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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
,
p
p
.
4
2
1
1
~
4
2
2
5
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
14
.i
5
.
p
p
4
2
1
1
-
4
2
2
5
4211
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Exploring
so
cia
l
media
sentime
nt
pa
tt
erns
for impr
o
v
ed
cy
berbully
ing
det
ection
W
a
e
l
M
.
S
.
Y
a
f
o
o
z
1
,
A
bd
ul
s
a
ma
d
E
b
r
a
h
i
m
Y
a
h
y
a
2
,
A
bd
u
ll
a
h
A
ls
a
e
ed
i
1
,
R
e
y
a
d
h
A
l
lu
h
a
i
b
i
3
,
F
a
is
a
l J
a
m
il
4
,
M
a
hm
o
ud
Sa
la
heldi
n E
ls
a
y
ed
2
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
C
o
l
l
e
g
e
o
f
C
o
mp
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
Ta
i
b
a
h
U
n
i
v
e
r
si
t
y
,
M
e
d
i
n
a
,
S
a
u
d
i
A
r
a
b
i
a
2
F
a
c
u
l
t
y
o
f
C
o
mp
u
t
i
n
g
a
n
d
I
n
f
o
r
mat
i
o
n
Te
c
h
n
o
l
o
g
y
,
N
o
r
t
h
e
r
n
B
o
r
d
e
r
U
n
i
v
e
r
si
t
y
,
R
a
f
h
a
,
S
a
u
d
i
A
r
a
b
i
a
3
D
e
p
a
r
t
me
n
t
o
f
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
a
n
d
D
a
t
a
S
c
i
e
n
c
e
,
C
o
l
l
e
g
e
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
Ta
i
b
a
h
U
n
i
v
e
r
s
i
t
y
,
M
e
d
i
n
a
,
S
a
u
d
i
A
r
a
b
i
a
4
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
S
c
h
o
o
l
o
f
C
o
m
p
u
t
i
n
g
,
E
n
g
i
n
e
e
r
i
n
g
a
n
d
I
n
t
e
l
l
i
g
e
n
t
S
y
s
t
e
ms
,
U
n
i
v
e
r
si
t
y
o
f
U
l
st
e
r
,
B
e
l
f
a
st
,
U
n
i
t
e
d
K
i
n
g
d
o
m
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 3
,
2
0
2
4
R
ev
is
ed
J
u
l 5
,
2
0
2
5
Acc
ep
ted
Au
g
6
,
2
0
2
5
Ca
se
s
o
f
o
n
li
n
e
b
u
ll
y
in
g
a
n
d
a
g
g
re
ss
iv
e
b
e
h
a
v
i
o
rs
d
irec
ted
a
t
so
c
ial
m
e
d
ia
u
se
rs
h
a
v
e
su
r
g
e
d
i
n
re
c
e
n
t
y
e
a
rs.
Th
e
se
b
e
h
a
v
i
o
rs
h
a
v
e
h
a
d
n
e
g
a
ti
v
e
imp
a
c
ts
o
n
v
ictims
fro
m
a
wid
e
r
a
n
g
e
o
f
d
e
m
o
g
ra
p
h
ic
g
ro
u
p
s.
Wh
il
e
e
ffo
rts
h
a
v
e
b
e
e
n
m
a
d
e
t
o
a
d
d
re
ss
p
e
rsiste
n
t
d
i
g
it
a
l
h
a
ra
ss
m
e
n
t,
th
e
e
x
p
e
c
ted
o
u
tco
m
e
h
a
s
b
e
e
n
li
m
it
e
d
d
u
e
to
th
e
lac
k
o
f
e
ffe
c
ti
v
e
t
o
o
ls
t
o
q
u
ick
l
y
id
e
n
ti
f
y
c
y
b
e
rb
u
ll
y
in
g
b
e
h
a
v
i
o
rs
a
n
d
c
e
n
so
r
t
h
e
m
a
c
c
o
rd
i
n
g
l
y
o
n
s
o
c
ial
m
e
d
ia
p
latfo
rm
s.
Th
is
stu
d
y
p
re
se
n
ts
a
sc
a
l
a
b
le
a
n
d
sy
ste
m
a
ti
c
m
e
th
o
d
to
d
e
tec
t
a
n
d
a
n
a
ly
z
e
o
ffe
n
siv
e
b
e
h
a
v
io
r
a
n
d
b
u
ll
y
i
n
g
o
n
Twi
tt
e
r
(n
o
w
k
n
o
w
n
a
s
X)
.
Ou
r
m
e
th
o
d
o
l
o
g
y
in
v
o
l
v
e
s
e
x
trac
ti
n
g
tex
t
u
a
l,
u
se
r
-
re
l
a
ted
,
a
n
d
n
e
two
rk
-
re
late
d
a
tt
ri
b
u
tes
to
u
n
d
e
rsta
n
d
th
e
traits
o
f
i
n
d
i
v
i
d
u
a
ls
i
n
v
o
lv
e
d
i
n
su
c
h
b
e
h
a
v
i
o
rs.
Th
is
a
p
p
r
o
a
c
h
a
ims
to
re
c
o
g
n
ize
d
isti
n
c
ti
v
e
c
h
a
r
a
c
teristics
th
a
t
se
t
th
e
m
a
p
a
rt
fr
o
m
re
g
u
lar
u
se
rs.
Th
is
stu
d
y
p
r
o
p
o
se
s
a
n
o
v
e
l
m
o
d
e
l
b
y
e
m
p
lo
y
i
n
g
a
n
i
n
teg
ra
ted
d
e
e
p
-
le
a
rn
in
g
m
o
d
e
l
,
c
o
m
b
i
n
i
n
g
th
e
b
i
d
irec
ti
o
n
a
l
g
a
ted
re
c
u
rre
n
t
u
n
it
(BiG
RU),
tran
sfo
rm
e
r
b
lo
c
k
,
a
n
d
c
o
n
v
o
lu
t
io
n
a
l
n
e
u
ra
l
n
e
two
rk
(CNN
).
T
h
is
m
o
d
e
l
a
ims
to
c
las
sify
X
c
o
m
m
e
n
ts
i
n
to
o
ffe
n
siv
e
a
n
d
non
-
o
ffe
n
si
v
e
c
a
teg
o
ries
.
Th
e
p
r
o
p
o
se
d
m
o
d
e
l’s
e
fficia
c
y
h
a
s b
e
e
n
e
v
a
lu
a
ted
th
ro
u
g
h
se
v
e
ra
l
e
x
p
e
rime
n
ts
b
y
c
o
m
b
in
i
n
g
t
h
re
e
wid
e
l
y
re
c
o
g
n
ize
d
d
a
tas
e
ts
o
f
h
a
te
s
p
e
e
c
h
.
T
h
e
p
r
o
p
o
se
d
m
o
d
e
l
a
c
h
iev
e
s
a
n
a
c
c
u
ra
c
y
ra
te
o
f
a
p
p
ro
x
ima
tely
9
8
.
9
5
%
,
s
h
o
wi
n
g
p
r
o
m
isin
g
re
su
lt
s
in
id
e
n
t
if
y
in
g
a
n
d
c
a
teg
o
rizin
g
o
ffe
n
si
v
e
b
e
h
a
v
io
r
i
n
c
y
b
e
r
b
u
ll
y
i
n
g
.
K
ey
w
o
r
d
s
:
C
y
b
er
b
u
lly
in
g
Dee
p
lear
n
in
g
Ma
ch
in
e
lear
n
in
g
So
cial
m
ed
ia
T
r
an
s
f
o
r
m
e
r
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
:
W
ae
l M
.
S
.
Yaf
o
o
z
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
C
o
lleg
e
o
f
C
o
m
p
u
ter
Sci
en
ce
an
d
E
n
g
in
ee
r
in
g
,
T
ai
b
ah
Un
iv
er
s
ity
Me
d
in
a
4
1
4
7
7
,
Sau
d
i A
r
ab
ia
E
m
ail: w
y
af
o
o
z@
taib
ah
u
.
e
d
u
.
s
a
1.
I
NT
RO
D
UCT
I
O
N
C
y
b
er
b
u
lly
in
g
is
a
s
ig
n
if
ican
t
th
r
ea
t
to
o
n
lin
e
u
s
er
s
.
I
t
p
ar
ticu
lar
ly
af
f
ec
ts
in
d
iv
id
u
als
wh
o
f
r
eq
u
e
n
tly
v
is
it
s
o
cial
m
ed
ia
p
latf
o
r
m
s
[
1
]
–
[
3
]
.
Un
lik
e
tr
ad
itio
n
al
b
u
lly
in
g
,
cy
b
er
b
u
lly
in
g
o
cc
u
r
s
elec
tr
o
n
ically
with
o
u
t
tem
p
o
r
al
o
r
s
p
atial
lim
ita
tio
n
s
[
4
]
–
[
7
]
.
Du
r
in
g
th
e
ea
r
ly
2
0
th
ce
n
tu
r
y
,
wh
en
s
o
cial
m
ed
ia
was in
its
n
ascen
t stag
e,
th
e
n
eg
ativ
e
im
p
ac
ts
o
f
cy
b
er
b
u
lly
in
g
wer
e
n
o
t f
u
lly
ac
k
n
o
wled
g
ed
.
Ho
wev
e
r
,
in
th
e
last
1
5
y
ea
r
s
,
ca
s
es
o
f
o
n
lin
e
h
ar
ass
m
en
t,
cy
b
er
s
talk
in
g
,
an
d
tr
o
llin
g
h
av
e
b
ee
n
o
n
t
h
e
r
is
e
[
8
]
.
I
n
2
0
1
7
,
4
1
%
o
f
US
r
esid
en
ts
ex
p
er
ien
ce
d
s
o
m
e
lev
el
o
f
o
n
lin
e
h
ar
a
s
s
m
en
t,
an
d
6
6
%
ex
p
er
ien
ce
d
s
o
m
e
ty
p
e
o
f
h
ate
s
p
ee
ch
,
wh
ich
was
d
ir
ec
ted
at
th
em
.
Ap
p
r
o
x
im
ately
5
0
%
o
f
y
o
u
n
g
p
eo
p
le
u
s
e
s
o
cial
m
ed
ia
to
d
ay
f
ac
e
cy
b
er
b
u
lly
in
g
o
n
p
latf
o
r
m
s
l
ik
e
T
witter
(
n
o
w
k
n
o
wn
as
X)
.
Desp
ite
attem
p
ts
to
tack
le
o
n
g
o
in
g
d
ig
ital
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
2
1
1
-
4
2
2
5
4212
h
ar
ass
m
en
t,
th
e
r
esu
lts
h
av
e
b
ee
n
co
n
s
tr
ain
ed
b
y
th
e
a
b
s
en
ce
o
f
ef
f
icien
t
to
o
ls
to
p
r
o
m
p
tly
id
e
n
tify
cy
b
er
b
u
lly
in
g
b
eh
a
v
io
r
s
an
d
t
ak
e
ap
p
r
o
p
r
iate
ce
n
s
o
r
s
h
ip
m
ea
s
u
r
es
[
9
]
,
[
1
0
]
.
T
h
er
e
h
as
b
ee
n
an
in
cr
ea
s
e
in
in
ter
est
in
d
etec
tin
g
cy
b
er
b
u
lly
in
g
in
t
h
e
f
ield
o
f
n
at
u
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
(
NL
P)
[
1
1
]
.
A
p
r
im
ar
y
g
o
al
o
f
cy
b
er
b
u
lly
in
g
d
etec
tio
n
is
to
p
r
e
-
p
r
o
ce
s
s
tex
tu
al
d
ata,
s
u
ch
as
twee
ts
clea
n
in
g
,
an
d
ex
tr
ac
t
r
elev
a
n
t
in
f
o
r
m
atio
n
s
o
th
at
m
ac
h
in
e
lear
n
in
g
(
ML
)
alg
o
r
ith
m
s
ca
n
b
e
u
s
ed
to
tr
ain
ed
m
o
d
els
to
u
n
d
er
s
tan
d
an
d
class
if
y
th
em
.
Ma
n
y
m
eth
o
d
s
ar
e
u
s
ed
to
s
im
p
lify
tex
t
r
ep
r
esen
tatio
n
an
d
tr
an
s
late
wo
r
d
s
in
to
n
u
m
er
ical
v
alu
es
in
te
x
t
class
if
icatio
n
/cate
g
o
r
izatio
n
,
s
u
ch
as
b
ag
o
f
wo
r
d
s
(
B
o
W
)
,
ter
m
f
r
eq
u
e
n
c
y
-
in
v
er
s
e
d
o
c
u
m
en
t
f
r
eq
u
e
n
cy
,
a
n
d
W
o
r
d
2
Vec
,
Gl
o
V,
an
d
f
ast
tex
t
.
T
h
ese
ty
p
es
o
f
n
u
m
er
ical
d
ata
ca
n
b
e
f
ed
in
to
ML
class
if
ier
s
s
u
ch
r
an
d
o
m
f
o
r
est
(
R
F),
d
e
cisi
o
n
tr
ee
(
DT
)
,
XGBo
o
s
t
(
XGB),
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
,
an
d
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
(
SVM)
,
an
d
r
ec
u
r
r
en
t
n
e
u
r
al
n
etwo
r
k
s
(
R
NNs)
[
1
2
]
.
Alth
o
u
g
h
c
o
n
v
e
n
tio
n
al
NL
P
m
eth
o
d
s
h
av
e
s
h
o
wn
c
o
n
s
id
er
ab
le
ef
f
ic
ac
y
in
id
en
tif
y
in
g
in
s
tan
ce
s
o
f
cy
b
er
b
u
lly
in
g
o
n
s
o
cial
m
ed
i
a
p
latf
o
r
m
s
,
s
ev
er
al
ch
allen
g
es
p
er
s
is
t.
