
8
1
0
.
9
3
0
.
8
7
0
.
9
3
D
e
c
i
s
i
o
n
t
r
e
e
c
l
a
ss
i
f
i
e
r
0
.
9
8
0
.
8
1
0
.
8
8
0
.
9
5
Ov
er
all,
th
ese
r
esu
lts
h
i
g
h
lig
h
t
th
e
a
d
v
an
ta
g
e
o
f
m
u
ltimo
d
al
f
ea
tu
r
e
in
teg
r
atio
n
in
f
a
k
e
r
ev
iew
d
etec
tio
n
.
B
y
co
m
b
in
in
g
tex
t
u
al,
b
eh
av
io
u
r
al,
tem
p
o
r
al,
an
d
n
etwo
r
k
-
b
ased
s
ig
n
als,
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
ca
p
tu
r
es
b
o
th
co
n
te
n
t
-
lev
el
d
e
ce
p
tio
n
an
d
co
o
r
d
in
ate
d
r
ev
ie
wer
ac
tiv
ity
th
at
s
in
g
le
-
m
o
d
ality
ap
p
r
o
ac
h
es
m
ay
o
v
er
lo
o
k
.
W
h
en
co
m
p
ar
e
d
with
r
ec
en
t
tex
t
-
ce
n
tr
ic
d
ee
p
lear
n
in
g
m
o
d
els
r
ep
o
r
ted
in
t
h
e
liter
atu
r
e,
wh
ich
ty
p
ically
ac
h
iev
e
F1
-
s
co
r
es
in
th
e
r
an
g
e
o
f
0
.
8
5
–
0
.
9
0
,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
d
em
o
n
s
tr
ates
co
m
p
etitiv
e
p
er
f
o
r
m
an
ce
(
F1
≈
0
.
8
7
)
wh
il
e
r
eq
u
ir
in
g
s
ig
n
if
ica
n
tly
lo
wer
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
an
d
o
f
f
er
in
g
e
n
h
an
ce
d
in
ter
p
r
etab
ilit
y
.
Fu
r
th
er
m
o
r
e,
SHAP
-
b
ased
an
aly
s
is
co
n
f
ir
m
s
th
at
n
etwo
r
k
c
o
n
n
ec
t
iv
ity
,
co
-
r
ev
iewe
r
in
ter
ac
tio
n
s
,
an
d
b
e
h
av
io
u
r
al
co
n
s
is
ten
cy
p
lay
a
cr
itical
r
o
le
in
im
p
r
o
v
in
g
d
etec
tio
n
ef
f
ec
tiv
en
ess
,
r
ein
f
o
r
cin
g
th
e
im
p
o
r
tan
ce
o
f
m
u
ltimo
d
al
lear
n
in
g
f
o
r
p
r
ac
tical
d
ep
lo
y
m
en
t.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
9
1
-
1
0
0
1
998
4
.
3
.
Co
m
pa
riso
n wit
h
ex
is
t
ing
a
nd
s
t
a
t
e
-
of
-
t
he
-
a
rt
m
et
h
o
ds
R
esear
ch
o
n
ea
r
ly
d
etec
tio
n
o
f
f
ak
e
r
ev
iews
b
eg
an
b
y
lo
o
k
in
g
at
tex
tu
al
c
h
ar
ac
ter
is
tics
an
d
u
s
in
g
m
an
u
ally
cr
ea
te
d
lan
g
u
ag
e
-
b
a
s
ed
cu
es.
Fo
r
ex
a
m
p
le,
Ott
et
a
l.
[
8
]
d
em
o
n
s
tr
ated
th
at
d
ec
e
p
tiv
e
r
ev
iews
h
av
e
d
is
tin
ct
ch
ar
ac
ter
is
tics
th
r
o
u
g
h
s
en
tim
en
t
p
o
lar
ity
an
d
b
ag
-
of
-
wo
r
d
s
ch
ar
ac
ter
is
tics
,
wh
il
e
J
in
d
al
an
d
L
iu
[
9
]
r
esear
ch
ed
o
p
in
io
n
s
p
a
m
u
s
in
g
r
u
le
-
b
ased
tech
n
i
q
u
es
to
m
o
d
el
r
ev
iewe
r
b
eh
a
v
io
u
r
.
W
h
ile
th
ese
s
tu
d
ies
p
r
o
v
id
e
d
im
p
o
r
ta
n
t
f
o
u
n
d
atio
n
al
k
n
o
wled
g
e,
b
o
th
o
f
th
ese
tech
n
iq
u
es
h
ad
th
eir
lim
itatio
n
s
r
eg
ar
d
in
g
th
eir
ab
ilit
y
to
id
en
tify
co
o
r
d
in
ated
r
ev
iewe
r
b
eh
a
v
io
u
r
s
a
n
d
d
etec
t tim
e
an
o
m
alies.
I
n
o
r
d
er
to
e
x
p
an
d
u
p
o
n
th
e
f
i
n
d
in
g
s
o
f
th
ese
s
tu
d
ies,
Mu
k
h
er
jee
et
a
l.
[
1
0
]
p
r
o
p
o
s
ed
two
n
ew
f
o
r
m
s
o
f
an
al
y
tical
f
r
am
ew
o
r
k
s
f
o
r
d
etec
tin
g
f
r
au
d
u
len
t
r
ev
iews
b
ased
o
n
b
eh
a
v
io
u
r
al
an
d
r
elatio
n
al
m
eth
o
d
s
.
