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.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
,
p
p
.
3
8
3
7
~
3
8
4
8
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
13
.i
4
.
p
p
3
8
3
7
-
3
8
4
8
3837
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
M
ulti
pla
t
forms
f
a
ke acco
unts de
te
ction ba
sed o
n
fe
dera
ted
lea
rning
M
a
rina
Aze
r
1
,
H
a
la
H
.
Z
a
y
e
d
1,
2
,
M
a
hm
o
ud
E
.
A.
Gad
a
lla
h
3
,
M
o
ha
m
ed
T
a
ha
1
1
F
a
c
u
l
t
y
o
f
C
o
mp
u
t
e
r
s a
n
d
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
B
e
n
h
a
U
n
i
v
e
r
s
i
t
y
,
B
e
n
h
a
,
Eg
y
p
t
2
F
a
c
u
l
t
y
o
f
En
g
i
n
e
e
r
i
n
g
,
E
g
y
p
t
U
n
i
v
e
r
si
t
y
o
f
I
n
f
o
r
mat
i
c
s
,
C
a
i
r
o
,
Eg
y
p
t
3
F
a
c
u
l
t
y
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
,
M
o
d
e
r
n
A
c
a
d
e
m
y
,
C
a
i
r
o
,
E
g
y
p
t
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
1
2
,
2
0
2
4
R
ev
is
ed
Ap
r
2
0
,
2
0
2
4
Acc
ep
ted
J
u
n
1
4
,
2
0
2
4
Id
e
n
ti
f
y
in
g
a
n
d
m
it
ig
a
ti
n
g
fa
k
e
p
ro
fil
e
s
is
a
n
u
r
g
e
n
t
issu
e
d
u
ri
n
g
t
h
e
a
g
e
o
f
wid
e
sp
re
a
d
i
n
teg
ra
ti
o
n
wi
th
so
c
ia
l
m
e
d
ia
p
latfo
rm
s.
t
h
is
st
u
d
y
a
d
d
re
ss
e
s
th
e
c
h
a
ll
e
n
g
e
o
f
fa
k
e
p
r
o
fil
e
d
e
tec
ti
o
n
o
n
m
a
jo
r
so
c
ial
p
latf
o
rm
s
-
F
a
c
e
b
o
o
k
,
In
sta
g
ra
m
,
a
n
d
X
(Twit
ter).
Emp
lo
y
in
g
a
two
-
sid
e
d
a
p
p
r
o
a
c
h
,
it
c
o
m
p
a
re
s
sta
c
k
in
g
m
o
d
e
l
o
f
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
with
t
h
e
fe
d
e
ra
ted
lea
rn
in
g
.
Th
e
re
se
a
rc
h
e
x
ten
d
s
to
fo
u
r
d
a
ta
se
ts,
two
In
sta
g
ra
m
d
a
tas
e
ts,
o
n
e
X
d
a
tas
e
t,
a
n
d
o
n
e
F
a
c
e
b
o
o
k
d
a
tas
e
t,
re
p
o
rt
in
g
imp
re
ss
iv
e
a
c
c
u
ra
c
y
m
e
tri
c
s.
F
e
d
e
ra
ted
lea
rn
in
g
sta
n
d
s
o
u
t
f
o
r
it
i
s
e
ffe
c
ti
v
e
n
e
ss
in
fa
k
e
p
r
o
fil
e
d
e
tec
ti
o
n
,
p
rio
ri
ti
z
in
g
u
se
r
d
a
ta
p
r
iv
a
c
y
.
Re
su
lt
s
re
v
e
a
l
In
sta
g
ra
m
fa
k
e
/
re
a
l
d
a
tas
e
t
a
c
h
iev
e
s
9
6
%
a
c
c
u
ra
c
y
wh
il
e
I
n
sta
g
ra
m
h
u
m
a
n
/
b
o
t
d
a
tas
e
t
re
a
c
h
e
s
9
5
%
a
c
c
u
ra
c
y
with
fe
d
e
ra
ted
lea
rn
in
g
.
u
sin
g
t
h
e
sta
c
k
in
g
m
o
d
e
l
X
’
s
fa
k
e
/
re
a
l
d
a
tas
e
t
a
c
h
iev
e
s
9
9
.
4
%
a
c
c
u
ra
c
y
,
a
n
d
F
a
c
e
b
o
o
k
fa
k
e
/
re
a
l
d
a
tas
e
t
re
a
c
h
e
s
9
9
.
8
%
a
c
c
u
ra
c
y
u
si
n
g
th
e
sa
m
e
m
o
d
e
l.
T
h
e
st
u
d
y
u
n
d
e
rsc
o
re
s
t
h
e
p
i
v
o
tal
ro
le
o
f
d
a
ta
p
ri
v
a
c
y
,
p
o
sit
io
n
in
g
fe
d
e
ra
ted
lea
rn
i
n
g
a
s
a
n
e
th
ica
l
c
h
o
ice
.
It
c
o
m
p
a
re
s
th
e
ti
m
e
e
fficie
n
c
y
o
f
sta
c
k
in
g
a
n
d
fe
d
e
ra
ted
lea
rn
in
g
,
with
th
e
fo
rm
e
r
p
ro
v
i
d
in
g
g
o
o
d
p
e
rf
o
rm
a
n
c
e
in
les
s
ti
m
e
a
n
d
t
h
e
latter
e
m
p
h
a
siz
in
g
d
a
ta
p
riv
a
c
y
b
u
t
c
o
n
su
m
in
g
m
o
re
ti
m
e
.
Re
su
lt
s
a
re
b
e
n
c
h
m
a
rk
e
d
a
g
a
in
st
re
late
d
wo
rk
s,
sh
o
wc
a
sin
g
s
u
p
e
rio
r
p
e
rf
o
rm
a
n
c
e
.
Th
e
st
u
d
y
c
o
n
tri
b
u
tes
sig
n
ifi
c
a
n
t
ly
t
o
fa
k
e
p
ro
f
il
e
d
e
tec
ti
o
n
,
o
ffe
rin
g
a
d
a
p
tab
le
so
l
u
ti
o
n
s
a
n
d
in
sig
h
ts.
K
ey
w
o
r
d
s
:
Data
p
r
iv
ac
y
Fak
e
ac
co
u
n
t
d
etec
tio
n
Fed
er
ated
lear
n
in
g
Ma
ch
in
e
lear
n
in
g
So
cial
m
ed
ia
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
:
Ma
r
in
a
Aze
r
Facu
lty
o
f
co
m
p
u
ter
s
an
d
Ar
ti
f
icial
in
tellig
en
ce
,
B
en
h
a
Un
iv
er
s
ity
B
en
h
a,
E
g
y
p
t
E
m
ail: m
ar
in
a.
ess
am
9
9
1
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
So
cial
m
ed
ia
’
s
r
ap
id
ex
p
an
s
io
n
p
r
esen
ts
b
o
th
o
p
p
o
r
tu
n
ities
an
d
ch
allen
g
es
f
o
r
m
ar
k
etin
g
ca
m
p
aig
n
s
an
d
ce
leb
r
ity
p
r
o
m
o
tio
n
.
Fak
e
p
r
o
f
iles
,
o
f
te
n
cr
ea
ted
t
o
m
is
r
ep
r
esen
t
in
d
iv
id
u
als
o
r
e
n
titi
es,
ca
n
d
am
ag
e
r
ep
u
tatio
n
s
a
n
d
d
is
to
r
t
en
g
ag
em
en
t
m
et
r
ics.
Mo
r
e
o
v
er
,
th
ey
co
n
tr
i
b
u
te
t
o
c
o
n
f
u
s
io
n
an
d
f
ac
ilit
ate
cy
b
er
b
u
lly
in
g
.
Priv
ac
y
co
n
ce
r
n
s
v
ar
y
am
o
n
g
u
s
er
s
in
th
e
o
n
lin
e
r
ea
lm
.
W
ith
th
e
p
r
o
lif
er
atio
n
o
f
b
i
g
d
ata
p
latf
o
r
m
s
lik
e
s
o
cial
m
ed
ia,
id
en
tity
f
r
au
d
h
as
b
ec
o
m
e
a
p
r
e
s
s
in
g
is
s
u
e.
So
cial
m
ed
ia
p
latf
o
r
m
s
h
av
e
b
ec
o
m
e
p
r
im
e
tar
g
ets
f
o
r
s
p
am
m
er
s
a
n
d
c
o
n
a
r
tis
ts
,
p
o
s
in
g
v
ar
io
u
s
th
r
ea
ts
.
I
n
s
tag
r
am
s
tan
d
s
o
u
t
a
s
a
d
o
m
i
n
an
t
s
o
cial
n
etwo
r
k
in
g
p
latf
o
r
m
,
b
u
t
alo
n
g
s
id
e
it
i
s
p
o
p
u
lar
ity
co
m
es
in
cr
ea
s
ed
ex
p
lo
itatio
n
o
f
u
s
er
s
.
X
(
T
witter
)
,
k
n
o
wn
f
o
r
its
s
im
p
licity
,
h
o
s
ts
a
m
ix
o
f
r
ea
l
u
s
er
s
an
d
b
o
ts
,
with
th
e
latter
o
f
ten
m
asq
u
e
r
ad
in
g
as
h
u
m
an
s
,
Face
b
o
o
k
,
as
a
lead
in
g
s
o
cial
n
etwo
r
k
in
g
p
latf
o
r
m
,
f
ac
es
s
im
ilar
ch
allen
g
es
with
th
e
p
r
o
life
r
atio
n
o
f
f
ak
e
p
r
o
f
iles
an
d
th
e
ex
p
lo
itatio
n
o
f
u
s
er
s
.
Dete
ctin
g
s
o
cial
m
ed
ia
b
o
ts
is
cr
u
cial
f
o
r
m
ain
tain
in
g
o
n
lin
e
d
is
co
u
r
s
e
in
teg
r
ity
,
p
r
o
m
p
tin
g
ex
ten
s
iv
e
r
esear
ch
in
to
r
elev
an
t
d
atasets
.
So
cial
b
o
ts
,
o
r
s
y
b
il
ac
c
o
u
n
ts
,
ar
e
au
to
m
ated
alg
o
r
ith
m
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.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
:
3
8
3
7
-
3
8
4
8
3838
th
at
en
g
a
g
e
with
u
s
er
s
o
n
s
o
c
ial
m
ed
ia.
W
h
ile
s
o
m
e
ar
e
h
a
r
m
less
o
r
ev
e
n
e
n
ter
tain
in
g
,
o
th
er
s
ar
e
u
s
ed
f
o
r
d
ec
ep
tiv
e
p
u
r
p
o
s
es,
s
u
ch
as
c
r
ea
tin
g
f
a
k
e
g
r
ass
r
o
o
ts
s
u
p
p
o
r
t
f
o
r
p
o
liti
ca
l
ag
en
d
as.
Ad
d
r
ess
in
g
th
e
is
s
u
e
o
f
f
ak
e
h
u
m
an
ac
co
u
n
ts
o
n
s
o
cia
l
m
ed
ia
p
latf
o
r
m
s
r
e
q
u
ir
es
a
m
u
ltifa
ce
ted
ap
p
r
o
ac
h
,
in
clu
d
i
n
g
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
f
o
r
d
etec
tio
n
an
d
s
tr
icter
v
er
if
icatio
n
p
r
o
ce
d
u
r
es.
R
eso
lv
in
g
th
is
p
r
o
b
lem
is
ess
en
tial
f
o
r
en
s
u
r
in
g
a
s
af
e
an
d
tr
u
s
two
r
th
y
o
n
lin
e
en
v
ir
o
n
m
en
t f
o
r
all
u
s
er
s
[
1
]
–
[
3
]
.
T
h
e
o
b
jectiv
e
o
f
th
is
s
tu
d
y
is
to
d
ev
elo
p
a
n
d
e
v
alu
ate
ef
f
ec
tiv
e
s
tr
ateg
ies
f
o
r
id
en
ti
f
y
in
g
a
n
d
m
itig
atin
g
f
ak
e
p
r
o
f
iles
with
in
th
ese
d
iv
er
s
e
an
d
d
y
n
am
ic
o
n
lin
e
en
v
ir
o
n
m
en
ts
.
I
n
th
e
p
u
r
s
u
it
o
f
th
is
g
o
al,
we
em
p
lo
y
a
two
-
f
o
l
d
ap
p
r
o
ac
h
es
.
Firstl
y
,
we
h
ar
n
ess
th
e
p
o
we
r
o
f
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
o
lo
g
ies,
lev
er
ag
i
n
g
v
ar
io
u
s
alg
o
r
ith
m
s
to
c
o
n
s
tr
u
ct
a
r
o
b
u
s
t
en
s
em
b
le
m
o
d
el.
T
h
is
m
o
d
el
is
d
esig
n
e
d
to
e
n
h
an
ce
th
e
ab
ilit
y
to
d
if
f
er
en
tiate
b
etwe
en
g
en
u
i
n
e
an
d
f
r
a
u
d
u
le
n
t
p
r
o
f
iles
[
4
]
.
S
ec
o
n
d
ly
,
we
e
x
p
lo
r
e
th
e
in
n
o
v
ativ
e
p
ar
ad
i
g
m
o
f
f
ed
er
ated
lear
n
in
g
.
Fed
e
r
ated
lear
n
in
g
p
r
esen
ts
an
i
n
tr
ig
u
i
n
g
o
p
p
o
r
tu
n
ity
to
ad
d
r
ess
f
ak
e
p
r
o
f
ile
d
etec
tio
n
wh
ile
r
esp
ec
tin
g
th
e
p
ar
a
m
o
u
n
t
im
p
o
r
tan
ce
o
f
d
ata
p
r
iv
a
cy
to
c
o
n
s
id
er
th
e
s
en
s
itiv
e
n
atu
r
e
o
f
th
e
d
ata.
f
ed
er
ated
lear
n
in
g
o
f
f
er
s
a
co
m
p
ellin
g
s
o
lu
tio
n
to
t
h
e
e
th
ical
an
d
tech
n
ical
ch
allen
g
es
o
f
f
ak
e
p
r
o
f
ile
d
etec
tio
n
in
a
p
r
iv
ac
y
-
awa
r
e
wo
r
ld
b
y
en
ab
lin
g
m
o
d
el
tr
ai
n
in
g
with
o
u
t
ce
n
tr
alize
d
ac
ce
s
s
to
s
en
s
itiv
e
u
s
er
d
ata.
O
f
f
er
s
a
co
m
p
ar
ativ
e
a
n
aly
s
is
b
etwe
en
th
e
en
s
em
b
le
m
ac
h
in
e
lear
n
in
g
m
o
d
el
an
d
t
h
e
f
ed
er
ated
lear
n
in
g
ap
p
r
o
ac
h
ac
r
o
s
s
th
e
f
o
u
r
d
atasets
s
p
an
n
in
g
th
e
t
h
r
ee
p
latf
o
r
m
s
.
T
h
is
co
m
p
a
r
is
o
n
s
cr
u
tin
izes
th
ei
r
p
er
f
o
r
m
an
ce
in
te
r
m
s
o
f
ac
c
u
r
ac
y
m
etr
ics an
d
tim
e
c
o
n
s
u
m
p
tio
n
[
5
]
.
T
h
e
ch
allen
g
es
in
h
er
en
t
in
th
is
wo
r
k
s
tem
f
r
o
m
v
ar
io
u
s
f
ac
to
r
s
.
Firstl
y
,
th
e
r
elia
b
ilit
y
an
d
r
ep
r
esen
tativ
en
ess
o
f
th
e
d
atasets
u
tili
ze
d
g
r
ea
tly
in
f
lu
en
ce
th
e
v
alid
ity
o
f
th
e
s
tu
d
y
’
s
f
in
d
in
g
s
.
Seco
n
d
l
y
,
th
e
s
u
b
s
tan
tial
co
m
p
u
tatio
n
al
r
es
o
u
r
ce
s
d
em
a
n
d
ed
f
o
r
t
r
ain
in
g
an
d
ass
ess
in
g
in
tr
icate
m
ac
h
i
n
e
lear
n
in
g
m
o
d
els,
p
ar
ticu
lar
ly
in
th
e
co
n
tex
t
o
f
f
ed
er
ated
lear
n
in
g
,
p
r
esen
t
p
r
ac
tical
lim
itatio
n
s
.