T
h
ese
e
n
co
m
p
ass
lim
itatio
n
s
im
p
o
s
ed
b
y
ch
ar
ac
ter
r
estrictio
n
s
o
n
s
o
cial
m
ed
ia,
d
if
f
er
en
ce
s
b
etwe
en
o
f
f
e
n
s
iv
e
an
d
n
o
n
-
o
f
f
e
n
s
iv
e
co
m
m
en
ts
,
in
h
er
en
t a
m
b
ig
u
ities
in
n
atu
r
al
lan
g
u
ag
e,
an
d
th
e
ex
ten
s
iv
e
u
s
e
o
f
s
lan
g
[
1
3
]
.
Ov
er
th
e
y
ea
r
s
,
n
e
u
r
al
n
etwo
r
k
-
b
ased
m
o
d
els
h
a
v
e
o
u
tp
er
f
o
r
m
ed
co
n
v
en
tio
n
al
ML
tec
h
n
iq
u
es
in
v
ar
io
u
s
NL
P
task
s
.
T
h
is
s
u
p
e
r
io
r
ity
o
wes
m
u
ch
to
th
e
r
ich
v
ec
to
r
r
ep
r
esen
tatio
n
s
o
f
f
er
ed
b
y
n
eu
r
al
n
etwo
r
k
s
,
p
ar
ticu
lar
ly
th
e
s
ig
n
if
ican
t
s
tr
id
es
m
ad
e
in
wo
r
d
em
b
e
d
d
in
g
s
[
1
4
]
.
Un
lik
e
co
n
v
en
tio
n
al
ML
m
eth
o
d
s
r
ely
in
g
h
ea
v
ily
o
n
m
an
u
ally
cr
af
ted
,
p
o
ten
tially
in
co
m
p
lete
f
ea
tu
r
es
th
at
ar
e
tim
e
-
co
n
s
u
m
in
g
,
d
ee
p
lear
n
in
g
(
DL
)
tech
n
iq
u
es
em
p
lo
y
h
ier
a
r
ch
i
ca
l
au
to
m
atic
f
ea
tu
r
e
e
x
tr
ac
t
io
n
to
u
n
d
er
s
tan
d
i
n
p
u
t
c
h
a
r
ac
ter
is
tics
.
R
ec
en
t
ad
v
an
ce
s
in
NL
P
h
av
e
b
ee
n
d
r
iv
en
b
y
th
e
u
s
e
o
f
n
eu
r
al
n
et
wo
r
k
m
o
d
els
lik
e
m
u
ltil
ay
er
p
er
ce
p
tr
o
n
s
(
ML
Ps
)
,
R
NNs,
an
d
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs)
.
T
h
es
e
m
o
d
els
h
av
e
d
em
o
n
s
tr
ated
p
r
o
m
is
in
g
r
esu
lts
ac
r
o
s
s
v
ar
io
u
s
NL
P
task
s
.
X
iao
an
d
C
h
o
[
1
5
]
p
r
o
p
o
s
ed
a
n
o
v
el
m
eth
o
d
o
lo
g
y
f
o
r
te
x
t
class
if
icatio
n
th
at
o
p
er
ated
at
th
e
ch
ar
ac
ter
lev
e
l.
T
h
ey
in
co
r
p
o
r
ated
a
h
y
b
r
id
m
o
d
el,
co
m
b
in
in
g
a
C
NN
f
r
am
ewo
r
k
with
an
R
NN
ar
ch
itectu
r
e.
T
ai
et
a
l
.
[
1
6
]
p
r
o
p
o
s
ed
an
o
th
e
r
in
n
o
v
ativ
e
m
eth
o
d
.
T
h
ey
em
p
lo
y
ed
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
f
o
r
s
en
ten
ce
-
by
-
s
en
ten
ce
class
if
icatio
n
to
co
m
p
r
eh
e
n
d
th
e
s
em
an
tics
o
f
th
e
tex
t
an
d
u
s
ed
a
C
NN
to
ex
tr
ac
t lo
ca
l f
ea
tu
r
es
f
r
o
m
th
e
ex
p
r
ess
io
n
s
.
Ma
n
y
r
ese
ar
ch
er
s
h
a
v
e
e
x
p
lo
r
ed
t
h
e
d
e
te
cti
o
n
o
f
c
y
b
e
r
b
u
l
ly
i
n
g
i
n
A
r
ab
ic
[
6
]
,
[
1
7
]
–
[
2
3
]
.
Fo
r
i
n
s
t
an
ce
,
Hai
d
a
r
et
a
l
.
[
1
7
]
i
n
i
tia
ll
y
e
x
p
l
o
r
e
d
c
y
b
er
b
u
ll
y
i
n
g
d
et
ec
t
io
n
i
n
Ar
ab
ic
,
wh
ile
S
ala
wu
et
a
l
.
[
1
8
]
c
lass
i
f
i
ed
d
et
ec
ti
o
n
a
p
p
r
o
a
c
h
es
in
to
c
ate
g
o
r
i
es
l
ik
e
s
u
p
e
r
v
is
e
d
lea
r
n
i
n
g
,
le
x
i
c
o
n
-
b
as
ed
,
r
u
le
-
b
as
ed
,
a
n
d
m
ix
e
d
-
in
iti
ati
v
e
tec
h
n
i
q
u
es.
R
o
s
a
et
a
l
.
[
6
]
i
d
e
n
ti
f
i
ed
m
is
i
n
t
er
p
r
et
ati
o
n
s
a
n
d
lac
k
o
f
s
ta
n
d
a
r
d
iz
e
d
ev
al
u
a
ti
o
n
m
e
th
o
d
o
l
o
g
ies
i
n
p
r
e
v
i
o
u
s
r
esea
r
ch
.
I
n
c
o
n
t
r
as
t,
A
l
-
G
a
r
a
d
i
et
a
l
.
[
1
9
]
e
m
p
h
asi
ze
d
cr
u
cial
f
ac
t
o
r
s
i
n
d
e
tect
in
g
a
g
g
r
ess
iv
e
b
e
h
a
v
i
o
r
,
a
n
d
E
ls
af
o
u
r
y
e
t
a
l
.
[
2
0
]
h
i
g
h
li
g
h
te
d
t
h
e
i
m
p
o
r
t
a
n
c
e
o
f
s
la
n
g
-
b
as
ed
w
o
r
d
em
b
ed
d
i
n
g
tec
h
n
i
q
u
es
.
Kim
et
a
l
.
[
2
1
]
s
t
r
ess
ed
h
u
m
a
n
i
n
v
o
lv
em
e
n
t
i
n
al
g
o
r
i
th
m
d
e
v
el
o
p
m
e
n
t
,
w
h
il
e
Al
-
H
ar
ig
y
et
a
l
.
[
2
3
]
f
o
c
u
s
e
d
o
n
p
r
o
m
p
t
an
d
e
f
f
ec
t
iv
e
cy
b
er
b
u
l
ly
i
n
g
d
e
tec
ti
o
n
.
T
a
b
l
e
1
p
r
es
e
n
ts
a
co
m
p
r
e
h
e
n
s
i
v
e
c
o
m
p
a
r
at
iv
e
o
v
er
v
i
ew
o
f
th
e
af
o
r
e
m
e
n
ti
o
n
e
d
s
tu
d
i
es,
e
lu
ci
d
ati
n
g
n
o
ta
b
l
e
ch
ar
ac
te
r
is
ti
cs
a
n
d
m
e
th
o
d
o
l
o
g
ica
l s
tr
ate
g
i
es.
T
ab
le
1
.
C
r
itical
an
aly
s
is
o
f
ex
is
tin
g
cy
b
er
b
u
lly
in
g
a
p
p
r
o
ac
h
es with
p
r
o
p
o
s
ed
m
o
d
el
R
e
f
e
r
e
n
c
e
s
DL
a
p
p
r
o
a
c
h
e
s
D
a
t
a
r
e
p
r
e
se
n
t
a
t
i
o
n
D
a
t
a
a
v
a
i
l
a
b
i
l
i
t
y
C
y
b
e
r
b
u
l
l
y
i
n
g
a
p
p
l
i
c
a
t
i
o
n
Li
mi
t
a
t
i
o
n
a
n
d
s
t
r
e
n
g
t
h
Te
x
t
I
mag
e
s
[
2
1
]
✓
✓
✓
✓
[
2
2
]
✓
✓
✓
[
2
5
]
✓
✓
✓
[
2
4
]
✓
✓
[
2
0
]
✓
[
2
3
]
✓
✓
✓
[
2
4
]
✓
✓
✓
P
r
o
p
o
se
d
✓
✓
✓
✓
✓
T
h
e
cu
r
r
en
t
s
u
r
v
e
y
s
o
n
th
e
d
e
tectio
n
o
f
c
y
b
er
b
u
lly
in
g
u
s
in
g
DL
tech
n
iq
u
es
h
av
e
ce
r
tain
r
estra
in
ts
,
as
th
er
e
is
n
o
co
m
p
r
eh
en
s
iv
e
s
u
r
v
ey
s
p
ec
if
ically
f
o
cu
s
ed
o
n
th
is
f
ield
.
T
h
e
lac
k
o
f
co
m
p
r
eh
en
s
iv
e
a
n
aly
s
is
r
eg
ar
d
in
g
th
e
s
tr
en
g
th
s
an
d
wea
k
n
ess
es
o
f
DL
m
o
d
els
in
ac
cu
r
ately
class
if
y
in
g
cy
b
er
b
u
lly
in
g
in
s
tan
ce
s
is
attr
ib
u
ted
to
th
e
in
ef
f
icien
cy
o
f
m
o
s
t
s
u
r
v
e
y
p
ap
er
s
.
T
h
e
c
u
r
r
en
t
s
u
r
v
ey
s
d
o
n
o
t
en
c
o
m
p
ass
th
e
tax
o
n
o
m
y
o
f
DL
-
b
ased
cy
b
er
b
u
lly
in
g
class
if
icatio
n
,
wh
ich
is
cr
u
cial
f
o
r
o
r
g
an
izin
g
an
d
ex
p
an
d
in
g
co
m
p
lex
co
n
ce
p
ts
.
T
h
is
s
tu
d
y
ex
am
in
es
im
ag
e
-
b
ased
d
ata
r
ep
r
esen
tatio
n
tech
n
iq
u
es,
ac
k
n
o
wled
g
in
g
th
ei
r
im
p
o
r
tan
c
e
in
cy
b
er
b
u
lly
in
g
class
if
y
in
g
.
T
h
e
ca
r
ef
u
l
c
h
o
ice
o
f
a
s
u
itab
le
f
r
am
ewo
r
k
is
ess
en
tial
f
o
r
th
e
e
f
f
ec
tiv
e
im
p
lem
en
tatio
n
o
f
a
r
o
b
u
s
t
m
o
d
el
f
o
r
class
if
icatio
n
.
T
h
e
av
ailab
ilit
y
o
f
ea
s
ily
o
b
tain
ab
le
d
atasets
i
s
cr
u
cial
f
o
r
r
esear
ch
e
r
s
to
ass
ess
th
e
p
lau
s
ib
ilit
y
o
f
th
eir
h
y
p
o
t
h
eses
.
T
h
e
p
a
p
er
also
ex
p
lo
r
es
ad
d
itio
n
al
f
ac
to
r
s
s
u
ch
as
cu
ltu
r
al
d
iv
er
s
ity
,
d
ata
r
e
p
r
esen
tatio
n
,
m
u
ltime
d
ia
an
d
m
u
ltil
in
g
u
al
co
n
ten
t,
a
n
d
th
e
im
p
licatio
n
s
f
o
r
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
E
xp
lo
r
in
g
s
o
cia
l m
ed
ia
s
en
tim
en
t p
a
tter
n
s
fo
r
imp
r
o
ve
d
cy
b
e
r
b
u
llyin
g
…
(
Wa
el
Mo
h
a
me
d
S
h
a
h
er Ya
fo
o
z
)
4213
m
en
tal
well
-
b
ein
g
.
E
n
g
ag
i
n
g
in
th
ese
d
is
cu
s
s
io
n
s
is
ess
en
ti
al
to
d
ev
elo
p
a
co
m
p
r
eh
e
n
s
iv
e
u
n
d
er
s
tan
d
in
g
o
f
th
e
ch
allen
g
es a
n
d
f
u
tu
r
e
tr
en
d
s
with
in
th
is
f
ield
.
T
h
e
r
e
f
o
r
e
,
t
h
i
s
s
t
u
d
y
a
i
m
s
to
in
v
e
s
t
ig
a
t
e
th
e
p
o
s
s
i
b
i
l
i
t
i
es
o
f
c
a
t
eg
o
r
i
z
i
n
g
b
r
ie
f
t
e
x
tu
a
l
c
o
n
t
en
t
.
T
h
i
s
i
n
v
e
s
t
i
g
a
t
i
o
n
i
s
d
r
iv
e
n
b
y
th
e
n
o
t
a
b
l
e
a
ch
i
e
v
e
m
en
t
s
w
i
t
n
e
s
s
e
d
in
d
i
v
er
s
e
D
L
m
o
d
e
l
s
,
s
p
e
c
i
f
i
ca
l
l
y
i
n
t
a
s
k
s
i
n
v
o
lv
i
n
g
ex
t
e
n
s
i
v
e
t
ex
t
u
a
l
co
n
t
en
t
.