Mu
k
h
er
jee
et
a
l.
p
r
o
v
id
ed
em
p
ir
ical
ev
id
en
ce
s
h
o
win
g
h
o
w
s
ites
s
u
ch
as
Yelp
f
ilter
o
u
t
f
r
au
d
u
le
n
t
r
ev
iews
u
s
in
g
th
e
in
te
r
ac
tio
n
s
o
f
r
ev
i
ewe
r
s
o
r
n
etwo
r
k
s
o
f
r
ev
iew
au
th
o
r
s
,
t
h
u
s
lead
in
g
to
b
etter
d
etec
tio
n
r
esu
lts
;
h
o
wev
er
,
t
h
e
m
eth
o
d
s
d
is
cu
s
s
ed
in
th
is
s
tu
d
y
ar
e
s
till
p
latf
o
r
m
-
s
p
ec
if
ic
an
d
d
o
es
n
o
t
ea
s
il
y
ad
ap
t
to
e
m
er
g
in
g
s
p
am
m
in
g
s
tr
ateg
ies.
Ov
er
th
e
p
ast
f
ew
y
ea
r
s
,
t
h
er
e
h
as
b
e
en
an
in
c
r
ea
s
ed
f
o
cu
s
o
n
u
tili
s
in
g
d
ee
p
lear
n
in
g
an
d
tr
an
s
f
o
r
m
er
-
b
ased
ar
ch
ite
ctu
r
es
in
v
ar
io
u
s
f
ield
s
.
Fo
r
in
s
tan
ce
,
L
i
et
a
l.
[
2
4
]
u
s
ed
B
E
R
T
to
cr
ea
te
a
co
n
tex
tu
al
s
em
an
tic
r
e
p
r
esen
t
atio
n
o
f
r
ev
iews
f
r
o
m
w
h
ich
th
ey
co
u
ld
s
u
cc
ess
f
u
lly
class
if
y
th
e
s
ev
er
ity
o
f
d
ec
eiv
in
g
r
e
v
iews.
L
ik
ewise,
R
en
an
d
J
i
[
2
5
]
h
av
e
u
tili
s
ed
co
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs
)
an
d
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
-
ty
p
e
NN
m
o
d
els
to
p
er
f
o
r
m
d
ec
ep
tiv
e
o
p
in
io
n
s
p
am
d
etec
ti
o
n
.
Mo
r
e
r
ec
en
tly
,
Z
h
an
g
et
a
l.
[
2
6
]
i
n
co
r
p
o
r
at
ed
th
e
u
tili
s
atio
n
o
f
tr
an
s
f
o
r
m
er
-
b
ased
m
o
d
els
in
to
th
eir
m
eth
o
d
o
l
o
g
y
f
o
r
d
etec
tin
g
f
ak
e
r
ev
iews b
y
p
r
o
v
id
in
g
d
ee
p
er
s
em
an
tic
r
e
p
r
es
en
tatio
n
s
o
f
th
e
r
ev
iews b
ein
g
ev
alu
ated
.
Desp
ite
th
e
im
p
r
o
v
ed
p
r
e
d
ictiv
e
ca
p
ab
ilit
y
p
r
o
v
id
e
d
b
y
th
ese
ty
p
es
o
f
d
ee
p
lear
n
in
g
tech
n
iq
u
es,
th
e
m
o
d
els
ar
e
p
r
ed
o
m
in
an
tly
b
a
s
ed
o
n
te
x
t
d
ata,
r
eq
u
i
r
e
s
u
b
s
tan
tial
am
o
u
n
ts
o
f
co
m
p
u
te
an
d
ar
e
d
if
f
ic
u
lt
to
in
ter
p
r
et.
I
n
c
o
n
tr
ast,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
in
clu
d
es a
m
u
lt
im
o
d
al
m
o
d
el
th
at
in
teg
r
ates te
x
tu
al,
b
eh
av
i
o
u
r
al,
ch
r
o
n
o
lo
g
ical,
an
d
n
etwo
r
k
/c
o
m
p
u
ter
is
ed
f
ea
t
u
r
e
s
ets
to
p
r
o
d
u
ce
co
m
p
etitiv
e
r
esu
lts
th
r
o
u
g
h
im
p
r
o
v
e
d
r
o
b
u
s
tn
ess
an
d
in
ter
p
r
etab
ilit
y
o
f
th
e
m
u
ltimo
d
al
m
o
d
el.
4
.
4
.
P
r
a
ct
ica
l
deplo
y
m
ent
i
m
pli
ca
t
io
ns
Fro
m
a
d
ep
lo
y
m
e
n
t
p
er
s
p
ec
t
iv
e,
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
o
f
f
er
s
s
ig
n
if
ican
t
ad
v
an
ta
g
es
o
v
er
d
ee
p
lear
n
in
g
-
b
ased
m
eth
o
d
s
.
T
r
a
n
s
f
o
r
m
er
m
o
d
els
ty
p
ically
r
eq
u
ir
e
s
u
b
s
tan
tial
co
m
p
u
tatio
n
a
l
r
eso
u
r
ce
s
,
lar
g
e
-
s
ca
le
d
ata,
an
d
f
r
eq
u
en
t
r
etr
a
in
in
g
.
I
n
co
n
tr
ast,
th
e
p
r
esen
t
ed
f
r
a
m
ewo
r
k
is
lig
h
tweig
h
t,
in
ter
p
r
eta
b
le,
a
n
d
ad
ap
tab
le,
m
ak
i
n
g
it
s
u
itab
le
f
o
r
r
ea
l
-
tim
e
d
e
p
lo
y
m
e
n
t
in
lar
g
e
-
s
ca
le
r
ev
iew
p
latf
o
r
m
s
.
T
h
e
ex
p
lain
ab
le
n
atu
r
e
o
f
th
e
p
r
ed
ictio
n
s
f
u
r
th
er
s
u
p
p
o
r
ts
tr
u
s
t,
ac
co
u
n
ta
b
ilit
y
,
an
d
r
eg
u
lato
r
y
co
m
p
lian
ce
in
au
to
m
ate
d
r
ev
iew
m
o
d
e
r
atio
n
s
y
s
tem
s
.