T
h
es
e
c
o
n
s
t
r
a
i
n
t
s
c
o
u
l
d
h
i
n
d
e
r
t
h
e
s
c
a
l
a
b
i
l
i
t
y
a
n
d
r
e
a
l
-
w
o
r
l
d
a
p
p
l
i
c
a
b
i
l
it
y
o
f
t
h
e
p
r
o
p
o
s
e
d
a
p
p
r
o
a
c
h
,
m
o
r
e
o
v
e
r
,
e
x
t
e
r
n
a
l
f
a
c
t
o
r
s
,
i
n
c
l
u
d
i
n
g
a
l
t
e
r
a
ti
o
n
s
i
n
p
l
at
f
o
r
m
r
e
g
u
l
a
t
io
n
s
,
s
h
i
f
ts
i
n
u
s
e
r
b
e
h
a
v
i
o
r
,
an
d
a
d
v
a
n
c
e
m
e
n
t
s
i
n
t
e
c
h
n
o
l
o
g
y
,
m
a
y
i
m
p
a
c
t
t
h
e
e
f
f
e
c
t
i
v
e
n
es
s
o
f
t
h
e
p
r
o
p
o
s
e
d
m
e
t
h
o
d
o
l
o
g
y
o
v
e
r
t
i
m
e
.
T
h
u
s
,
o
n
g
o
i
n
g
m
o
n
i
t
o
r
i
n
g
a
n
d
a
d
a
p
t
a
t
i
o
n
a
r
e
i
m
p
e
r
a
t
i
v
e
t
o
n
a
v
i
g
a
t
e
t
h
es
e
e
x
t
e
r
n
al
i
n
f
l
u
e
n
c
e
s
a
n
d
e
n
s
u
r
e
t
h
e
c
o
n
t
i
n
u
e
d
r
e
l
e
v
a
n
c
e
a
n
d
e
f
f
i
c
a
c
y
o
f
t
h
e
ap
p
r
o
a
c
h
.
T
h
is
s
tu
d
y
co
n
tr
ib
u
tes
to
th
e
ex
is
tin
g
b
o
d
y
o
f
r
esear
c
h
b
y
i
n
v
esti
g
atin
g
th
e
ef
f
ec
ts
o
f
f
ak
e
ac
co
u
n
t
d
etec
tio
n
u
s
in
g
two
d
if
f
e
r
en
t
ap
p
r
o
ac
h
es
o
n
t
h
r
ee
d
i
f
f
er
en
t
p
latf
o
r
m
s
.
Mo
s
t
ea
r
lier
s
tu
d
ies
h
av
e
n
o
t
ex
p
licitly
ad
d
r
ess
ed
th
e
in
f
lu
en
ce
o
f
f
a
k
e
ac
co
u
n
t d
etec
tio
n
ac
r
o
s
s
m
u
ltip
le
p
latf
o
r
m
s
alwa
y
s
f
o
cu
s
o
n
s
p
ec
if
ic
p
latf
o
r
m
an
d
h
av
e
n
o
t
f
o
cu
s
ed
o
n
d
at
a
p
r
iv
ac
y
a
p
p
r
o
ac
h
es
in
th
is
to
p
ic.
T
h
er
e
f
o
r
e,
b
y
ex
a
m
in
in
g
th
e
ef
f
icac
y
o
f
d
etec
tio
n
m
eth
o
d
s
ac
r
o
s
s
v
ar
i
o
u
s
p
latf
o
r
m
s
an
d
em
p
h
asizin
g
th
e
im
p
o
r
ta
n
ce
o
f
d
ata
p
r
i
v
ac
y
,
th
is
r
esear
ch
ex
p
an
d
s
o
u
r
u
n
d
er
s
tan
d
in
g
o
f
f
a
k
e
p
r
o
f
ile
d
etec
tio
n
s
tr
ateg
ies
an
d
t
h
eir
im
p
licatio
n
s
in
d
iv
e
r
s
e
o
n
lin
e
en
v
ir
o
n
m
en
ts
.
Sar
h
an
an
d
M
attar
[
6
]
co
m
p
ar
e
m
ac
h
in
e
le
ar
n
in
g
alg
o
r
ith
m
s
p
er
f
o
r
m
an
c
e
to
ca
teg
o
r
ize
u
s
er
ac
co
u
n
ts
o
n
th
e
I
n
s
tag
r
a
m
p
latf
o
r
m
,
R
o
s
tam
i
[
7
]
f
o
cu
s
ed
o
n
f
ea
tu
r
e
s
elec
tio
n
to
en
h
an
ce
class
if
icatio
n
ac
cu
r
ac
y
.
Utilizin
g
a
m
u
lti
-
o
b
jectiv
e
h
y
b
r
id
f
ea
tu
r
e
s
elec
tio
n
ap
p
r
o
ac
h
,
th
e
r
esear
ch
er
s
id
en
tify
a
ca
n
d
id
ate
f
ea
tu
r
e
s
et
with
h
ig
h
r
elev
an
c
e
to
th
e
tar
g
et
class
an
d
m
in
im
al
r
ed
u
n
d
a
n
cy
u
s
in
g
th
e
m
in
im
u
m
r
ed
u
n
d
a
n
cy
m
ax
im
u
m
r
elev
a
n
ce
(
m
R
MR
)
alg
o
r
ith
m
.
T
h
e
f
in
al
f
ea
tu
r
e
s
et
is
ch
o
s
en
f
o
r
d
etec
tio
n
o
p
er
atio
n
s
,
y
ield
in
g
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
co
m
p
ar
ed
to
ex
is
tin
g
m
eth
o
d
s
wh
e
n
test
ed
o
n
T
witter
d
atasets
.
Hak
im
i
et
a
l.
[
8
]
estab
lis
h
ed
a
co
m
p
r
eh
en
s
iv
e
s
et
o
f
f
iv
e
cr
itical
ch
ar
ac
ter
is
tics
p
iv
o
tal
in
d
is
ce
r
n
in
g
f
ak
e
f
r
o
m
g
en
u
in
e
u
s
er
s
o
n
Face
b
o
o
k
.
Su
b
s
eq
u
en
tly
,
we
em
p
lo
y
th
ese
attr
ib
u
tes
to
id
en
tify
k
ey
class
if
ier
s
in
m
a
ch
in
e
lear
n
in
g
,
with
p
ar
ticu
lar
f
o
cu
s
o
n
K
-
n
ea
r
est
n
eig
h
b
o
r
(
KNN)
,
s
u
p
p
o
r
t
v
e
cto
r
m
ac
h
in
e
(
SVM)
,
an
d
n
e
u
r
al
n
etwo
r
k
(
NN)
.
R
esu
lts
in
d
icate
th
at
KNN
em
er
g
es
as
th
e
to
p
p
er
f
o
r
m
er
,
ac
h
iev
in
g
an
8
2
%
class
if
icatio
n
p
r
ec
is
io
n
r
ate
am
o
n
g
t
h
e
class
if
ier
s
ev
alu
ated
.
I
n
co
n
clu
s
io
n
,
th
is
r
esear
ch
is
p
o
is
ed
t
o
s
ig
n
if
ica
n
tly
co
n
tr
i
b
u
te
to
th
e
o
n
g
o
in
g
ef
f
o
r
ts
to
f
ig
h
t
f
ak
e
p
r
o
f
iles
o
n
s
o
cial
m
ed
ia.
B
y
u
s
in
g
m
ac
h
in
e
lear
n
in
g
tech
n
iq
u
es
an
d
th
e
in
n
o
v
ativ
e
p
r
iv
ac
y
-
p
r
eser
v
i
n
g
ca
p
ab
ilit
ies
o
f
f
ed
er
ated
lear
n
in
g
,
we
aim
to
f
o
r
tify
th
e
s
e
cu
r
ity
an
d
tr
u
s
two
r
th
in
ess
o
f
o
n
lin
e
in
ter
ac
tio
n
s
wh
ile
s
u
p
p
o
r
tin
g
th
e
p
r
in
cip
le
s
o
f
d
ata
p
r
iv
ac
y
i
n
to
d
a
y
’
s
d
i
g
ital
lan
d
s
ca
p
e.
T
h
e
s
tr
u
ctu
r
e
o
f
th
is
p
ap
e
r
is
as
f
o
llo
ws:
s
ec
tio
n
2
o
u
tlin
es
th
e
r
esear
ch
m
eth
o
d
o
lo
g
y
.
Sectio
n
3
p
r
esen
ts
th
e
r
esu
lts
an
d
d
is
cu
s
s
io
n
.
Fin
ally
,
s
ec
tio
n
4
co
n
clu
d
es th
e
p
a
p
er
an
d
p
r
o
v
id
in
g
s
u
g
g
esti
o
n
s
f
o
r
f
u
tu
r
e
r
esear
c
h
ar
ea
s
.
2.
M
E
T
H
O
D
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
f
o
r
f
a
k
e
p
r
o
f
ile
d
etec
tio
n
,
as
s
h
o
wn
in
Fig
u
r
e
1
,
tak
es
a
s
y
s
tem
atic,
m
u
lti
-
p
h
ased
ap
p
r
o
ac
h
to
ac
h
iev
e
r
o
b
u
s
t
r
esu
lts
.
I
t
b
eg
i
n
s
with
co
m
p
r
eh
e
n
s
iv
e
d
ata
g
ath
er
in
g
f
r
o
m
t
h
r
ee
m
ajo
r
s
o
cial
m
ed
ia
p
latf
o
r
m
s
:
Face
b
o
o
k
,
I
n
s
tag
r
a
m
,
an
d
X
(
T
witter
)
.
T
h
is
p
r
o
ce
s
s
en
tails
ac
q
u
ir
in
g
u
s
er
p
r
o
f
iles
,
th
eir
ass
o
ciate
d
ac
tiv
ities
,
an
d
an
y
r
elev
an
t
m
etad
a
ta.
Su
b
s
eq
u
en
tly
,
m
eticu
l
o
u
s
d
ata
p
r
e
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
ar
e
e
m
p
lo
y
ed
to
c
lean
an
d
e
n
s
u
r
e
c
o
n
s
is
ten
cy
with
in
th
e
d
ata.
Featu
r
e
en
g
in
ee
r
in
g
s
er
v
es
as
a
cr
u
cial
s
tep
wh
er
e
d
o
m
ain
-
s
p
e
cif
ic
f
ea
tu
r
es
ar
e
ca
r
ef
u
lly
cr
a
f
ted
to
ca
p
tu
r
e
th
e
s
u
b
tle
b
eh
a
v
io
r
al
p
atter
n
s
th
at
d
if
f
er
en
tiate
f
ak
e
p
r
o
f
iles
f
r
o
m
g
en
u
in
e
o
n
es.
T
o
f
ac
ilit
ate
m
o
d
el
ev
alu
atio
n
o
n
u
n
s
ee
n
d
ata,
th
e
d
ata
is
th
en
s
tr
ateg
ically
s
p
lit
in
to
tr
ain
in
g
an
d
test
in
g
s
ets.
Fo
llo
win
g
th
is
,
th
e
m
o
d
el
d
iv
er
g
es
in
to
tw
o
p
ath
s
,
o
f
f
e
r
in
g
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
Mu
lti p
la
tfo
r
ms fa
ke
a
cc
o
u
n
ts
d
etec
tio
n
b
a
s
ed
o
n
fed
era
te
d
l
ea
r
n
in
g
(
Ma
r
in
a
A
z
er
)
3839
co
m
p
r
eh
e
n
s
iv
e
ap
p
r
o
ac
h
:
s
ev
er
al
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
lev
er
ag
e
th
e
en
g
in
ee
r
ed
f
e
atu
r
es
to
co
n
s
tr
u
ct
p
r
ed
ictiv
e
m
o
d
els b
y
tr
ain
in
g
th
e
m
o
d
el
o
n
th
e
tr
ain
in
g
d
ata
an
d
m
ak
e
a
p
r
ed
ictio
n
s
o
n
th
e
test
s
et
,
th
en
in
th
e
m
eta_
m
o
d
el’
s
tr
ain
in
g
p
h
ase
th
e
m
o
d
el
is
f
u
n
ctio
n
in
g
as
a
s
tack
in
g
m
o
d
el
wh
ich
is
d
esig
n
ed
to
an
aly
ze
an
d
co
m
b
in
e
th
e
p
r
ed
ictio
n
s
g
en
e
r
ated
b
y
th
e
f
iv
e
m
o
s
t
ac
cu
r
ate
m
ac
h
in
e
lear
n
in
g
m
o
d
els
T
h
e
m
eta
-
m
o
d
el’
s
o
u
tp
u
t
is
th
en
c
o
n
s
id
er
ed
th
e
f
in
al
p
r
ed
ictio
n
o
f
th
e
s
y
s
tem
in
th
e
f
ir
s
t
p
ath
,
wh
ile
th
e
s
e
co
n
d
p
at
h
ex
p
l
o
r
es
f
ed
er
ated
lear
n
in
g
as
a
p
r
iv
ac
y
-
p
r
eser
v
in
g
alter
n
ativ
e,
en
a
b
lin
g
m
o
d
el
tr
ain
in
g
with
o
u
t
h
av
in
g
to
s
h
ar
e
th
e
d
ata
its
elf
an
d
with
o
u
t c
e
n
tr
alizin
g
u
s
er
d
ata
f
r
o
m
th
e
th
r
ee
p
latf
o
r
m
s
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
f
o
r
f
ak
e
p
r
o
f
ile
d
etec
tio
n
en
tails
n
av
ig
atin
g
th
r
o
u
g
h
v
ar
io
u
s
co
m
p
lex
ities
in
h
er
en
t
in
th
e
p
r
o
ce
s
s
.
Firstl
y
,
co
m
p
r
eh
en
s
iv
e
d
ata
g
at
h
er
in
g
f
r
o
m
m
ajo
r
s
o
cial
m
ed
ia
p
latf
o
r
m
s
lik
e
Face
b
o
o
k
,
I
n
s
tag
r
am
,
a
n
d
X
(
T
witter
)
p
r
esen
ts
a
s
ig
n
if
i
ca
n
t
ch
allen
g
e
d
u
e
to
d
if
f
er
in
g
d
ata
f
o
r
m
ats.
Fo
llo
win
g
d
ata
co
llectio
n
p
r
ep
r
o
ce
s
s
in
g
tech
n
iq
u
es
a
r
e
e
s
s
en
tial
to
en
s
u
r
e
d
ata
q
u
ality
an
d
co
n
s
is
ten
cy
,
in
v
o
lv
in
g
task
s
s
u
ch
as
m
is
s
in
g
v
alu
es
h
an
d
lin
g
an
d
n
o
r
m
aliza
tio
n
.
Mo
r
eo
v
er
,
cr
af
tin
g
d
o
m
ain
-
s
p
ec
if
ic
f
ea
tu
r
es
f
o
r
ef
f
ec
tiv
e
d
if
f
er
en
tiatio
n
b
etwe
en
f
ak
e
an
d
g
e
n
u
in
e
p
r
o
f
iles
ad
d
s
an
o
th
e
r
lay
er
o
f
co
m
p
lex
ity
to
th
e
m
o
d
el.
Selectin
g
r
elev
an
t
f
ea
tu
r
es
an
d
tr
a
n
s
f
o
r
m
in
g
r
a
w
d
ata
in
to
m
ea
n
in
g
f
u
l
i
n
p
u
t
s
d
em
an
d
s
a
d
ee
p
u
n
d
er
s
tan
d
i
n
g
o
f
s
o
cial
m
e
d
i
a
b
eh
av
i
o
r
a
n
d
d
ec
ep
tio
n
tactics.
Su
b
s
eq
u
en
tly
,
e
v
alu
atin
g
th
e
p
er
f
o
r
m
an
ce
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els
o
n
u
n
s
ee
n
d
ata
b
ec
o
m
es
cr
u
cial,
n
ec
ess
itatin
g
ca
r
ef
u
l
s
elec
tio
n
o
f
ev
alu
atio
n
m
etr
ics
an
d
s
tr
ateg
ies
to
av
o
id
o
v
er
f
itti
n
g
.
Fin
ally
,
th
e
im
p
lem
en
tatio
n
o
f
th
e
f
ed
er
atin
g
lear
n
in
g
ap
p
r
o
ac
h
ac
r
o
s
s
m
u
ltip
le
p
latf
o
r
m
s
in
v
o
lv
es tec
h
n
ical
ch
allen
g
es su
ch
as c
o
m
p
atib
ilit
y
is
s
u
es,
s
ca
la
b
ilit
y
co
n
ce
r
n
s
,
an
d
p
r
iv
ac
y
co
n
s
id
er
atio
n
s
.
Desp
ite
th
ese
d
if
f
ic
u
lties
,
th
e
m
o
d
el
aim
s
to
ac
h
iev
e
r
o
b
u
s
t
an
d
ef
f
ec
tiv
e
f
ak
e
p
r
o
f
ile
d
etec
tio
n
ca
p
ab
ilit
ies.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
2
.
1
.