O
u
r
i
n
v
e
s
t
i
g
a
t
io
n
c
e
n
t
er
s
o
n
a
m
u
l
t
i
c
h
an
n
e
l
D
L
m
o
d
e
l,
wh
i
c
h
i
n
c
o
r
p
o
r
a
t
e
s
e
le
m
en
t
s
f
r
o
m
t
h
r
e
e
s
o
p
h
i
s
t
i
c
a
t
e
d
D
L
ar
c
h
i
t
e
c
tu
r
e
s
:
b
i
d
ir
e
c
t
io
n
a
l
L
S
T
M
(
B
i
L
S
T
M
)
,
t
r
a
n
s
f
o
r
m
e
r
b
lo
ck
,
a
n
d
C
N
N
.
E
m
p
lo
y
in
g
ad
v
an
c
ed
NL
P
a
n
d
M
L
c
l
a
s
s
i
f
i
er
s
o
n
s
o
c
i
a
l
n
e
t
wo
r
k
s
l
ik
e
X
p
l
a
y
s
a
c
r
u
c
i
a
l
r
o
l
e
i
n
au
t
o
m
a
t
i
n
g
c
y
b
e
r
b
u
l
ly
i
n
g
id
en
t
i
f
i
c
a
t
i
o
n
.
H
o
w
ev
e
r
,
th
e
av
a
i
la
b
i
l
i
t
y
o
f
s
u
i
t
ab
l
e
d
a
t
a
s
e
t
s
f
o
r
t
r
a
in
i
n
g
ML
c
l
a
s
s
i
f
i
e
r
s
l
i
m
i
t
s
t
h
i
s
p
o
s
s
i
b
i
l
i
ty
.
T
h
i
s
l
im
i
t
a
t
i
o
n
u
n
d
e
r
s
co
r
e
s
t
h
e
i
m
p
o
r
t
an
c
e
o
f
o
b
t
a
in
i
n
g
a
n
ex
t
e
n
s
i
v
e
d
a
t
a
s
et
e
n
c
o
m
p
a
s
s
i
n
g
v
a
r
io
u
s
cy
b
er
b
u
l
ly
i
n
g
in
s
t
a
n
ce
s
.
T
h
e
p
u
r
p
o
s
e
o
f
th
i
s
p
ap
e
r
i
s
t
o
p
r
e
s
en
t
a
m
u
l
t
i
ch
a
n
n
e
l
t
ec
h
n
iq
u
e,
a
f
u
s
i
o
n
o
f
th
r
e
e
DL
m
o
d
e
l
s
,
to
e
n
h
a
n
ce
p
r
ed
i
ct
i
o
n
a
c
cu
r
a
cy
.
T
o
e
v
a
l
u
a
t
e
th
e
p
r
o
p
o
s
e
d
a
p
p
r
o
a
ch
an
d
a
s
s
e
s
s
t
h
e
a
c
cu
r
a
cy
l
ev
e
l
s
,
w
e
co
m
b
i
n
e
t
h
r
e
e
ex
t
en
s
i
v
e
l
y
a
c
k
n
o
w
l
ed
g
ed
d
a
t
a
s
e
t
s
o
n
h
a
te
s
p
e
e
ch
[
2
4
]
–
[
2
6
]
.
T
h
e
f
in
d
in
g
s
in
d
icate
th
at
th
e
p
r
o
p
o
s
ed
m
eth
o
d
ca
n
ac
h
iev
e
n
o
tab
le
ac
cu
r
ac
y
.
T
h
e
co
n
tr
ib
u
tio
n
s
o
f
th
is
r
esear
ch
ar
e
as
f
o
llo
ws:
t
h
is
s
tu
d
y
in
tr
o
d
u
ce
s
a
n
o
v
el
h
y
b
r
id
m
o
d
el
wh
ich
co
n
s
is
ts
o
f
:
tr
an
s
f
o
r
m
er
b
lo
ck
,
B
iG
R
U,
an
d
C
NN
all
with
in
t
h
e
m
u
ltich
a
n
n
el
tech
n
iq
u
e,
to
s
ig
n
if
ican
tly
e
n
h
an
ce
th
e
p
r
e
d
ictio
n
ac
cu
r
ac
y
i
n
d
etec
tio
n
o
f
f
en
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
twee
ts
.
T
h
is
s
tu
d
y
d
e
m
o
n
s
tr
ates
th
e
ac
h
iev
e
m
en
t
o
f
n
o
tab
le
lev
els
o
f
p
r
ed
ictio
n
ac
cu
r
ac
y
b
y
in
teg
r
atin
g
th
r
ee
wid
ely
r
ec
o
g
n
ized
h
ate
s
p
ee
ch
d
atasets
.
T
h
is
s
tu
d
y
en
h
an
ce
s
o
n
lin
e
p
latf
o
r
m
s
’
d
etec
tio
n
ca
p
a
b
ilit
ies,
em
p
o
wer
in
g
th
em
to
s
tr
en
g
th
en
cy
b
e
r
s
ec
u
r
ity
m
ea
s
u
r
es
an
d
p
r
o
tect
u
s
er
s
f
r
o
m
c
y
b
er
b
u
lly
in
g
.
T
h
e
r
em
i
n
d
er
o
f
th
e
p
ap
e
r
is
o
r
g
an
ized
as
f
o
llo
ws:
s
ec
tio
n
3
in
tr
o
d
u
ce
s
t
h
e
in
teg
r
ated
D
L
ap
p
r
o
ac
h
.
Sectio
n
4
o
u
tlin
es
th
e
d
ataset
u
s
ed
f
o
r
ev
al
u
atio
n
p
u
r
p
o
s
es.
Sectio
n
5
p
r
esen
ts
,
an
aly
ze
s
an
d
c
o
m
p
ar
es
t
h
e
r
esu
lts
with
clo
s
e
ly
r
elate
d
ap
p
r
o
ac
h
es.
Sectio
n
6
en
ca
p
s
u
la
tes
th
e
co
n
clu
s
io
n
o
f
th
e
r
esear
ch
p
ap
er
,
f
o
llo
wed
b
y
a
d
is
cu
s
s
io
n
o
f
p
o
ten
tial a
v
en
u
es f
o
r
f
u
tu
r
e
r
esear
ch
in
s
ec
tio
n
7
.
2.
M
E
T
H
O
DO
L
O
G
Y
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
p
r
o
p
o
s
ed
in
tellig
en
t
cy
b
er
b
u
ll
y
in
g
d
etec
tio
n
p
latf
o
r
m
.
T
h
e
p
latf
o
r
m
is
o
r
g
an
ized
i
n
to
two
m
ain
c
o
m
p
o
n
e
n
ts
:
d
ata
an
aly
s
is
an
d
p
r
ed
ictiv
e
an
aly
tics
.
T
h
e
p
latf
o
r
m
f
u
n
ctio
n
s
as
a
r
o
b
u
s
tar
tific
ial
in
tellig
en
ce
an
d
tex
t
-
m
in
in
g
s
o
lu
tio
n
s
p
ec
if
ically
d
ev
elo
p
e
d
f
o
r
class
if
y
in
g
o
f
c
y
b
er
b
u
lly
in
g
in
to
o
f
f
en
s
iv
e
an
d
n
o
n
-
o
f
f
e
n
s
iv
e
twee
ts
.
Fig
u
r
e
1
d
ep
icts
th
e
m
o
d
el’
s
s
tr
u
ctu
r
e
f
o
r
ca
teg
o
r
izin
g
cy
b
er
b
u
lly
in
g
twee
ts
an
d
th
e
u
s
er
d
etec
tio
n
s
y
s
tem
wh
ich
was
d
ev
elo
p
ed
to
f
o
r
ec
ast
o
f
f
en
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
twee
ts
.
Fu
r
th
er
m
o
r
e
,
th
e
s
y
s
tem
tack
l
es
th
e
p
r
o
b
lem
o
f
an
o
n
y
m
ity
in
cy
b
er
b
u
lly
in
g
th
r
o
u
g
h
th
e
im
p
lem
en
tatio
n
o
f
u
s
er
id
en
tific
atio
n
tech
n
iq
u
es
.
T
h
is
s
o
lu
tio
n
in
clu
d
es
th
e
ex
am
in
atio
n
o
f
u
s
er
b
eh
av
io
r
p
atter
n
s
,
lin
g
u
is
tic
s
ty
le,
an
d
ad
d
itio
n
al
co
n
tex
tu
al
cu
es
to
es
tab
lis
h
d
is
tin
ct
u
s
er
p
r
o
f
iles
.
B
y
im
p
lem
en
tin
g
th
is
ap
p
r
o
ac
h
,
th
e
s
y
s
tem
ca
n
ef
f
ec
tiv
ely
m
o
n
it
o
r
a
n
d
tr
ac
k
p
o
ten
tial
o
r
ex
is
t
in
g
c
y
b
er
b
u
llies
,
d
eter
th
e
o
c
cu
r
r
en
ce
o
f
ab
u
s
iv
e
co
n
d
u
ct,
an
d
f
o
s
ter
th
e
d
e
v
elo
p
m
en
t o
f
an
o
n
lin
e
co
m
m
u
n
it
y
ch
ar
ac
ter
ized
b
y
e
n
h
an
ce
d
l
ev
els o
f
r
esp
ec
t.
Fig
u
r
e
1
.
Ar
c
h
itectu
r
e
o
f
p
r
o
p
o
s
ed
in
tellig
en
t d
etec
tio
n
o
f
c
y
b
er
b
u
lly
in
g
m
o
d
el
T
h
e
p
latf
o
r
m
em
p
lo
y
s
a
h
y
b
r
id
ap
p
r
o
ac
h
,
le
v
er
ag
in
g
s
em
an
tic
em
b
ed
d
in
g
s
,
co
n
tex
t
u
al
s
im
ilar
ity
s
co
r
es,
an
d
p
r
ed
ictiv
e
m
o
d
els to
en
h
a
n
ce
ac
cu
r
a
cy
in
class
if
y
in
g
twee
ts
.
T
h
e
d
ata
an
al
y
s
is
m
o
d
u
le
co
n
s
is
ts
o
f
d
ata
ac
q
u
is
itio
n
,
p
r
ep
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
an
d
f
ea
tu
r
e
s
elec
tio
n
s
tep
s
.
I
t
in
co
r
p
o
r
ates
em
b
ed
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
2
1
1
-
4
2
2
5
4214
b
ased
s
em
an
tic
s
im
ilar
ity
m
ea
s
u
r
es,
in
clu
d
in
g
co
n
tex
tu
al
em
b
ed
d
in
g
o
v
er
la
p
s
im
ilar
ity
(
C
E
OS)
,
wh
ich
en
h
an
ce
s
th
e
p
latf
o
r
m
’
s
ab
ilit
y
to
d
is
ce
r
n
s
u
b
tle
n
u
an
ce
s
in
twee
ts
.
T
h
e
p
r
ed
ictiv
e
an
aly
tics
m
o
d
u
le,
o
n
th
e
o
th
er
h
an
d
,
in
clu
d
es
tr
ain
in
g
a
n
d
test
in
g
ML
an
d
DL
m
o
d
els,
o
p
tim
ized
f
o
r
h
ig
h
p
r
ec
is
io
n
in
d
etec
tio
n
task
s
.
T
h
e
m
eth
o
d
o
lo
g
y
i
n
teg
r
ates
DL
class
if
icatio
n
tech
n
iq
u
e
s
to
ca
teg
o
r
ize
t
h
e
twee
ts
a
n
d
u
s
er
s
in
to
two
o
f
f
en
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
c
ateg
o
r
ies.
T
h
er
e
ar
e
f
iv
e
s
tep
s
in
th
e
d
ata
an
aly
s
i
s
m
o
d
u
le:
twee
t
ac
q
u
is
it
io
n
f
r
o
m
th
e
T
witter
API
(
twee
tI
n
v
i)
,
n
o
w
k
n
o
wn
as
X
API
,
twee
t
p
r
e
-
p
r
o
ce
s
s
in
g
,
f
ea
tu
r
e
ex
tr
ac
tio
n
,
f
ea
tu
r
e
n
o
r
m
aliza
tio
n
,
a
n
d
f
ea
tu
r
e
s
el
ec
tio
n
.
Pre
d
ictiv
e
an
aly
tics
in
v
o
lv
es
tr
ain
in
g
an
d
test
in
g
v
ar
io
u
s
DL
m
o
d
els,
it
also
in
v
o
lv
es o
p
tim
izin
g
h
y
p
e
r
p
ar
am
eter
s
.
2
.
1
.
X
da
t
a
a
cquis
it
io
n
T
h
e
d
ata
f
r
o
m
X
was
o
b
tain
e
d
u
s
in
g
th
e
twee
tI
n
v
i
API
to
d
etec
t
o
n
lin
e
cy
b
er
b
u
lly
i
n
g
.
T
h
e
d
atasets
wer
e
co
m
p
r
is
ed
o
f
twee
ts
an
d
u
s
er
n
etwo
r
k
s
.
T
h
ese
wer
e
e
x
tr
ac
ted
f
r
o
m
tex
tu
al
d
ata,
w
h
ich
o
n
ly
ar
e
twee
ts
,
h
astag
es,
co
m
m
en
ts
,
an
d
p
h
o
to
s
d
escr
ip
tio
n
.
A
d
ataset
o
f
2
0
,
0
0
0
twee
ts
as
s
h
o
wn
in
Fig
u
r
e
2
,
alo
n
g
with
u
s
er
I
Ds,
h
ash
tag
s
,
d
ateT
im
e,
an
d
lo
ca
tio
n
o
f
o
r
ig
i
n
,
was
co
llected
in
J
SON
f
o
r
m
at.