4
.
4
.
1
.
M
o
del
ex
pla
ina
bil
it
y
(
SH
AP
a
na
ly
s
is
)
Sh
ap
ley
ad
d
itiv
e
ex
p
la
n
atio
n
s
(
SHAP)
wer
e
u
s
ed
to
in
t
er
p
r
et
m
o
d
el
p
r
ed
ictio
n
s
.
T
h
e
an
aly
s
is
r
ev
ea
led
th
at
th
e
f
o
llo
win
g
f
ea
tu
r
es
h
a
d
th
e
g
r
ea
test
i
n
f
lu
en
ce
:
d
eg
r
ee
_
ce
n
tr
ality
,
n
u
m
_
c
o
_
r
ev
iewe
r
s
,
ac
co
u
n
t_
a
g
e_
d
a
y
s
,
r
atin
g
,
a
n
d
s
en
tim
en
t_
s
co
r
e.
SHAP
p
lo
ts
s
h
o
wed
th
at
h
ig
h
er
r
e
v
ie
wer
co
n
n
ec
tiv
ity
,
u
n
u
s
u
al
r
ev
iewin
g
p
atter
n
s
,
an
d
m
is
m
atch
es
b
etwe
en
s
e
n
tim
en
t
an
d
r
ati
n
g
s
ig
n
i
f
ican
tly
in
cr
ea
s
ed
th
e
lik
elih
o
o
d
o
f
a
r
e
v
iew
b
ei
n
g
class
if
ied
as
f
ak
e.
Fig
u
r
e
5
illu
s
tr
ates
th
e
SHAP
-
b
ased
ex
p
lan
atio
n
f
o
r
an
in
d
iv
id
u
al
p
r
ed
ictio
n
,
s
h
o
win
g
h
o
w
k
ey
f
ea
tu
r
es
s
u
ch
as
th
e
n
u
m
b
er
o
f
co
-
r
ev
iewe
r
s
,
d
e
g
r
ee
ce
n
tr
ality
,
an
d
r
ev
iew
len
g
th
co
n
tr
i
b
u
te
p
o
s
itiv
ely
o
r
n
e
g
ativ
ely
to
th
e
m
o
d
el’
s
d
ec
is
io
n
.
I
m
p
o
r
tan
ce
:
E
x
p
lain
ab
ilit
y
in
cr
ea
s
ed
tr
u
s
t
in
th
e
s
y
s
tem
a
n
d
en
s
u
r
ed
th
at
th
e
m
o
d
els
we
r
e
lear
n
in
g
m
ea
n
i
n
g
f
u
l
p
atter
n
s
r
ath
er
th
a
n
n
o
is
e
o
r
b
ias.
Fig
u
r
e
5
.
Featu
r
e
co
n
tr
ib
u
tio
n
f
o
r
th
e
r
ev
iew
5.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
p
r
o
p
o
s
es
a
m
u
ltimo
d
al
m
ac
h
in
e
lear
n
in
g
f
r
a
m
ewo
r
k
f
o
r
f
a
k
e
r
ev
iew
d
et
ec
tio
n
b
y
in
teg
r
atin
g
b
eh
av
io
u
r
al,
tem
p
o
r
al,
tex
tu
al,
n
etwo
r
k
,
an
d
s
y
n
th
etic
f
ea
tu
r
es.
E
x
p
er
im
en
ta
l
ev
alu
atio
n
s
h
o
ws
th
at
en
s
em
b
le
m
o
d
els,
p
ar
tic
u
lar
ly
XGBo
o
s
t
an
d
th
e
s
tac
k
in
g
en
s
em
b
le,
ac
h
iev
e
c
o
n
s
is
ten
t
an
d
b
alan
ce
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Mu
ltimo
d
a
l m
a
ch
in
e
lea
r
n
i
n
g
fr
a
mewo
r
k
fo
r
fa
ke
r
ev
ie
w
d
etec
tio
n
(
R
a
s
h
mi
R
.
)
999
p
er
f
o
r
m
an
ce
ac
r
o
s
s
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e,
wh
ile
S
HAP
-
b
ased
an
aly
s
is
im
p
r
o
v
es
in
ter
p
r
etab
ilit
y
b
y
h
ig
h
lig
h
tin
g
k
e
y
f
ea
tu
r
e
co
n
t
r
ib
u
tio
n
s
.
B
ey
o
n
d
s
tr
o
n
g
p
r
e
d
i
ctiv
e
p
er
f
o
r
m
a
n
ce
,
t
h
e
f
r
am
e
wo
r
k
h
as
p
r
ac
tical
d
ep
lo
y
m
e
n
t
p
o
ten
tial
f
o
r
r
ea
l
-
wo
r
ld
e
-
co
m
m
er
ce
a
n
d
h
o
s
p
it
ality
p
latf
o
r
m
s
,
wh
e
r
e
it
ca
n
s
u
p
p
o
r
t
lar
g
e
-
s
ca
le
r
ev
iew
m
o
d
er
atio
n
,
r
ed
u
ce
f
r
a
u
d
u
len
t
ac
tiv
ity
,
an
d
en
h
an
ce
co
n
s
u
m
er
tr
u
s
t.
Ho
wev
er
,
th
e
cu
r
r
en
t
ap
p
r
o
ac
h
is
lim
ited
to
E
n
g
lis
h
-
lan
g
u
a
g
e
d
atasets
an
d
r
em
ain
s
s
en
s
itiv
e
to
ev
o
lv
in
g
d
ec
ep
tiv
e
p
r
ac
tices.