Da
t
a
co
llect
io
n
Fo
r
th
is
co
m
p
r
eh
e
n
s
iv
e
r
esear
ch
o
n
f
a
k
e
p
r
o
f
ile
d
etec
tio
n
,
we
m
eticu
lo
u
s
ly
cu
r
ated
an
d
u
tili
ze
d
f
o
u
r
d
is
tin
ct
d
atasets
,
ea
ch
s
o
u
r
ce
d
f
r
o
m
th
e
p
r
o
m
in
e
n
t
s
o
cial
m
ed
ia
p
latf
o
r
m
s
o
f
I
n
s
tag
r
am
,
T
witter
,
an
d
Face
b
o
o
k
.
T
h
ese
d
atasets
co
llectiv
ely
r
ep
r
esen
t
a
d
iv
er
s
e
an
d
co
m
p
r
e
h
en
s
iv
e
cr
o
s
s
-
s
ec
tio
n
o
f
u
s
er
b
eh
av
i
o
r
s
an
d
p
r
o
f
iles
f
r
o
m
th
e
e
v
er
-
ev
o
lv
in
g
s
o
cial
m
e
d
ia
lan
d
s
ca
p
e
.
T
h
e
I
n
s
tag
r
am
d
ataset
en
ca
p
s
u
lates
th
e
d
y
n
am
ic
v
is
u
al
in
ter
ac
tio
n
s
o
f
u
s
er
s
,
wh
ile
th
e
T
witter
d
ataset
b
r
in
g
s
f
o
r
th
th
e
r
ea
lm
o
f
c
o
n
cise
an
d
r
ap
id
-
f
ir
e
co
m
m
u
n
icatio
n
.
T
h
e
Face
b
o
o
k
d
ataset
p
r
o
v
i
d
es
in
s
ig
h
ts
in
t
o
th
e
m
u
ltifa
ce
ted
ac
tiv
ities
o
f
u
s
er
s
with
in
th
is
ex
p
an
s
iv
e
s
o
cial
n
etwo
r
k
.
C
o
llectiv
ely
,
th
ese
d
atasets
en
ab
le
a
co
m
p
r
eh
en
s
iv
e
an
al
y
s
is
o
f
f
ak
e
p
r
o
f
ile
d
etec
tio
n
ac
r
o
s
s
a
s
p
ec
tr
u
m
o
f
p
latf
o
r
m
s
,
f
ac
ilit
atin
g
a
r
o
b
u
s
t e
v
alu
atio
n
o
f
th
e
p
r
o
p
o
s
ed
d
etec
tio
n
m
o
d
els.
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.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
:
3
8
3
7
-
3
8
4
8
3840
2
.
1
.
1
.
T
he
I
ns
t
a
g
ra
m
da
t
a
s
et
s
I
n
t
h
is
s
t
u
d
y
,
t
wo
d
i
v
e
r
s
e
d
ata
s
ets
w
er
e
em
p
l
o
y
e
d
t
o
c
o
m
p
r
e
h
e
n
s
i
v
e
ly
a
n
a
ly
ze
p
r
o
f
il
es
o
n
I
n
s
ta
g
r
a
m
.
First
ly
,
we
u
tili
ze
d
th
e
“
I
n
s
ta
g
r
am
f
a
k
e
a
cc
o
u
n
ts
d
at
ase
t
”
s
o
u
r
ce
d
f
r
o
m
Ka
g
g
le
,
it
is
n
a
m
e
d
Dat
aset
1
.
T
h
is
d
at
ase
t
e
n
co
m
p
ass
es
a
t
o
ta
l
o
f
7
8
6
I
n
s
t
ag
r
am
p
r
o
f
iles
,
m
e
tic
u
lo
u
s
l
y
ca
t
eg
o
r
i
ze
d
i
n
t
o
6
9
3
f
a
k
e
ac
c
o
u
n
ts
a
n
d
9
4
au
t
h
en
tic
p
r
o
f
i
les.
T
h
ese
p
r
o
f
iles
a
r
e
ac
c
o
m
p
a
n
ie
d
b
y
a
r
i
ch
s
e
t
o
f
1
2
f
e
at
u
r
es
.
F
u
r
t
h
e
r
m
o
r
e
,
o
u
r
a
n
a
ly
s
is
ex
t
e
n
d
s
t
o
a
s
ec
o
n
d
d
atas
et
,
d
e
n
o
te
d
as Da
tase
t
2
,
wh
ic
h
s
p
ec
if
i
ca
l
ly
ta
r
g
ets
th
e
d
et
ec
ti
o
n
o
f
b
o
t
-
d
r
i
v
e
n
ac
t
iv
it
y
o
n
I
n
s
t
ag
r
a
m
.
T
h
is
d
a
tase
t
c
o
m
p
r
is
es
a
b
a
la
n
c
ed
s
et
o
f
7
0
0
r
e
al
a
cc
o
u
n
ts
al
o
n
g
s
i
d
e
7
0
0
a
u
t
o
m
ate
d
ac
co
u
n
ts
,
to
t
ali
n
g
1
,
4
0
0
p
r
o
f
iles
.
E
a
ch
p
r
o
f
i
le
is
c
h
a
r
a
cte
r
i
ze
d
b
y
a
c
o
m
p
r
e
h
en
s
iv
e
s
et
o
f
1
6
f
ea
t
u
r
es
.
2
.
1
.
2
.
T
he
X
(
T
wit
t
er
)
da
t
a
s
et
I
n
ad
d
itio
n
t
o
o
u
r
e
x
p
lo
r
atio
n
o
f
I
n
s
tag
r
a
m
p
r
o
f
iles
,
th
e
s
tu
d
y
en
co
m
p
ass
es
an
in
v
esti
g
atio
n
in
to
T
witter
ac
co
u
n
t b
eh
av
i
o
r
u
s
in
g
th
e
‘
MI
B
’
d
ataset
(
Data
s
et
3
)
T
h
is
d
ataset
am
alg
a
m
ates
th
r
ee
d
is
tin
ct
d
atasets
co
m
p
iled
f
r
o
m
T
witter
ac
co
u
n
ts
,
co
m
p
r
is
in
g
a
to
tal
o
f
5
,
3
0
1
p
r
o
f
iles
.
Am
o
n
g
t
h
ese,
3
,
3
5
1
p
r
o
f
iles
ar
e
f
lag
g
ed
as
f
ak
e
ac
co
u
n
ts
,
wh
ile
1
,
9
5
0
p
r
o
f
iles
ar
e
au
th
en
ti
ca
ted
as
g
en
u
in
e.
E
ac
h
p
r
o
f
il
e
in
th
is
d
ataset
i
s
ass
o
ciate
d
with
a
d
etailed
s
e
t
o
f
3
4
f
ea
tu
r
es,
p
r
o
v
id
i
n
g
i
n
s
ig
h
ts
in
to
v
ar
io
u
s
asp
ec
ts
o
f
ac
co
u
n
t
ac
tiv
ity
,
in
clu
d
in
g
twee
t f
r
eq
u
e
n
cy
,
r
et
wee
t b
eh
av
io
r
,
f
o
llo
wer
d
em
o
g
r
ap
h
ics,
an
d
lin
g
u
is
tic
p
atter
n
s
.
2
.
1
.
3
.
T
he
F
a
ce
bo
o
k
da
t
a
s
et
Ob
tain
in
g
a
d
ataset
,
Data
s
et
4
,
f
o
r
s
tu
d
y
in
g
Face
b
o
o
k
u
s
er
p
r
o
f
iles
is
ch
allen
g
in
g
d
u
e
to
p
r
iv
ac
y
an
d
ac
ce
s
s
r
estrictio
n
s
.
Ho
wev
er
,
th
is
r
esear
ch
ac
q
u
ir
e
d
a
d
at
aset
co
n
s
is
tin
g
o
f
2
,
8
2
0
p
r
o
f
iles
.
Am
o
n
g
th
ese,
1
,
4
8
2
we
r
e
co
n
f
ir
m
ed
as
r
ea
l
u
s
er
s
,
wh
ile
1
,
3
3
8
wer
e
s
u
s
p
ec
ted
to
b
e
f
ak
e
ac
c
o
u
n
ts
.
T
h
e
d
ataset
en
co
m
p
ass
es
3
4
d
is
tin
ct
f
ea
tu
r
es,
m
ak
in
g
it
a
v
alu
ab
le
r
e
s
o
u
r
ce
f
o
r
i
n
v
esti
g
atin
g
th
e
c
o
m
p
lex
ities
o
f
f
ak
e
ac
co
u
n
t
d
etec
tio
n
o
n
Face
b
o
o
k
.
T
h
is
h
ig
h
lig
h
ts
th
e
im
p
o
r
t
an
ce
o
f
eth
ical
d
ata
co
llectio
n
p
r
ac
tices
an
d
th
e
s
ig
n
if
ican
ce
o
f
s
u
c
h
d
atasets
in
r
esear
ch
en
d
ea
v
o
r
s
in
th
is
d
o
m
ain
.
2
.
2
.
P
re
pro
ce
s
s
ing
T
h
e
p
r
ep
r
o
ce
s
s
in
g
p
h
ase
in
o
u
r
r
esear
ch
is
a
cr
u
cial
s
tep
th
at
ad
ap
ts
to
ea
ch
d
ataset
’
s
u
n
iq
u
e
ch
ar
ac
ter
is
tics
an
d
id
io
s
y
n
cr
a
s
ies
s
o
u
r
ce
d
f
r
o
m
I
n
s
tag
r
am
,
T
witter
,
an
d
Face
b
o
o
k
.
T
h
is
p
h
ase
en
s
u
r
es
th
e
d
ata
is
s
tan
d
ar
d
ized
an
d
clea
n
ed
,
s
ettin
g
th
e
s
tag
e
f
o
r
s
u
b
s
eq
u
en
t
an
aly
s
is
.
T
h
r
o
u
g
h
o
u
t
ea
ch
d
ataset
’
s
p
r
ep
r
o
ce
s
s
in
g
jo
u
r
n
e
y
,
we
m
e
ticu
lo
u
s
ly
ad
d
r
ess
ed
m
is
s
in
g
d
ata,
o
u
tlier
s
,
an
d
in
co
n
s
is
ten
cies,
th
u
s
p
r
ep
ar
in
g
th
e
d
ata
f
o
r
th
e
s
u
b
s
eq
u
en
t f
ea
tu
r
e
en
g
in
ee
r
in
g
a
n
d
m
o
d
el
tr
a
in
in
g
s
tep
s
.
I
n
th
e
co
n
te
x
t
o
f
o
u
r
I
n
s
tag
r
a
m
f
ak
e
an
d
r
ea
l
d
ataset,
it
b
e
ca
m
e
ev
id
e
n
t
th
at
we
wer
e
d
e
alin
g
with
im
b
alan
ce
d
d
ata,
wh
er
e
th
e
n
u
m
b
er
o
f
f
a
k
e
an
d
r
ea
l
p
r
o
f
ile
s
d
em
o
n
s
tr
ated
a
s
ig
n
if
ican
t
v
ar
iatio
n
.
T
o
ad
d
r
ess
th
is
is
s
u
e
an
d
en
s
u
r
e
t
h
e
r
o
b
u
s
tn
ess
o
f
o
u
r
an
aly
s
is
,
we
e
m
p
lo
y
ed
an
o
v
er
s
am
p
lin
g
te
ch
n
iq
u
e
k
n
o
wn
as
s
y
n
th
etic
m
in
o
r
ity
o
v
er
-
s
am
p
lin
g
tech
n
i
q
u
e
f
o
r
n
o
m
in
al
a
n
d
c
o
n
tin
u
o
u
s
f
ea
tu
r
es
(
SM
OT
E
-
NC
)
[
9
]
.
T
h
is
tech
n
iq
u
e
was
ch
o
s
en
d
u
e
to
its
d
em
o
n
s
tr
ated
ef
f
icac
y
in
s
itu
atio
n
s
in
v
o
lv
in
g
im
b
alan
ce
d
d
ata
co
m
p
r
is
in
g
n
u
m
er
ical
an
d
ca
teg
o
r
ical
f
ea
t
u
r
es [
1
0
]
.
I
n
th
e
T
witter
d
ataset,
c
h
ec
k
n
u
ll
v
alu
es
u
s
in
g
th
e
n
u
ll
f
u
n
ctio
n
,
o
u
tlier
s
,
a
n
d
class
b
alan
ce
an
d
co
n
v
er
t
ca
teg
o
r
ical
f
ea
tu
r
es
to
n
u
m
e
r
ic
v
alu
es.
I
n
th
e
Face
b
o
o
k
d
ataset,
th
er
e
ar
e
3
4
f
ea
tu
r
es;
s
o
m
e
o
f
th
em
,
lik
e
v
er
if
ied
an
d
p
r
o
tecte
d
,
ar
e
em
p
ty
with
o
u
t
v
al
u
es,
s
o
th
ey
ar
e
d
r
o
p
p
e
d
,
g
eo
-
en
a
b
led
,
d
ef
a
u
lt
p
r
o
f
ile
im
ag
e,
p
r
o
f
ile_
b
ac
k
g
r
o
u
n
d
_
ti
tle,
u
tc
-
o
f
f
s
et,
an
d
tim
e
z
o
n
e
h
av
e
ju
s
t
1
%
v
alu
es
o
f
all
u
s
er
s
s
o
h
er
e
th
e
im
p
u
tatio
n
is
n
o
t
a
g
o
o
d
s
o
lu
t
io
n
,
s
o
th
e
y
ar
e
also
d
r
o
p
p
e
d
.
T
h
er
e
wer
e
n
u
ll
v
al
u
es
in
o
th
er
f
ea
tu
r
es
im
p
u
ted
u
s
in
g
m
ea
n
with
n
u
m
er
ic
v
alu
es.
2
.
2
.
1
.
F
ea
t
ure
eng
ineering
T
h
e
f
ea
tu
r
e
en
g
in
ee
r
in
g
p
h
ase
is
a
cr
u
cial
s
tep
in
th
e
d
ata
p
r
ep
r
o
ce
s
s
in
g
p
ip
elin
e,
wh
er
e
r
aw
d
ata
u
n
d
er
g
o
es
tr
an
s
f
o
r
m
atio
n
an
d
m
an
ip
u
latio
n
to
cr
ea
te
n
ew,
in
f
o
r
m
ativ
e
f
ea
tu
r
es
th
at
e
n
h
an
c
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
m
ac
h
in
e
lea
r
n
in
g
m
o
d
els.
T
h
is
p
h
ase
in
v
o
lv
es
e
x
tr
ac
tin
g
r
elev
a
n
t
in
f
o
r
m
atio
n
f
r
o
m
t
h
e
av
ailab
le
d
ataset,
s
elec
tin
g
an
d
c
o
m
b
in
in
g
f
ea
t
u
r
es,
an
d
g
en
e
r
atin
g
n
ew
r
e
p
r
esen
tatio
n
s
th
at
ca
p
tu
r
e
m
ea
n
in
g
f
u
l
p
atter
n
s
an
d
r
elatio
n
s
h
ip
s
with
in
th
e
d
ata.
Featu
r
e
en
g
in
ee
r
in
g
aim
s
to
im
p
r
o
v
e
th
e
p
r
ed
ictiv
e
p
o
wer
o
f
m
o
d
els b
y
r
ef
i
n
in
g
th
e
in
p
u
t
d
ata
[
1
1
]
.
‒
I
n
s
tag
r
am
f
a
k
e
-
r
ea
l
d
ataset
T
h
e
f
ea
tu
r
e
e
n
g
in
ee
r
in
g
p
h
a
s
e
en
co
m
p
ass
es
th
e
u
tili
za
tio
n
o
f
1
2
d
is
tin
ct
f
ea
tu
r
es
to
an
aly
ze
en
g
ag
em
e
n
t
d
y
n
am
ics
an
d
p
r
o
f
ile
au
th
en
ticity
.
T
h
ese
f
ea
t
u
r
es
in
clu
d
e
p
ar
am
ete
r
s
s
u
ch
as
“
ed
g
e
f
o
llo
wed
by,
”
“
e
d
g
e
f
o
llo
w,
”
“
u
s
er
n
am
e
len
g
th
,
”
“
u
s
er
n
am
e
co
n
tain
i
n
g
a
n
u
m
b
er
,
”
“
f
u
ll
n
am
e
co
n
t
ain
in
g
a
n
u
m
b
er
,
”
“
f
u
ll
n
am
e
len
g
t
h
,
”
“
is
p
r
iv
ate
,
”
“
is
jo
in
ed
r
ec
en
tly
,
”
“
h
as
a
ch
an
n
el,
”
“
is
a
b
u
s
in
ess
ac
co
u
n
t,
”
“
h
as
g
u
id
es,
”
an
d
“
h
as
an
ex
ter
n
al
UR
L
”.