T
wo
s
ep
ar
ate
lab
els
f
o
r
twee
ts
wer
e
cr
ea
ted
to
d
is
tin
g
u
is
h
b
etwe
en
o
f
f
e
n
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
twee
ts
.
T
ab
le
2
p
r
o
v
id
es
a
d
etailed
b
r
ea
k
d
o
wn
o
f
th
e
d
ataset'
s
k
ey
attr
ib
u
tes.
Fig
u
r
e
2
.
C
u
b
e
r
b
u
lly
in
g
d
atas
et
in
clu
d
es two
class
es n
o
n
-
o
f
f
en
s
iv
e
an
d
o
f
f
e
n
s
iv
e
T
ab
le
2
.
Data
s
et
d
escr
ip
tio
n
th
at
in
clu
d
es f
ea
tu
r
es a
n
d
d
ata
c
o
u
n
t
S
r
n
o
F
e
a
t
u
r
e
s
A
p
p
l
i
c
a
t
i
o
n
p
r
o
g
r
a
m
m
i
n
g
i
n
t
e
r
f
a
c
e
D
a
t
a
c
o
u
n
t
D
a
t
a
d
e
s
c
r
i
p
t
i
o
n
1
T
w
e
e
t
/
t
w
e
e
t
i
n
v
i
/
t
w
i
t
t
e
r
-
a
p
i
/
t
w
e
e
t
s
5
,
0
0
0
T
h
e
i
n
f
o
r
m
a
t
i
o
n
w
h
i
c
h
u
s
e
r
s
s
h
a
r
e
t
o
g
e
t
h
e
r
2
H
a
s
h
T
a
g
s
/
t
w
e
e
t
i
n
v
i
/
t
w
i
t
t
e
r
-
a
p
i
/
H
a
s
h
t
a
g
s
1
,
0
0
0
S
h
a
r
e
d
h
a
s
h
t
a
g
s
p
r
o
j
e
c
t
s
o
p
e
n
i
n
g
o
n
.
3
C
o
m
m
e
n
t
s
/
t
w
e
e
t
i
n
v
i
/
t
w
i
t
t
e
r
-
a
p
i
/
C
o
m
m
e
n
t
s
5
,
000
C
o
m
m
e
n
t
s
f
o
r
s
h
a
r
e
d
t
o
p
i
c
4
P
h
o
t
o
s
/
t
w
e
e
t
i
n
v
i
/
t
w
i
t
t
e
r
-
a
p
i
/
P
h
o
t
o
s
9
,
000
S
h
a
r
e
d
p
h
o
t
o
s
b
y
v
a
r
i
o
u
s
u
s
e
r
s
2
.
2
.
P
re
-
pro
ce
s
s
ing
o
f
X
da
t
a
Data
p
r
e
-
p
r
o
ce
s
s
in
g
p
lay
s
a
v
i
tal
r
o
le
in
g
u
ar
an
teein
g
th
e
ac
cu
r
ac
y
an
d
d
ep
en
d
ab
ilit
y
o
f
u
n
p
r
o
ce
s
s
ed
d
ata.
Acc
u
r
ac
y
is
en
h
a
n
ce
d
b
y
tr
an
s
f
o
r
m
i
n
g
u
n
p
r
o
ce
s
s
ed
J
SON
f
iles
in
to
a
s
tan
d
ar
d
ized
f
o
r
m
at.
Af
ter
war
d
,
th
e
tex
tu
al
d
ata
is
s
u
b
jecte
d
to
v
ar
io
u
s
p
r
e
-
p
r
o
ce
s
s
in
g
tech
n
iq
u
es
i
n
NL
P.
T
h
e
X
d
ata
u
n
d
er
g
o
es
th
e
f
o
llo
win
g
s
tep
s
.
−
No
r
m
aliza
tio
n
:
i
n
th
is
s
tep
,
te
x
ts
wer
e
co
n
v
er
ted
in
to
a
s
tan
d
ar
d
ized
f
o
r
m
at
to
im
p
r
o
v
e
d
a
ta
q
u
ality
an
d
en
ab
le
ef
f
icien
t
s
y
s
tem
p
r
o
ce
s
s
in
g
.
No
r
m
aliza
tio
n
in
o
u
r
s
y
s
tem
was
ac
co
m
p
lis
h
ed
th
r
o
u
g
h
s
ev
er
al
tech
n
iq
u
es.
First,
th
e
d
u
p
licate
wh
ite
s
p
ac
es
wer
e
elim
in
ated
.
Nex
t,
th
e
tex
t
was
co
n
v
er
ted
to
lo
wer
ca
s
e.
Af
ter
th
at,
co
n
tr
ac
tio
n
s
wer
e
ex
p
an
d
ed
,
an
d
wo
r
d
n
u
m
er
als
wer
e
co
n
v
er
ted
to
th
eir
co
r
r
esp
o
n
d
in
g
n
u
m
er
ical
v
alu
es.
T
h
e
n
o
r
m
ali
ze
d
was th
en
in
p
u
tted
in
to
th
e
to
k
en
izatio
n
m
o
d
u
le.
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
E
xp
lo
r
in
g
s
o
cia
l m
ed
ia
s
en
tim
en
t p
a
tter
n
s
fo
r
imp
r
o
ve
d
cy
b
e
r
b
u
llyin
g
…
(
Wa
el
Mo
h
a
me
d
S
h
a
h
er Ya
fo
o
z
)
4215
−
T
o
k
en
izatio
n
:
i
n
th
is
s
tep
,
p
ar
ag
r
ap
h
s
an
d
s
en
ten
ce
s
wer
e
d
iv
id
ed
in
to
in
d
iv
id
u
al
to
k
en
tex
ts
.
T
h
is
d
iv
is
io
n
f
ac
ilit
ates
th
e
as
s
ig
n
m
en
t
o
f
s
em
an
tic
m
ea
n
in
g
to
th
e
to
k
en
s
,
th
er
eb
y
en
h
an
ci
n
g
th
e
o
v
er
all
u
n
d
er
s
tan
d
i
n
g
an
d
an
aly
s
is
o
f
th
e
tex
t.
Ker
as,
an
o
p
en
-
s
o
u
r
ce
d
ee
p
-
lear
n
in
g
lib
r
ar
y
,
was
em
p
lo
y
ed
f
o
r
th
e
p
u
r
p
o
s
e
o
f
tex
t
p
r
e
-
p
r
o
c
ess
in
g
,
s
p
ec
if
ically
f
o
r
t
o
k
e
n
izatio
n
.
T
h
e
lis
t
o
f
t
o
k
en
s
o
b
tain
e
d
was
s
u
b
s
eq
u
en
tly
cr
u
cial
f
o
r
t
h
e
n
e
x
t p
r
e
-
p
r
o
ce
s
s
in
g
s
tep
.
−
Pu
n
ctu
atio
n
a
n
d
n
u
m
b
er
r
em
o
v
al:
p
u
n
ctu
atio
n
s
a
n
d
n
u
m
b
e
r
s
wer
e
r
em
o
v
ed
in
t
h
is
s
tep
.
Su
b
s
eq
u
en
tly
,
th
e
cu
r
ated
lis
t w
as tr
an
s
f
er
r
ed
to
th
e
s
u
b
s
eq
u
e
n
t p
h
ase.
−
L
em
m
atiza
tio
n
:
i
n
t
h
is
s
tep
,
wo
r
d
s
wer
e
tr
a
n
s
f
o
r
m
ed
in
to
th
eir
b
ase
f
o
r
m
s
b
y
co
n
s
id
er
in
g
m
o
r
p
h
o
lo
g
ical
an
aly
s
is
.
Fo
r
in
s
tan
ce
,
th
e
ter
m
‘
co
m
p
u
ter
s
’
was
tr
an
s
f
o
r
m
ed
in
to
th
e
s
in
g
u
lar
f
o
r
m
‘
co
m
p
u
te
r
’
.
As
p
ar
t
o
f
o
u
r
m
o
d
el,
we
u
s
ed
t
h
e
W
o
r
d
Net
le
m
m
atize
r
p
r
o
v
id
e
d
b
y
t
h
e
n
atu
r
al
lan
g
u
ag
e
to
o
lk
it
(
NL
T
K)
.
−
Par
t
-
of
-
s
p
ee
ch
(
POS):
t
h
is
s
tep
in
v
o
lv
ed
ass
ig
n
in
g
POS
tag
s
wo
r
d
s
to
e
n
h
an
ce
s
em
an
tic
s
im
ilar
ity
.
I
n
o
u
r
p
r
o
p
o
s
ed
m
o
d
el,
we
em
p
l
o
y
ed
POS tag
g
in
g
to
im
p
r
o
v
e
th
e
p
r
ec
is
io
n
o
f
s
em
an
tic
s
im
ilar
ity
.
−
E
lim
in
atio
n
o
f
s
to
p
wo
r
d
s
:
i
n
th
is
s
tep
,
f
r
eq
u
en
tly
u
s
ed
s
tr
u
ctu
r
al
wo
r
d
s
wer
e
r
em
o
v
e
d
to
o
p
tim
ize
te
x
t
m
in
in
g
.
I
n
th
e
I
SE
m
o
d
el,
th
e
r
em
o
v
al
o
f
s
to
p
wo
r
d
s
was
im
p
lem
en
ted
to
p
r
io
r
itize
r
elev
an
t
in
f
o
r
m
atio
n
,
o
p
tim
ize
tex
t m
in
in
g
p
r
o
ce
d
u
r
es,
an
d
m
itig
ate
i
n
tr
icac
y
.
2
.
3
.
F
e
a
t
ure
e
ng
ineering
W
ith
in
th
is
s
ec
tio
n
,
th
e
u
tili
za
tio
n
o
f
th
e
f
ea
tu
r
e
en
g
in
ee
r
i
n
g
tech
n
i
q
u
e
h
as
b
ee
n
im
p
le
m
en
ted
to
p
r
o
d
u
ce
ad
d
itio
n
al
d
ata
f
r
o
m
th
e
g
iv
en
d
ataset.
T
h
e
laten
t
f
ea
tu
r
es
th
at
h
av
e
b
ee
n
id
e
n
tifie
d
an
d
d
er
i
v
ed
with
in
th
e
p
r
o
p
o
s
ed
p
latf
o
r
m
.
T
h
ese
d
er
iv
ed
f
ea
tu
r
es a
r
e
o
u
t
lin
ed
in
T
ab
le
3
.
T
ab
le
3
.
Der
iv
e
d
f
ea
tu
r
es a
n
d
th
eir
d
escr
ip
tio
n
Sr
n
o
Ty
p
e
F
e
a
t
u
r
e
n
a
me
D
e
scri
p
t
i
o
n
R
e
f
e
r
e
n
c
e
1
Emb
e
d
d
i
n
g
f
e
a
t
u
r
e
v
e
c
t
o
r
W
o
r
k
2
V
e
c
Th
e
t
w
e
e
t
-
l
e
v
e
l
f
e
a
t
u
r
e
r
e
p
r
e
se
n
t
a
t
i
o
n
i
s
d
e
r
i
v
e
d
b
y
e
m
p
l
o
y
i
n
g
p
r
e
t
r
a
i
n
e
d
W
o
r
d
2
V
e
c
e
mb
e
d
d
i
n
g
s.
[
2
7
]
2
S
e
n
t
i
me
n
t
f
e
a
t
u
r
e
v
e
c
t
o
r
S
e
n
t
i
S
t
r
e
n
g
t
h
Th
i
s
f
e
a
t
u
r
e
i
s
u
t
i
l
i
z
e
d
t
o
e
v
a
l
u
a
t
e
e
a
c
h
t
w
e
e
t
’
s
p
o
si
t
i
v
e
a
n
d
n
e
g
a
t
i
v
e
se
n
t
i
me
n
t
sc
o
r
e
s.
[
2
8
]
3
M
P
Q
A
s
u
b
j
e
c
t
i
v
i
t
y
l
e
x
i
c
o
n
D
e
r
i
v
e
s
l
e
x
i
c
o
n
f
e
a
t
u
r
e
s
a
t
t
h
e
p
h
r
a
se
l
e
v
e
l
t
o
d
e
t
e
r
mi
n
e
t
h
e
p
o
s
i
t
i
v
e
a
n
d
n
e
g
a
t
i
v
e
c
o
n
t
e
x
t
u
a
l
p
o
l
a
r
i
t
y
o
f
s
e
n
t
i
m
e
n
t
e
x
p
r
e
ssi
o
n
i
n
t
w
e
e
t
s
.
S
e
n
t
i
m
e
n
t
1
4
0
i
s
a
c
o
l
l
e
c
t
i
o
n
o
f
t
w
e
e
t
s
l
a
b
e
l
e
d
p
o
si
t
i
v
e
,
n
e
g
a
t
i
v
e
,
o
r
n
e
u
t
r
a
l
f
o
r
s
e
n
t
i
me
n
t
a
n
a
l
y
si
s
.
[
2
9
]
4
Le
x
i
c
o
n
f
e
a
t
u
r
e
v
e
c
t
o
r
S
e
n
t
i
me
n
t
-
1
4
0
[
3
0
]
5
B
i
n
g
Li
u
U
si
n
g
f
e
a
t
u
r
e
mi
n
i
n
g
,
c
u
st
o
mer
r
e
v
i
e
w
s
a
r
e
c
a
t
e
g
o
r
i
z
e
d
a
s
f
a
v
o
r
a
b
l
e
o
r
n
e
g
a
t
i
v
e
w
i
t
h
o
u
t
se
n
t
e
n
c
e
se
l
e
c
t
i
o
n
.