Fu
tu
r
e
wo
r
k
s
h
o
u
ld
f
o
c
u
s
o
n
e
x
ten
d
in
g
m
u
ltil
in
g
u
al
ca
p
ab
ilit
ies,
in
c
o
r
p
o
r
ati
n
g
ad
a
p
tiv
e
lear
n
in
g
to
ad
d
r
ess
em
er
g
in
g
m
an
ip
u
latio
n
s
tr
ateg
ies,
an
d
v
alid
atin
g
th
e
f
r
a
m
ewo
r
k
th
r
o
u
g
h
in
teg
r
atio
n
in
t
o
o
p
er
ati
o
n
al
p
latf
o
r
m
s
to
im
p
r
o
v
e
t
h
e
r
eliab
ilit
y
o
f
u
s
er
-
g
en
er
ated
c
o
n
ten
t a
t scale
.
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
R
ash
m
i R
.
✓
✓
✓
✓
✓
✓
✓
✓
✓
Sh
o
b
h
a
T
.
✓
✓
✓
✓
✓
J
ee
v
ita
S
.
Dev
ad
ig
✓
✓
✓
✓
✓
✓
✓
✓
Gay
atr
i S
.
San
ti
✓
✓
✓
✓
Dh
an
u
s
h
r
ee
C
.
S
.
✓
✓
✓
✓
✓
Har
s
h
ita
L
.
V
.
✓
✓
✓
✓
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
ata
th
at
s
u
p
p
o
r
t th
e
f
in
d
i
n
g
s
o
f
th
is
s
tu
d
y
a
r
e
o
p
en
ly
a
v
ailab
le
in
b
elo
w
r
ep
o
s
ito
r
ies:
1
.
Hu
g
g
i
n
g
Face
Data
s
et:
h
ttp
s
:
//h
u
g
g
in
g
f
a
ce
.
co
/d
a
ta
s
ets/
d
eb
o
jit0
1
/fa
ke
-
r
ev
iew
-
d
a
ta
s
et
2
.
Yelp
Data
s
et:
GitHu
b
-
h
ttp
s
:
//g
ith
u
b
.
co
m/ch
ir
a
g
d
a
r
ya
n
i/
fa
ke
-
r
ev
iew
-
d
etec
tio
n
3
.
Fin
al
Data
s
et:
GitHu
b
-
h
ttp
s
:
//g
ith
u
b
.
co
m/Dh
a
n
u
0
7
4
6
/M
LG
-
Da
ta
s
et.
g
it
RE
F
E
R
E
NC
E
S
[
1
]
S
.
A
.
A
s
h
r
a
f
,
A
.
F
.
Ja
v
e
d
,
S
.
B
e
l
l
a
r
y
,
P
.
K
.
B
a
l
a
,
a
n
d
P
.
K
.
P
a
n
i
g
r
a
h
i
,
“
L
e
v
e
r
a
g
i
n
g
st
a
c
k
i
n
g
f
r
a
m
e
w
o
r
k
f
o
r
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
i
n
t
h
e
h
o
s
p
i
t
a
l
i
t
y
s
e
c
t
o
r
,
”
J
o
u
r
n
a
l
o
f
T
h
e
o
re
t
i
c
a
l
a
n
d
Ap
p
l
i
e
d
E
l
e
c
t
r
o
n
i
c
C
o
m
m
e
rc
e
Re
s
e
a
r
c
h
,
v
o
l
.
1
9
,
n
o
.
2
,
p
.
1
5
1
7
,
J
u
n
.
2
0
2
4
,
d
o
i
:
1
0
.
3
3
9
0
/
j
t
a
e
r
1
9
0
2
0
0
7
5
.
[
2
]
W
.
H
.
A
sa
a
d
,
R
.
A
l
l
a
mi
,
a
n
d
Y
.
H
.
A
l
i
,
“
F
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
u
s
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
Re
v
u
e
d
’
i
n
t
e
l
l
i
g
e
n
c
e
a
r
t
i
f
i
c
i
e
l
l
e
,
v
o
l
.
3
7
,
n
o
.
5
,
O
c
t
.
2
0
2
3
,
d
o
i
:
1
0
.
1
8
2
8
0
/
r
i
a
.
3
7
0
5
0
7
.
[
3
]
M
.
L
e
e
,
Y
.
S
o
n
g
,
L.
L
i
,
K
.
Le
e
,
a
n
d
S
.
-
B
.
Y
a
n
g
,
“
D
e
t
e
c
t
i
n
g
f
a
k
e
r
e
v
i
e
w
s
w
i
t
h
su
p
e
r
v
i
s
e
d
ma
c
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms
,
”
S
e
r
v
i
c
e
I
n
d
u
st
r
i
e
s
J
o
u
r
n
a
l
,
v
o
l
.
4
2
,
n
o
.
1
3
–
1
4
,
p
p
.
1
1
0
1
–
1
1
2
1
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
8
0
/
0
2
6
4
2
0
6
9
.
2
0
2
2
.
2
0
5
4
9
9
6
.
[
4
]
A
.
H
.
A
l
s
h
e
h
r
i
,
“
A
n
o
n
l
i
n
e
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
a
p
p
r
o
a
c
h
u
si
n
g
f
a
mo
u
s
mac
h
i
n
e
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
ms,
”
C
o
m
p
u
t
e
rs,
M
a
t
e
ri
a
l
s
&
C
o
n
t
i
n
u
a
,
v
o
l
.
7
8
,
n
o
.
2
,
p
.
2
7
6
7
,
Ja
n
.
2
0
2
4
,
d
o
i
:
1
0
.
3
2
6
0
4
/
c
mc
.
2
0
2
3
.
0
4
6
8
3
8
.
[
5
]
Z.
W
a
n
g
,
A
.
Y
a
o
,
G
.
