I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
13
,
N
o.
4
,
D
e
c
e
m
be
r
2024
, pp.
4526
~
4536
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
13
.i
4
.pp
4526
-
4536
4526
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
Ph
i
sh
i
n
g
d
e
t
e
c
t
i
on
u
si
n
g
c
l
u
st
e
r
i
n
g
a
n
d
m
ac
h
i
n
e
l
e
ar
n
i
n
g
L
u
ai
Al
-
S
h
al
ab
i,
Y
ah
ia
H
as
an
Jaz
yah
F
a
c
ul
t
y
of
C
om
put
e
r
S
t
udi
e
s
,
A
r
a
b
O
pe
n
U
ni
ve
r
s
i
t
y,
A
r
di
ya
,
K
uw
a
i
t
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
D
e
c
28
,
2023
R
e
vi
s
e
d
M
a
y
16
,
2024
A
c
c
e
pt
e
d
J
un
1
,
2024
Phishing
is
a
prevalent
and
evolvin
g
cyber
threat
that
continu
es
to
exploit
human
vulnerability
to
deceive
individuals
and
organizations
into
revealing
sensitive
information.
Phishing
attacks
encompass
a
range
of
tactics,
from
deceptive
emails
and
fraudulent
websites
to
social
engineerin
g
tech
niques.
Traditional
methods
of
detection,
such
as
signature
-
based
approach
es
and
rule
-
based
filtering,
have
proven
to
be
limited
in
their
effectiven
ess,
as
attackers
frequently
adapt
and
create
new,
previously
unseen
p
hishing
campaigns
.
Conseque
ntly,
there
is
a
growing
need
for
more
sophisticat
ed
and
adaptable
detection
methods.
In
recent
years,
machine
learning
(M
L)
and
artificial
intelligence
(AI)
have
played
a
significant
role
in
enhancing
p
hishing
detection.
These
technologie
s
leverag
e
large
datasets
to
train
models
c
apable
of
recognizing
subtle
patterns
and
anomalies
in
both
email
content
and
website
behavior.
This
research
proposes
a
hybrid
algorit
hm
to
detect
phishing
attacks
based
on
clusteri
ng
and
classifi
cation
machine
learning
methods
(CMLM):
deep
learning
(DL)
and
decision
tree
(DT).
Simulation
results
show
that
the
proposed
technique
achieves
a
high
percentage
of
accuracy
in
detecting
phishing.
K
e
y
w
o
r
d
s
:
A
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
D
e
c
is
io
n
t
r
e
e
D
e
e
p
l
e
a
r
ni
ng
M
a
c
hi
ne
l
e
a
r
ni
ng
P
hi
s
hi
ng
This
is
an
open
access
article
under
the
CC
BY
-
SA
license.
C
or
r
e
s
pon
di
n
g
A
u
th
or
:
Y
a
hi
a
H
a
s
a
n
J
a
z
y
a
h
F
a
c
ul
ty
of
C
om
put
e
r
S
tu
di
e
s
,
A
r
a
b
O
pe
n
U
ni
ve
r
s
it
y
S
t.
M
oha
m
m
e
d
N
a
z
z
a
l
Al
-
M
o
a
s
s
a
b,
A
r
di
ya
,
K
uw
a
it
E
m
a
il
:
ya
hi
a
@
a
ou.e
du.kw
1.
I
N
T
R
O
D
U
C
T
I
O
N
P
hi
s
hi
ng
is
a
pe
r
va
s
iv
e
a
nd
in
s
id
io
u
s
f
or
m
of
c
ybe
r
c
r
im
e
th
a
t
pr
e
ys
on
hum
a
n
ps
yc
hol
ogy
a
nd
te
c
hni
c
a
l
vul
ne
r
a
bi
li
ti
e
s
.
It
in
vol
ve
s
th
e
u
s
e
of
de
c
e
pt
iv
e
te
c
hni
que
s
to
tr
ic
k
in
di
vi
dua
ls
or
or
ga
ni
z
a
ti
ons
in
to
di
vul
gi
ng
s
e
ns
it
iv
e
in
f
or
m
a
ti
on,
s
uc
h
as
lo
gi
n
c
r
e
d
e
nt
ia
ls
,
f
in
a
n
c
ia
l
de
ta
il
s
,
or
pe
r
s
ona
l
da
t
a
.
P
hi
s
hi
ng
a
tt
a
c
k
s
a
r
e
of
te
n
th
e
in
it
ia
l
e
nt
r
y
poi
nt
f
o
r
br
oa
de
r
c
ybe
r
th
r
e
a
ts
,
in
c
lu
di
ng
id
e
nt
it
y
th
e
f
t,
f
r
a
ud,
a
nd
m
a
lwa
r
e
in
f
e
c
ti
ons
.
To
c
om
ba
t
th
is
gr
ow
in
g
m
e
na
c
e
,
e
f
f
e
c
ti
ve
phi
s
hi
ng
de
te
c
ti
on
m
e
th
ods
ha
ve
be
c
om
e
in
di
s
pe
n
s
a
bl
e
.
T
he
s
ophi
s
ti
c
a
ti
on
of
phi
s
hi
ng
a
tt
a
c
k
s
c
ont
in
ue
s
to
e
vol
v
e
,
m
a
ki
ng
it
a
c
ha
ll
e
ngi
ng
ta
s
k
to
th
w
a
r
t
th
e
s
e
th
r
e
a
ts
.
C
ybe
r
c
r
im
in
a
ls
ut
il
iz
e
a
v
a
r
ie
ty
of
ta
c
ti
c
s
,
in
c
lu
di
ng
m
is
le
a
di
ng
e
m
a
il
s
,
f
r
a
udul
e
nt
w
e
b
s
it
e
s
,
a
nd
s
oc
ia
l
e
ngi
ne
e
r
in
g
s
tr
a
te
gi
e
s
th
a
t
e
xpl
oi
t
hum
a
n
tr
us
t
a
nd
c
ur
io
s
it
y.
T
he
dyna
m
ic
na
tu
r
e
of
th
e
s
e
a
tt
a
c
ks
m
e
a
ns
th
a
t
tr
a
di
ti
ona
l,
s
ta
ti
c
s
e
c
ur
it
y
m
e
a
s
ur
e
s
a
r
e
of
te
n
in
e
f
f
e
c
ti
ve
.
T
hi
s
ha
s
le
d
to
th
e
de
v
e
lo
pm
e
nt
of
a
dva
nc
e
d
a
nd
a
da
pt
iv
e
te
c
hni
que
s
f
or
de
te
c
ti
ng
a
nd
m
it
ig
a
ti
ng
phi
s
hi
ng
a
tt
e
m
pt
s
.
P
hi
s
hi
ng
de
te
c
ti
on
in
vol
ve
s
th
e
id
e
nt
if
ic
a
ti
on
a
nd
pr
e
v
e
nt
io
n
of
de
c
e
pt
iv
e
or
m
a
li
c
io
us
c
ont
e
nt
w
it
hi
n
e
m
a
il
s
,
w
e
bs
it
e
s
,
or
ot
he
r
di
gi
ta
l
c
om
m
uni
c
a
ti
on
c
ha
nne
ls
.
It
e
nc
om
pa
s
s
e
s
a
br
oa
d
s
pe
c
tr
um
of
m
e
th
ods
,
r
a
ngi
ng
f
r
om
r
ul
e
-
ba
s
e
d
f
il
te
r
s
a
nd
s
ig
na
tu
r
e
-
ba
s
e
d
s
ys
te
m
s
to
m
or
e
a
dv
a
nc
e
d
a
ppr
oa
c
he
s
th
a
t
l
e
ve
r
a
ge
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
,
m
a
c
hi
ne
le
a
r
ni
ng
(
M
L
)
,
a
nd
be
ha
vi
or
a
l
a
na
ly
s
is
.