[
2
9
]
6
A
F
I
N
N
En
g
l
i
sh
w
o
r
d
s
f
o
r
e
mo
t
i
o
n
p
o
l
a
r
i
t
y
:
p
o
s
i
t
i
v
e
,
n
e
g
a
t
i
v
e
,
o
r
n
e
u
t
r
a
l
.
I
t
sco
r
e
s
w
o
r
d
s
n
u
m
e
r
i
c
a
l
l
y
f
o
r
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s
a
n
d
t
e
x
t
mi
n
i
n
g
.
[
3
1
]
7
Ex
p
a
n
d
e
d
N
R
C
-
10
D
e
p
l
o
y
e
d
t
o
c
l
a
ss
i
f
y
t
w
e
e
t
s’
e
m
o
t
i
o
n
s
o
r
d
o
i
n
g
s
e
n
t
i
m
e
n
t
a
n
a
l
y
ses.
[
3
2
]
8
N
R
C
h
a
s
h
t
a
g
sen
t
i
m
e
n
t
l
e
x
i
c
o
n
N
R
C
h
a
s
h
t
a
g
se
n
t
i
me
n
t
l
e
x
i
c
o
n
f
o
r
h
a
sh
t
a
g
-
f
o
c
u
se
d
s
o
c
i
a
l
me
d
i
a
sen
t
i
m
e
n
t
r
e
s
e
a
r
c
h
.
I
t
i
d
e
n
t
i
f
i
e
s
f
r
e
q
u
e
n
t
s
o
c
i
a
l
m
e
d
i
a
t
e
r
ms
a
s
f
a
v
o
r
a
b
l
e
,
b
a
d
,
o
r
n
e
u
t
r
a
l
.
[
3
3
]
9
S
e
n
t
i
W
o
r
d
n
e
t
S
e
n
t
i
W
o
r
d
N
e
t
r
a
n
k
s
E
n
g
l
i
sh
w
o
r
d
s’
me
a
n
i
n
g
s
a
s
p
o
si
t
i
v
e
,
n
e
g
a
t
i
v
e
,
o
r
n
e
u
t
r
a
l
.
I
t
si
m
p
l
i
f
i
e
s
N
LP
sen
t
i
m
e
n
t
a
n
a
l
y
si
s
w
i
t
h
W
o
r
d
N
e
t
s
y
n
se
t
s a
n
d
s
e
n
t
i
m
e
n
t
s
c
o
r
e
s.
[
3
4
]
10
N
R
C
-
10
Th
e
N
R
C
e
m
o
t
i
o
n
l
e
x
i
c
o
n
c
l
a
ss
i
f
i
e
s
w
o
r
d
s
i
n
t
o
1
0
f
u
n
d
a
m
e
n
t
a
l
e
mo
t
i
o
n
s fo
r
t
e
x
t
d
a
t
a
s
e
n
t
i
m
e
n
t
a
n
a
l
y
si
s a
n
d
c
l
a
ss
i
f
i
c
a
t
i
o
n
.
[
3
5
]
11
N
R
C
h
a
s
h
t
a
g
e
mo
t
i
o
n
a
sso
c
i
a
t
i
o
n
l
e
x
i
c
o
n
Th
e
N
R
C
h
a
s
h
t
a
g
e
m
o
t
i
o
n
a
ss
o
c
i
a
t
i
o
n
L
e
x
i
c
o
n
h
e
l
p
s
so
c
i
a
l
me
d
i
a
d
a
t
a
a
n
a
l
y
st
s
a
n
a
l
y
z
e
e
m
o
t
i
o
n
s
b
y
l
i
n
k
i
n
g
h
a
s
h
t
a
g
s
t
o
e
mo
t
i
o
n
s.
[
3
6
]
12
K
n
o
w
l
e
d
g
e
b
a
s
e
d
s
i
m
i
l
a
r
i
t
y
C
EO
S
C
o
m
p
u
t
e
s
s
i
mi
l
a
r
i
t
y
b
e
t
w
e
e
n
t
w
e
e
t
s
e
m
b
e
d
d
i
n
g
s
a
n
d
r
e
f
e
r
e
n
c
e
e
mb
e
d
d
i
n
g
s f
o
r
o
f
f
e
n
s
i
v
e
a
n
d
n
o
n
-
o
f
f
e
n
s
i
v
e
c
o
n
t
e
n
t
.
P
r
o
p
o
se
d
T
h
e
p
r
o
p
o
s
ed
m
eth
o
d
ev
alu
ates
s
em
an
tic
s
im
ilar
ity
b
y
lev
er
ag
in
g
co
n
tex
tu
al
em
b
ed
d
in
g
s
to
m
ea
s
u
r
e
th
e
o
v
er
la
p
b
etwe
en
a
twee
t
an
d
p
r
ed
ef
in
e
d
r
ef
e
r
en
ce
v
ec
to
r
s
,
r
e
p
r
esen
tin
g
o
f
f
en
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
lan
g
u
ag
e.
T
h
is
ap
p
r
o
ac
h
elim
in
ates
r
elian
ce
o
n
f
r
eq
u
en
cy
-
b
ased
m
ec
h
a
n
is
m
s
an
d
in
s
tead
f
o
c
u
s
es
o
n
th
e
s
em
an
tic
r
elatio
n
s
h
ip
s
i
n
a
h
ig
h
-
d
im
en
s
io
n
al
em
b
ed
d
i
n
g
s
p
ac
e.
T
o
ca
lcu
late
th
e
s
im
ilar
ity
,
em
b
e
d
d
in
g
v
ec
to
r
s
ar
e
g
e
n
er
ated
f
o
r
th
e
in
p
u
t
twee
t.
L
et
⃗
r
ep
r
ese
n
t
th
e
em
b
e
d
d
in
g
v
ec
to
r
f
o
r
th
e
twee
t,
ag
g
r
eg
ated
f
r
o
m
wo
r
d
-
lev
el
e
m
b
ed
d
in
g
s
as
in
(
1
)
,
w
h
er
e
⃗
(
)
i
s
th
e
em
b
ed
d
in
g
o
f
th
e
it
-
th
wo
r
d
s
an
d
wo
r
d
s
in
th
e
twee
ts
.
T
wo
r
ef
e
r
en
ce
v
ec
to
r
s
ar
e
p
r
ec
o
m
p
u
te
d
:
o
f
f
e
n
s
iv
e
⃗
f
o
r
o
f
f
en
s
iv
e
lan
g
u
ag
e,
as
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
2
1
1
-
4
2
2
5
4216
s
h
o
wn
in
(
2
)
,
a
n
d
n
o
n
-
o
f
f
en
s
i
v
e
⃗
−
f
o
r
n
o
n
-
o
f
f
en
s
iv
e
lan
g
u
ag
e
as
p
r
esen
te
d
in
(
3
)
.
T
h
e
s
im
ilar
ity
b
etwe
en
th
e
twee
t
em
b
ed
d
in
g
an
d
ea
ch
r
ef
e
r
en
ce
v
ec
to
r
is
ca
lcu
lated
u
s
in
g
co
s
in
e
s
i
m
ilar
ity
.
Fin
aly
,
th
e
C
E
OS
f
o
r
a
twee
t
is
th
en
ca
lcu
lated
as
in
(
4
)
.
A
p
o
s
itiv
e
C
E
OS
s
u
g
g
ests
a
h
ig
h
er
lik
elih
o
o
d
o
f
o
f
f
en
s
iv
e
co
n
ten
t,
w
h
ile
a
n
eg
ativ
e
s
co
r
e
in
d
icate
s
n
o
n
-
o
f
f
e
n
s
iv
e
ten
d
en
cies.
T
h
is
m
eth
o
d
f
o
cu
s
es
o
n
s
em
an
tic
r
elatio
n
s
h
ip
s
,
o
f
f
er
i
n
g
a
r
o
b
u
s
t
an
d
c
o
n
tex
t
-
awa
r
e
m
ec
h
a
n
is
m
f
o
r
d
is
tin
g
u
is
h
in
g
o
f
f
e
n
s
iv
e
an
d
non
-
o
f
f
en
s
iv
e
la
n
g
u
a
g
e.
⃗
=
1
∑
⃗
(
)
=
1
(
1
)
=
⃗
.
⃗
‖
⃗
‖
‖
⃗
‖
(
2
)
−
=
⃗
.
⃗
⃗
−
‖
⃗
‖
‖
⃗
−
‖
(
3
)
C
E
OS
=
−
−
(
4
)
2
.
4
.
F
e
a
t
ure
s
elec
t
io
n
T
o
id
en
tif
y
th
e
m
o
s
t
r
elev
a
n
t
f
ea
tu
r
es
f
o
r
t
h
e
class
if
icatio
n
m
ec
h
a
n
is
m
,
we
co
n
d
u
cte
d
a
th
o
r
o
u
g
h
an
aly
s
is
o
f
th
e
p
r
e
p
ar
ed
d
ata
s
et
in
th
is
m
o
d
u
le.
T
h
e
in
clu
s
io
n
o
f
th
is
s
tep
is
o
f
u
tm
o
s
t
im
p
o
r
tan
ce
in
th
e
p
r
o
ce
s
s
o
f
g
ettin
g
r
i
d
o
f
a
p
o
t
en
tially
ex
p
an
s
iv
e
f
ea
tu
r
e
s
p
a
ce
.
B
y
r
em
o
v
in
g
u
n
n
ec
ess
ar
y
ch
ar
ac
ter
is
tics
th
at
d
o
n
o
t
s
u
b
s
tan
tially
co
n
tr
ib
u
t
e
to
th
e
an
aly
s
is
p
r
o
ce
d
u
r
e,
th
is
p
r
o
ce
s
s
o
p
tim
izes
d
ata
p
r
o
ce
s
s
in
g
ef
f
icien
cy
.
Fu
r
th
er
m
o
r
e
,
it f
ac
ilit
ates
th
e
i
d
en
tific
atio
n
o
f
d
ata
f
ea
t
u
r
es with
th
e
h
i
g
h
est
an
d
lo
west
le
v
els
o
f
i
n
f
lu
en
ce
,
as
d
eter
m
in
ed
b
y
th
eir
r
esp
ec
tiv
e
weig
h
ts
.
I
n
th
i
s
s
t
u
d
y
,
we
u
t
i
l
i
ze
d
tw
o
w
e
l
l
-
e
s
t
ab
l
i
s
h
ed
m
e
t
h
o
d
o
l
o
g
i
e
s
k
n
o
wn
a
s
p
r
i
n
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
l
y
s
i
s
(
P
C
A
)
a
n
d
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
co
e
f
f
i
c
i
en
t
(
P
C
C
)
t
o
e
n
h
an
c
e
th
e
s
e
l
e
c
t
io
n
o
f
i
n
i
t
i
a
l
f
ea
t
u
r
e
s
,
a
s
d
e
s
c
r
ib
e
d
i
n
T
ab
l
e
3
.
I
n
a
d
d
it
i
o
n
,
t
h
e
n
e
w
l
y
in
t
r
o
d
u
c
e
d
C
E
O
S
s
c
o
r
e
,
d
er
i
v
ed
f
r
o
m
p
r
e
-
t
r
a
in
e
d
e
m
b
e
d
d
i
n
g
m
o
d
e
l
s
s
u
c
h
a
s
B
E
R
T
o
r
Gl
o
V
e
,
h
a
s
e
m
e
r
g
e
d
a
s
a
c
r
i
ti
c
a
l
f
e
a
tu
r
e
,
o
f
f
e
r
in
g
s
e
m
a
n
ti
c
c
o
n
t
ex
t
t
o
th
e
c
l
a
s
s
i
f
i
c
a
t
i
o
n
p
r
o
c
e
s
s
.
P
C
A
a
n
d
P
C
C
p
r
im
a
r
i
ly
e
m
p
h
a
s
iz
e
t
h
e
s
e
l
e
c
t
io
n
o
f
l
in
e
ar
v
a
r
i
a
b
l
e
s
.
P
C
A
i
s
a
t
e
c
h
n
iq
u
e
t
h
a
t
i
s
u
s
e
d
to
c
a
l
cu
l
a
t
e
t
h
e
f
u
n
d
a
m
en
t
a
l
co
m
p
o
n
e
n
t
s
,
wh
i
ch
a
i
d
s
in
r
ed
u
c
i
n
g
t
h
e
d
im
e
n
s
io
n
a
l
i
ty
o
f
a
co
m
p
l
e
x
f
e
a
tu
r
e
s
p
ac
e
t
o
a
m
o
r
e
m
a
n
ag
e
ab
l
e
o
n
e
.
T
h
i
s
p
r
o
c
e
s
s
en
s
u
r
e
s
th
a
t
cr
u
c
ia
l
f
e
a
t
u
r
e
i
n
f
o
r
m
a
t
io
n
i
s
r
e
t
a
in
e
d
,
a
s
d
ep
i
c
t
e
d
in
F
i
g
u
r
e
3
.
T
h
e
C
E
OS
s
c
o
r
e
d
e
m
o
n
s
t
r
a
te
s
a
h
ig
h
c
o
r
r
e
l
a
t
i
o
n
w
i
t
h
t
h
e
t
a
r
g
e
t
v
ar
i
a
b
l
e
i
n
P
C
C
a
n
a
ly
s
i
s
(
F
i
g
u
r
e
4
)
,
r
e
in
f
o
r
c
in
g
i
t
s
s
i
g
n
i
f
i
c
an
c
e
a
s
a
f
e
a
t
u
r
e
f
o
r
d
i
s
t
i
n
g
u
i
s
h
i
n
g
o
f
f
en
s
i
v
e
a
n
d
n
o
n
-
o
f
f
e
n
s
i
v
e
tw
e
e
t
s
.