X
u
,
a
n
d
M
.
R
e
n
,
“
A
l
a
r
g
e
l
a
n
g
u
a
g
e
mo
d
e
l
-
b
a
s
e
d
a
p
p
r
o
a
c
h
f
o
r
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
:
t
h
e
i
m
p
l
i
c
i
t
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
p
e
r
s
p
e
c
t
i
v
e
,
”
I
n
f
o
rm
a
t
i
o
n
Pro
c
e
ss
i
n
g
&
M
a
n
a
g
e
m
e
n
t
,
v
o
l
.
6
3
,
n
o
.
1
,
p
.
1
0
4
3
5
2
,
A
u
g
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
p
m.
2
0
2
5
.
1
0
4
3
5
2
.
[
6
]
J.
S
a
l
m
i
n
e
n
,
C
.
K
a
n
d
p
a
l
,
A
.
M
.
K
a
mel
,
S
.
J
u
n
g
,
a
n
d
B
.
J.
Ja
n
se
n
,
“
C
r
e
a
t
i
n
g
a
n
d
d
e
t
e
c
t
i
n
g
f
a
k
e
r
e
v
i
e
w
s
o
f
o
n
l
i
n
e
p
r
o
d
u
c
t
s,”
J
o
u
rn
a
l
o
f
Re
t
a
i
l
i
n
g
a
n
d
C
o
n
su
m
e
r
S
e
rvi
c
e
s
,
v
o
l
.
6
4
,
p
.
1
0
2
7
7
1
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
r
e
t
c
o
n
ser.
2
0
2
1
.
1
0
2
7
7
1
.
[
7
]
R
.
G
u
p
t
a
,
V
.
Ji
n
d
a
l
,
a
n
d
I
.
K
a
sh
y
a
p
,
“
R
e
c
e
n
t
st
a
t
e
-
of
-
t
h
e
-
a
r
t
o
f
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
:
a
c
o
mp
r
e
h
e
n
s
i
v
e
r
e
v
i
e
w
,
”
K
n
o
w
l
e
d
g
e
En
g
i
n
e
e
ri
n
g
Re
v
i
e
w
,
v
o
l
.
3
9
,
p
.
e
8
,
N
o
v
.
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
7
/
S
0
2
6
9
8
8
8
9
2
4
0
0
0
0
6
7
.
[
8
]
M
.
O
t
t
,
Y
.
C
h
o
i
,
C
.
C
a
r
d
i
e
,
a
n
d
J.
T.
H
a
n
c
o
c
k
,
“
F
i
n
d
i
n
g
d
e
c
e
p
t
i
v
e
o
p
i
n
i
o
n
sp
a
m
b
y
a
n
y
s
t
r
e
t
c
h
o
f
t
h
e
i
ma
g
i
n
a
t
i
o
n
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
4
9
t
h
An
n
u
a
l
Me
e
t
i
n
g
o
f
t
h
e
Asso
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
st
i
c
s
(
AC
L
)
,
2
0
1
1
,
p
p
.
3
0
9
–
319.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
1
6
,
No
.
2
,
Ap
r
il
20
2
6
:
9
9
1
-
1
0
0
1
1000
[
9
]
N
.
Ji
n
d
a
l
a
n
d
B
.
Li
u
,
“
O
p
i
n
i
o
n
s
p
a
m
a
n
d
a
n
a
l
y
s
i
s,
”
i
n
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
1
st
AC
M
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
re
n
c
e
o
n
We
b
S
e
a
r
c
h
a
n
d
D
a
t
a
M
i
n
i
n
g
(
W
S
D
M)
,
2
0
0
8
,
p
p
.
2
1
9
–
2
3
0
,
d
o
i
:
1
0
.
1
1
4
5
/
1
3
4
1
5
3
1
.
1
3
4
1
5
6
0
.
[
1
0
]
A
.
M
u
k
h
e
r
j
e
e
,
V
.
V
e
n
k
a
t
a
r
a
ma
n
,
B
.
Li
u
,
a
n
d
N
.
S
.
G
l
a
n
c
e
,
“
W
h
a
t
Y
e
l
p
f
a
k
e
r
e
v
i
e
w
f
i
l
t
e
r
mi
g
h
t
b
e
d
o
i
n
g
?
,
”
i
n
Pro
c
e
e
d
i
n
g
s o
f
t
h
e
7
t
h
I
n
t
e
rn
a
t
i
o
n
a
l
AAAI
C
o
n
f
e
r
e
n
c
e
o
n
We
b
l
o
g
s
a
n
d
S
o
c
i
a
l
Me
d
i
a
(
I
C
WS
M)
,
2
0
1
3
,
p
p
.
4
0
9
–
4
1
8
.
[
1
1
]
S
.
H
e
,
B
.
H
o
l
l
e
n
b
e
c
k
,
G
.
O
v
e
r
g
o
o
r
,
D
.
P
r
o
s
e
r
p
i
o
,
a
n
d
A
.
T
o
s
y
a
l
i
,
“
D
e
t
e
c
t
i
n
g
f
a
k
e
-
r
e
v
i
e
w
b
u
y
e
r
s
u
s
i
n
g
n
e
t
w
o
r
k
s
t
r
u
c
t
u
r
e
:
d
i
r
e
c
t
e
v
i
d
e
n
c
e
f
r
o
m
A
maz
o
n
,
”
Pr
o
c
e
e
d
i
n
g
s
o
f
t
h
e
N
a
t
i
o
n
a
l
A
c
a
d
e
m
y
o
f
S
c
i
e
n
c
e
s
,
v
o
l
.
1
1
9
,
n
o
.
4
7
,
p
.
e
2
2
1
1
9
3
2
1
1
9
,
2
0
2
2
,
d
o
i
:
1
0
.