As
c
ybe
r
c
r
im
in
a
ls
c
ons
ta
nt
ly
r
e
f
in
e
th
e
ir
ta
c
ti
c
s
to
bypa
s
s
c
onve
nt
io
na
l
de
f
e
ns
e
s
,
th
e
ne
e
d
f
or
in
nova
ti
ve
a
nd
r
e
s
pons
iv
e
de
te
c
ti
on
m
e
c
ha
ni
s
m
s
h
a
s
be
c
om
e
in
c
r
e
a
s
in
gl
y
pr
e
s
s
in
g.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
P
hi
s
hi
ng de
te
c
ti
on us
in
g c
lu
s
te
r
in
g and mac
hi
ne
l
e
a
r
ni
ng
(
L
uai
A
l
-
Shal
abi
)
4527
In
th
is
c
ont
e
xt
,
th
is
pa
pe
r
e
xpl
or
e
s
th
e
la
nds
c
a
pe
of
phi
s
hi
ng
d
e
te
c
ti
on,
a
ddr
e
s
s
in
g
bot
h
th
e
e
xi
s
ti
ng
c
ha
ll
e
nge
s
a
nd
th
e
la
te
s
t
a
dva
n
c
e
m
e
nt
s
in
th
e
f
ie
ld
.
A
nd
pr
op
os
in
g
a
hybr
id
a
lg
or
it
hm
th
a
t
m
e
r
ge
s
be
tw
e
e
n
c
lu
s
te
r
in
g
a
nd
c
la
s
s
if
ic
a
ti
on
us
in
g
de
e
p
le
a
r
ni
ng
(
DL
)
a
nd
de
c
is
io
n
tr
e
e
(
DT
)
f
or
two
di
f
f
e
r
e
nt
d
a
ta
s
e
ts
(
D
S
s
)
.
T
he
m
a
in
c
ont
r
ib
ut
io
ns
of
th
is
w
or
k
a
r
e
s
um
m
a
r
iz
e
d
in
th
r
e
e
-
f
ol
d:
−
A
r
obus
t
hybr
id
a
lg
or
it
hm
to
de
te
c
t
phi
s
hi
ng
a
tt
a
c
ks
us
in
g
c
lu
s
te
r
in
g,
c
la
s
s
if
ic
a
ti
on,
a
nd
s
ta
bi
li
ty
-
c
or
r
e
la
ti
on
a
nd
c
or
r
e
la
ti
on
(
S
c
C
)
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
w
a
s
pr
opos
e
d
c
ons
id
e
r
in
g
th
e
s
pe
e
d
a
nd
s
im
pl
ic
it
y.
−
T
hor
ough
s
ta
ti
s
ti
c
a
l
a
n
a
ly
s
is
of
th
e
pr
opos
e
d
m
e
th
od
us
in
g
w
e
ll
-
known
phi
s
hi
ng
da
ta
s
e
ts
.
−
A
c
om
pa
r
is
on
be
twe
e
n
th
e
pr
opos
e
d
m
e
th
od
a
nd
ot
he
r
w
e
ll
-
known
m
e
th
ods
f
or
de
te
c
ti
ng
phi
s
hi
ng
w
e
bs
it
e
s
th
a
t
a
r
e
pr
e
s
e
nt
e
d
in
th
e
li
te
r
a
tu
r
e
,
m
e
th
ods
th
a
t
u
s
e
DL
a
nd
DT
to
ge
th
e
r
w
it
h
S
c
C
,
r
ough
s
e
t
(
RS
)
,
a
nd
pr
in
c
ip
a
l
c
om
pone
nt
a
na
ly
s
is
(
P
C
A
)
,
a
nd
m
e
th
ods
th
a
t
us
e
DL
a
nd
DT
to
ge
th
e
r
w
it
h
RS
-
K
-
m
e
a
n
a
nd
P
C
A
-
K
-
m
e
a
n.
T
he
r
e
m
a
in
in
g
of
th
is
a
r
ti
c
le
is
or
ga
ni
z
e
d
as
f
ol
lo
w
s
:
s
e
c
ti
on
2
pr
e
s
e
nt
s
c
om
pa
r
is
ons
be
tw
e
e
n
phi
s
hi
ng
de
te
c
ti
on
A
lg
or
it
hm
s
,
s
e
c
ti
on
3
is
pr
e
li
m
in
a
r
ie
s
a
bout
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
,
s
e
c
ti
on
4
pr
e
s
e
nt
s
th
e
pr
opos
e
d
a
lg
or
it
hm
,
s
e
c
ti
on
5
pr
e
s
e
nt
s
th
e
c
om
pl
e
xi
ty
of
t
he
pr
opos
e
d
a
lg
or
it
hm
,
a
nd
s
e
c
ti
on
6
is
th
e
c
onc
lu
s
io
n.
2.
P
H
I
S
H
I
N
G
D
E
T
E
C
T
I
O
N
A
L
G
O
R
I
T
H
M
S
’
C
O
M
P
A
R
I
S
O
N
S
In
li
te
r
a
tu
r
e
,
pl
e
nt
y
of
phi
s
hi
ng
de
te
c
ti
on
a
lg
or
it
hm
s
w
e
r
e
de
ve
l
ope
d.
It
is
im
por
ta
nt
to
unde
r
s
ta
nd
th
e
a
dva
nt
a
ge
s
a
nd
di
s
a
dva
nt
a
ge
s
of
th
e
m
.
T
a
bl
e
1
s
um
m
a
r
is
e
s
th
e
se
a
dva
nt
a
ge
s
a
nd
di
s
a
dva
nt
a
g
e
s
,
w
hi
le
T
a
bl
e
2
pr
e
s
e
nt
s
a
c
om
pa
r
is
on
b
e
twe
e
n
di
f
f
e
r
e
nt
m
e
th
ods
f
or
p
hi
s
hi
ng
w
e
bs
it
e
d
e
te
c
ti
on
m
e
th
ods
in
te
r
m
s
of
a
c
c
ur
a
c
y,
w
hi
c
h
is
a
m
e
tr
ic
th
a
t
m
e
a
s
ur
e
s
how
w
e
ll
a
phi
s
hi
ng
de
te
c
ti
on
s
ys
t
e
m
or
a
lg
or
it
hm
c
or
r
e
c
tl
y
id
e
nt
if
ie
s
a
nd
c
la
s
s
if
ie
s
phi
s
hi
ng
e
m
a
il
s
or
w
e
b
s
it
e
s
.
3.
P
R
E
L
I
M
I
N
A
R
I
E
S
3.1.
F
e
at
u
r
e
s
e
le
c
t
io
n
m
e
t
h
od
s
F
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
s
f
or
r
e
duc
in
g
th
e
s
iz
e
of
da
t
a
s
e
t
s
a
r
e
c
la
s
s
if
ie
d
in
to
th
r
e
e
gr
oups
:
f
il
te
r
,
w
r
a
ppe
r
,
a
nd
e
m
be
dde
d
[
1
],
[
2
].
T
he
f
il
te
r
m
e
th
od
c
a
lc
ul
a
te
s
a
s
c
or
e
f
or
each
f
e
a
tu
r
e
a
nd
a
ll
f
e
a
tu
r
e
s
w
it
h
s
c
or
e
s
m
or
e
th
a
n
a
pr
e
-
de
f
in
e
d
th
r
e
s
hol
d
v
a
lu
e
a
r
e
c
hos
e
n.
On
t
he
ot
he
r
ha
nd,
w
r
a
ppe
r
m
e
th
od
s
us
e
a
c
l
a
s
s
if
ie
r
to
e
va
lu
a
te
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
va
r
io
us
r
e
duc
ts
a
nd
c
hoos
e
th
e
be
s
t
of
th
e
m
.
It
is
m
or
e
pow
e
r
f
ul
th
a
n
f
il
te
r
m
e
th
ods
,
but
it
is
a
l
s
o
m
or
e
c
om
pl
e
x.