Fig
u
r
e
3
.
Featu
r
e
im
p
o
r
ta
n
ce
an
aly
s
is
b
ased
o
n
PC
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
E
xp
lo
r
in
g
s
o
cia
l m
ed
ia
s
en
tim
en
t p
a
tter
n
s
fo
r
imp
r
o
ve
d
cy
b
e
r
b
u
llyin
g
…
(
Wa
el
Mo
h
a
me
d
S
h
a
h
er Ya
fo
o
z
)
4217
Fig
u
r
e
4
.
Pear
s
o
n
co
r
r
elatio
n
an
aly
s
is
o
f
in
p
u
t
f
ea
tu
r
es
2
.
5
.
P
re
dict
iv
e
a
na
ly
s
is
o
f
X
da
t
a
f
o
r
c
y
berbull
y
cla
s
s
if
ic
a
t
io
n ba
s
ed
o
n
ML
a
lg
o
rit
hm
s
T
h
is
s
ec
tio
n
f
o
cu
s
es
o
n
o
u
r
p
r
o
p
o
s
ed
c
y
b
er
b
u
lly
in
g
d
etec
tio
n
s
y
s
tem
’
s
ML
class
if
ier
s
.
T
h
e
p
u
r
p
o
s
e
o
f
th
is
s
tu
d
y
is
to
p
er
f
o
r
m
a
th
o
r
o
u
g
h
co
m
p
ar
ativ
e
an
al
y
s
is
o
f
th
e
ef
f
icac
y
o
f
v
ar
io
u
s
ML
class
if
ier
s
in
d
if
f
er
en
tiatin
g
b
etwe
en
o
f
f
e
n
s
iv
e
an
d
n
o
n
-
o
f
f
en
s
iv
e
twee
ts
.
T
h
e
p
r
o
p
o
s
ed
p
latf
o
r
m
in
co
r
p
o
r
ates
a
r
a
n
g
e
o
f
class
if
icatio
n
m
o
d
els f
o
r
s
u
p
er
v
is
ed
d
ata
class
if
icatio
n
,
as d
etailed
in
T
ab
le
4
.
I
n
o
r
d
er
to
th
o
r
o
u
g
h
ly
ass
ess
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
class
if
ier
,
we
im
p
lem
en
ted
a
k
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
m
eth
o
d
o
lo
g
y
,
u
tili
zin
g
a
v
alu
e
o
f
k
eq
u
al
to
1
0
.
T
h
is
m
eth
o
d
o
lo
g
y
s
y
s
tem
atica
lly
ev
alu
ates
an
d
co
m
p
ar
es
th
e
o
u
tco
m
es
o
f
th
e
ap
p
lied
class
if
ier
s
.
T
h
e
d
atas
et
was
d
iv
id
ed
in
to
k
eq
u
al
s
eg
m
en
ts
,
wh
er
e
ea
ch
s
eg
m
en
t w
as u
tili
ze
d
f
o
r
tr
ain
i
n
g
a
class
if
ier
.
T
ab
le
4
.
Ap
p
lied
class
if
ier
s
o
n
th
e
lab
el
d
ataset
S
R
n
o
#
M
L
c
l
a
ssi
f
i
e
r
s
1
XGB
2
S
V
M
3
DT
4
Lo
g
i
s
t
i
c
r
e
g
r
e
ssi
o
n
(
L
R
)
5
RF
6
K
N
N
E
ac
h
iter
atio
n
u
s
ed
a
d
if
f
er
en
t
s
eg
m
en
t
f
o
r
test
in
g
,
wh
ile
th
e
r
em
ain
in
g
k
-
1
s
eg
m
en
ts
wer
e
allo
ca
ted
f
o
r
tr
ain
i
n
g
.
T
h
e
d
elib
er
ate
p
a
r
titi
o
n
in
g
o
f
th
e
d
ataset
in
to
d
is
tin
ct
tr
ain
in
g
an
d
test
in
g
s
ets
is
im
p
lem
en
ted
to
ad
d
r
ess
th
e
p
r
o
b
lem
o
f
o
v
e
r
f
i
ttin
g
an
d
im
p
r
o
v
e
th
e
ac
cu
r
ac
y
o
f
class
if
icatio
n
.
E
ac
h
class
if
ier
was
ev
alu
ate
d
b
y
m
ea
s
u
r
in
g
th
e
p
r
o
p
o
r
tio
n
o
f
co
r
r
ec
tly
class
if
ied
in
s
tan
ce
s
o
u
t
o
f
th
e
to
tal
n
u
m
b
e
r
o
f
class
if
icatio
n
s
p
er
f
o
r
m
ed
b
y
th
e
class
if
ier
.
I
n
th
is
s
tu
d
y
,
we
em
p
lo
y
ed
v
ar
io
u
s
ML
class
if
ier
s
,
s
u
ch
a
s
R
F,
DT
,
XG
B
,
KN
N,
an
d
SVM
[
3
7
]
,
[
3
8
]
with
q
u
ad
r
atic
an
d
lin
ea
r
k
er
n
els,
to
f
o
r
ec
ast
th
e
class
if
i
ca
tio
n
o
f
twee
ts
.
E
v
er
y
al
g
o
r
it
h
m
f
u
n
ctio
n
b
ased
o
n
th
e
p
r
in
cip
les
o
f
s
u
p
er
v
is
e
d
lear
n
in
g
,
wh
ich
r
eq
u
ir
es
a
d
ataset
f
o
r
tr
ain
in
g
an
d
allo
ws
f
o
r
th
e
p
r
ed
ictio
n
o
f
class
lab
els
f
o
r
in
s
tan
ce
s
th
a
t
ar
e
n
o
t
k
n
o
wn
.
I
n
co
n
clu
s
io
n
,
a
r
an
g
e
o
f
p
er
f
o
r
m
an
ce
m
etr
ics,
in
clu
d
in
g
p
r
ec
is
io
n
,
f
-
m
ea
s
u
r
e,
ac
c
u
r
a
cy
,
an
d
r
ec
all
,
wer
e
u
tili
ze
d
to
ev
al
u
ate
th
e
ef
f
icac
y
o
f
th
e
im
p
lem
en
ted
ML
class
if
ier
s
.
T
h
is
s
tu
d
y
i
n
tr
o
d
u
ce
s
a
m
u
ltich
an
n
el
DL
f
r
am
ewo
r
k
th
at
in
teg
r
ates
th
r
ee
a
d
v
an
ce
d
m
o
d
els:
tr
an
s
f
o
r
m
er
b
lo
ck
,
B
iGR
U,
an
d
C
NN
to
en
h
an
ce
th
e
ac
c
u
r
a
cy
o
f
c
y
b
er
b
u
lly
in
g
d
etec
tio
n
b
y
lev
er
a
g
in
g
th
eir
co
m
p
lem
en
tar
y
s
tr
en
g
th
s
as
s
h
o
wn
in
Fig
u
r
e
5
.
E
ac
h
c
o
m
p
o
n
en
t
p
lay
s
a
d
is
tin
ct
r
o
le
in
ca
p
tu
r
in
g
d
if
f
er
e
n
t
asp
ec
ts
o
f
th
e
tex
tu
al
d
ata,
co
n
tr
ib
u
tin
g
to
a
h
o
lis
tic
f
ea
tu
r
e
r
ep
r
esen
tatio
n
.
T
h
e
tr
an
s
f
o
r
m
er
b
lo
ck
is
d
esig
n
ed
to
ca
p
tu
r
e
lo
n
g
-
r
an
g
e
d
ep
en
d
e
n
cies
an
d
r
elatio
n
s
h
ip
s
with
in
tex
tu
al
d
ata
u
s
in
g
s
elf
-
atten
tio
n
m
ec
h
an
is
m
s
.
B
y
p
r
o
ce
s
s
in
g
s
eq
u
e
n
ce
s
in
p
ar
all
el,
th
e
t
r
an
s
f
o
r
m
e
r
e
f
f
ec
tiv
ely
ca
p
tu
r
es
g
lo
b
al
c
o
n
tex
tu
al
in
f
o
r
m
atio
n
,
allo
win
g
th
e
m
o
d
el
to
u
n
c
o
v
er
in
tr
icat
e
p
atter
n
s
in
th
e
in
p
u
t
tex
t.
T
h
is
ca
p
ab
ilit
y
is
cr
itical
f
o
r
u
n
d
er
s
tan
d
in
g
s
u
b
tle
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
2
1
1
-
4
2
2
5
4218
cu
es
in
twee
ts
th
at
m
ay
in
d
icate
o
f
f
en
s
iv
e
b
e
h
av
io
r
.
C
o
m
p
lem
en
tin
g
th
is
,
th
e
B
iGR
U
p
r
o
ce
s
s
es
s
eq
u
en
ce
s
b
id
ir
ec
tio
n
ally
,
r
etain
in
g
co
n
tex
tu
al
in
f
o
r
m
atio
n
f
r
o
m
b
o
th
p
ast
an
d
f
u
tu
r
e
wo
r
d
s
in
a
s
en
ten
ce
.
T
h
is
b
id
ir
ec
tio
n
al
a
p
p
r
o
ac
h
en
s
u
r
e
s
th
at
th
e
m
o
d
el
co
m
p
r
eh
e
n
d
s
th
e
b
r
o
a
d
er
co
n
tex
t
ar
o
u
n
d
ea
ch
wo
r
d
.
T
h
e
B
iG
R
U’
s
o
u
tp
u
ts
ar
e
f
u
r
th
e
r
r
ef
in
ed
t
h
r
o
u
g
h
s
p
atial
d
r
o
p
o
u
t,
g
lo
b
al
av
e
r
ag
e
p
o
o
lin
g
,
a
n
d
g
lo
b
al
m
ax
i
m
u
m
p
o
o
lin
g
,
wh
ich
e
n
h
an
ce
f
ea
tu
r
e
r
eten
tio
n
wh
ile
m
in
i
m
izin
g
n
o
is
e.
T
h
e
C
NN
co
m
p
o
n
en
t
f
o
cu
s
es
o
n
ex
tr
ac
tin
g
lo
ca
l
an
d
s
p
ati
al
f
ea
tu
r
es
f
r
o
m
th
e
in
p
u
t
tex
t,
s
u
ch
as
n
-
g
r
am
s
(
e.
g
.
,
b
ig
r
am
s
an
d
t
r
ig
r
am
s
)
,
wh
ich
ar
e
cr
itical
f
o
r
id
en
tify
in
g
o
f
f
en
s
iv
e
lan
g
u
ag
e.
B
y
ap
p
ly
in
g
co
n
v
o
l
u
tio
n
al
f
ilter
s
,
th
e
C
NN
id
en
tifie
s
p
atter
n
s
th
at
o
th
e
r
m
o
d
els
m
ig
h
t
m
is
s
,
s
u
ch
as
lo
ca
lized
p
h
r
ases
o
r
wo
r
d
g
r
o
u
p
in
g
s
.
R
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
ac
tiv
atio
n
in
tr
o
d
u
ce
s
n
o
n
-
lin
ea
r
ity
t
o
e
n
h
an
ce
th
e
m
o
d
el’
s
lear
n
in
g
ca
p
ac
it
y
,
wh
ile
a
5
0
% d
r
o
p
o
u
t r
ate
r
e
d
u
ce
s
o
v
e
r
f
itti
n
g
an
d
im
p
r
o
v
es g
en
er
aliza
ti
o
n
.
T
o
ac
h
iev
e
a
u
n
if
ied
r
ep
r
esen
tatio
n
,
th
e
o
u
tp
u
ts
o
f
th
e
tr
an
s
f
o
r
m
er
b
l
o
ck
,
B
iGR
U
,
an
d
C
NN
ar
e
co
n
ca
ten
ated
i
n
to
a
s
in
g
le
f
ea
tu
r
e
v
ec
to
r
,
d
e
n
o
ted
in
(
5
)
,
w
h
er
e
T
,
B
,
an
d
C
r
ep
r
esen
t
th
e
in
d
iv
id
u
al
o
u
t
p
u
ts
o
f
th
e
r
esp
ec
tiv
e
m
o
d
els.
T
h
i
s
co
n
ca
ten
ated
f
ea
tu
r
e
v
ec
to
r
is
th
en
p
r
o
ce
s
s
ed
th
r
o
u
g
h
t
wo
f
u
lly
co
n
n
ec
ted
d
en
s
e
lay
er
s
f
o
r
f
u
r
t
h
er
f
ea
t
u
r
e
in
teg
r
atio
n
a
n
d
d
im
en
s
io
n
ali
ty
r
ed
u
ctio
n
.
T
h
e
tr
an
s
f
o
r
m
atio
n
s
ap
p
lied
b
y
th
e
d
en
s
e
lay
er
s
a
r
e
g
i
v
en
i
n
(
6
)
an
d
(
7
)
,
wh
er
e
1
an
d
2
ar
e
weig
h
t
m
atr
ices,
1
an
d
1
ar
e
b
iases
,
an
d
(
f
)
r
ep
r
esen
ts
th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
.
=
[
,
,
]
(
5
)
1
=
(
1
.
1
+
2
)
(
6
)
2
=
(
2
.