T
h
r
o
u
g
h
ca
lcu
latin
g
th
e
r
at
io
o
f
f
o
llo
wer
s
to
ac
co
u
n
ts
f
o
llo
wed
,
p
o
ten
tial
im
b
alan
ce
s
in
f
o
llo
wer
-
f
o
llo
win
g
r
elatio
n
s
h
ip
s
ca
n
b
e
d
e
tecte
d
,
aid
in
g
in
th
e
id
en
tific
atio
n
o
f
s
u
s
p
icio
u
s
ac
co
u
n
ts
.
Ad
d
itio
n
ally
,
tech
n
i
q
u
es
lik
e
J
ac
ca
r
d
s
im
ilar
ity
[
1
2
]
ar
e
e
m
p
lo
y
e
d
to
ass
ess
s
i
m
ilar
ities
b
etwe
en
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
Mu
lti p
la
tfo
r
ms fa
ke
a
cc
o
u
n
ts
d
etec
tio
n
b
a
s
ed
o
n
fed
era
te
d
l
ea
r
n
in
g
(
Ma
r
in
a
A
z
er
)
3841
u
s
er
n
am
es
an
d
f
u
ll
n
a
m
es,
with
n
o
tab
le
d
if
f
er
en
ce
s
p
o
t
en
tially
s
ig
n
alin
g
ac
co
u
n
ts
war
r
an
tin
g
f
u
r
th
e
r
in
v
esti
g
atio
n
.
A
co
m
p
o
s
ite
f
e
atu
r
e
is
g
en
er
ate
d
,
am
alg
a
m
a
tin
g
“is
p
r
iv
ate,
”
“
d
escr
ip
tio
n
len
g
th
,
”
a
n
d
“
h
as
ex
ter
n
a
l
UR
L
,
”
o
f
f
er
in
g
in
s
ig
h
ts
in
to
th
e
co
m
p
leten
ess
o
f
b
io
in
f
o
r
m
atio
n
,
wh
e
r
e
m
u
ltip
le
lin
k
s
in
th
e
d
escr
ip
tio
n
co
u
ld
r
aise
s
u
s
p
icio
n
.
Fu
r
th
er
an
aly
s
is
in
v
o
lv
es
ca
lcu
latin
g
in
ter
ac
tio
n
r
atio
s
,
n
u
m
er
ic
r
atio
s
b
etwe
en
n
am
es
an
d
u
s
er
n
am
e
s
,
ex
am
in
in
g
n
o
n
-
alp
h
a
n
u
m
er
ic
ch
ar
ac
ter
s
,
n
am
e
an
d
u
s
er
n
am
e
len
g
th
r
atio
s
,
an
d
g
u
id
es
-
to
-
p
o
s
ts
r
atio
s
.
E
l
ev
ated
r
atio
s
m
ay
s
ig
n
i
f
y
s
u
s
p
icio
u
s
b
eh
a
v
io
r
.
Mo
r
e
o
v
er
,
f
ea
tu
r
es
in
d
icatin
g
u
s
er
n
am
e
s
im
ilar
ity
to
v
e
r
if
ied
ac
co
u
n
ts
an
d
th
e
p
r
esen
ce
o
f
ass
o
ciate
d
ex
ter
n
al
c
h
an
n
e
ls
o
r
p
latf
o
r
m
s
ar
e
cr
ea
ted
.
Fak
e
ac
co
u
n
ts
ar
e
less
lik
ely
to
b
e
co
n
n
ec
ted
to
e
x
ter
n
al
ch
an
n
els,
o
f
ten
p
r
io
r
itizi
n
g
s
p
am
p
r
o
m
o
tio
n
,
m
alicio
u
s
ac
tiv
ities
,
o
r
u
s
er
i
m
p
er
s
o
n
atio
n
.
T
h
r
o
u
g
h
c
o
m
p
r
eh
en
s
iv
e
f
ea
tu
r
e
en
g
in
ee
r
in
g
,
th
ese
p
ar
am
ete
r
s
co
llectiv
ely
co
n
tr
ib
u
te
to
th
e
r
o
b
u
s
t id
en
tific
atio
n
a
n
d
c
h
ar
a
cter
izatio
n
o
f
f
r
au
d
u
len
t p
r
o
f
il
es.
‒
I
n
s
tag
r
am
b
o
t
/
h
u
m
an
d
ataset
T
h
e
d
ataset
en
co
m
p
ass
es
1
6
d
is
tin
ct
f
ea
tu
r
es
aim
ed
at
ca
p
t
u
r
in
g
v
ar
io
u
s
asp
ec
ts
o
f
u
s
er
an
d
m
ed
ia
b
eh
av
io
r
o
n
th
e
p
latf
o
r
m
.
T
h
ese
f
ea
tu
r
es
in
clu
d
e
m
et
r
ics
s
u
ch
as
“
m
ed
ia
-
lik
e
n
u
m
b
er
s
,
”
“
m
ed
ia
co
m
m
en
t
n
u
m
b
er
s
,
”
“
m
ed
ia
c
o
m
m
e
n
ts
d
is
ab
led
,
”
“
m
ed
ia
h
ash
tag
n
u
m
b
er
s
,
”
“
m
ed
ia
u
p
lo
a
d
ti
m
es,
”
“
m
ed
ia
h
as
lo
ca
tio
n
in
f
o
,
”
“
u
s
er
f
o
llo
wer
co
u
n
t,
”
“
u
s
er
f
o
llo
win
g
co
u
n
t,
”
“
u
s
er
h
as
h
ig
h
r
ee
ls
,
”
“
u
s
er
h
as
ex
ter
n
al
UR
L
,
”
“
u
s
er
tag
s
co
u
n
t,
”
“
u
s
er
b
io
g
r
ap
h
y
len
g
th
,
”
“
u
s
er
n
a
m
e
len
g
th
,
”
“
u
s
er
n
am
e
d
i
g
it
co
u
n
t,
”
“
au
t
o
m
ated
b
eh
av
io
r
,
”
an
d
“
u
s
er
m
ed
ia
c
o
u
n
t
”.
T
h
ese
f
ea
tu
r
es
s
er
v
e
as
th
e
f
o
u
n
d
atio
n
f
o
r
co
m
p
u
tin
g
ess
en
tial
s
tatis
t
ics
s
u
ch
as
av
er
ag
e
lik
es
p
er
m
ed
ia,
av
e
r
ag
e
c
o
m
m
en
ts
p
er
m
ed
ia,
an
d
th
e
r
atio
o
f
lik
es
to
co
m
m
e
n
ts
.
Ad
d
itio
n
ally
,
t
h
e
d
ataset
f
ac
ilit
ates
th
e
an
aly
s
is
o
f
p
o
s
tin
g
ti
m
e
v
ar
ia
b
ilit
y
ac
r
o
s
s
all
m
ed
i
a,
th
e
f
r
eq
u
en
cy
o
f
m
ed
ia
u
p
lo
ad
s
o
v
er
a
s
p
ec
if
ie
d
p
er
io
d
,
an
d
th
e
d
iv
er
s
ity
o
f
h
ash
tag
s
an
d
lo
ca
tio
n
s
.
Fu
r
th
er
m
o
r
e,
k
e
y
r
atio
s
s
u
ch
as
f
o
llo
wer
-
to
-
lik
e
r
ati
o
an
d
f
o
llo
wer
-
to
-
co
m
m
en
t
r
atio
ca
n
b
e
ca
lcu
lated
,
s
h
e
d
d
in
g
lig
h
t
o
n
th
e
en
g
ag
em
e
n
t
d
y
n
am
ics
an
d
au
d
ien
ce
in
ter
ac
tio
n
s
ass
o
ciate
d
with
ea
ch
u
s
er
.
B
y
lev
er
ag
in
g
th
ese
co
m
p
r
eh
e
n
s
iv
e
f
ea
tu
r
es,
we
c
an
g
ain
v
alu
ab
le
in
s
ig
h
ts
in
t
o
u
s
er
b
eh
av
io
r
p
atter
n
s
an
d
m
ed
ia
en
g
a
g
em
en
t
tr
en
d
s
,
u
ltima
tely
en
a
b
lin
g
m
o
r
e
in
f
o
r
m
ed
d
ec
is
io
n
-
m
ak
in
g
an
d
an
aly
s
is
o
n
t
h
e
p
latf
o
r
m
.
‒
T
witter
f
ak
e
/
r
ea
l d
ataset
T
witter
d
ata
co
n
s
is
ts
o
f
3
4
f
e
atu
r
es,
wh
ich
ar
e
(
p
r
o
f
ile
lin
k
co
lo
r
,
p
r
o
f
ile
b
ac
k
g
r
o
u
n
d
co
l
o
r
,
p
r
o
f
ile
s
id
eb
ar
f
ill
co
l
o
r
,
p
r
o
f
ile
b
ac
k
g
r
o
u
n
d
titl
e,
p
r
o
f
ile
b
an
n
er
U
R
L
,
p
r
o
f
ile
te
x
t
co
lo
r
,
u
n
iv
er
s
al
tim
e
co
o
r
d
in
ate
d
(
UT
C
)
o
f
f
s
et,
d
ef
au
lt
p
r
o
f
ile
i
m
ag
e,
d
ef
au
lt
p
r
o
f
ile,
g
eo
-
en
a
b
led
,
lis
ted
co
u
n
t,
f
av
o
r
ites
co
u
n
t,
f
r
ien
d
s
co
u
n
t,
f
o
llo
wer
s
co
u
n
t,
s
tatu
s
es
co
u
n
t,
p
r
o
f
ile
b
ac
k
g
r
o
u
n
d
im
ag
e
UR
L
h
ttp
s
,
p
r
o
f
ile
s
id
eb
ar
b
o
r
d
er
c
o
lo
r
,
s
cr
ee
n
n
am
e,
p
r
o
tecte
d
,
v
e
r
if
i
ed
,
d
escr
ip
tio
n
,
u
p
d
ate
d
,
d
ata
s
et,
cr
ea
ted
at,
UR
L
,
lan
g
,
tim
e
zo
n
e,
lo
ca
tio
n
,
p
r
o
f
ile
im
a
g
e
UR
L
,
n
am
e,
id
,
p
r
o
f
ile
im
a
g
e
UR
L
h
ttp
s
,
p
r
o
f
ile
b
ac
k
g
r
o
u
n
d
im
ag
e
U
R
L
an
d
p
r
o
f
ile
u
s
e
b
ac
k
g
r
o
u
n
d
)
.
b
ec
a
u
s
e
o
f
th
e
h
ig
h
n
u
m
b
er
o
f
f
ea
tu
r
es,
ess
en
tial
f
ea
tu
r
es
s
h
o
u
ld
b
e
s
elec
ted
.
B
y
u
tili
zin
g
v
ar
io
u
s
b
ase
m
o
d
els
s
u
ch
as
l
o
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
,
r
an
d
o
m
f
o
r
est
(
R
F),
SVM,
g
r
ad
ie
n
t
b
o
o
s
tin
g
(
GB
)
,
an
d
ar
tific
ial
n
eu
r
al
n
etwo
r
k
(
AN
N)
o
n
th
e
d
ataset
to
ex
tr
ac
t f
ea
tu
r
e
im
p
o
r
tan
ce
s
co
r
es.
E
ac
h
m
o
d
el
will
ev
alu
ate
an
d
ass
ig
n
s
co
r
es
to
in
d
iv
id
u
al
f
ea
tu
r
es
b
ased
o
n
its
p
er
ce
p
tio
n
o
f
th
eir
s
ig
n
if
ican
ce
in
id
en
tify
i
n
g
f
a
k
e
p
r
o
f
iles
.
T
h
is
p
r
o
ce
s
s
allo
ws
f
o
r
a
co
m
p
r
eh
e
n
s
iv
e
u
n
d
e
r
s
tan
d
in
g
o
f
f
ea
tu
r
e
r
elev
an
ce
ac
r
o
s
s
d
if
f
er
e
n
t
m
o
d
elin
g
ap
p
r
o
ac
h
es.
T
h
e
ty
p
ical
r
an
g
e
f
o
r
im
p
o
r
tan
ce
s
co
r
es
lies
b
etwe
en
0
an
d
1
,
wh
er
e
0
s
ig
n
if
ies
f
ea
tu
r
es
o
f
lo
w
o
r
n
e
g
lig
ib
le
im
p
o
r
tan
ce
,
wh
ile
1
d
e
n
o
te
s
f
ea
tu
r
es
o
f
h
ig
h
s
ig
n
if
ican
ce
in
d
etec
tin
g
f
ak
e
p
r
o
f
iles
.
T
h
is
s
tan
d
ar
d
ize
d
s
ca
le
f
ac
ilit
ates
th
e
in
ter
p
r
etatio
n
o
f
f
ea
tu
r
e
im
p
o
r
tan
ce
ac
r
o
s
s
d
if
f
er
en
t
m
o
d
els
an
d
d
atasets
b
ased
o
n
th
e
ag
g
r
e
g
ated
im
p
o
r
tan
ce
s
co
r
es
o
b
tai
n
ed
f
r
o
m
v
ar
io
u
s
m
o
d
els,
th
e
to
p
twen
t
y
s
elec
ted
f
ea
tu
r
es
[
1
3
]
as
th
e
y
ar
e
lik
ely
to
o
f
f
er
th
e
m
o
s
t
d
is
cr
im
in
ativ
e
p
o
wer
in
d
is
tin
g
u
is
h
in
g
b
etwe
en
f
a
k
e
an
d
g
en
u
in
e
p
r
o
f
iles
ar
e
id
e
n
tifie
d
as
f
o
llo
ws:
h
ig
h
im
p
o
r
tan
ce
(
s
co
r
e
:
0
.
6
-
1
)
p
r
o
f
ile
lin
k
co
lo
r
,
p
r
o
f
ile
b
ac
k
g
r
o
u
n
d
co
lo
r
,
p
r
o
f
ile
s
id
eb
ar
co
lo
r
,
p
r
o
f
ile
b
an
n
er
UR
L
,
p
r
o
f
ile
tex
t
co
lo
r
,
d
ef
au
lt
p
r
o
f
ile
im
ag
e
,
lis
ted
c
o
u
n
t,
f
r
ien
d
s
co
u
n
t,
f
o
llo
wer
s
co
u
n
t,
s
tatu
s
es
co
u
n
t,
s
cr
ee
n
n
am
e,
d
escr
ip
tio
n
.
Me
d
iu
m
im
p
o
r
ta
n
ce
(
s
co
r
e:
0
.
3
-
0
.
6
)
:
g
e
o
-
en
a
b
led
,
p
r
o
tec
ted
s
tatu
s
,
v
er
if
ied
s
tatu
s
,
u
p
d
ated
tim
estam
p
,
cr
ea
ted
at
tim
estam
p
,
UR
L
,
lan
g
u
ag
e
.
L
o
w
im
p
o
r
tan
ce
(
s
co
r
e:
0
.
0
-
0
.
3
)
l
o
ca
tio
n
.
Ho
wev
er
,
to
f
ac
ilit
ate
f
air
c
o
m
p
ar
is
o
n
an
d
ag
g
r
eg
atio
n
o
f
f
ea
tu
r
e
im
p
o
r
tan
ce
ac
r
o
s
s
m
o
d
els,
it
is
ess
en
tial to
n
o
r
m
alize
th
e
o
b
tain
ed
im
p
o
r
tan
ce
s
co
r
es.
No
r
m
alizin
g
th
e
im
p
o
r
tan
ce
s
co
r
es o
b
tain
ed
f
r
o
m
ea
ch
m
o
d
el
to
en
s
u
r
e
co
n
s
is
ten
cy
i
n
s
ca
le
ac
r
o
s
s
all
f
ea
t
u
r
es.
T
e
ch
n
iq
u
es
lik
e
m
in
-
m
ax
s
ca
lin
g
ca
n
b
e
a
p
p
lied
to
r
escale
th
e
s
co
r
es
to
a
co
m
m
o
n
r
an
g
e,
f
ac
ilit
atin
g
f
air
co
m
p
ar
is
o
n
an
d
ag
g
r
eg
atio
n
o
f
f
ea
tu
r
e
im
p
o
r
tan
ce
ac
r
o
s
s
m
o
d
els.
Min
-
m
ax
s
ca
lin
g
is
a
tech
n
iq
u
e
u
s
ed
to
n
o
r
m
alize
d
ata
b
y
tr
a
n
s
f
o
r
m
i
n
g
it
to
a
co
m
m
o
n
s
ca
le.
T
h
is
m
eth
o
d
r
escales
th
e
v
al
u
es
o
f
a
f
ea
t
u
r
e
t
o
a
s
p
ec
if
ie
d
r
an
g
e,
ty
p
ically
b
etwe
en
0
an
d
1
.