1
0
7
3
/
p
n
a
s.
2
2
1
1
9
3
2
1
1
9
.
[
1
2
]
J.
S
a
l
m
i
n
e
n
,
M
.
M
u
s
t
a
k
,
S
.
J
u
n
g
,
H
.
M
a
k
k
o
n
e
n
,
a
n
d
B
.
J.
J
a
n
s
e
n
,
“
D
e
c
o
d
i
n
g
d
e
c
e
p
t
i
o
n
i
n
t
h
e
o
n
l
i
n
e
m
a
r
k
e
t
p
l
a
c
e
:
e
n
h
a
n
c
i
n
g
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
w
i
t
h
p
s
y
c
h
o
l
i
n
g
u
i
st
i
c
s
a
n
d
t
r
a
n
sf
o
r
mer
m
o
d
e
l
s,”
J
o
u
r
n
a
l
o
f
M
a
rk
e
t
i
n
g
A
n
a
l
y
t
i
c
s
,
M
a
r
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
5
7
/
s
4
1
2
7
0
-
0
2
5
-
0
0
3
9
3
-
8.
[
1
3
]
J.
C
h
e
n
,
T
.
Zh
a
n
g
,
Z
.
Y
a
n
,
Z
.
Zh
e
n
g
,
W
.
Z
h
a
n
g
,
a
n
d
J.
Z
h
a
n
g
,
“
A
t
t
e
n
t
i
o
n
-
b
a
se
d
B
i
LSTM
w
i
t
h
p
o
s
i
t
i
o
n
a
l
e
m
b
e
d
d
i
n
g
s
f
o
r
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
,
”
J
o
u
r
n
a
l
of
B
i
g
D
a
t
a
,
v
o
l
.
1
2
,
n
o
.
1
,
A
p
r
.
2
0
2
5
,
d
o
i
:
1
0
.
1
1
8
6
/
s4
0
5
3
7
-
025
-
0
1
1
3
0
-
9.
[
1
4
]
S
.
G
e
e
t
h
a
,
E.
E
l
a
k
i
y
a
,
R
.
S
.
K
a
n
m
a
n
i
,
a
n
d
M
.
D
a
s,
“
H
i
g
h
p
e
r
f
o
r
ma
n
c
e
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
u
s
i
n
g
p
r
e
t
r
a
i
n
e
d
D
e
B
ER
T
a
o
p
t
i
m
i
z
e
d
w
i
t
h
M
o
n
a
r
c
h
B
u
t
t
e
r
f
l
y
p
a
r
a
d
i
g
m
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
rt
s
,
v
o
l
.
1
5
,
n
o
.
1
,
p
.
7
4
4
5
,
2
0
2
5
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
8
-
0
2
5
-
8
9
4
5
3
-
8.
[
1
5
]
L.
Zh
a
n
g
,
M
.
L
e
e
,
S
.
P
a
t
e
l
,
a
n
d
J
.
S
e
o
,
“
A
st
u
d
y
o
n
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
b
a
s
e
d
o
n
R
o
B
E
R
Ta
,
”
Pr
o
c
e
d
i
a
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
v
o
l
.
2
4
2
,
p
p
.
1
3
2
3
–
1
3
3
0
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
p
r
o
c
s
.
2
0
2
4
.
0
8
.
1
3
1
.
[
1
6
]
Z.
S
u
,
M
.
Y
a
n
g
,
Q
.
Z
h
a
i
,
K
.
G
u
o
,
Y
.
H
u
a
n
g
,
a
n
d
Y
.
C
o
n
g
,
“
A
mu
l
t
i
g
r
a
i
n
e
d
p
r
e
f
e
r
e
n
c
e
a
n
a
l
y
s
i
s
me
t
h
o
d
f
o
r
p
r
o
d
u
c
t
i
t
e
r
a
t
i
v
e
d
e
s
i
g
n
i
n
c
o
r
p
o
r
a
t
i
n
g
A
I
-
g
e
n
e
r
a
t
e
d
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
,
”
S
c
i
e
n
t
i
f
i
c
Re
p
o
r
t
s
,
v
o
l
.
1
5
,
n
o
.
1
,
p
.
2
5
2
8
,
Ja
n
.
2
0
2
5
,
d
o
i
:
1
0
.
1
0
3
8
/
s
4
1
5
9
8
-
0
2
5
-
8
6
5
5
1
-
5.
[
1
7
]
N
.
C
a
o
,
S
.
Ji
,
D
.
K
.
W
.
C
h
i
u
,
a
n
d
M
.
G
o
n
g
,
“
A
d
e
c
e
p
t
i
v
e
r
e
v
i
e
w
s
d
e
t
e
c
t
i
o
n
mo
d
e
l
:
se
p
a
r
a
t
e
d
t
r
a
i
n
i
n
g
o
f
m
u
l
t
i
-
f
e
a
t
u
r
e
l
e
a
r
n
i
n
g
a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
,
”
Ex
p
e
r
t
S
y
s
t
e
m
s
w
i
t
h
Ap
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
8
7
,
p
.
1
1
5
9
7
7
,
J
a
n
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
2
1
.
1
1
5
9
7
7
.
[
1
8
]
R
.
M
o
h
a
w
e
sh
,
S
.
X
u
,
M
.
S
p
r
i
n
g
e
r
,
Y
.
Jararw
e
h
,
M
.
A
l
-
H
a
w
a
w
r
e
h
,
a
n
d
S
.
M
a
q
s
o
o
d
,
“
A
n
e
x
p
l
a
i
n
a
b
l
e
e
n
s
e
m
b
l
e
o
f
M
u
l
t
i
v
i
e
w
d
e
e
p
l
e
a
r
n
i
n
g
mo
d
e
l
f
o
r
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
,
”
J
o
u
r
n
a
l
o
f
K
i
n
g
S
a
u
d
U
n
i
v
e
rsi
t
y
--
C
o
m
p
u
t
e
r a
n
d
I
n
f
o
rm
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
3
5
,
n
o
.