C
onv
e
r
s
e
ly
,
to
w
r
a
ppe
r
m
e
th
ods
,
e
m
be
dd
e
d
m
e
th
ods
ju
dge
f
e
a
tu
r
e
s
e
le
c
ti
on
in
th
e
tr
a
in
in
g
pr
oc
e
dur
e
.
F
i
lt
e
r
m
e
th
ods
w
e
r
e
a
pp
li
e
d
f
or
f
e
a
tu
r
e
s
e
le
c
ti
on
to
s
e
le
c
t
th
e
m
os
t
pr
e
s
e
nt
in
g
a
tt
r
ib
ut
e
s
th
a
t
ha
ve
th
e
hi
ghe
s
t
in
f
or
m
a
ti
on
in
a
da
ta
s
e
t
th
a
t
can
di
s
ti
ngui
s
h
be
twe
e
n
c
la
s
s
e
s
.
T
he
a
ppl
ie
d
m
e
th
ods
a
r
e
e
xpl
a
in
e
d
ne
xt
.
V
a
r
io
us
s
im
pl
e
ML
a
lg
or
it
hm
s
m
a
y
w
or
k
jo
in
tl
y
(
known
as
hybr
id
)
to
c
om
pl
e
m
e
nt
a
nd
e
nha
nc
e
each
ot
he
r
,
T
a
bl
e
3
pr
e
s
e
nt
s
a
c
om
pa
r
is
on
be
twe
e
n
di
f
f
e
r
e
nt
hybr
id
m
e
th
ods
f
o
r
phi
s
hi
ng
w
e
bs
it
e
de
te
c
ti
on
m
e
th
ods
in
te
r
m
s
of
a
c
c
ur
a
c
y.
T
a
bl
e
1. P
r
os
a
nd c
on
s
of
phi
s
hi
ng de
te
c
ti
on a
lg
or
it
hm
s
A
l
gor
i
t
hm
A
dva
nt
a
ge
D
i
s
a
dva
nt
a
ge
R
ul
e
-
ba
s
e
d de
t
e
c
t
i
on
S
i
m
pl
e
t
o
i
m
pl
e
m
e
nt
,
c
a
n
be
e
f
f
e
c
t
i
ve
f
or
know
n
phi
s
hi
ng pa
t
t
e
r
ns
L
i
m
i
t
e
d
t
o
pr
e
de
f
i
ne
d
r
u
l
e
s
,
s
t
r
uggl
e
s
w
i
t
h
ne
w
a
nd e
vol
vi
ng phi
s
hi
ng t
a
c
t
i
c
s
S
i
gna
t
ur
e
-
ba
s
e
d de
t
e
c
t
i
on
E
f
f
e
c
t
i
ve
f
or
know
n
phi
s
hi
ng
t
hr
e
a
t
s
,
c
a
n
qui
c
kl
y
i
de
nt
i
f
y know
n pa
t
t
e
r
ns
I
ne
f
f
e
c
t
i
ve
a
ga
i
ns
t
z
e
r
o
-
da
y
a
t
t
a
c
k
s
,
c
a
nnot
a
da
pt
t
o ne
w
t
a
c
t
i
c
s
M
L
a
nd
A
I
-
ba
s
e
d de
t
e
c
t
i
on
E
f
f
e
c
t
i
ve
a
t
de
t
e
c
t
i
ng
e
vol
vi
ng
a
nd
ne
w
phi
s
hi
ng
t
hr
e
a
t
s
, c
a
n a
da
pt
t
o c
ha
ngi
ng t
a
c
t
i
c
s
, c
a
n
a
na
l
yz
e
l
a
r
ge
da
t
a
s
e
t
s
f
or
pa
t
t
e
r
ns
R
e
qui
r
e
s
s
ubs
t
a
nt
i
a
l
da
t
a
f
or
t
r
a
i
ni
ng, m
a
y be
vul
ne
r
a
bl
e
t
o a
dve
r
s
a
r
i
a
l
a
t
t
a
c
ks
B
e
ha
vi
our
a
l
a
na
l
ys
i
s
E
f
f
e
c
t
i
ve
a
t
i
de
nt
i
f
yi
ng a
nom
a
l
ous
be
ha
vi
our
, c
a
n
de
t
e
c
t
z
e
r
o
-
da
y a
t
t
a
c
ks
M
a
y
pr
oduc
e
f
a
l
s
e
pos
i
t
i
ve
s
,
c
a
n
be
c
om
pl
e
x
t
o i
m
pl
e
m
e
nt
U
R
L
a
na
l
y
s
i
s
C
a
n de
t
e
c
t
de
c
e
pt
i
ve
U
R
L
s
a
nd dom
a
i
n
s
poof
i
ng
L
i
m
i
t
e
d
t
o
U
R
L
a
na
l
ys
i
s
,
m
a
y
not
de
t
e
c
t
ot
he
r
a
s
pe
c
t
s
of
phi
s
hi
ng.
C
ont
e
nt
a
na
l
ys
i
s
E
f
f
e
c
t
i
ve
a
t
de
t
e
c
t
i
ng
de
c
e
pt
i
ve
l
a
ngu
a
ge
a
nd
t
a
c
t
i
c
s
i
n e
m
a
i
l
s
a
nd w
e
bs
i
t
e
s
M
a
y
not
de
t
e
c
t
m
or
e
s
ophi
s
t
i
c
a
t
e
d
phi
s
hi
ng
a
t
t
a
c
ks
B
l
a
c
kl
i
s
t
s
a
nd
r
e
put
a
t
i
on
-
ba
s
e
d
a
ppr
oa
c
he
s
Q
ui
c
k
t
o
i
m
pl
e
m
e
nt
,
c
a
n
bl
oc
k
know
n
m
a
l
i
c
i
ous
e
nt
i
t
i
e
s
I
ne
f
f
e
c
t
i
ve
a
ga
i
ns
t
n
e
w
t
hr
e
a
t
s
,
m
a
y
pr
oduc
e
f
a
l
s
e
pos
i
t
i
ve
s
H
e
ur
i
s
t
i
c
-
ba
s
e
d de
t
e
c
t
i
on
E
f
f
e
c
t
i
ve
a
t
i
de
nt
i
f
yi
ng
s
us
pi
c
i
ous
f
or
m
s
a
nd
r
e
que
s
t
s
f
or
s
e
ns
i
t
i
ve
i
nf
or
m
a
t
i
on
M
a
y
pr
oduc
e
f
a
l
s
e
pos
i
t
i
ve
s
,
l
i
m
i
t
e
d
t
o
he
ur
i
s
t
i
c
-
ba
s
e
d r
ul
e
s
R
e
a
l
-
t
i
m
e
a
na
l
ys
i
s
C
a
n de
t
e
c
t
a
nd bl
oc
k phi
s
hi
ng a
t
t
a
c
k
s
i
n r
e
a
l
t
i
m
e
M
a
y
r
e
qui
r
e
s
ubs
t
a
nt
i
a
l
c
om
put
a
t
i
ona
l
r
e
s
our
c
e
s
, c
a
n be
r
e
s
our
c
e
-
i
nt
e
ns
i
ve
H
ybr
i
d a
ppr
oa
c
he
s
C
om
bi
ne
m
ul
t
i
pl
e
de
t
e
c
t
i
on
m
e
t
hods
t
o
i
m
pr
ove
a
c
c
ur
a
c
y
M
a
y
be
m
or
e
c
om
pl
e
x
t
o
i
m
pl
e
m
e
nt
,
r
e
qui
r
e
ongoi
ng t
uni
ng
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4526
-
4536
4528
T
a
bl
e
2.
A
c
c
ur
a
c
y
of
non
-
tr
a
di
ti
ona
l
m
e
th
ods
f
or
phi
s
hi
ng
w
e
bs
it
e
s
de
te
c
ti
on
T
a
bl
e
3.