T
h
e
p
r
o
ce
s
s
in
v
o
lv
es
s
u
b
tr
ac
ti
n
g
t
h
e
m
i
n
im
u
m
v
alu
e
o
f
t
h
e
f
ea
t
u
r
e
f
r
o
m
ea
ch
d
ata
p
o
i
n
t
an
d
th
e
n
d
iv
id
in
g
it
b
y
th
e
d
if
f
er
en
ce
b
etwe
en
th
e
m
ax
i
m
u
m
an
d
m
in
im
u
m
v
alu
es.
As
a
r
esu
lt,
th
e
im
p
o
r
ta
n
ce
s
co
r
es
o
b
tain
ed
f
r
o
m
d
if
f
er
en
t
m
o
d
els
ar
e
ad
ju
s
ted
to
en
s
u
r
e
co
n
s
is
ten
cy
in
s
ca
le
ac
r
o
s
s
all
f
ea
tu
r
es.
T
h
is
n
o
r
m
aliza
tio
n
f
ac
ilit
ates
f
air
co
m
p
ar
is
o
n
a
n
d
a
g
g
r
e
g
a
tio
n
o
f
f
ea
tu
r
e
im
p
o
r
tan
ce
a
cr
o
s
s
m
o
d
els,
m
ak
i
n
g
it
ea
s
ier
to
in
te
r
p
r
et
a
n
d
an
aly
ze
th
e
r
elativ
e
im
p
o
r
tan
c
e
o
f
ea
ch
f
ea
tu
r
e
[
1
4
]
.
‒
Face
b
o
o
k
f
a
k
e
/
r
ea
l d
ataset
F
o
l
l
o
w
i
n
g
d
a
t
a
p
r
e
p
r
o
c
e
s
s
i
n
g
,
t
h
e
d
a
t
a
s
e
t
i
s
r
e
f
i
n
e
d
t
o
i
n
c
l
u
d
e
e
s
s
e
n
t
i
a
l
f
e
a
t
u
r
e
s
c
r
u
c
i
a
l
f
o
r
s
u
b
s
e
q
u
e
n
t
a
n
a
l
y
s
i
s
.
T
h
e
s
e
f
e
a
t
u
r
e
s
e
n
c
o
m
p
a
s
s
v
a
r
i
o
u
s
a
s
p
e
c
t
s
s
u
c
h
a
s
u
s
e
r
a
t
t
r
i
b
u
t
e
s
a
n
d
p
r
o
f
i
l
e
c
h
a
r
a
c
t
e
r
i
s
t
i
c
s
,
i
n
c
l
u
d
i
n
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.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
:
3
8
3
7
-
3
8
4
8
3842
'
i
d
,
'
'
n
a
m
e
,
'
'
s
c
r
e
e
n
_
n
a
m
e
,
'
'
s
t
a
t
u
s
e
s
_
c
o
u
n
t
,
'
'
f
o
l
l
o
w
e
r
s
_
c
o
u
n
t
,
'
'
f
r
i
e
n
d
s
_
c
o
u
n
t
,
'
'
f
a
v
o
u
r
i
t
e
s
_
c
o
u
n
t
,
'
'
l
i
s
t
e
d
_
c
o
u
n
t
,
'
'
c
r
e
a
t
e
d
_
a
t
,
'
'
U
R
L
,
'
'
l
a
n
g
,
'
'
l
o
c
a
t
i
o
n
,
'
'
d
e
f
a
u
l
t
_
p
r
o
f
i
l
e
,
'
'
p
r
o
f
i
l
e
_
i
m
a
g
e
_
u
r
l
,
'
'
p
r
o
f
i
l
e
_
b
a
n
n
e
r
_
u
r
l
,
'
'
p
r
o
f
i
l
e
_
u
s
e
_
b
a
c
k
g
r
o
u
n
d
_
i
m
a
g
e
,
”
p
r
o
f
i
l
e
_
b
a
c
k
g
r
o
u
n
d
_
i
m
a
g
e
_
u
r
l
_
h
t
t
p
s
,
'
'
p
r
o
f
i
l
e
_
t
e
x
t
_
c
o
l
o
r
,
”
p
r
o
f
i
l
e
_
i
m
a
g
e
_
u
r
l
_
h
t
t
p
s
,
”
p
r
o
f
i
l
e
_
s
i
d
e
b
a
r
_
b
o
r
d
e
r
_
c
o
l
o
r
,
'
'
p
r
o
f
i
l
e
_
s
i
d
e
b
a
r
_
f
i
l
l
_
c
o
l
o
r
,
'
'
p
r
o
f
i
l
e
_
b
a
c
k
g
r
o
u
n
d
_
i
m
a
g
e
_
u
r
l
,
'
'
p
r
o
f
i
l
e
_
b
a
c
k
g
r
o
u
n
d
_
c
o
l
o
r
,
'
'
p
r
o
f
i
l
e
_
l
i
n
k
_
c
o
l
o
r
,
'
'
d
e
s
c
r
i
p
t
i
o
n
,
'
'
u
p
d
a
t
e
d
,
'
a
n
d
'
d
a
t
a
s
e
t
.
'
F
r
o
m
t
h
i
s
c
o
m
p
r
e
h
e
n
s
i
v
e
p
o
o
l
o
f
f
e
a
t
u
r
e
s
,
t
h
e
t
o
p
t
w
e
n
t
y
a
r
e
c
a
r
e
f
u
l
l
y
s
e
l
e
c
t
e
d
b
a
s
e
d
o
n
t
h
e
i
r
s
i
g
n
i
f
i
c
a
n
c
e
f
o
r
s
u
b
s
e
q
u
e
n
t
a
n
a
l
y
s
i
s
.
T
h
e
s
e
i
n
c
l
u
d
e
'
s
t
a
t
u
s
c
o
u
n
t
,
'
'
f
o
l
l
o
w
e
r
s
c
o
u
n
t
,
'
'
f
r
i
e
n
d
s
c
o
u
n
t
,
'
'
f
a
v
o
r
i
t
e
s
c
o
u
n
t
,
'
'
l
i
s
t
e
d
c
o
u
n
t
,
'
'
c
r
e
a
t
e
d
a
t
,
'
'
UR
L
,
'
'
l
a
n
g
,
'
'
l
o
c
a
t
i
o
n
,
'
'
d
e
f
a
u
l
t
p
r
o
f
i
l
e
,
'
'
d
e
s
c
r
i
p
t
i
o
n
,
'
'
p
r
o
f
i
l
e
i
m
a
g
e
U
R
L
,
'
'
p
r
o
f
i
l
e
b
a
n
n
e
r
U
R
L
,
'
'
p
r
o
f
i
l
e
u
s
e
a
b
a
c
k
g
r
o
u
n
d
i
m
a
g
e
,
'
'
p
r
o
f
i
l
e
b
a
c
k
g
r
o
u
n
d
i
m
a
g
e
U
R
L
h
t
t
p
s
,
'
'
p
r
o
f
i
l
e
t
e
x
t
c
o
l
o
r
,
'
'
p
r
o
f
i
l
e
i
m
a
g
e
U
R
L
h
t
t
p
s
,
'
'
p
r
o
f
i
l
e
s
i
d
e
b
a
r
b
o
r
d
e
r
c
o
l
o
r
,
'
'
p
r
o
f
i
l
e
s
i
d
e
b
a
r
f
i
l
l
c
o
l
o
r
,
'
a
n
d
'
p
r
o
f
i
l
e
l
i
n
k
c
o
l
o
r
.
'
T
h
e
s
e
l
e
c
t
i
o
n
p
r
o
c
e
s
s
e
m
p
l
o
y
s
t
h
e
k
b
e
s
t
m
e
t
h
o
d
t
e
c
h
n
i
q
u
e
,
u
t
i
l
i
z
i
n
g
t
h
e
A
N
O
V
A
F
-
s
t
a
t
i
s
t
i
c
(
f
_
c
l
a
s
s
i
f
)
s
c
o
r
i
n
g
f
u
n
c
t
i
o
n
.
T
h
i
s
t
e
c
h
n
i
q
u
e
m
e
a
s
u
r
e
s
t
h
e
r
a
t
i
o
o
f
v
a
r
i
a
n
c
e
s
b
e
t
w
e
e
n
c
l
a
s
s
e
s
t
o
t
h
e
v
a
r
i
a
n
c
e
w
i
t
h
i
n
c
l
a
s
s
e
s
,
m
a
k
i
n
g
i
t
s
u
i
t
a
b
l
e
f
o
r
a
s
s
e
s
s
i
n
g
t
h
e
s
t
a
t
i
s
t
i
c
a
l
s
i
g
n
i
f
i
c
a
n
c
e
o
f
n
u
m
e
r
i
c
a
l
f
e
a
t
u
r
e
s
c
o
n
c
e
r
n
i
n
g
a
c
a
t
e
g
o
r
i
c
a
l
t
a
r
g
e
t
v
a
r
i
a
b
l
e
.
F
e
a
t
u
r
e
s
w
i
t
h
h
i
g
h
e
r
F
-
s
t
a
t
i
s
t
i
c
s
c
o
r
e
s
a
r
e
p
r
i
o
r
i
t
i
z
e
d
a
s
t
h
e
y
o
f
f
e
r
g
r
e
a
t
e
r
i
n
f
o
r
m
a
t
i
v
e
n
e
s
s
f
o
r
c
l
a
s
s
i
f
i
c
a
t
i
o
n
t
a
s
k
s
.
2
.
3
.
Cla
s
s
if
ica
t
io
n a
pp
ro
a
ch
es
T
wo
d
is
ti
n
ct
a
p
p
r
o
a
ch
es
h
a
v
e
em
e
r
g
e
d
as
p
r
o
m
i
n
e
n
t
s
o
l
u
tio
n
s
i
n
t
ac
k
l
in
g
c
o
m
p
le
x
c
la
s
s
if
i
ca
t
io
n
task
s
:
m
a
c
h
i
n
e
le
ar
n
i
n
g
a
p
p
r
o
ac
h
an
d
f
e
d
e
r
a
te
d
le
ar
n
i
n
g
a
p
p
r
o
ac
h
.
m
a
c
h
i
n
e
le
ar
n
i
n
g
m
et
h
o
d
s
[
1
5
]
e
n
c
o
m
p
ass
a
w
i
d
e
s
p
ec
tr
u
m
o
f
est
a
b
lis
h
ed
t
ec
h
n
i
q
u
es
s
u
c
h
as
d
ec
is
i
o
n
tr
ee
s
(
DT
)
,
RF
,
SVM
,
LR
,
an
d
AN
N
.
T
h
ese
m
et
h
o
d
s
h
a
v
e
d
e
m
o
n
s
tr
ate
d
t
h
ei
r
p
r
o
wess
i
n
d
i
v
e
r
s
e
d
o
m
ain
s
,
l
e
v
e
r
a
g
i
n
g
h
is
t
o
r
ica
l
d
at
a
p
att
er
n
s
t
o
m
a
k
e
p
r
e
d
ic
ti
o
n
s
o
r
c
lass
i
f
y
d
a
ta
i
n
t
o
p
r
e
d
e
f
i
n
e
d
c
ate
g
o
r
ies
.
I
n
co
n
tr
as
t,
f
e
d
e
r
at
e
d
l
ea
r
n
i
n
g
is
a
m
o
r
e
r
ec
e
n
t
p
a
r
a
d
i
g
m
wit
h
a
p
r
iv
ac
y
-
c
en
tr
ic
a
n
d
d
ec
e
n
t
r
a
liz
e
d
a
p
p
r
o
ac
h
.
I
t
e
n
a
b
les
m
o
d
e
l
tr
ai
n
i
n
g
a
c
r
o
s
s
m
u
lti
p
l
e
d
is
t
r
i
b
u
te
d
d
e
v
i
ce
s
o
r
s
e
r
v
er
s
wit
h
o
u
t
s
h
ar
i
n
g
r
aw
d
a
ta
.
I
n
s
t
ea
d
,
o
n
l
y
m
o
d
el
u
p
d
ates
ar
e
e
x
c
h
an
g
e
d
,
p
r
es
er
v
i
n
g
d
at
a
p
r
iv
ac
y
.
T
h
es
e
t
wo
a
p
p
r
o
a
ch
e
s
o
f
f
er
u
n
i
q
u
e
ad
v
an
ta
g
es
a
n
d
t
r
a
d
e
-
o
f
f
s
,
a
ll
o
wi
n
g
p
r
ac
t
iti
o
n
e
r
s
t
o
c
h
o
o
s
e
t
h
e
m
o
s
t s
u
it
a
b
le
m
et
h
o
d
b
as
ed
o
n
t
h
ei
r
s
p
ec
i
f
i
c
u
s
e
c
ases
,
d
ata
a
v
ai
la
b
il
it
y
,
a
n
d
p
r
i
v
a
cy
r
e
q
u
ir
e
m
e
n
ts
.
2
.
3
.
1
.
M
a
chine le
a
rning
I
n
th
e
p
u
r
s
u
it
o
f
ac
h
iev
in
g
s
u
p
er
io
r
p
r
ed
ictiv
e
p
er
f
o
r
m
an
ce
,
a
co
m
p
r
eh
en
s
iv
e
s
tr
ateg
y
o
f
ten
in
v
o
lv
es
em
p
lo
y
in
g
a
s
u
ite
o
f
tr
ad
itio
n
al
m
ac
h
in
e
lear
n
in
g
m
o
d
els,
i
n
clu
d
in
g
DT
th
at
p
a
r
titi
o
n
th
e
f
ea
tu
r
e
s
p
ac
e
b
ased
o
n
d
ec
is
io
n
r
u
les,
m
ak
in
g
th
e
m
ea
s
y
to
in
ter
p
r
et.
T
h
ey
ca
n
h
an
d
le
b
o
th
n
u
m
er
ical
an
d
c
ateg
o
r
ical
d
ata
an
d
ar
e
r
o
b
u
s
t
to
o
u
tlier
s
,
L
R
m
o
d
els
th
e
p
r
o
b
ab
ilit
y
o
f
a
b
in
a
r
y
o
u
tco
m
e
an
d
p
r
o
v
id
es
in
ter
p
r
etab
le
r
esu
lts
in
ter
m
s
o
f
o
d
d
s
r
atio
s
.
I
t
is
ef
f
i
cien
t
f
o
r
b
in
ar
y
class
if
icatio
n
task
s
an
d
wo
r
k
s
well
w
ith
lar
g
e
d
atasets
,
R
F
ar
e
an
en
s
em
b
le
o
f
DT
th
at
ar
e
r
o
b
u
s
t
to
o
v
er
f
itti
n
g
.
T
h
e
y
h
an
d
l
e
h
ig
h
-
d
im
en
s
io
n
al
d
ata
well
an
d
p
r
o
v
id
e
f
ea
tu
r
e
im
p
o
r
tan
ce
s
co
r
es,
m
a
k
in
g
t
h
e
m
u
s
ef
u
l f
o
r
f
ea
tu
r
e
s
elec
tio
n
,
ANN
in
s
p
ir
ed
b
y
th
e
h
u
m
an
b
r
ain
,
ANN
ca
p
tu
r
e
co
m
p
lex
r
elatio
n
s
h
ip
s
in
d
ata.
T
h
ey
ar
e
e
f
f
ec
tiv
e
f
o
r
lar
g
e
-
s
ca
le
p
r
o
b
lem
s
.
N
aiv
e
B
ay
es
(
NB
)
is
a
p
r
o
b
a
b
ilis
tic
clas
s
if
ier
b
ased
o
n
th
e
a
s
s
u
m
p
tio
n
o
f
co
n
d
itio
n
al
in
d
e
p
en
d
en
ce
b
etwe
en
f
ea
tu
r
es.
I
t
is
s
im
p
le,
p
er
f
o
r
m
s
well
with
s
m
all
tr
ain
in
g
d
ata,
an
d
is
p
ar
ticu
lar
ly
ef
f
ec
tiv
e
in
tex
t
class
if
icatio
n
task
s
,
SVM
th
at
f
in
d
th
e
h
y
p
er
p
lan
e
th
at
b
est
s
ep
ar
ates
clas
s
es
in
th
e
f
ea
tu
r
e
s
p
ac
e.
T
h
ey
ar
e
ef
f
ec
tiv
e
in
h
ig
h
-
d
im
en
s
io
n
al
s
p
ac
es,
wo
r
k
with
lin
ea
r
an
d
n
o
n
-
lin
ea
r
d
ata,
an
d
ar
e
r
o
b
u
s
t
to
o
v
er
f
itti
n
g
,
an
d
GB
b
u
ild
s
an
en
s
em
b
le
o
f
s
eq
u
en
tially
tr
ain
ed
DT
,
lea
d
in
g
to
h
ig
h
ly
ac
cu
r
ate
p
r
e
d
ictio
n
s
.