8
,
p
.
1
0
1
6
4
4
,
S
e
p
.
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
k
su
c
i
.
2
0
2
3
.
1
0
1
6
4
4
.
[
1
9
]
G
.
M
.
S
h
a
h
a
r
i
a
r
,
M
.
T.
R
.
S
h
a
w
o
n
,
F
.
M
.
S
h
a
h
,
M
.
S
.
A
l
a
m,
a
n
d
M
.
S
.
M
a
h
b
u
b
,
“
B
e
n
g
a
l
i
f
a
k
e
r
e
v
i
e
w
s:
a
b
e
n
c
h
mark
d
a
t
a
se
t
a
n
d
d
e
t
e
c
t
i
o
n
s
y
st
e
m,
”
N
e
u
ro
c
o
m
p
u
t
i
n
g
,
v
o
l
.
5
9
2
,
p
.
1
2
7
7
3
2
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
e
u
c
o
m.
2
0
2
4
.
1
2
7
7
3
2
.
[
2
0
]
S
.
Ö
B
i
r
i
m,
I
.
K
a
z
a
n
c
o
g
l
u
,
S
.
K
.
M
a
n
g
l
a
,
A
.
K
a
h
r
a
m
a
n
,
S
.
K
u
mar,
a
n
d
Y
.
K
a
z
a
n
c
o
g
l
u
,
“
D
e
t
e
c
t
i
n
g
f
a
k
e
r
e
v
i
e
w
s
t
h
r
o
u
g
h
t
o
p
i
c
mo
d
e
l
l
i
n
g
,
”
J
o
u
r
n
a
l
o
f
B
u
s
i
n
e
ss R
e
se
a
rc
h
,
v
o
l
.
1
4
9
,
p
p
.
8
8
4
–
9
0
0
,
O
c
t
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
b
u
sr
e
s.
2
0
2
2
.
0
5
.
0
8
1
.
[
2
1
]
J.
Y
a
o
,
L.
Ji
a
n
g
,
C
.
S
h
i
,
a
n
d
S
.
Y
a
n
,
“
F
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
w
i
t
h
l
a
b
e
l
-
c
o
n
si
st
e
n
t
a
n
d
h
i
e
r
a
r
c
h
i
c
a
l
-
r
e
l
a
t
i
o
n
-
a
w
a
r
e
g
r
a
p
h
c
o
n
t
r
a
s
t
i
v
e
l
e
a
r
n
i
n
g
,
”
K
n
o
w
l
e
d
g
e
-
Ba
s
e
d
S
y
st
e
m
s
,
v
o
l
.
3
0
2
,
p
.
1
1
2
3
8
5
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
k
n
o
s
y
s.
2
0
2
4
.
1
1
2
3
8
5
.
[
2
2
]
M
.
W
a
l
t
h
e
r
,
T.
Ja
k
o
b
i
,
S
.
J
.
W
a
t
s
o
n
,
a
n
d
G
.
S
t
e
v
e
n
s,
“
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
a
b
o
u
t
t
h
e
c
o
n
s
u
mers
’
si
d
e
o
f
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
–
W
h
i
c
h
c
u
e
s
d
o
c
o
n
s
u
m
e
r
s
u
s
e
t
o
d
e
t
e
r
m
i
n
e
t
h
e
v
e
r
a
c
i
t
y
o
f
o
n
l
i
n
e
u
s
e
r
r
e
v
i
e
w
s?
,
”
C
o
m
p
u
t
e
rs
i
n
H
u
m
a
n
Be
h
a
v
i
o
u
r
Re
p
o
r
t
s
,
v
o
l
.
1
0
,
p
.
1
0
0
2
7
8
,
M
a
y
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
c
h
b
r
.
2
0
2
3
.
1
0
0
2
7
8
.
[
2
3
]
S
.
B
.
J
a
b
e
u
r
,
H
.
B
a
l
l
o
u
k
,
W
.
B
.
A
r
f
i
,
a
n
d
J.
-
M
.
S
a
h
u
t
,
“
A
r
t
i
f
i
c
i
a
l
i
n
t
e
l
l
i
g
e
n
c
e
a
p
p
l
i
c
a
t
i
o
n
s
i
n
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
:
b
i
b
l
i
o
me
t
r
i
c
a
n
a
l
y
si
s
a
n
d
f
u
t
u
r
e
a
v
e
n
u
e
s
f
o
r
r
e
sea
r
c
h
,
”
J
o
u
r
n
a
l
o
f
B
u
s
i
n
e
ss
Re
se
a
rc
h
,
v
o
l
.
1
5
8
,
p
.
1
1
3
6
3
1
,
2
0
2
3
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
j
b
u
sr
e
s.
2
0
2
2
.
1
1
3
6
3
1
.
[
2
4
]
J.
L
i
,
M
.
H
u
a
n
g
,
X
.
Y
a
n
g
,
a
n
d
X
.
Z
h
u
,
“
E
x
p
l
o
i
t
i
n
g
B
ER
T
f
o
r
f
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
,
”
I
E
EE
Ac
c
e
ss
,
v
o
l
.
7
,
p
p
.
1
5
0
0
0
0
–
1
5
0
0
1
0
,
2
0
1
9
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
1
9
.
2
9
4
5
0
0
2
.
[
2
5
]
Y
.
R
e
n
a
n
d
D
.