1
+
2
)
(
7
)
T
h
e
co
n
f
ig
u
r
atio
n
o
f
th
e
two
d
en
s
e
lay
er
s
,
with
6
0
an
d
3
0
n
eu
r
o
n
s
r
esp
ec
tiv
ely
,
was
d
eter
m
in
ed
th
r
o
u
g
h
em
p
ir
ical
ex
p
er
im
en
t
atio
n
an
d
h
y
p
er
p
a
r
am
eter
t
u
n
i
n
g
.
Var
io
u
s
co
n
f
ig
u
r
atio
n
s
we
r
e
test
ed
to
ac
h
iev
e
an
o
p
tim
al
b
alan
ce
b
etwe
en
m
o
d
el
c
o
m
p
lex
ity
an
d
p
er
f
o
r
m
an
ce
.
T
h
is
co
n
f
ig
u
r
atio
n
p
r
o
v
id
ed
t
h
e
b
est
r
esu
lts
in
ter
m
s
o
f
ac
c
u
r
ac
y
an
d
g
en
er
aliza
tio
n
,
as
r
ef
lecte
d
i
n
th
e
ev
alu
atio
n
m
etr
ics.
T
h
e
f
in
al
class
if
icatio
n
in
to
o
f
f
en
s
iv
e
o
r
n
o
n
-
o
f
f
en
s
iv
e
ca
t
eg
o
r
ies
is
p
er
f
o
r
m
ed
b
y
a
s
o
f
t
m
ax
ac
tiv
atio
n
lay
e
r
,
d
e
f
in
ed
in
(
8
)
.
T
o
o
p
tim
ize
th
e
m
o
d
el’
s
p
r
ed
ictio
n
s
,
th
e
b
in
ar
y
cr
o
s
s
-
en
tr
o
p
y
lo
s
s
f
u
n
ctio
n
B
C
E
is
em
p
lo
y
ed
,
wh
ich
q
u
an
tifie
s
th
e
class
if
icatio
n
er
r
o
r
.
T
h
is
is
f
o
r
m
u
lated
in
(
9
)
.
(
)
=
(
3
.
3
+
3
)
(
8
)
=
∑
l
og
(
0
(
)
)
(
9
)
T
h
e
en
d
-
to
-
e
n
d
w
o
r
k
f
l
o
w
b
e
g
in
s
with
p
r
e
p
r
o
ce
s
s
in
g
a
n
d
to
k
en
izatio
n
o
f
in
p
u
t
twee
ts
to
en
s
u
r
e
co
m
p
atib
ilit
y
with
t
h
e
DL
m
o
d
els.
T
h
e
p
r
e
p
r
o
ce
s
s
ed
tex
t
is
s
im
u
ltan
eo
u
s
ly
f
ed
in
to
th
e
tr
an
s
f
o
r
m
er
b
lo
ck
,
B
iG
R
U,
an
d
C
NN,
ea
ch
ex
tr
a
ctin
g
u
n
iq
u
e
f
ea
tu
r
es
th
at
ar
e
f
u
s
ed
in
to
th
e
co
n
ca
ten
ate
d
v
ec
to
r
v
.
T
h
is
v
ec
to
r
is
r
ef
in
ed
th
r
o
u
g
h
t
h
e
d
e
n
s
e
lay
er
s
,
an
d
th
e
f
in
al
class
if
icatio
n
is
ac
h
iev
e
d
in
th
e
s
o
f
tm
ax
lay
er
.
B
y
in
teg
r
atin
g
g
l
o
b
al,
s
eq
u
e
n
tial,
an
d
lo
ca
l
f
ea
tu
r
es,
th
is
m
u
l
tich
an
n
el
f
r
am
ewo
r
k
ca
p
t
u
r
es
a
co
m
p
r
e
h
en
s
iv
e
r
ep
r
esen
tatio
n
o
f
th
e
d
ata,
r
es
u
ltin
g
in
im
p
r
o
v
e
d
p
r
ed
ictio
n
ac
cu
r
ac
y
.
T
h
e
o
v
er
all
ar
ch
ite
ctu
r
e,
illu
s
tr
ated
in
Fig
u
r
e
5
,
h
ig
h
lig
h
ts
t
h
e
in
te
r
a
ctio
n
s
am
o
n
g
t
h
e
c
o
m
p
o
n
en
ts
an
d
d
em
o
n
s
tr
ates
h
o
w
th
eir
i
n
teg
r
atio
n
en
h
a
n
ce
s
th
e
m
o
d
el’
s
p
e
r
f
o
r
m
an
ce
in
d
e
tectin
g
o
f
f
e
n
s
iv
e
b
eh
av
i
o
r
o
n
s
o
cial
m
ed
ia
p
latf
o
r
m
s
.
Fig
u
r
e
5
.
Pro
p
o
s
ed
DL
f
r
am
e
wo
r
k
em
p
l
o
y
in
g
a
co
m
b
in
atio
n
o
f
B
iGR
U,
tr
an
s
f
o
r
m
er
u
n
it
an
d
C
NN
with
in
a
m
u
ltich
an
n
el
s
tr
u
ctu
r
e
f
o
r
in
te
llig
en
t c
y
b
er
b
u
lly
in
g
d
etec
tio
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
E
xp
lo
r
in
g
s
o
cia
l m
ed
ia
s
en
tim
en
t p
a
tter
n
s
fo
r
imp
r
o
ve
d
cy
b
e
r
b
u
llyin
g
…
(
Wa
el
Mo
h
a
me
d
S
h
a
h
er Ya
fo
o
z
)
4219
3.
I
M
P
L
E
M
E
NT
A
T
I
O
N
E
NV
I
RO
NM
E
N
T
T
h
is
s
ec
tio
n
p
r
esen
ts
a
co
m
p
r
eh
en
s
iv
e
o
v
er
v
iew
o
f
th
e
v
a
r
io
u
s
to
o
ls
an
d
tech
n
o
lo
g
ies
u
tili
ze
d
in
cr
ea
tin
g
th
e
p
r
o
p
o
s
ed
cy
b
e
r
b
u
lly
in
g
p
latf
o
r
m
(
T
ab
le
5
)
.
Py
th
o
n
v
er
s
io
n
3
.
6
was
u
tili
ze
d
as
th
e
f
u
n
d
am
en
ta
l
p
r
o
g
r
a
m
m
in
g
lan
g
u
ag
e
to
d
e
v
elo
p
th
e
p
latf
o
r
m
.
T
h
e
p
o
wer
f
u
l
in
teg
r
ated
d
e
v
elo
p
m
e
n
t
en
v
ir
o
n
m
en
t
(
I
DE
)
k
n
o
wn
as
Vis
u
al
Stu
d
io
C
o
d
e
2
0
2
0
was
em
p
lo
y
e
d
to
e
n
h
a
n
ce
th
e
f
u
n
ctio
n
ality
o
f
h
i
g
h
-
lev
el
p
r
o
g
r
am
m
in
g
in
ter
f
ac
es.
T
h
is
en
v
ir
o
n
m
en
t
p
r
o
v
id
e
d
s
u
p
p
o
r
t
f
o
r
th
e
ex
p
an
s
io
n
o
f
f
u
n
d
am
e
n
tal
lib
r
ar
i
es
s
u
ch
as
Ker
as,
T
en
s
o
r
Flo
w,
an
d
Scik
it
-
lear
n
.
T
h
e
p
r
o
ce
s
s
was in
itiated
b
y
r
etr
iev
in
g
twee
ts
f
r
o
m
th
e
X
API
u
s
in
g
th
e
t
wee
tI
n
v
i lib
r
ar
y
.
Af
ter
th
at,
th
e
twee
ts
o
b
tain
e
d
wer
e
s
u
b
j
ec
ted
to
to
k
e
n
izatio
n
u
s
in
g
t
h
e
n
ltk
.
to
k
en
ize
lib
r
ar
y
.
T
h
is
p
r
o
ce
s
s
r
esu
lted
in
t
h
e
g
en
er
atio
n
o
f
a
lis
t
o
f
to
k
en
s
.
T
h
e
to
k
en
lis
ts
wer
e
s
u
b
s
eq
u
en
tly
u
tili
ze
d
t
o
g
e
n
er
ate
n
-
g
r
am
s
(
s
p
ec
if
ically
,
uni
-
g
r
a
m
,
b
i
-
g
r
am
,
an
d
tr
i
-
g
r
am
)
b
y
em
p
lo
y
in
g
th
e
n
ltk
.
n
g
r
am
lib
r
ar
y
.
Af
ter
co
m
p
u
tin
g
th
ese
f
ea
tu
r
es,
th
e
PC
A
l
ib
r
ar
ies
wer
e
im
p
o
r
ted
in
to
th
e
d
esig
n
ated
p
latf
o
r
m
.
T
h
e
h
ig
h
est
-
p
er
f
o
r
m
i
n
g
ch
ar
a
cter
is
tics
wer
e
th
en
em
p
lo
y
ed
to
tr
ai
n
an
d
ev
al
u
at
e
d
if
f
er
e
n
t
ML
class
if
ier
s
an
d
DL
m
o
d
els
b
y
u
tili
zin
g
th
e
Ker
as,
T
en
s
o
r
Flo
w,
an
d
Scik
it
-
lear
n
lib
r
ar
ies.
Var
io
u
s
class
if
icatio
n
alg
o
r
ith
m
s
b
ased
o
n
ML
h
av
e
b
ee
n
u
tili
ze
d
to
ass
e
s
s
th
e
ef
f
icac
y
o
f
th
e
cy
b
er
b
u
ll
y
in
g
p
latf
o
r
m
u
n
d
er
co
n
s
id
er
atio
n
.
T
h
e
o
u
tco
m
es
o
f
ea
ch
class
if
icatio
n
alg
o
r
ith
m
u
n
d
er
g
o
a
r
i
g
o
r
o
u
s
v
alid
atio
n
p
r
o
ce
s
s
u
s
in
g
k
-
f
o
ld
c
r
o
s
s
-
v
alid
atio
n
.
T
ab
le
5
.
Pro
p
o
s
ed
im
p
lem
e
n
ta
tio
n
en
v
ir
o
n
m
en
t
o
f
c
y
b
er
b
u
lly
in
g
d
etec
tio
n
C
o
m
p
o
n
e
n
t
To
o
l
s
a
n
d
t
e
c
h
n
o
l
o
g
i
e
s
D
e
scri
p
t
i
o
n
U
sed
h
a
r
d
w
a
r
e
P
r
o
g
r
a
m
mi
n
g
l
a
n
g
u
a
g
e
P
y
t
h
o
n
v
e
r
s
i
o
n
3
.
6
I
D
E
V
S
C
o
d
e
1
.
7
4
V
R
A
M
1
8
G
B
0
OS
W
i
n
d
o
w
s
10
C
P
U
I
n
t
e
l
(
R
)
C
o
r
e
(
T
M
)
i
7
-
6
7
0
0
C
P
U
3
.
4
0
G
H
z
C
o
r
e
l
i
b
r
a
r
i
e
s
M
a
t
p
l
o
t
l
i
b
,
N
L
K
T
,
K
e
r
a
s
D
a
t
a
v
i
su
a
l
i
z
a
t
i
o
n
l
i
b
r
a
r
i
e
s t
e
x
t
a
n
a
l
y
s
i
s s
u
p
p
o
r
t
t
o
o
l
k
i
t
a
n
d
l
i
b
r
a
r
i
e
s NN
l
i
b
r
a
r
i
e
s s
u
p
p
o
r
t
P
a
n
d
a
s
D
a
t
a
p
r
e
p
a
r
a
t
i
o
n
P
a
n
d
a
s
l
i
b
r
a
r
i
e
s
S
k
l
e
a
r
n
ML
s
u
p
p
o
r
t
l
i
b
r
a
r
y
f
o
r
c
l
a
ss
i
f
i
c
a
t
i
o
n
Tw
e
e
t
I
n
v
i
X
A
P
I
su
p
p
o
r
t
l
i
b
r
a
r
y
4.
E
XP
E
R
I
M
E
N
T
A
L
RE
SUL
T
S
T
h
is
s
ec
tio
n
p
r
esen
ts
th
e
em
p
ir
ical
f
in
d
in
g
s
o
f
th
e
p
r
o
p
o
s
e
d
cy
b
e
r
b
u
lly
in
g
d
etec
tio
n
p
lat
f
o
r
m
.
T
h
e
p
latf
o
r
m
was
o
r
g
an
ized
b
ased
o
n
f
o
u
r
d
is
tin
ct
m
o
d
els.
T
h
e
f
ac
tu
al
d
ec
is
io
n
attr
ib
u
tes
wer
e
u
s
ed
to
f
o
r
ec
ast
twee
ts
’
ca
teg
o
r
ies
with
in
th
e
tr
ad
itio
n
al
ML
class
if
icatio
n
m
o
d
el.
Su
b
s
eq
u
en
tl
y
,
th
e
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
o
l
o
g
y
was
u
tili
ze
d
to
id
en
tify
r
ele
v
an
t
f
ea
t
u
r
es
f
o
r
f
o
r
ec
asti
n
g
o
f
f
en
s
iv
e
a
n
d
n
o
n
-
o
f
f
e
n
s
iv
e
twee
ts
.
T
h
is
id
en
tific
atio
n
was
ac
co
m
p
lis
h
ed
b
y
e
m
p
lo
y
i
n
g
th
e
af
o
r
em
en
tio
n
ed
m
et
h
o
d
s
.
T
h
e
r
esu
lts
o
f
th
ese
ex
p
er
im
en
ts
ar
e
p
r
esen
ted
in
F
ig
u
r
e
6
.