I
t
h
an
d
les
h
eter
o
g
en
e
o
u
s
d
ata
well
an
d
c
an
b
e
u
s
ed
f
o
r
b
o
th
class
if
icatio
n
an
d
r
eg
r
ess
io
n
task
s
.
T
h
e
p
r
o
c
e
s
s
e
x
t
e
n
d
s
b
e
y
o
n
d
m
e
r
e
m
o
d
e
l
s
e
l
e
ct
i
o
n
a
s
it
d
e
lv
e
s
i
n
t
o
e
n
s
e
m
b
le
l
e
a
r
n
i
n
g
.
T
h
i
s
a
p
p
r
o
a
c
h
e
n
t
a
i
ls
c
h
o
o
s
i
n
g
t
h
e
m
o
s
t
e
f
f
ec
t
i
v
e
c
o
m
b
i
n
a
ti
o
n
o
f
m
o
d
e
l
s
,
s
u
c
h
a
s
L
R
,
R
F
,
SV
M
,
GB
,
an
d
A
N
N
,
a
n
d
u
s
i
n
g
s
t
a
c
k
i
n
g
as
t
h
e
e
n
s
e
m
b
l
e
te
c
h
n
i
q
u
e
.
W
h
e
n
c
o
m
b
i
n
e
d
i
n
a
s
t
a
ck
i
n
g
e
n
s
e
m
b
l
e
,
t
h
es
e
m
o
d
e
ls
o
f
f
e
r
a
d
i
v
e
r
s
e
s
e
t
o
f
s
t
r
e
n
g
t
h
s
t
h
a
t
c
a
n
c
o
ll
e
c
ti
v
e
l
y
c
o
n
t
r
i
b
u
t
e
t
o
s
u
p
e
r
i
o
r
p
r
e
d
ic
ti
v
e
p
e
r
f
o
r
m
a
n
c
e
.
L
R
p
r
o
v
i
d
e
s
i
n
t
e
r
p
r
e
t
a
b
il
i
t
y
,
R
F
a
n
d
S
V
M
o
f
f
e
r
r
o
b
u
s
t
n
e
s
s
a
n
d
f
e
a
t
u
r
e
h
a
n
d
l
i
n
g
c
a
p
a
b
i
l
it
i
es
,
GB
p
r
o
v
i
d
es
a
c
c
u
r
a
c
y
t
h
r
o
u
g
h
s
e
q
u
e
n
t
i
a
l
t
r
ee
b
o
o
s
t
i
n
g
,
a
n
d
A
N
N
a
d
d
s
t
h
e
c
a
p
a
b
i
l
i
t
y
t
o
c
a
p
t
u
r
e
c
o
m
p
l
e
x
r
e
l
a
t
i
o
n
s
h
i
p
s
.
B
y
l
e
v
e
r
a
g
i
n
g
t
h
e
s
t
r
e
n
g
t
h
s
o
f
e
a
c
h
m
o
d
e
l
,
t
h
e
s
ta
c
k
i
n
g
e
n
s
e
m
b
le
ca
n
m
i
t
i
g
a
t
e
i
n
d
i
v
i
d
u
a
l
we
a
k
n
e
s
s
e
s
a
n
d
a
c
h
i
e
v
e
e
n
h
a
n
c
e
d
p
r
e
d
ic
t
i
v
e
p
o
w
e
r
.
E
ac
h
b
ase
m
o
d
el
is
in
d
iv
id
u
a
lly
tr
ain
ed
o
n
th
e
t
r
ain
in
g
d
at
a
in
s
tack
in
g
,
an
d
th
ei
r
p
r
ed
i
ctio
n
s
ar
e
co
llected
f
o
r
th
e
v
alid
atio
n
d
ataset.
T
h
ese
m
o
d
el
p
r
e
d
ictio
n
s
,
alo
n
g
with
t
h
e
o
r
ig
in
al
f
ea
tu
r
es
an
d
tar
g
et
v
ar
iab
le,
a
r
e
f
e
d
in
to
a
m
eta
-
lear
n
er
,
t
y
p
ically
a
g
r
a
d
ien
t
-
b
o
o
s
tin
g
m
o
d
el.
T
h
e
g
o
al
is
to
h
ar
n
ess
th
e
d
iv
er
s
e
s
tr
en
g
th
s
o
f
ea
ch
b
ase
m
o
d
el
to
en
h
an
ce
p
r
ed
ictiv
e
ac
cu
r
a
cy
an
d
g
e
n
er
aliza
tio
n
,
r
esu
lti
n
g
in
a
r
o
b
u
s
t
an
d
p
o
wer
f
u
l e
n
s
em
b
le
ap
p
r
o
ac
h
th
at
o
f
ten
o
u
tp
er
f
o
r
m
s
th
e
i
n
d
i
v
id
u
al
m
o
d
els [
1
6
]
,
[
1
7
]
.
Ag
g
r
eg
atio
n
tech
n
iq
u
es
:
Ag
g
r
eg
atio
n
tech
n
iq
u
es
ar
e
m
eth
o
d
s
u
s
ed
to
c
o
m
b
in
e
th
e
p
r
ed
ictio
n
s
o
f
m
u
ltip
le
b
ase
m
o
d
els
to
en
h
an
ce
p
r
e
d
ictiv
e
ac
cu
r
ac
y
.
On
e
co
m
m
o
n
ty
p
e
o
f
ag
g
r
e
g
atio
n
tech
n
iq
u
e
is
s
im
p
le
av
er
ag
in
g
,
wh
er
e
th
e
p
r
ed
ictio
n
s
o
f
all
b
ase
m
o
d
els
ar
e
av
er
a
g
ed
t
o
o
b
tain
th
e
f
in
al
p
r
e
d
ictio
n
.
W
eig
h
te
d
av
er
ag
in
g
is
an
o
th
e
r
ap
p
r
o
ac
h
,
wh
er
e
ea
c
h
b
ase
m
o
d
el
’
s
p
r
ed
ictio
n
is
weig
h
ted
d
if
f
er
en
tly
b
ef
o
r
e
a
v
er
ag
i
n
g
,
allo
win
g
f
o
r
m
o
r
e
in
f
lu
en
ce
f
r
o
m
ce
r
tain
m
o
d
els.
Ho
wev
er
,
a
m
o
r
e
s
o
p
h
is
ticated
ag
g
r
e
g
atio
n
tech
n
i
q
u
e
k
n
o
wn
as
s
tack
in
g
th
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
Mu
lti p
la
tfo
r
ms fa
ke
a
cc
o
u
n
ts
d
etec
tio
n
b
a
s
ed
o
n
fed
era
te
d
l
ea
r
n
in
g
(
Ma
r
in
a
A
z
er
)
3843
p
r
o
p
o
s
ed
en
s
em
b
le
m
eth
o
d
in
o
u
r
wo
r
k
h
as
g
ain
ed
p
o
p
u
lar
it
y
in
r
ec
en
t
y
ea
r
s
[
1
8
]
.
Stack
in
g
in
v
o
lv
es
tr
ain
in
g
m
u
ltip
le
b
ase
m
o
d
els
an
d
th
e
n
co
m
b
in
in
g
th
eir
p
r
e
d
ictio
n
s
u
s
in
g
a
m
eta
-
m
o
d
el.
Un
lik
e
s
im
p
le
av
er
ag
in
g
,
s
tack
in
g
in
tr
o
d
u
ce
s
a
n
a
d
d
itio
n
al
lay
er
o
f
lear
n
in
g
wh
er
e
th
e
m
eta
-
m
o
d
el
lear
n
s
to
co
m
b
in
e
th
e
p
r
ed
ictio
n
s
o
f
th
e
b
ase
m
o
d
els.
T
h
is
ap
p
r
o
a
ch
allo
ws
f
o
r
lev
er
ag
i
n
g
th
e
d
iv
er
s
e
p
er
s
p
ec
tiv
es
p
r
o
v
id
e
d
b
y
th
e
b
ase
m
o
d
els,
wh
ich
m
ay
u
s
e
d
if
f
er
en
t
alg
o
r
ith
m
s
o
r
f
ea
tu
r
e
s
u
b
s
ets.
B
y
ca
p
italizin
g
o
n
th
e
co
m
p
lem
en
tar
y
s
tr
en
g
th
s
o
f
in
d
iv
id
u
al
m
o
d
els,
s
tack
in
g
h
as
th
e
p
o
ten
tial
to
s
ig
n
if
ican
tly
im
p
r
o
v
e
p
r
ed
ictiv
e
ac
cu
r
a
cy
.
Ho
wev
er
,
it
is
im
p
o
r
tan
t
to
n
o
te
th
at
s
tack
in
g
m
ay
r
eq
u
ir
e
a
d
d
itio
n
al
co
m
p
u
tatio
n
al
r
eso
u
r
ce
s
an
d
ca
r
ef
u
l
tu
n
in
g
o
f
m
eta
-
p
ar
am
eter
s
to
ac
h
iev
e
o
p
tim
al
p
er
f
o
r
m
a
n
ce
.
2
.
3
.
2
.
F
eder
a
t
ed
lea
rning
I
n
th
e
co
n
te
x
t
o
f
task
s
in
v
o
lv
in
g
u
s
er
p
e
r
s
o
n
al
d
ata,
f
ed
er
ated
lea
r
n
in
g
h
as
em
e
r
g
ed
as
a
g
r
o
u
n
d
b
r
ea
k
in
g
ap
p
r
o
ac
h
t
h
at
m
ar
r
ies
th
e
p
o
wer
o
f
m
a
ch
in
e
lear
n
in
g
with
d
ata
p
r
i
v
ac
y
p
r
eser
v
atio
n
.
Fed
er
ated
lear
n
in
g
allo
ws
f
o
r
th
e
co
llab
o
r
ativ
e
tr
ain
in
g
o
f
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els
ac
r
o
s
s
m
u
ltip
le
d
ev
ices
o
r
s
er
v
er
s
wh
ile
k
ee
p
in
g
s
en
s
itiv
e
u
s
er
d
ata
d
ec
en
tr
alize
d
a
n
d
s
ec
u
r
e.
I
n
s
u
ch
task
s
,
wh
e
r
e
p
r
iv
ac
y
an
d
d
ata
p
r
o
tectio
n
ar
e
p
ar
am
o
u
n
t
co
n
ce
r
n
s
,
f
e
d
er
ated
lear
n
in
g
s
h
in
es
b
y
en
s
u
r
in
g
th
at
u
s
er
p
er
s
o
n
al
d
ata
r
em
ain
s
o
n
th
e
lo
ca
l
d
ev
ices
o
r
s
er
v
er
s
,
n
ev
er
leav
in
g
th
e
u
s
er
’
s
co
n
t
r
o
l.
I
n
s
tead
o
f
tr
a
n
s
m
itti
n
g
r
a
w
d
ata
to
a
ce
n
tr
al
s
er
v
er
f
o
r
m
o
d
el
tr
ain
in
g
,
o
n
ly
m
o
d
el
u
p
d
ates
ar
e
s
h
ar
ed
,
g
r
ea
tly
r
ed
u
cin
g
th
e
r
is
k
o
f
d
ata
b
r
ea
c
h
es
o
r
p
r
iv
ac
y
v
io
latio
n
s
.
T
h
is
in
n
o
v
ativ
e
tech
n
iq
u
e
allo
ws
f
o
r
d
ev
elo
p
in
g
h
ig
h
ly
p
er
s
o
n
alize
d
a
n
d
ef
f
ec
tiv
e
m
o
d
els
with
o
u
t
co
m
p
r
o
m
is
in
g
th
e
co
n
f
id
en
tiality
o
f
u
s
er
in
f
o
r
m
atio
n
.
Fed
er
ate
d
lear
n
i
n
g
th
u
s
r
ep
r
esen
ts
a
s
ig
n
if
ican
t
s
tep
f
o
r
war
d
in
ad
d
r
ess
in
g
p
r
i
v
ac
y
a
n
d
s
ec
u
r
ity
ch
allen
g
es
w
h
ile
u
tili
zin
g
m
ac
h
in
e
lear
n
in
g
to
d
eliv
er
cu
s
to
m
ized
an
d
ac
cu
r
ate
r
esu
lts
in
task
s
r
elian
t
o
n
u
s
er
s
’
p
er
s
o
n
al
d
ata.
Div
id
in
g
d
ata
in
to
d
if
f
er
e
n
t
v
ir
tu
al
clien
ts
d
ep
en
d
s
o
n
th
e
s
ize
o
f
th
e
d
ata.
in
I
n
s
tag
r
a
m
d
atasets
,
we
u
s
ed
two
clien
ts
o
n
ly
b
ec
a
u
s
e
o
f
th
e
s
m
all
d
ata
s
ize,
wh
ile
in
T
witter
an
d
Face
b
o
o
k
d
atasets
,
we
u
s
ed
th
r
ee
clien
ts
,
wh
ich
r
ef
er
to
th
r
ee
s
u
b
s
ets
o
f
d
ata.
T
h
e
s
tr
u
ctu
r
e
o
f
f
ed
er
ate
d
lear
n
in
g
is
s
h
o
wn
in
Fig
u
r
e
2
[
1
9
]
,
[
2
0
]
.
Fig
u
r
e
2
.
Fed
er
ate
d
lear
n
in
g
s
tr
u
ctu
r
e
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
is
s
tu
d
y
d
el
v
ed
i
n
to
th
e
im
p
ac
t
o
f
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
es
o
n
d
etec
tin
g
f
a
k
e
p
r
o
f
iles
ac
r
o
s
s
m
u
ltip
le
p
latf
o
r
m
s
,
wh
ile
ac
k
n
o
wled
g
i
n
g
th
at
m
o
s
t
o
f
p
r
e
v
io
u
s
s
tu
d
ies
p
r
im
ar
ily
f
o
cu
s
ed
o
n
ass
ess
in
g
m
eth
o
d
s
o
n
in
d
iv
id
u
al
p
latf
o
r
m
s
an
d
o
f
te
n
o
v
er
lo
o
k
ed
d
ata
p
r
iv
ac
y
c
o
n
ce
r
n
s
in
th
is
r
ea
lm
.
T
h
e
e
x
p
er
im
en
ts
en
tailed
an
aly
zin
g
th
e
p
er
f
o
r
m
a
n
ce
o
f
b
o
th
m
ac
h
in
e
lear
n
in
g
an
d
f
ed
er
ated
lear
n
in
g
ap
p
r
o
ac
h
es
ac
r
o
s
s
f
o
u
r
d
atasets
f
r
o
m
th
r
ee
p
latf
o
r
m
s
.
Sev
en
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
wer
e
r
ig
o
r
o
u
s
ly
ev
alu
ated
,
in
clu
d
in
g
DT
,
LR
,
RF
,
SVM
,
NB
,
GB
,
an
d
AN
N
.
Per
f
o
r
m
an
ce
m
etr
ics
s
u
ch
as
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
wer
e
u
tili
ze
d
to
g
a
u
g
e
p
r
ed
ictiv
e
ca
p
ab
ilit
ies
o
n
p
latf
o
r
m
-
s
p
ec
if
ic
d
atasets
,
r
ev
ea
lin
g
in
d
iv
i
d
u
al
s
tr
en
g
th
s
a
n
d
wea
k
n
ess
es.
Su
b
s
eq
u
en
tly
,
th
e
s
tack
in
g
m
o
d
el,
a
n
en
s
em
b
le
tech
n
iq
u
e
co
m
b
in
in
g
b
ase
m
o
d
el
p
r
e
d
ictio
n
s
with
o
r
ig
in
al
f
ea
tu
r
es
a
n
d
ta
r
g
et
v
ar
iab
les,
was
in
tr
o
d
u
ce
d
an
d
its
im
p
ac
t
o
n
o
v
e
r
all
p
er
f
o
r
m
an
ce
d
is
cu
s
s
ed
.
T
h
e
d
is
cu
s
s
io
n
d
elv
ed
in
to
n
u
an
ce
s
with
in
ea
ch
d
ataset,
s
h
o
wca
s
in
g
wh
er
e
s
p
ec
if
ic
alg
o
r
ith
m
s
ex
ce
lled
an
d
h
o
w
s
tack
in
g
e
n
h
an
ce
d
p
r
ed
ic
tiv
e
ca
p
ab
ilit
ies.
T
h
e
s
tack
in
g
m
o
d
el
ag
g
r
eg
ates
p
r
ed
ictio
n
s
f
r
o
m
th
e
f
iv
e
m
o
s
t
ac
cu
r
ate
b
ase
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els,
lev
er
a
g
in
g
th
e
co
m
p
lem
en
tar
y
s
tr
en
g
t
h
s
o
f
in
d
iv
id
u
al
m
o
d
els
to
p
o
ten
tially
en
h
a
n
ce
p
e
r
f
o
r
m
an
ce
co
m
p
ar
e
d
to
an
y
s
in
g
le
m
o
d
el.