A
c
c
ur
a
c
y
of
hybr
id
m
e
th
ods
f
or
phi
s
hi
ng
w
e
bs
it
e
s
de
t
e
c
ti
on
A
nt
i
-
phi
s
hi
ng
m
e
t
ho
d
A
ut
hor
s
T
e
c
hni
que
s
D
a
t
a
s
e
t
A
c
c
ur
a
c
y
(%)
H
ybr
i
d
M
e
t
hods
A
l
i
a
nd
A
hm
e
d
[
23]
D
eep
ne
ur
a
l
ne
t
w
or
ks
(
D
N
N
s
)
a
nd
ge
ne
t
i
c
a
l
gor
i
t
hm
(
G
A
)
U
C
I
phi
s
hi
ng
w
e
bs
i
t
e
s
[
12]
91.13
Z
hu
et
al
.
[
24]
DT
a
nd
opt
i
m
a
l
f
e
a
t
ur
e
s
ba
s
e
d
a
r
t
i
f
i
c
i
a
l
ne
ur
a
l
ne
t
w
or
k
(
A
N
N
)
,
K
-
m
e
doi
ds
c
l
us
t
e
r
i
ng
a
l
gor
i
t
hm
U
C
I
[
11]
,
[
12]
,
P
hi
s
h
T
a
nk
[
6]
,
A
l
e
xa
[
4]
95.76
S
ul
e
m
a
n
a
nd
A
w
a
n
[
25]
I
t
e
r
a
t
i
ve
d
i
c
hot
om
i
s
e
r
-
3
(
I
D
3
)
a
nd
ye
t
a
not
he
r
ge
ne
r
a
t
i
ng
ge
ne
t
i
c
a
l
gor
i
t
hm
(
Y
A
G
G
A
)
U
C
I
ML
w
e
b
s
i
t
e
[
11]
,
[
12]
95
V
r
ba
nč
i
č
et
al
.
[
26]
B
at
a
l
gor
i
t
hm
(
B
A
)
a
nd
hybr
i
d
ba
t
a
l
gor
i
t
hm
(
H
B
A
)
U
C
I
[
12]
96.5
C
hi
n
et
a
l
.
[
27]
D
e
e
p
pa
c
ke
t
i
ns
p
e
c
t
i
on
(
D
P
I
)
,
s
of
t
w
a
r
e
-
de
f
i
ne
d
ne
t
w
or
ki
ng
(
S
D
N
)
a
nd
ANN
U
C
I
[
12]
98.39
C
he
n
et
al
.
[
28]
P
a
r
t
i
c
l
e
s
w
a
r
m
opt
i
m
i
z
a
t
i
on
(
P
S
O
)
a
nd
ba
c
k
pr
opa
ga
t
i
on (
BP
)
ne
ur
a
l
ne
t
w
or
k
P
hi
s
ht
a
nk
[
6]
98.95
3.
1.1
.
R
ou
gh
s
et
t
h
e
or
y
[
29
]
It
is
a
m
a
th
e
m
a
ti
c
a
l
f
r
a
m
e
w
or
k
a
nd
s
e
t
of
pr
in
c
ip
le
s
us
e
d
f
or
da
ta
a
na
ly
s
is
a
nd
f
e
a
tu
r
e
s
e
le
c
ti
on.
T
h
e
c
or
e
id
e
a
of
RS
is
to
ha
ndl
e
unc
e
r
ta
in
ty
a
nd
va
gue
n
e
s
s
in
d
a
ta
.
It
w
or
ks
w
it
h
in
c
om
pl
e
te
,
im
pr
e
c
i
s
e
,
or
in
c
ons
is
te
nt
in
f
or
m
a
ti
on
to
a
ppr
oxi
m
a
te
a
nd
r
e
a
s
on
a
bout
da
ta
.
It
id
e
nt
if
ie
s
th
e
m
os
t
r
e
le
va
nt
f
e
a
tu
r
e
s
in
a
DS
w
hi
le
m
in
im
iz
in
g
in
f
or
m
a
ti
on
lo
s
s
.
T
h
e
pr
im
a
r
y
c
onc
e
pt
be
hi
nd
RS
f
e
a
tu
r
e
s
e
le
c
ti
on
is
to
pa
r
ti
ti
on
th
e
da
t
a
in
to
e
qui
va
le
nc
e
c
la
s
s
e
s
ba
s
e
d
on
th
e
v
a
lu
e
s
of
a
pa
r
ti
c
ul
a
r
f
e
a
tu
r
e
a
nd
a
na
ly
z
e
th
e
d
e
pe
nde
nc
y
of
th
e
c
la
s
s
la
be
ls
on
th
a
t
f
e
a
tu
r
e
.
T
he
pr
im
a
r
y
(
1)
in
vo
lv
e
d
in
th
e
r
ough
s
e
t
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
is
th
e
de
pe
nde
nc
y
s
c
or
e
, w
hi
c
h m
e
a
s
ur
e
s
t
he
i
m
por
ta
nc
e
of
a
f
e
a
tu
r
e
i
n c
la
s
s
if
yi
ng
da
ta
.
(
,
)
=
|
SI
|
|
U
|
∗
|
SI
|
−
|
SI
|
_
A
|
SI
|
(
1)
W
he
r
e
D
e
pe
nde
nc
y
(
S
I
,
A)
is
th
e
de
pe
nde
nc
y
s
c
or
e
of
f
e
a
tu
r
e
A
c
onc
e
r
ni
ng
th
e
s
e
t
of
in
s
ta
nc
e
s
S
I
,
SI
is
th
e
s
e
t
of
in
s
ta
nc
e
s
f
or
w
hi
c
h
th
e
f
e
a
tu
r
e
is
e
va
lu
a
te
d,
A
is
th
e
f
e
a
t
ur
e
f
or
w
hi
c
h
th
e
de
pe
nde
nc
y
is
m
e
a
s
ur
e
d,
|S
I
|
is
th
e
num
be
r
of
in
s
ta
nc
e
s
in
S
I
,
|U
|
is
th
e
to
ta
l
num
be
r
of
in
s
ta
nc
e
s
in
th
e
da
ta
s
e
t,
|S
I
|_A
is
th
e
num
be
r
of
di
s
ti
nc
t
va
lu
e
s
of
f
e
a
tu
r
e
A
in
S
I
.
T
he
de
pe
nde
nc
y
s
c
or
e
m
e
a
s
ur
e
s
th
e
s
ig
ni
f
ic
a
nc
e
of
f
e
a
tu
r
e
A
in
di
s
c
r
im
in
a
ti
ng
be
twe
e
n
di
f
f
e
r
e
nt
c
la
s
s
e
s
or
va
lu
e
s
of
th
e
ta
r
ge
t
va
r
ia
bl
e
w
it
hi
n
th
e
s
e
t
of
in
s
ta
nc
e
s
S
I
.
A
hi
ghe
r
D
e
pe
nde
nc
y
S
c
or
e
in
di
c
a
te
s
a
s
tr
onge
r
de
pe
nde
n
c
y,
a
nd
th
e
r
e
f
or
e
,
th
e
f
e
a
tu
r
e
is
c
ons
id
e
r
e
d
m
or
e
im
por
ta
nt
f
or
c
la
s
s
if
ic
a
ti
on.
3.
1
.
2.
S
t
ab
il
it
y
-
c
or
r
e
la
t
io
n
an
d
c
or
r
e
la
t
io
n
[
30
]
It
is
a
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
us
e
d
in
ML
a
nd
da
ta
a
na
ly
s
i
s
.
it
is
de
s
ig
ne
d
to
id
e
nt
if
y
r
e
le
va
nt
f
e
a
tu
r
e
s
by
c
ons
id
e
r
in
g
bot
h
s
ta
bi
li
ty
(
S
)
w
hi
c
h
r
e
pr
e
s
e
nt
s
th
e
c
ons
is
te
nc
y
of
f
e
a
tu
r
e
i
m
por
ta
nc
e
ba
s
e
d
on
th
e
va
r
ie
ty
of
th
e
f
e
a
tu
r
e
'
s
va
lu
e
s
(
hi
gh
va
r
ie
ty
of
va
lu
e
s
r
e
pr
e
s
e
nt
s
hi
gh
s
ta
bi
li
ty
of
th
e
f
e
a
tu
r
e
)
,
a
nd
c
or
r
e
la
ti
on
(
r
)
w
hi
c
h
m
e
a
s
ur
e
s
ho
w
c
lo
s
e
ly
r
e
la
te
d
a
f
e
a
tu
r
e
is
to
th
e
ta
r
ge
t
va
r
ia
bl
e
or
not
.