Ji
,
“
N
e
u
r
a
l
n
e
t
w
o
r
k
s
f
o
r
d
e
c
e
p
t
i
v
e
o
p
i
n
i
o
n
s
p
a
m
d
e
t
e
c
t
i
o
n
,
”
Ex
p
e
rt
S
y
s
t
e
m
s
w
i
t
h
A
p
p
l
i
c
a
t
i
o
n
s
,
v
o
l
.
1
1
3
,
p
p
.
1
4
0
–
1
5
1
,
J
a
n
.
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
sw
a
.
2
0
1
8
.
0
6
.
0
2
7
.
[
2
6
]
Y
.
Zh
a
n
g
,
G
.
Zh
o
n
g
,
J.
C
h
e
n
,
a
n
d
H
.
Li
,
“
F
a
k
e
r
e
v
i
e
w
d
e
t
e
c
t
i
o
n
u
s
i
n
g
t
r
a
n
sf
o
r
mer
-
b
a
se
d
m
o
d
e
l
s,”
I
n
f
o
rm
a
t
i
o
n
Pr
o
c
e
ssi
n
g
&
Ma
n
a
g
e
m
e
n
t
,
v
o
l
.
5
8
,
n
o
.
3
,
p
.
1
0
2
4
7
2
,
M
a
y
2
0
2
1
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
p
m
.
2
0
2
1
.
1
0
2
4
7
2
.
B
I
O
G
RAP
H
I
E
S O
F
AUTH
O
RS
Ra
shm
i
R
.
re
c
e
iv
e
d
h
e
r
B.
E.
d
e
g
re
e
in
c
o
m
p
u
ter
sc
ien
c
e
fro
m
Ap
p
a
In
stit
u
te
o
f
En
g
i
n
e
e
rin
g
a
n
d
Tec
h
n
o
l
o
g
y
,
Ka
rn
a
tak
a
,
in
2
0
0
7
a
n
d
h
e
r
M
.
Tec
h
.
d
e
g
re
e
in
c
o
m
p
u
te
r
n
e
two
rk
s
a
n
d
e
n
g
i
n
e
e
rin
g
fro
m
B.
M
.
S
.
C
o
ll
e
g
e
o
f
En
g
in
e
e
rin
g
,
Be
n
g
a
lu
r
u
,
in
2
0
1
9
.
S
h
e
is
c
u
rre
n
tl
y
a
n
a
ss
istan
t
p
r
o
fe
ss
o
r
i
n
th
e
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
S
c
ien
c
e
a
n
d
En
g
i
n
e
e
rin
g
a
t
B.
M
.
S
.
Co
ll
e
g
e
o
f
En
g
in
e
e
ri
n
g
.
He
r
re
se
a
rc
h
in
tere
sts
in
c
lu
d
e
m
a
c
h
in
e
lea
rn
in
g
a
n
d
d
e
e
p
lea
rn
in
g
.
S
h
e
h
a
s
o
v
e
r
1
0
y
e
a
rs
o
f
i
n
d
u
stry
e
x
p
e
rien
c
e
i
n
s
o
ftwa
re
d
e
v
e
lo
p
m
e
n
t
a
n
d
c
o
n
tri
b
u
tes
to
t
h
is
stu
d
y
th
r
o
u
g
h
h
e
r
e
x
p
e
rti
se
in
in
tell
ig
e
n
t
s
y
ste
m
s.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
ra
sh
m
ir.
ise
@b
m
sc
e
.
a
c
.
in
.
S
h
o
b
h
a
T
.
re
c
e
iv
e
d
h
e
r
B.
E.
d
e
g
re
e
i
n
i
n
fo
rm
a
ti
o
n
sc
ien
c
e
a
n
d
e
n
g
i
n
e
e
rin
g
in
2
0
0
4
a
n
d
th
e
M
.
Tec
h
.
d
e
g
re
e
i
n
c
o
m
p
u
ter
sc
ien
c
e
a
n
d
e
n
g
i
n
e
e
rin
g
in
2
0
1
0
,
b
o
t
h
fr
o
m
Visv
e
sv
a
ra
y
a
Tec
h
n
o
l
o
g
ica
l
Un
i
v
e
rsity
,
Ka
rn
a
tak
a
.
S
h
e
c
o
m
p
let
e
d
h
e
r
P
h
.
D.
i
n
m
a
c
h
in
e
lea
rn
in
g
fro
m
Visv
e
sv
a
ra
y
a
Tec
h
n
o
l
o
g
ica
l
Un
iv
e
rsit
y
,
Be
lag
a
v
i
,
i
n
2
0
2
2
.
S
h
e
is
c
u
r
re
n
tl
y
a
n
a
ss
o
c
iate
p
ro
fe
ss
o
r
in
th
e
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
S
c
ie
n
c
e
a
n
d
En
g
in
e
e
ri
n
g
a
t
B.
M
.
S
.
Co
ll
e
g
e
o
f
En
g
in
e
e
ri
n
g
,
Be
n
g
a
l
u
ru
.
He
r
re
se
a
rc
h
i
n
tere
sts
i
n
c
lu
d
e
d
a
ta
m
in
in
g
,
m
a
c
h
in
e
lea
rn
in
g
,
a
n
d
a
rti
ficia
l
i
n
telli
g
e
n
c
e
.
S
h
e
h
a
s
p
u
b
li
sh
e
d
e
x
ten
si
v
e
ly
in
S
c
o
p
u
s
a
n
d
Web
o
f
S
c
ien
c
e
–
in
d
e
x
e
d
j
o
u
r
n
a
ls an
d
c
o
n
fe
re
n
c
e
s.
S
h
e
c
a
n
b
e
c
o
n
tac
ted
a
t
sh
o
b
h
a
.
ise
@b
m
sc
e
.
a
c
.
in
.
Evaluation Warning : The document was created with Spire.PDF for Python.