W
e
f
in
d
th
at
alth
o
u
g
h
th
er
e
ar
e
s
ev
er
al
en
s
em
b
le
m
eth
o
d
s
ex
is
t
in
m
ac
h
in
e
lear
n
in
g
,
s
u
ch
as
b
ag
g
in
g
an
d
b
o
o
s
tin
g
,
b
u
t
s
tack
in
g
h
as
ad
v
a
n
tag
es
o
v
er
t
h
ese
m
eth
o
d
s
.
T
h
ese
in
clu
d
e
m
o
d
el
f
lex
ib
ilit
y
,
im
p
r
o
v
e
d
p
er
f
o
r
m
an
ce
,
an
d
th
e
ab
ilit
y
to
ca
p
tu
r
e
co
m
p
l
ex
d
ata
r
elatio
n
s
h
ip
s
in
a
n
o
n
-
lin
ea
r
m
an
n
er
.
Fu
r
th
er
m
o
r
e
,
th
e
s
elec
tio
n
o
f
p
r
o
f
icien
t
b
ase
lear
n
er
s
s
ig
n
if
ican
tly
in
f
lu
en
ce
s
th
e
o
v
er
all
p
er
f
o
r
m
an
ce
o
f
th
e
s
tack
ed
m
o
d
el,
p
r
o
m
p
tin
g
ex
t
en
s
iv
e
ex
p
er
im
e
n
tatio
n
in
th
is
p
h
ase.
Ad
d
itio
n
ally
,
f
e
d
er
ate
d
lear
n
in
g
em
er
g
ed
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.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
:
3
8
3
7
-
3
8
4
8
3844
as
a
p
r
o
m
is
in
g
ap
p
r
o
ac
h
,
p
ar
t
icu
lar
ly
in
s
ce
n
ar
io
s
in
v
o
l
v
in
g
p
er
s
o
n
al
an
d
s
en
s
itiv
e
d
ata,
as
it
p
r
eser
v
es
d
ata
p
r
iv
ac
y
w
h
ile
s
till
ac
h
iev
in
g
co
m
m
en
d
ab
le
r
esu
lts
.
Ho
wev
er
,
it
is
wo
r
th
n
o
ti
n
g
th
at
f
ed
er
ated
lear
n
in
g
ty
p
ically
r
eq
u
i
r
es a
h
ig
h
e
r
ex
e
cu
tio
n
tim
e
co
m
p
ar
ed
to
tr
ad
it
io
n
al
m
ac
h
in
e
lea
r
n
in
g
alg
o
r
it
h
m
s
.
Ou
r
ex
p
er
im
en
ts
f
i
n
d
in
g
s
u
n
d
er
s
co
r
e
th
e
ef
f
icac
y
o
f
th
e
s
tack
in
g
m
eth
o
d
wh
en
lev
er
a
g
in
g
ef
f
ec
tiv
e
b
ase
lear
n
er
s
f
r
o
m
m
ac
h
in
e
le
ar
n
in
g
alg
o
r
ith
m
s
in
tack
lin
g
class
if
icatio
n
ch
allen
g
es,
as
d
em
o
n
s
tr
ated
in
o
u
r
s
tu
d
y
f
o
cu
s
ed
o
n
id
en
tify
i
n
g
f
ak
e
ac
co
u
n
ts
ac
r
o
s
s
v
ar
io
u
s
s
o
cial
p
latf
o
r
m
s
u
s
in
g
f
o
u
r
d
is
tin
ct
d
atasets
.
Ov
er
co
m
in
g
d
iv
er
s
e
ch
alle
n
g
es,
r
an
g
in
g
f
r
o
m
d
ata
b
ias
t
o
f
ea
tu
r
e
s
elec
tio
n
,
v
alu
ab
le
less
o
n
s
h
av
e
b
ee
n
lear
n
ed
f
r
o
m
th
e
m
is
tak
es
en
co
u
n
ter
e
d
.
B
iases
in
tr
ain
in
g
d
ata
an
d
in
co
m
p
lete
f
ea
tu
r
e
en
g
in
ee
r
in
g
ca
n
s
ig
n
if
ican
tly
im
p
ac
t
m
o
d
el
p
e
r
f
o
r
m
a
n
ce
.
T
o
ad
d
r
ess
b
iases
,
r
esear
ch
er
s
m
u
s
t
b
e
v
ig
ilan
t
an
d
m
itig
ate
t
h
em
d
u
r
in
g
p
r
ep
r
o
ce
s
s
in
g
.
Ad
d
itio
n
ally
,
p
r
io
r
itizin
g
co
m
p
r
eh
e
n
s
iv
e
f
ea
tu
r
e
en
g
in
ee
r
i
n
g
is
cr
u
cial
to
ca
p
tu
r
e
r
elev
an
t
in
f
o
r
m
atio
n
ef
f
ec
tiv
e
ly
.
I
m
p
lem
e
n
tin
g
c
o
r
r
ec
tiv
e
m
ea
s
u
r
es
b
ased
o
n
th
ese
less
o
n
s
aim
s
to
en
h
an
ce
m
o
d
el
r
o
b
u
s
tn
ess
an
d
r
eliab
ili
ty
in
r
ea
l
-
wo
r
l
d
s
ce
n
ar
io
s
.
T
h
i
s
ap
p
r
o
ac
h
is
ex
ac
tly
wh
at
we
ad
o
p
t in
o
u
r
s
tu
d
y
.
3
.
1
.
E
v
a
lua
t
i
o
n m
et
rics
W
e
p
r
esen
t
th
e
m
etr
ics
u
s
ed
to
ev
alu
ate
t
h
e
r
esu
lts
to
s
elec
t
th
e
b
est
-
s
u
p
er
v
is
ed
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
ith
m
.
W
e
s
h
o
w
th
e
m
o
d
e
l a
cc
u
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
a
n
d
F_
Sco
r
e,
w
h
ich
ar
e
ca
lc
u
lated
as [
2
1
]
,
[
2
2
]
:
−
A
cc
u
r
ac
y
:
ac
cu
r
ac
y
m
ea
s
u
r
es
th
e
p
r
o
p
o
r
tio
n
o
f
co
r
r
ec
tl
y
p
r
ed
icted
in
s
tan
ce
s
o
u
t
o
f
th
e
to
tal
n
u
m
b
er
o
f
in
s
tan
ce
s
in
th
e
d
ataset
as d
escr
ib
ed
in
(
1
)
.
+
+
+
+
(
1
)
W
h
er
e
,
tr
u
e
p
o
s
itiv
es
(
T
P)
ar
e
th
e
n
u
m
b
er
o
f
co
r
r
ec
tly
p
r
e
d
icted
p
o
s
itiv
e
in
s
tan
ce
s
,
tr
u
e
n
eg
ativ
es
(
T
N)
ar
e
th
e
n
u
m
b
er
o
f
co
r
r
ec
tly
p
r
ed
icted
n
e
g
ativ
e
in
s
tan
ce
s
,
f
al
s
e
p
o
s
itiv
es
(
FP
)
ar
e
th
e
n
u
m
b
er
o
f
in
s
tan
ce
s
in
co
r
r
ec
tly
p
r
ed
icted
as
p
o
s
itiv
e
,
an
d
f
alse
n
eg
ativ
es
(
FN)
is
th
e
n
u
m
b
er
o
f
in
s
tan
ce
s
in
co
r
r
ec
tly
p
r
ed
icted
as n
eg
ativ
e.
−
Pre
cisi
o
n
:
p
r
ec
is
io
n
q
u
a
n
tifie
s
th
e
ac
cu
r
ac
y
o
f
p
o
s
itiv
e
p
r
e
d
ictio
n
s
ca
lcu
lated
as d
escr
ib
ed
in
(
2
)
.
+
(
2
)
−
R
ec
all
(
s
en
s
itiv
ity
)
:
r
ec
all
m
ea
s
u
r
es
th
e
m
o
d
el'
s
ab
ilit
y
to
id
en
tify
all
p
o
s
itiv
e
in
s
tan
ce
s
co
r
r
ec
tly
.
ca
lcu
lated
as d
escr
ib
ed
in
(
3
)
.
+
(
3
)
−
F
-
s
co
r
e
(
F1
_
Sco
r
e
)
:
th
e
F1
s
c
o
r
e
is
th
e
h
ar
m
o
n
ic
m
ea
n
o
f
p
r
ec
is
io
n
an
d
r
ec
all
an
d
p
r
o
v
i
d
es
a
b
alan
ce
d
m
ea
s
u
r
e
o
f
th
e
m
o
d
el
’
s
p
er
f
o
r
m
an
ce
as sh
o
wn
in
(
4
)
.
2
∗
∗
+
(
4
)
T
h
ese
f
o
r
m
u
las
o
f
f
er
p
r
ec
is
e
ca
lcu
latio
n
s
f
o
r
ea
c
h
m
etr
ic
,
f
ac
ilit
atin
g
a
co
m
p
r
eh
en
s
iv
e
ev
alu
atio
n
o
f
th
e
m
o
d
el
’
s
p
er
f
o
r
m
a
n
ce
in
class
if
icatio
n
task
s
.
3
.
2
.
Co
m
pa
ra
t
iv
e
a
na
ly
s
is
o
f
re
s
ults
T
h
e
ex
p
e
r
im
en
ts
with
I
n
s
tag
r
am
d
atasets
ar
e
d
ep
icted
as
f
o
llo
ws:
T
ab
le
1
illu
s
tr
ates
th
e
m
ac
h
in
e
lear
n
in
g
r
esu
lts
o
b
tain
e
d
f
o
r
Data
s
et
1
,
wh
ile
T
ab
le
2
s
h
o
w
ca
s
es
th
e
o
u
tco
m
es
f
o
r
Data
s
et
2
.
I
t
is
n
o
tewo
r
th
y
th
at
th
e
s
tack
in
g
m
o
d
el
wh
i
ch
co
m
b
in
es
p
r
ed
ictio
n
s
f
r
o
m
th
e
b
est
f
iv
e
b
ase
m
o
d
els,
o
u
tp
er
f
o
r
m
s
ea
ch
in
d
iv
id
u
al
m
o
d
el.
T
h
is
en
h
an
ce
m
en
t
in
p
er
f
o
r
m
an
ce
h
ig
h
lig
h
ts
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
s
tack
in
g
a
p
p
r
o
ac
h
in
lev
er
ag
in
g
t
h
e
co
llectiv
e
s
tr
en
g
th
s
o
f
d
iv
e
r
s
e
m
o
d
els to
ac
h
iev
e
s
u
p
er
io
r
p
r
e
d
ictiv
e
ac
cu
r
a
cy
.
T
a
b
l
e
3
d
is
p
l
a
y
s
t
h
e
r
es
u
l
ts
o
f
o
u
r
p
r
o
p
o
s
e
d
m
o
d
e
l
,
w
h
ic
h
s
u
r
p
a
s
s
es
t
h
e
a
p
p
r
o
a
c
h
o
u
tl
i
n
e
d
i
n
r
e
f
e
r
e
n
ce
[
6
]
.
T
h
e
a
u
t
h
o
r
s
o
f
t
h
a
t
s
t
u
d
y
ac
h
i
e
v
e
d
a
n
a
c
c
u
r
a
c
y
o
f
9
2
.
9
%
b
y
e
m
p
l
o
y
i
n
g
t
h
e
R
F
a
l
g
o
r
i
t
h
m
a
n
d
o
v
e
r
s
a
m
p
l
i
n
g
t
h
e
d
a
t
a
u
s
i
n
g
t
h
e
S
MO
T
E
-
NC
al
g
o
r
i
t
h
m
.
A
d
d
i
t
i
o
n
a
ll
y
,
o
u
r
a
p
p
r
o
a
c
h
o
u
t
p
e
r
f
o
r
m
s
t
h
e
r
esu
l
t
s
r
e
p
o
r
t
e
d
i
n
[
8
]
,
w
h
e
r
e
a
n
a
c
c
u
r
a
c
y
o
f
8
2
%
w
a
s
a
c
h
i
e
v
e
d
u
s
i
n
g
t
h
e
K
NN
a
l
g
o
r
i
t
h
m
o
n
D
a
t
a
s
e
t
1
.
T
ab
le
4
p
r
esen
ts
a
co
m
p
ar
ativ
e
an
aly
s
is
u
s
in
g
Data
s
et
2
,
wh
er
ein
o
u
r
ap
p
r
o
ac
h
is
ju
x
tap
o
s
ed
with
th
e
m
eth
o
d
o
u
tlin
e
d
in
r
ef
er
en
ce
[
2
3
]
.
T
h
e
r
ef
er
e
n
c
e
s
tu
d
y
u
tili
ze
d
v
a
r
io
u
s
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
an
d
ac
h
iev
ed
a
n
8
6
%
F
-
s
co
r
e
u
s
in
g
SVM
an
d
NN.
Ho
wev
er
,
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
attain
ed
s
u
p
er
io
r
p
er
f
o
r
m
a
n
ce
in
co
m
p
a
r
is
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
Mu
lti p
la
tfo
r
ms fa
ke
a
cc
o
u
n
ts
d
etec
tio
n
b
a
s
ed
o
n
fed
era
te
d
l
ea
r
n
in
g
(
Ma
r
in
a
A
z
er
)
3845
T
ab
le
1
.
Data
s
et
1
r
esu
lts
u
s
in
g
m
ac
h
in
e
lear
n
in
g
C
l
a
s
si
f
i
e
r
A
c
c
u
r
a
c
y
(
%)
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
F
-
sco
r
e
(
%)
S
V
M
9
2
.
4
9
0
.
3
91
8
8
.
3
RF
9
2
.
9
8
9
.
2
8
8
.
1
8
9
.
2
GB
9
3
90
90
92
LR
9
1
.
2
8
7
.
3
89
88
NB
8
9
.
1
8
5
.
2
8
6
.
2
8
6
.
1
DT
8
9
.
8
8
6
.
3
8
2
.
3
8
8
.
1
ANN
91
89
88
87
S
t
a
c
k
i
n
g
M
o
d
e
l
9
4
.
8
9
3
.
2
9
2
.
1
9
4
.
2
T
ab
le
2
.
Data
s
et
2
r
esu
lts
u
s
in
g
m
ac
h
in
e
lear
n
in
g
C
l
a
s
si
f
i
e
r
A
c
c
u
r
a
y
(
%)
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
F
sco
r
e
(
%)
S
V
M
9
1
.
3
9
1
.
3
82
8
6
.
3
RF
9
2
.
2
8
8
.
2
7
9
.
9
8
4
.
2
GB
9
2
.
8
9
0
.
3
9
1
.
2
8
8
.
3
LR
8
3
.
4
8
0
.
3
70
75
NB
8
6
.
2
8
5
.
2
6
8
.
2
7
8
.
1
DT
8
9
.
3
8
6
.
3
8
2
.
3
8
8
.
1
ANN
9
0
.
4
89
84
86
S
t
a
c
k
i
n
g
M
o
d
e
l
9
3
.
5
9
3
.
2
8
7
.
1
8
9
.
2
T
ab
le
3
.
C
o
m
p
a
r
in
g
p
r
o
p
o
s
ed
m
o
d
el
r
esu
lts
with
[
6
]
,
[
8
]
o
n
Data
s
et
1
Ref
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
-
sco
r
e
(
%)
[
6
]
R
F
w
i
t
h
S
M
O
T
E
-
NC
9
2
.
9
-
-
-
[
8
]
K
N
N
82
-
-
-
P
r
o
p
o
se
d
M
o
d
e
l
9
4
.
8
%
94
96
96
T
ab
le
4
.
C
o
m
p
a
r
in
g
p
r
o
p
o
s
ed
m
o
d
el
r
esu
lts
with
[
2
3
]
o
n
Data
s
et
2
Ref
M
o
d
e
l
F
-
sco
r
e
(
%)
[
2
3
]
S
V
M
,
N
B
,
N
N
,
LR
86
Th
i
s
w
o
r
k
P
r
o
p
o
se
d
M
o
d
e
l
89
T
ab
le
5
in
o
u
r
s
tu
d
y
p
r
esen
t
s
a
co
m
p
a
r
is
o
n
o
f
r
esu
lts
o
b
tain
ed
f
r
o
m
d
if
f
er
e
n
t
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
alo
n
g
s
id
e
th
e
p
er
f
o
r
m
an
ce
o
f
o
u
r
s
tack
in
g
m
o
d
el.