S
c
C
ha
s
two
di
s
ti
n
c
t
m
e
th
ods
f
or
s
e
le
c
ti
ng
A
nt
i
-
phi
s
hi
ng
m
e
t
hod
A
ut
hor
s
T
e
c
hni
que
s
D
a
t
a
s
e
t
A
c
c
ur
a
c
y
(%)
C
ont
e
nt
-
ba
s
e
d
J
a
i
n
et
al
.
[
3]
M
odi
f
i
e
d
t
e
r
m
f
r
e
que
nc
y
-
i
nve
r
s
e
doc
um
e
nt
f
r
e
que
nc
y (
TF
-
I
D
F
)
A
l
e
xa
da
t
a
s
e
t
[
4
]
,
O
pe
nP
hi
s
h
[
5]
,
a
nd
P
hi
s
h
T
a
nk
[
6]
89
S
onow
a
l
a
nd
K
uppus
a
m
y
[
7]
P
hi
s
hi
ng de
t
e
c
t
i
on us
i
ng m
ul
t
i
-
f
i
l
t
e
r
a
ppr
oa
c
h
(
P
hi
D
M
A
)
f
r
a
m
e
w
or
k
i
nc
or
por
a
t
e
s
f
i
ve
l
a
ye
r
s
P
hi
s
hl
oa
d
a
nd
L
e
gi
t
i
m
a
t
e
U
R
L
da
t
a
s
e
t
[
8]
92.72
H
e
ur
i
s
t
i
c
s
-
ba
s
e
d
R
a
o
et
al
.
[
9]
T
w
i
n s
uppor
t
ve
c
t
or
m
a
c
hi
ne
s
(
T
W
S
V
M
)
P
hi
s
h
T
a
nk
[
6]
a
nd
A
l
e
xa
da
t
a
s
e
t
[
4]
98.05
B
a
ba
gol
i
et
al
.
[
10]
m
e
t
a
-
he
ur
i
s
t
i
c
s
(
H
S
,
S
V
M
)
U
C
I
phi
s
hi
ng
d
a
t
a
s
e
t
s
[
11]
,
[
12]
92.80
ML
C
hi
e
w
et
al
.
[
13]
C
um
ul
a
t
i
ve
di
s
t
r
i
but
i
on
f
unc
t
i
on
gr
a
di
e
nt
(
C
D
F
-
g)
,
r
a
ndom
f
o
r
e
s
t
,
S
V
M
,
na
i
ve
ba
ye
s
,
C
4.5,
J
R
i
p,
a
nd
P
A
R
T
U
C
I
phi
s
hi
ng
d
a
t
a
s
e
t
s
[
11]
,
[
12]
94.6
Y
a
dol
l
a
hi
[
14]
X
C
S
R
e
a
l
U
R
L
s
98.39
DL
S
m
a
di
et
al
.
[
15]
R
e
i
nf
or
c
e
m
e
nt
l
e
a
r
ni
ng,
ne
ur
a
l
ne
t
w
or
k
P
hi
s
hi
ng
C
or
pus
[
16]
,
S
pa
m
A
s
s
a
s
s
i
n
[
17]
,
a
nd
P
hi
s
h
T
a
nk
[
6]
97
Wei
et
al
.
[
18]
C
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
ks
P
hi
s
h
T
a
nk
[
6]
a
nd
C
om
m
on
C
r
a
w
l
F
ounda
t
i
on
[
19]
99.98
D
a
t
a
M
i
ni
ng
S
m
a
di
et
al
.
[
20]
J
48
a
l
gor
i
t
hm
a
nd
C
4.5
a
l
gor
i
t
hm
P
hi
s
hi
ng
C
or
pus
[
16]
a
nd
S
pa
m
A
s
s
a
s
s
i
n
[
17]
98.87
S
uba
s
i
[
21]
R
a
ndom
f
or
e
s
t
U
C
I
[
12]
a
nd
W
E
K
A
[
22]
97.36
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
P
hi
s
hi
ng de
te
c
ti
on us
in
g c
lu
s
te
r
in
g and mac
hi
ne
l
e
a
r
ni
ng
(
L
uai
A
l
-
Shal
abi
)
4529
r
e
le
va
nt
f
e
a
tu
r
e
s
,
th
e
f
ir
s
t
m
e
th
od
is
s
ta
bi
li
ty
-
c
or
r
e
la
ti
on
(
S
c
)
f
e
a
tu
r
e
s
e
le
c
ti
on
w
hi
c
h
c
om
bi
ne
s
bot
h
s
ta
bi
li
ty
-
ba
s
e
d
a
nd
c
or
r
e
la
ti
on
-
ba
s
e
d
c
r
it
e
r
ia
to
s
e
le
c
t
f
e
a
tu
r
e
s
,
it
a
im
s
to
id
e
nt
if
y
f
e
a
tu
r
e
s
th
a
t
a
r
e
s
ta
bl
e
a
c
r
os
s
di
f
f
e
r
e
nt
s
ubs
e
ts
of
th
e
da
ta
a
nd
hi
ghl
y
c
or
r
e
la
te
d
w
it
h
th
e
ta
r
ge
t
va
r
ia
bl
e
or
c
la
s
s
la
be
ls
.
T
he
s
ta
bi
li
ty
-
c
or
r
e
la
ti
on
s
c
or
e
is
c
a
lc
ul
a
t
e
d
us
in
g
(
2
).
=
(
)
(
2)
W
he
r
e
:
xi
is
th
e
f
e
a
tu
r
e
,
n
is
th
e
num
be
r
of
r
ow
s
in
th
e
da
ta
s
e
t.
T
he
s
e
c
ond
m
e
th
od
is
c
or
r
e
la
ti
on
-
ba
s
e
d
(
C
B
)
f
e
a
tu
r
e
s
e
le
c
t
io
n
w
hi
c
h
is
a
ppl
ie
d
to
th
e
da
ta
s
e
t
ge
ne
r
a
te
d
by
th
e
f
ir
s
t
m
e
th
od.
CB
f
oc
u
s
e
s
on
s
e
le
c
ti
ng
f
e
a
tu
r
e
s
th
a
t
a
r
e
hi
ghl
y
c
or
r
e
la
te
d
w
it
h
th
e
ta
r
ge
t
va
r
ia
bl
e
w
hi
le
pot
e
nt
ia
ll
y
a
voi
di
ng
m
ul
ti
c
ol
li
ne
a
r
it
y
a
m
ong
th
e
s
e
le
c
te
d
f
e
a
tu
r
e
s
.
T
h
e
e
qua
ti
on
f
or
a
s
s
e
s
s
in
g
th
e
c
or
r
e
la
ti
on
be
twe
e
n
a
f
e
a
tu
r
e
X
a
nd
th
e
ta
r
ge
t
va
r
ia
bl
e
Y
is
th
e
P
e
a
r
s
on
c
or
r
e
la
ti
on
c
oe
f
f
ic
ie
nt
(
P
C
C
)
as
s
how
n
in
(
3)
to
(
5)
:
=
1
−
1
∗
∑
∑
(
−
x
̅
)
(
−
ȳ
)
(
3)
=
√
∑
(
−
x
̅
)
2
−
1
(
4)
=
√
∑
(
−
ȳ
)
2
−
1
(
5)
w
he
r
e
:
n
is
th
e
num
be
r
of
pa
ir
s
of
da
ta
us
e
d,
Σ
is
s
ig
m
a
w
hi
c
h
r
e
pr
e
s
e
nt
s
th
e
s
um
m
a
ti
on,
̅
=
th
e
m
e
a
n
of
a
ll
x
-
va
lu
e
s
,
ȳ
is
th
e
m
e
a
n
of
a
ll
y
-
va
lu
e
s
,
S
tx
is
th
e
s
ta
nd
a
r
d
de
vi
a
ti
on
of
va
r
ia
bl
e
x,
S
ty
is
th
e
s
ta
nda
r
d
de
vi
a
ti
on
of
va
r
ia
bl
e
y.