C
r
itical
an
aly
s
is
o
f
th
e
g
r
ap
h
s
h
o
ws
th
at
ea
ch
class
if
icatio
n
m
o
d
el
u
tili
zin
g
th
e
c
u
r
r
en
t
s
et
o
f
f
ea
tu
r
es
ac
h
iev
ed
a
tr
u
e
class
if
icatio
n
ac
cu
r
ac
y
r
ate
e
x
ce
e
d
in
g
7
7
%.
T
h
e
SVM
an
d
K
NN
class
if
ier
s
wer
e
n
o
tewo
r
th
y
d
u
e
to
th
eir
r
em
ar
k
ab
le
p
r
ec
is
io
n
in
p
r
e
d
ictin
g
o
u
tco
m
es.
T
h
e
SVM
class
if
ier
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
4
%,
wh
ile
th
e
KNN
clas
s
if
ier
ac
h
iev
ed
an
ac
cu
r
ac
y
o
f
9
1
.
5
%.
C
o
m
p
ar
ati
v
e
an
aly
s
is
am
o
n
g
d
if
f
er
en
t
m
o
d
els
r
ev
ea
led
an
a
cc
u
r
ac
y
o
f
8
9
%,
8
8
.
7
%,
8
7
.
2
%,
an
d
8
3
.
8
%
f
o
r
DT
,
R
F,
an
d
XGB,
r
esp
ec
tiv
ely
as p
r
esen
ted
in
T
ab
le
6
.
Fu
r
th
er
m
o
r
e
,
wh
e
n
u
tili
zin
g
a
s
u
b
s
et
o
f
f
ea
t
u
r
es,
th
e
p
r
e
d
ictiv
e
ac
cu
r
ac
y
o
f
all
ML
class
if
ier
s
s
u
r
p
ass
ed
a
th
r
esh
o
ld
o
f
7
8
.
7
%.
T
o
tr
ain
an
d
test
th
e
ML
m
o
d
els,
PC
A
an
d
PC
C
wer
e
u
s
ed
to
s
elec
t
th
e
r
elev
an
t
s
u
b
s
et
o
f
d
ec
is
io
n
f
ea
tu
r
es.
Mo
d
el
co
m
p
lex
ity
a
n
d
o
v
er
f
itti
n
g
wer
e
ef
f
ec
tiv
el
y
m
itig
ated
b
y
th
e
f
ea
tu
r
e
r
ed
u
ctio
n
ap
p
r
o
ac
h
.
T
h
e
g
r
ap
h
s
h
o
ws
th
e
p
o
s
itiv
e
i
m
p
ac
t
o
f
im
p
lem
e
n
tin
g
th
is
s
tr
ateg
y
o
n
th
e
o
v
er
all
ac
cu
r
ac
y
o
f
t
h
e
ML
m
o
d
els
.
T
h
m
o
d
el’
s
p
e
r
f
o
r
m
an
ce
was
s
u
p
er
io
r
to
o
th
er
class
if
ier
s
,
ac
h
iev
in
g
an
im
p
r
ess
iv
e
ac
cu
r
ac
y
r
ate
o
f
9
8
%.
Similar
ly
,
th
e
KNN,
DT
,
R
F,
XGB,
an
d
L
R
m
o
d
els
h
ad
an
ac
cu
r
ac
y
o
f
9
2
%,
9
2
%,
9
0
.
3
%,
8
8
%,
an
d
8
4
.
5
%,
r
esp
ec
tiv
ely
.
Mu
ltip
l
e
ass
e
s
s
m
en
t
m
etr
ics
wer
e
ca
lcu
lated
to
ev
alu
ate
th
e
m
o
d
els’
p
e
r
f
o
r
m
an
ce
,
s
u
c
h
as
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e
[
3
9
]
.
Acc
u
r
ac
y
is
d
e
f
in
ed
as
th
e
p
r
o
p
o
r
tio
n
o
f
co
r
r
ec
tly
p
r
ed
icte
d
o
cc
u
r
r
en
ce
s
(
tr
u
e
p
o
s
itiv
e)
o
u
t
o
f
th
e
to
tal
n
u
m
b
er
o
f
in
s
tan
ce
s
.
T
h
e
m
ea
s
u
r
e
o
f
ac
c
u
r
ac
y
f
o
r
th
e
m
u
lti
-
class
if
icatio
n
p
r
o
b
lem
is
s
h
o
wn
in
(
1
0
)
.
Usi
n
g
th
e
n
u
m
b
er
o
f
tr
u
e
p
o
s
itiv
e
in
s
tan
ce
s
d
iv
id
ed
b
y
th
e
to
tal
n
u
m
b
er
o
f
in
s
tan
ce
s
,
we
ca
n
ca
lcu
late
th
e
ac
cu
r
ac
y
o
f
a
m
o
d
el.
R
ec
all
q
u
an
tifie
s
th
e
p
r
o
p
o
r
tio
n
o
f
ex
p
ec
ted
p
o
s
itiv
e
o
u
tco
m
es
th
at
ar
e
ac
cu
r
ately
id
en
tifie
d
as
p
o
s
itiv
e.
I
n
a
m
u
lti
-
class
if
icat
io
n
is
s
u
e,
th
e
i
-
th
lab
el
o
f
th
e
class
is
d
eter
m
in
ed
b
y
s
u
m
m
in
g
th
e
v
alu
es in
a
co
lu
m
n
o
f
th
e
co
n
f
u
s
io
n
m
atr
ix
.
T
h
e
ca
lcu
latio
n
o
f
r
ec
all
is
d
eter
m
in
e
d
in
(
1
1)
.
A
c
c
ura
c
y
=
(
10
)
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.
1
4
,
No
.
5
,
Octo
b
er
2
0
2
5
:
4
2
1
1
-
4
2
2
5
4220
=
(1
1
)
Div
id
e
th
e
n
u
m
b
er
o
f
tr
u
e
p
o
s
itiv
es
b
y
th
e
to
tal
n
u
m
b
e
r
o
f
ac
tu
al
p
o
s
itiv
es
to
ca
lcu
late
th
e
r
ec
all.
Pre
cisi
o
n
r
ef
er
s
t
o
th
e
p
r
o
p
o
r
t
io
n
o
f
ac
c
u
r
ately
an
ticip
ated
p
o
s
itiv
e
ca
s
es
o
u
t
o
f
all
th
e
a
ctu
al
p
o
s
itiv
e
ca
s
es.
Su
m
m
in
g
th
e
v
al
u
es
in
a
co
l
u
m
n
o
f
t
h
e
co
n
f
u
s
io
n
m
atr
ix
d
eter
m
in
es
th
e
j
-
th
lab
el
o
f
t
h
e
class
in
a
m
u
lti
-
class
if
icatio
n
is
s
u
e.
Acc
u
r
ac
y
is
d
ef
in
ed
as
th
e
r
atio
o
f
th
e
n
u
m
b
er
o
f
tr
u
e
p
o
s
itiv
e
p
r
e
d
ictio
n
s
to
th
e
n
u
m
b
er
o
f
tr
u
e
n
eg
ativ
e
p
r
ed
ictio
n
s
.
T
h
e
(
T
r
u
ePo
s
itiv
e
j
)
p
r
e
d
icted
p
o
s
itiv
e
in
s
tan
ce
s
as
a
p
r
o
p
o
r
tio
n
o
f
th
e
to
tal
(
T
o
talPre
d
icted
Po
s
itiv
e
j
)
as
c
alcu
lated
in
(
1
2)
.
L
et
m
b
e
a
m
atr
ix
,
w
h
er
e
i
r
ep
r
esen
ts
t
h
e
r
o
ws
(
p
r
ed
icted
lab
el)
an
d
j
r
ep
r
esen
ts
th
e
co
lu
m
n
s
(
ac
tu
al
lab
el)
.
=
(1
2
)
T
o
ev
alu
ate
t
h
e
r
o
b
u
s
tn
ess
an
d
g
en
e
r
aliza
tio
n
ca
p
a
b
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
m
o
d
el,
we
em
p
l
o
y
ed
k
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
k
=1
0
.
I
n
th
i
s
m
eth
o
d
,
th
e
d
ataset
was
d
i
v
id
ed
i
n
to
1
0
eq
u
al
s
u
b
s
ets,
o
r
f
o
ld
s
.
Fo
r
ea
ch
iter
atio
n
,
o
n
e
-
f
o
ld
was
u
s
ed
as
th
e
tes
tin
g
s
et,
wh
ile
th
e
r
em
ain
in
g
k
-
1
f
o
ld
s
wer
e
u
s
ed
f
o
r
tr
ain
in
g
.
T
h
is
p
r
o
ce
s
s
was
r
ep
ea
ted
1
0
tim
e
s
,
en
s
u
r
in
g
th
at
e
v
er
y
d
ata
p
o
in
t
was
u
s
ed
f
o
r
b
o
th
t
r
ain
in
g
an
d
test
in
g
ex
ac
tly
o
n
ce
.
T
h
e
p
er
f
o
r
m
a
n
ce
m
etr
ics,
in
clu
d
in
g
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
an
d
F1
-
s
co
r
e,
wer
e
av
er
ag
ed
ac
r
o
s
s
all
f
o
ld
s
to
p
r
o
v
id
e
a
c
o
m
p
r
e
h
en
s
iv
e
ev
alu
atio
n
o
f
t
h
e
m
o
d
el'
s
p
er
f
o
r
m
an
ce
.
T
h
is
ap
p
r
o
ac
h
m
in
im
izes
th
e
r
is
k
o
f
o
v
er
f
itti
n
g
a
n
d
p
r
o
v
i
d
es a
r
eliab
le
esti
m
ate
o
f
th
e
m
o
d
el'
s
ef
f
ec
tiv
en
ess
ac
r
o
s
s
d
if
f
er
en
t d
a
ta
s
p
lits
.
I
n
T
ab
le
6
,
b
o
th
th
e
wh
o
le
an
d
s
im
p
lifie
d
f
ea
tu
r
e
s
ets
ar
e
ill
u
s
tr
ated
f
o
r
th
e
ex
p
er
im
e
n
tal
m
o
d
els.
I
n
ter
m
s
o
f
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
a
n
d
r
ec
all,
t
h
e
SVM
co
n
s
is
ten
tly
o
u
tp
e
r
f
o
r
m
ed
o
th
er
class
if
icatio
n
m
o
d
els
in
all
class
es.
On
av
er
ag
e,
th
e
SV
M
co
n
s
is
ten
tly
s
h
o
ws
th
e
h
i
g
h
est
lev
el
o
f
ac
c
u
r
ac
y
.
I
n
co
n
tr
ast,
th
e
KNN
ex
h
ib
ited
im
p
r
o
v
e
d
p
r
e
d
ictio
n
o
u
tco
m
es,
s
p
ec
if
ically
in
r
e
latio
n
to
m
em
o
r
y
.
Ho
wev
e
r
,
L
R
h
ad
th
e
lo
west
p
er
f
o
r
m
an
ce
c
o
m
p
a
r
ed
to
th
e
o
th
er
class
if
icatio
n
m
o
d
els.
I
t
s
h
o
wed
p
o
o
r
p
e
r
f
o
r
m
an
ce
ac
r
o
s
s
all
ass
ess
m
en
t
m
ea
s
u
r
es,
ir
r
esp
ec
tiv
e
o
f
e
m
p
l
o
y
in
g
f
u
ll o
r
r
ed
u
ce
d
d
ec
is
io
n
f
ea
tu
r
e
s
ets.
Fig
u
r
e
6
.
T
wee
ts
class
if
icatio
n
(
o
f
f
e
n
s
iv
e
an
d
n
o
n
-
o
f
f
e
n
s
iv
e)
r
esu
lts
u
s
in
g
ac
tu
al
an
d
r
ed
u
c
ed
f
ea
tu
r
es
s
et
T
ab
le
6
.
E
v
alu
atin
g
class
if
icatio
n
m
o
d
els b
ased
o
n
b
o
th
f
u
ll
an
d
r
ed
u
ce
d
f
ea
tu
r
es
C
l
a
s
si
f
i
e
r
s
\
m
e
t
r
i
c
s
A
c
t
u
a
l
f
e
a
t
u
r
e
s se
t
p
e
r
f
o
r
ma
n
c
e
me
a
s
u
r
e
s
R
e
d
u
c
e
d
f
e
a
t
u
r
e
s s
e
t
p
e
r
f
o
r
ma
n
c
e
m
e
a
su
r
e
s
F1
A
c
c
u
r
a
c
y
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
F1
A
c
c
u
r
a
c
y
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
S
V
M
0
.
9
4
94
0
.
9
3
5
0
.
9
2
1
0
.
9
8
98
0
.
9
6
1
0
.
9
8
1
K
N
N
0
.
9
3
9
1
.
5
0
.
9
0
3
0
.
8
9
5
0
.
9
2
92
0
.
9
0
5
0
.
9
3
2
DT
0
.
9
0
89
0
.
9
0
1
0
.
8
9
2
0
.
9
2
92
0
.
9
0
1
0
.
8
8
1
RF
0
.
9
1
8
8
.
7
0
.
8
9
0
.
8
6
0
.
9
1
9
0
.
3
0
.
8
9
0
0
.
8
9
1
XGB
0
.
8
9
8
7
.
2
0
.
8
8
1
0
.
8
7
5
2
0
.
8
9
88
0
.
8
5
3
0
.
8
6
5
LR
0
.
8
4
8
3
.
8
0
.
8
4
2
0
.
8
2
1
0
.
8
6
8
4
.
5
0
.
8
2
4
0
.
8
2
3
Evaluation Warning : The document was created with Spire.PDF for Python.