T
h
is
tab
le
o
f
f
e
r
s
v
alu
ab
l
e
in
s
ig
h
ts
in
to
th
e
ef
f
icac
y
o
f
en
s
em
b
le
tech
n
iq
u
es
in
en
h
an
cin
g
p
r
e
d
ictiv
e
ac
cu
r
ac
y
ac
r
o
s
s
v
ar
io
u
s
d
ata
s
ets
an
d
s
ce
n
ar
io
s
.
Fu
r
th
er
m
o
r
e
,
T
ab
le
6
p
r
o
v
id
es
an
an
aly
s
is
o
f
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
’
s
p
er
f
o
r
m
an
ce
co
m
p
ar
e
d
to
th
e
m
eth
o
d
o
l
o
g
ies
d
escr
ib
ed
in
r
ef
er
en
ce
s
[
2
4
]
,
[
2
5
]
.
Sp
ec
if
ically
f
o
cu
s
in
g
o
n
X
(
T
witter
)
Data
s
et
3
,
th
is
co
m
p
ar
is
o
n
h
i
g
h
lig
h
ts
th
e
co
m
p
etitiv
e
ad
v
an
tag
e
o
f
o
u
r
m
o
d
el
in
ad
d
r
ess
in
g
th
e
u
n
i
q
u
e
ch
allen
g
es
p
o
s
ed
b
y
th
is
d
ataset.
C
o
n
s
eq
u
en
tly
,
it
co
n
tr
ib
u
tes
s
ig
n
if
ican
tly
to
th
e
b
r
o
ad
er
co
m
p
r
eh
e
n
s
io
n
o
f
ef
f
ec
tiv
e
m
ac
h
in
e
-
lear
n
in
g
s
tr
ateg
ies f
o
r
th
e
g
iv
e
n
task
.
T
ab
le
5
.
Data
s
et
3
r
esu
lts
u
s
in
g
m
ac
h
in
e
lear
n
in
g
C
l
a
s
si
f
i
e
r
A
c
c
u
r
a
c
y
(
%)
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
F
-
s
c
o
r
e
(
%)
S
V
M
9
1
.
3
9
1
.
3
82
8
6
.
3
RF
9
8
.
2
8
8
.
2
7
9
.
9
8
4
.
2
GB
9
8
.
8
9
0
.
3
9
1
.
2
8
8
.
3
LR
9
2
.
4
8
0
.
3
70
75
NB
9
4
.
2
8
5
.
2
6
8
.
2
7
8
.
1
DT
9
1
.
3
8
6
.
3
8
2
.
3
8
8
.
1
ANN
9
6
.
4
89
84
86
S
t
a
c
k
i
n
g
m
o
d
e
l
9
9
.
4
9
2
.
2
8
7
.
1
8
9
.
2
T
ab
le
6
.
C
o
m
p
a
r
in
g
p
r
o
p
o
s
ed
m
o
d
el
r
esu
lts
with
[
2
4
]
,
[
2
5
]
o
n
Data
s
et
3
Ref
M
o
d
e
l
A
c
c
u
r
a
c
y
(
%)
[
2
4
]
RF
98
[
2
5
]
S
V
M
9
5
.
7
Th
i
s
w
o
r
k
P
r
o
p
o
se
d
m
o
d
e
l
9
9
.
4
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.
13
,
No
.
4
,
Dec
em
b
er
2
0
2
4
:
3
8
3
7
-
3
8
4
8
3846
T
ab
le
7
p
r
esen
ts
th
e
p
er
f
o
r
m
an
ce
r
esu
lts
o
f
v
ar
io
u
s
m
ac
h
i
n
e
lear
n
in
g
alg
o
r
ith
m
s
an
d
th
e
s
tack
in
g
m
o
d
el
wh
e
n
ap
p
lied
to
f
ac
e
b
o
o
k
d
ata
(
Data
s
et
4
)
.
T
h
is
tab
l
e
p
r
o
v
id
es
v
alu
a
b
le
in
s
ig
h
ts
in
to
th
e
ef
f
ec
tiv
en
ess
o
f
th
ese
ap
p
r
o
ac
h
es
in
tack
lin
g
th
e
u
n
i
q
u
e
ch
a
r
ac
ter
is
tics
o
f
Data
s
et
4
.
T
ab
le
8
co
m
p
ar
es
o
u
r
p
r
o
p
o
s
ed
m
o
d
e
l
an
d
th
e
m
o
d
el
p
r
esen
te
d
in
[
2
6
]
.
T
h
is
co
m
p
ar
is
o
n
is
s
p
ec
if
ically
ca
r
r
ied
o
u
t
o
n
Data
s
et
4
,
s
h
ed
d
in
g
lig
h
t
o
n
ea
ch
ap
p
r
o
ac
h
’
s
r
elativ
e
s
tr
en
g
th
s
an
d
wea
k
n
ess
es with
in
th
e
co
n
tex
t o
f
th
is
d
ataset.
T
ab
le
7
.
R
esu
lts
u
s
in
g
m
ac
h
in
e
lear
n
in
g
o
n
Data
s
et
4
C
l
a
s
si
f
i
e
r
A
c
c
u
r
a
c
y
(
%)
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
F
-
s
c
o
r
e
(
%)
S
V
M
99
8
9
.
5
9
2
.
3
9
2
.
3
RF
9
9
.
6
9
2
.
5
9
3
.
2
9
5
.
3
GB
8
7
.
3
8
8
.
3
8
5
.
4
8
8
.
4
LR
9
5
.
1
9
0
.
2
8
5
.
5
9
1
.
5
NB
9
0
.
4
8
7
.
5
8
9
.
4
9
0
.
3
DT
8
0
.
2
7
4
.
7
7
9
.
5
80
ANN
9
3
.
4
7
8
.
3
8
0
.
2
7
6
.
4
S
t
a
c
k
i
n
g
m
o
d
e
l
9
9
.
3
9
3
.
2
9
5
.
4
9
5
.
5
T
ab
le
8
.
C
o
m
p
a
r
in
g
th
e
p
r
o
p
o
s
ed
m
o
d
el
with
[
2
6
]
o
n
Data
s
et
4
Ref
M
o
d
e
l
A
c
c
u
r
a
c
y
(
%)
[
2
6
]
N
B
,
j
4
8
a
n
d
R
F
9
9
.
6
Th
i
s
w
o
r
k
P
r
o
p
o
se
d
M
o
d
e
l
9
9
.
8
I
n
th
e
co
m
p
ar
ativ
e
an
aly
s
is
o
f
f
e
d
er
ated
lear
n
in
g
a
n
d
th
e
s
tack
in
g
m
o
d
el
o
f
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
,
we
d
iv
id
e
d
d
ata
f
r
o
m
I
n
s
tag
r
am
,
T
witter
,
a
n
d
Face
b
o
o
k
in
to
v
ir
tu
al
r
eso
u
r
c
es
to
ac
co
m
m
o
d
at
e
d
if
f
er
en
t d
ataset
s
izes.
T
ab
le
9
ev
alu
ated
th
e
two
ap
p
r
o
ac
h
es
b
ased
o
n
ex
ec
u
tio
n
tim
e
an
d
ac
cu
r
ac
y
m
ea
s
u
r
es.
No
tab
ly
,
f
ed
er
ate
d
lear
n
in
g
co
n
s
u
m
ed
m
o
r
e
tim
e
th
a
n
th
e
s
tack
in
g
m
o
d
el
d
u
e
to
s
ev
er
al
f
ac
to
r
s
,
in
clu
d
in
g
th
e
n
ee
d
f
o
r
d
ata
ag
g
r
e
g
atio
n
ac
r
o
s
s
r
eso
u
r
ce
s
,
co
m
m
u
n
icatio
n
o
v
e
r
h
ea
d
f
o
r
m
o
d
el
u
p
d
ates,
an
d
co
m
p
lex
ities
ass
o
ciate
d
with
f
ed
er
ated
o
p
t
im
izatio
n
.
T
h
i
s
o
b
s
e
r
v
a
t
i
o
n
h
i
g
h
l
i
g
h
t
s
t
h
e
t
r
a
d
e
-
o
f
f
s
b
e
tw
e
en
d
a
t
a
p
r
i
v
a
c
y
,
t
h
e
s
t
r
e
n
g
t
h
o
f
f
e
d
e
r
a
t
e
d
l
e
a
r
n
i
n
g
,
a
n
d
i
t
s
c
o
m
p
u
t
a
t
i
o
n
a
ll
y
i
n
t
e
n
s
iv
e
n
a
t
u
r
e
.
T
h
e
c
h
o
i
c
e
b
e
t
w
e
e
n
t
h
e
t
w
o
a
p
p
r
o
a
c
h
e
s
s
h
o
u
l
d
b
e
b
a
s
e
d
o
n
t
h
e
s
p
ec
i
f
i
c
u
s
e
c
as
e
a
v
a
i
la
b
l
e
c
o
m
p
u
t
a
ti
o
n
a
l
r
e
s
o
u
r
c
e
s
a
n
d
t
h
e
n
e
e
d
f
o
r
s
e
c
u
r
i
t
y
.
T
ab
le
9
.
A
co
m
p
ar
is
o
n
b
etwe
en
f
ed
er
ate
d
lear
n
in
g
v
er
s
u
s
s
tack
in
g
m
o
d
el
D
a
t
a
s
e
t
F
e
d
e
r
a
t
e
d
l
e
a
r
n
i
n
g
Ex
e
c
u
t
i
o
n
t
i
m
e
F
e
d
e
r
a
t
e
d
l
e
a
r
n
i
n
g
_
a
c
c
u
r
a
c
y
(
%)
S
t
a
c
k
i
n
g
m
o
d
e
l
Ex
e
c
u
t
i
o
n
t
i
m
e
(
s)
S
t
a
c
k
i
n
g
_
m
o
d
e
l
_
a
c
c
u
r
a
c
y
(
%)
D
a
t
a
s
e
t
1
(
2
r
e
s
o
u
r
c
e
s)
1
.
3
s
96
0
.
8
9
4
.
8
D
a
t
a
s
e
t
2
(
2
r
e
s
o
u
r
c
e
s)
1
.
2
s
95
0
.
7
9
3
.
5
D
a
t
a
s
e
t
3
(
3
r
e
s
o
u
r
c
e
s)
2
.
3
s
98
1
.
4
9
9
.
4
D
a
t
a
s
e
t
4
(
3
r
e
s
o
u
r
c
e
s)
2
.
1
s
9
9
1
.
5
9
9
.
8
Ad
d
r
ess
in
g
lim
itatio
n
s
an
d
th
eir
p
o
ten
tial
im
p
ac
t
o
n
th
e
r
es
u
lts
is
cr
u
cial
f
o
r
en
s
u
r
in
g
th
e
cr
ed
ib
ilit
y
an
d
r
eliab
ilit
y
o
f
th
e
s
tu
d
y
f
in
d
in
g
s
.
W
h
ile
th
e
r
ec
o
m
m
en
d
e
d
ap
p
r
o
ac
h
es
in
d
e
ed
o
u
tp
er
f
o
r
m
o
th
er
m
eth
o
d
s
,
it's
cr
u
cial
to
ac
k
n
o
wled
g
e
th
e
lim
itatio
n
s
in
h
er
e
n
t
in
th
e
m
o
d
el.
On
e
s
u
c
h
lim
itatio
n
p
er
tain
s
to
g
en
er
aliza
b
ilit
y
.
W
h
ile
o
u
r
s
tu
d
y
d
em
o
n
s
tr
ates
th
e
ef
f
icac
y
o
f
th
e
s
tack
in
g
m
o
d
el
in
i
d
en
t
if
y
in
g
f
a
k
e
p
r
o
f
iles
ac
r
o
s
s
v
ar
io
u
s
s
o
cial
m
ed
ia
p
latf
o
r
m
s
,
it
i
s
es
s
en
tial to
r
ec
o
g
n
ize
p
o
ten
tial lim
itatio
n
s
in
its
g
en
er
aliza
b
ilit
y
to
o
th
er
d
o
m
ain
s
.
Dif
f
er
en
t
d
o
m
ain
s
p
o
s
s
ess
d
is
tin
ct
ch
ar
a
cter
is
tics
an
d
ch
allen
g
es
th
a
t
m
ig
h
t
af
f
ec
t
th
e
s
tack
in
g
m
o
d
el
’
s
p
er
f
o
r
m
an
c
e
d
if
f
er
e
n
tly
.
Mo
r
eo
v
e
r
,
th
e
p
er
f
o
r
m
an
ce
o
f
m
ac
h
in
e
lear
n
in
g
m
o
d
els,
in
cl
u
d
in
g
th
e
s
tack
in
g
m
o
d
el,
is
h
ig
h
ly
d
ep
en
d
e
n
t
o
n
th
e
q
u
ality
an
d
d
iv
er
s
ity
o
f
th
e
tr
ain
in
g
d
ata.
Ou
r
s
tu
d
y
u
tili
ze
d
d
atasets
s
p
ec
if
ic
to
I
n
s
tag
r
am
,
T
witter
,
an
d
Face
b
o
o
k
,
w
h
ich
m
ay
n
o
t
f
u
lly
ca
p
tu
r
e
th
e
d
iv
er
s
ity
o
f
f
a
k
e
p
r
o
f
iles
ac
r
o
s
s
all
s
o
cial
m
e
d
ia
p
latf
o
r
m
s
s
u
ch
as
L
in
k
e
d
I
n
,
Sn
ap
ch
at,
T
ik
T
o
k
,
R
ed
d
it,
an
d
Pin
ter
est.
Ad
d
itio
n
ally
,
b
iases
p
r
esen
t
in
th
e
tr
ain
in
g
d
ata
co
u
ld
lead
to
b
iased
p
r
ed
ictio
n
s
,
p
ar
ticu
lar
ly
in
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
wh
er
e
d
ata
m
a
y
b
e
in
co
m
p
lete
o
r
s
k
ewe
d
.
A
d
d
r
ess
in
g
th
ese
co
n
ce
r
n
s
is
cr
u
cial
f
o
r
e
n
s
u
r
in
g
th
e
r
eliab
ilit
y
an
d
ap
p
licab
ilit
y
o
f
th
e
s
tack
in
g
m
o
d
el
in
b
r
o
ad
er
co
n
tex
ts
.
F
o
r
f
u
t
u
r
e
r
es
e
a
r
c
h
,
f
u
r
t
h
e
r
r
e
f
i
n
e
m
e
n
t
o
f
t
h
e
s
t
a
c
k
i
n
g
m
o
d
e
l
w
a
r
r
a
n
ts
e
x
p
l
o
r
at
i
o
n
,
wh
i
c
h
c
o
u
l
d
i
n
v
o
l
v
e
i
n
v
e
s
t
i
g
at
i
n
g
v
a
r
i
o
u
s
e
n
s
e
m
b
l
e
t
ec
h
n
i
q
u
e
s
a
n
d
m
e
t
a
-
m
o
d
e
l
a
r
c
h
i
te
c
t
u
r
e
s
t
o
p
o
t
e
n
t
i
a
l
l
y
el
e
v
a
t
e
p
r
e
d
i
c
t
i
v
e
p
e
r
f
o
r
m
a
n
c
e
t
o
e
v
e
n
g
r
e
a
t
e
r
h
e
i
g
h
t
s
.
M
o
r
e
o
v
e
r
,
i
n
co
r
p
o
r
a
t
i
n
g
a
d
v
a
n
c
e
d
f
e
a
t
u
r
e
e
n
g
i
n
e
e
r
i
n
g
m
e
t
h
o
d
s
a
n
d
i
n
t
e
g
r
a
t
i
n
g
a
d
d
i
t
i
o
n
al
d
at
a
s
o
u
r
c
es
m
a
y
y
i
e
l
d
v
a
l
u
a
b
l
e
i
n
s
i
g
h
t
s
,
e
n
h
a
n
c
i
n
g
t
h
e
m
o
d
e
l
’
s
r
o
b
u
s
t
n
es
s
.
A
d
d
i
t
i
o
n
a
l
l
y
,
s
u
b
j
e
ct
i
n
g
t
h
e
d
e
v
e
l
o
p
e
d
f
r
a
m
e
w
o
r
k
t
o
r
e
a
l
-
t
im
e
d
a
t
a
s
t
r
e
a
m
s
a
n
d
a
s
s
es
s
i
n
g
i
t
s
p
e
r
f
o
r
m
a
n
c
e
i
n
d
y
n
a
m
i
c
e
n
v
i
r
o
n
m
e
n
t
s
c
o
u
l
d
p
r
o
v
i
d
e
i
n
v
a
l
u
a
b
l
e
i
n
s
i
g
h
t
s
i
n
t
o
i
ts
p
r
a
c
t
i
ca
l
v
i
a
b
i
l
it
y
a
n
d
s
c
a
l
a
b
i
l
it
y
.
L
a
s
t
l
y
,
Evaluation Warning : The document was created with Spire.PDF for Python.