3.
1
.
3.
P
r
in
c
ip
al
c
om
p
on
e
n
t
an
al
ys
is
[3
5
]
It
is
a
m
e
th
od
f
or
r
e
duc
in
g
th
e
di
m
e
ns
io
n
a
li
ty
of
da
ta
w
hi
le
pr
e
s
e
r
vi
ng
as
m
uc
h
of
th
e
va
r
ia
nc
e
in
th
e
da
ta
as
pos
s
ib
le
.
It
a
c
c
om
pl
is
he
s
th
is
by
tr
a
ns
f
or
m
in
g
th
e
or
ig
in
a
l
f
e
a
tu
r
e
s
(
va
r
ia
bl
e
s
)
in
to
a
ne
w
s
e
t
of
li
ne
a
r
ly
unc
or
r
e
la
te
d
va
r
ia
bl
e
s
(
pr
in
c
ip
a
l
c
om
pone
nt
s
)
.T
he
r
e
la
te
d
c
on
c
e
pt
to
P
C
A
is
th
e
e
xpl
a
in
e
d
va
r
ia
nc
e
(
EV
),
w
hi
c
h
is
us
e
d
to
id
e
nt
if
y
th
e
im
por
ta
nc
e
of
each
pr
in
c
ip
a
l
c
om
pone
nt
.
T
he
a
m
ount
of
va
r
ia
nc
e
e
xpl
a
in
e
d
by
each
pr
in
c
ip
a
l
c
om
pone
nt
is
a
m
e
a
s
ur
e
of
f
e
a
tu
r
e
im
por
ta
nc
e
in
a
P
C
A
-
ba
s
e
d
f
e
a
tu
r
e
s
e
le
c
ti
on
c
ont
e
xt
.
In
(
6)
pr
e
s
e
nt
s
th
e
EV
f
or
a
pr
in
c
ip
a
l
c
om
pone
nt
k.
(
)
=
Eig
e
n
v
a
l
u
e
_
k
(
6)
w
he
r
e
:
E
V
(
P
C
_k)
is
th
e
pr
opor
ti
on
of
th
e
to
ta
l
va
r
ia
nc
e
e
xpl
a
in
e
d
by
th
e
kt
h
p
r
in
c
ip
a
l
c
om
pone
nt
(
PC
),
e
ig
e
nva
lu
e
_k
is
th
e
e
ig
e
nva
lu
e
a
s
s
oc
ia
te
d
w
it
h
th
e
kt
h
pr
in
c
ip
a
l
c
om
pone
nt
,
a
nd
to
ta
l
e
ig
e
n
va
lu
e
s
is
th
e
s
um
of
a
ll
e
ig
e
nva
lu
e
s
obt
a
in
e
d
f
r
om
P
C
A
.
3.
2
.
C
la
s
s
if
i
c
at
io
n
m
e
t
h
od
s
ML
a
nd
AI
te
c
hni
qu
e
s
,
s
uc
h
as
s
upe
r
vi
s
e
d
a
nd
uns
up
e
r
vi
s
e
d
le
a
r
ni
ng,
a
r
e
e
m
pl
oye
d
to
bui
ld
m
ode
ls
th
a
t
can
id
e
nt
if
y
phi
s
hi
ng
a
tt
e
m
pt
s
b
a
s
e
d
on
hi
s
to
r
ic
a
l
da
ta
a
n
d
pa
tt
e
r
ns
.
C
la
s
s
if
ic
a
ti
on
is
a
s
upe
r
vi
s
e
d
ML
pr
oc
e
s
s
of
gr
oupi
ng
a
gi
ve
n
da
ta
s
e
t
in
to
c
la
s
s
e
s
b
a
s
e
d
on
one
or
m
or
e
f
e
a
tu
r
e
s
.
S
om
e
c
om
m
on
ML
a
lg
or
it
hm
s
in
c
lu
de
[
9]
,
[
11]
.
3.
2
.
1
D
e
c
is
io
n
t
r
e
e
s
It
is
a
s
upe
r
vi
s
e
d
ML
a
lg
or
it
hm
us
e
d
f
or
bot
h
c
la
s
s
if
ic
a
ti
on
a
nd
r
e
gr
e
s
s
io
n
ta
s
ks
.
It
is
a
popula
r
m
e
th
od
f
or
m
a
ki
ng
de
c
is
io
ns
a
nd
s
ol
vi
ng
pr
obl
e
m
s
by
vi
s
ua
ll
y
r
e
pr
e
s
e
nt
in
g
a
de
c
is
io
n
-
m
a
ki
ng
pr
oc
e
s
s
as
a
tr
e
e
-
li
ke
s
tr
uc
tu
r
e
.
E
a
c
h
node
in
th
e
tr
e
e
r
e
pr
e
s
e
nt
s
a
de
c
is
io
n
or
a
te
s
t
on
a
pa
r
ti
c
ul
a
r
a
tt
r
ib
ut
e
,
a
nd
each
br
a
nc
h
r
e
pr
e
s
e
nt
s
th
e
out
c
om
e
of
th
a
t
te
s
t.
T
he
le
a
ve
s
of
th
e
tr
e
e
c
ont
a
in
th
e
f
in
a
l
de
c
is
io
n
or
th
e
pr
e
di
c
te
d
va
lu
e
.
D
T
s
ha
ve
s
e
ve
r
a
l
a
dva
nt
a
ge
s
,
s
u
c
h
as
s
im
pl
ic
it
y,
in
te
r
pr
e
ta
bi
li
t
y,
a
nd
e
a
s
e
of
vi
s
ua
li
z
a
ti
on.
T
he
y
can
ha
ndl
e
bot
h
c
a
te
gor
ic
a
l
a
nd
num
e
r
ic
a
l
da
ta
,
a
nd
th
e
y
a
r
e
c
a
p
a
b
le
of
ha
ndl
in
g
m
is
s
in
g
va
lu
e
s
.
H
ow
e
ve
r
,
th
e
y
can
be
pr
one
to
ove
r
f
it
ti
ng,
a
nd
th
e
s
tr
uc
tu
r
e
of
th
e
tr
e
e
m
a
y
not
a
lwa
ys
ge
ne
r
a
li
z
e
w
e
ll
to
ne
w
da
ta
.
To
m
it
ig
a
te
th
e
s
e
i
s
s
ue
s
,
te
c
hni
qu
e
s
li
ke
pr
uni
ng
a
nd
u
s
in
g
e
n
s
e
m
bl
e
m
e
th
ods
,
s
u
c
h
as
r
a
ndom
f
or
e
s
t
s
,
a
r
e
of
te
n
e
m
pl
oye
d.
3.
2
.
1
.
D
e
e
p
le
ar
n
in
g
It
r
e
pr
e
s
e
nt
s
a
s
ub
s
e
t
of
A
N
N
s
th
a
t
c
on
s
is
t
of
m
ul
ti
pl
e
la
ye
r
s
of
in
te
r
c
onne
c
te
d
ne
ur
ons
or
node
s
.
T
he
s
e
ne
twor
k
s
a
r
e
c
a
pa
bl
e
of
le
a
r
ni
ng
c
om
pl
e
x
pa
tt
e
r
ns
a
nd
r
e
pr
e
s
e
nt
a
ti
ons
f
r
om
la
r
ge
a
nd
hi
gh
-
di
m
e
ns
io
na
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4526
-
4536
4530
da
ta
s
e
ts
,
m
a
ki
ng
th
e
m
a
po
w
e
r
f
ul
to
ol
f
or
va
r
io
us
ML
a
nd
AI
ta
s
ks
.
DL
,
li
ke
c
onvolut
io
na
l
a
nd
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
ks
,
can
a
na
ly
s
e
e
m
a
il
c
ont
e
nt
a
nd
pa
tt
e
r
ns
in
ne
twor
k
tr
a
f
f
ic
to
de
te
c
t
phi
s
hi
ng.
H
ow
e
ve
r
,
th
e
y
can
be
c
om
put
a
ti
ona
ll
y
in
te
ns
iv
e
a
nd
r
e
qui
r
e
s
ubs
ta
nt
ia
l
a
m
ount
s
of
la
be
le
d
da
ta
f
or
tr
a
in
in
g.
P
r
ope
r
a
r
c
hi
te
c
tu
r
e
s
e
le
c
ti
on,
hype
r
pa
r
a
m
e
t
e
r
tu
ni
ng,
a
nd
da
ta
pr
e
pr
oc
e
s
s
in
g
a
r
e
c
r
uc
ia
l
f
or
th
e
s
u
c
c
e
s
s
f
ul
de
pl
oym
e
nt
of
DL
m
ode
ls
.
3.
3
.
K
-
m
e
an
c
lu
s
t
e
r
in
g
m
e
t
h
od
It
is
a
c
lu
s
te
r
in
g
a
lg
or
it
hm
,
a
nd
an
un
s
upe
r
vi
s
e
d
le
a
r
ni
ng
te
c
hni
que
de
s
ig
ne
d
to
pa
r
ti
ti
on
a
da
ta
s
e
t
in
to
K
di
s
ti
nc
t,
non
-
ove
r
la
ppi
ng
c
lu
s
te
r
s
.
T
he
s
e
c
lu
s
te
r
s
a
r
e
c
ha
r
a
c
t
e
r
iz
e
d
by
th
e
ir
c
e
nt
r
oi
d,
w
hi
c
h
is
th
e
m
e
a
n
of
th
e
da
ta
poi
nt
s
w
it
hi
n
e
a
c
h
c
lu
s
te
r
.
K
-
m
e
a
n
s
di
vi
de
s
th
e
da
t
a
s
e
t
in
to
c
lu
s
te
r
s
w
it
hout
a
ny
hi
e
r
a
r
c
hi
c
a
l
s
tr
uc
tu
r
e
.
In
our
w
or
k,
we
u
s
e
k=
2
s
in
c
e
th
e
two
da
ta
s
e
ts
a
r
e
u
s
e
d
to
di
s
ti
ngui
s
h
be
tw
e
e
n
phi
s
hi
ng
a
nd
non
-
phi
s
hi
ng
c
a
s
e
s
.
T
he
f
ol
lo
w
in
g
a
r
e
s
om
e
r
e
a
s
on
s
w
hy
c
lu
s
te
r
in
g
pe
r
f
or
m
s
be
tt
e
r
in
phi
s
hi
ng
de
te
c
ti
on
.
C
lu
s
te
r
in
g
is
a
te
c
hni
que
u
s
e
d
f
or
di
s
c
ove
r
in
g
in
he
r
e
nt
pa
tt
e
r
ns
a
nd
gr
oupe
d
s
tr
uc
tu
r
e
s
of
da
ta
.
If
your
da
ta
na
tu
r
a
ll
y
e
xhi
bi
ts
c
lu
s
te
r
s
or
gr
oups
,
th
e
DS
in
th
is
te
s
ti
ng
is
s
ue
can
be
gr
oupe
d
ba
s
e
d
on
c
e
r
ta
in
c
r
it
e
r
ia
.
C
lu
s
te
r
in
g
is
an
un
s
upe
r
vi
s
e
d
le
a
r
ni
ng
te
c
hni
que
,
m
e
a
ni
ng
it
do
e
s
not
r
e
qui
r
e
pr
io
r
knowle
dge
or
la
be
le
d
da
ta
.
It
can
be
us
e
d
f
or
a
nom
a
ly
de
te
c
ti
on
by
id
e
nt
if
yi
ng
da
ta
po
in
ts
th
a
t
don'
t
be
lo
ng
to
a
ny
of
th
e
e
s
ta
bl
is
he
d
c
lu
s
te
r
s
.
A
ll
th
e
a
bove
a
r
e
va
li
d
c
ha
r
a
c
te
r
is
ti
c
s
of
D
S
s
be
lo
ngi
n
g
to
phi
s
hi
ng
de
te
c
ti
on.
3.
4
.
P
e
r
f
or
m
an
c
e
m
e
as
u
r
e
m
e
n
t
s
P
e
r
f
or
m
a
nc
e
m
e
a
s
ur
e
m
e
nt
s
a
r
e
s
uc
c
e
s
s
in
di
c
a
to
r
s
th
a
t
e
xp
r
e
s
s
how
w
e
ll
ML
a
lg
or
it
hm
s
a
r
e
f
unc
ti
oni
ng.
To
e
va
lu
a
te
th
e
pe
r
f
or
m
a
nc
e
of
th
e
pr
opos
e
d
a
lg
or
it
hm
,
a
r
a
nge
of
m
e
a
s
ur
e
m
e
nt
s
can
be
us
e
d.
In
th
e
w
or
k,
f
iv
e
of
th
e
m
w
il
l
be
us
e
d
a
nd
e
xpl
a
in
e
d.
−
A
c
c
ur
a
c
y
:
i
t
is
c
a
lc
ul
a
te
d
as
a
r
a
ti
o
r
e
pr
e
s
e
nt
e
d
in
(
7)
.
=
+
+
+
+
(
7)
W
he
r
e
:
tr
ue
pos
it
iv
e
s
(
TP
)
is
th
e
num
be
r
of
phi
s
hi
ng
e
m
a
il
s
or
w
e
bs
it
e
s
c
or
r
e
c
tl
y
id
e
nt
if
ie
d
as
phi
s
hi
ng,
tr
ue
ne
ga
ti
ve
s
(
TN
)
is
th
e
num
be
r
of
l
e
gi
ti
m
a
te
(
non
-
phi
s
hi
ng)
e
m
a
il
s
or
w
e
bs
it
e
s
c
or
r
e
c
tl
y
id
e
nt
if
ie
d
as
non
-
phi
s
hi
ng,
f
a
ls
e
pos
it
iv
e
s
(
FP
)
is
th
e
num
be
r
of
le
gi
ti
m
a
te
e
m
a
il
s
or
w
e
b
s
it
e
s
in
c
or
r
e
c
tl
y
id
e
nt
if
ie
d
as
phi
s
hi
ng
(a
ty
pe
of
e
r
r
o
r
)
,
f
a
ls
e
ne
ga
ti
ve
s
(
FN
)
is
th
e
num
b
e
r
of
phi
s
hi
ng
e
m
a
il
s
or
w
e
bs
it
e
s
in
c
or
r
e
c
tl
y
id
e
nt
if
ie
d
as
le
gi
ti
m
a
te
(
a
not
he
r
ty
pe
of
e
r
r
or
)
.
−
A
r
e
a
un
de
r
c
ur
v
e
(
A
U
C
)
:
i
t
is
o
ne
of
t
he
w
e
l
l
-
kn
ow
n
m
e
a
s
ur
e
m
e
n
ts
in
ML
a
r
e
a
.
It
is
u
s
e
d
to
a
s
s
e
s
s
t
he
pe
r
f
or
m
a
n
c
e
of
b
in
a
r
y
c
l
a
s
s
if
i
c
a
ti
o
n
m
o
de
l
s
.
It
qu
a
nt
if
i
e
s
th
e
a
b
il
it
y
of
a
m
o
de
l
to
di
s
t
in
g
ui
s
h
be
tw
e
e
n
t
w
o
c
l
a
s
s
e
s
(
po
s
i
ti
v
e
a
n
d
ne
ga
ti
v
e
)
by
m
e
a
s
ur
i
ng
th
e
a
r
e
a
u
nd
e
r
th
e
r
e
c
e
iv
e
r
op
e
r
a
ti
n
g
c
h
a
r
a
c
te
r
is
ti
c
(
R
O
C
)
c
ur
v
e
.
−
P
r
e
c
is
io
n
:
i
t
is
a
m
e
a
s
ur
e
of
th
e
a
c
c
ur
a
c
y
of
a
m
od
e
l
in
c
or
r
e
c
tl
y
id
e
nt
if
yi
ng
pos
it
iv
e
in
s
ta
nc
e
s
a
m
ong
th
e
in
s
ta
nc
e
s
th
a
t
it
ha
s
c
l
a
s
s
if
ie
d
as
pos
it
iv
e
T
P
.
It
pr
ovi
de
s
in
f
or
m
a
ti
on
a
bout
th
e
a
bi
li
ty
of
m
ode
l
to
a
voi
d
FP.
P
r
e
c
is
io
n
is
c
a
lc
ul
a
te
d
by
(
8)
.
=
TP
TP
+
FP
(
8)
−
R
e
c
a
ll
:
i
t
is
a
not
he
r
vi
ta
l
m
e
tr
ic
th
a
t
e
va
lu
a
t
e
s
ML
a
lg
or
it
hm
s
.
It
is
known
as
s
e
ns
it
iv
it
y
or
tr
ue
po
s
it
iv
e
r
a
te
,
w
hi
c
h
is
us
e
d
to
a
s
s
e
s
s
th
e
pe
r
f
or
m
a
nc
e
of
a
bi
na
r
y
c
la
s
s
i
f
ic
a
ti
on
m
ode
l
by
m
e
a
s
ur
in
g
th
e
a
bi
li
ty
of
m
ode
l
to
c
or
r
e
c
tl
y
id
e
nt
if
y
a
ll
r
e
le
va
nt
in
s
ta
nc
e
s
f
r
om
th
e
pos
it
iv
e
c
la
s
s
.
It
is
c
a
lc
ul
a
te
d
u
s
in
g
(
9)
.
=
TP
TP
+
FN
(
9)
−
F
-
m
e
a
s
ur
e
:
i
t
is
an
im
por
ta
nt
m
e
a
s
ur
e
to
e
va
lu
a
te
ML
a
lg
or
it
hm
s
.
It
pr
ovi
de
s
a
s
in
gl
e
m
e
a
s
ur
e
of
a
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
of
m
ode
l
by
c
om
bi
ni
ng
bot
h
pr
e
c
is
io
n
a
nd
r
e
c
a
ll
in
to
a
s
in
gl
e
s
c
or
e
.
It
is
c
a
lc
ul
a
te
d
us
in
g
(
10)
.
−
=
2
∗
P
r
e
ci
s
io
n
∗
R
e
ca
l
l
P
r
e
ci
s
io
n
+
R
e
ca
l
l
(
10)
4.
P
R
O
P
O
S
E
D
A
L
G
O
R
I
T
H
M
AND
S
I
M
U
L
A
T
I
O
N
R
E
S
U
L
T
S
4.1.
D
at
as
e
t
s
T
w
o
phi
s
hi
ng
da
ta
s
e
ts
[
5
]
,
[3
2
]
a
r
e
us
e
d
f
or
tr
a
in
in
g
a
nd
te
s
ti
ng.
T
he
y
a
r
e
c
le
a
ne
d
by
r
e
m
ovi
ng
r
e
dunda
nc
y
a
nd
m
i
s
s
in
g
v
a
lu
e
s
.
T
he
num
be
r
of
r
ow
s
in
each
D
S
1
a
nd
D
S
2
a
r
e
11
,
050
a
nd
10
,
000
r
e
s
p
e
c
ti
ve
ly
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
P
hi
s
hi
ng de
te
c
ti
on us
in
g c
lu
s
te
r
in
g and mac
hi
ne
l
e
a
r
ni
ng
(
L
uai
A
l
-
Shal
abi
)
4531
4.2.
P
r
op
os
e
d
al
gor
it
h
m
T
he
pr
opos
e
d
a
lg
or
it
hm
is
a
hybr
id
m
e
th
od
th
a
t
m
e
r
ge
s
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
,
c
lu
s
t
e
r
in
g
[
33
]
,
[
34
]
(K
-
m
e
a
n)
,
a
nd
c
la
s
s
if
ic
a
ti
on
ML
m
e
th
ods
u
s
in
g
th
e
DL
(
H
2O
)
[
35
]
a
lg
or
it
hm
,
w
hi
c
h
is
an
ope
n
-
s
our
c
e
ML
pl
a
tf
or
m
th
a
t
is
de
s
ig
ne
d
f
or
s
c
a
la
bl
e
a
nd
di
s
tr
ib
ut
e
d
da
ta
a
na
ly
s
is
.
It
can
pe
r
f
or
m
a
w
id
e
r
a
nge
of
ML
ta
s
ks
e
f
f
ic
ie
nt
ly
a
nd
e
f
f
e
c
ti
ve
ly
,
pa
r
ti
c
ul
a
r
ly
f
or
la
r
ge
da
ta
s
e
ts
.
T
he
s
te
p
s
of
th
e
pr
oc
e
s
s
a
r
e
de
s
c
r
ib
e
d
in
th
e
f
ol
lo
w
in
g:
S
te
p
1.
P
r
e
pa
r
in
g
th
e
da
ta
s
e
ts
S
te
p
2.
A
ppl
yi
ng
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
(
R
S
,
S
c
C
,
a
nd
P
C
A
)
to
th
e
or
ig
in
a
l
da
ta
s
e
t
a
nd
pr
oduc
e
th
e
r
e
duc
e
d
da
ta
s
e
ts
.
T
he
n
ge
n
e
r
a
te
th
e
c
la
s
s
if
ic
a
ti
on
m
ode
l
of
each
u
s
in
g
DL
a
nd
D
T
.
S
te
p
3.
T
e
s
ti
ng
th
e
pe
r
f
or
m
a
nc
e
of
th
e
m
ode
ls
ge
n
e
r
a
te
d
in
s
te
p
1
us
in
g
a
ll
th
e
m
e
tr
ic
s
e
xpl
a
in
e
d
e
a
r
li
e
r
.
S
te
p
4.
R
e
m
ove
th
e
c
la
s
s
if
ic
a
ti
on
a
tt
r
ib
ut
e
f
r
om
each
r
e
duc
e
d
d
a
ta
s
e
ts
in
s
te
p
2.
S
te
p
5.
A
ppl
y
th
e
K
-
m
e
a
n
to
each
ne
w
da
t
a
s
e
t
ge
n
e
r
a
te
d
in
s
te
p
4.
S
te
p
6.
A
ppl
y
m
a
c
hi
ne
le
a
r
ni
ng
c
la
s
s
if
ic
a
ti
on
m
e
th
ods
(
M
L
a
nd
D
T
)
to
th
e
r
e
s
ul
te
d
da
ta
s
e
t
s
in
s
te
p
5
a
nd
ge
ne
r
a
te
th
e
m
ode
ls
.
S
te
p
7.
T
e
s
ti
ng
th
e
pe
r
f
or
m
a
nc
e
of
th
e
ge
ne
r
a
te
d
hybr
id
m
ode
ls
in
s
te
p
6
us
in
g
a
ll
th
e
m
e
a
s
ur
e
m
e
nt
s
e
xpl
a
in
e
d
e
a
r
li
e
r
.
S
te
p
8.
C
om
pa
r
e
a
ll
th
e
r
e
s
ul
ts
of
a
ll
th
e
m
ode
ls
.
S
te
p
9.
C
hoos
e
th
e
be
s
t
of
th
e
m
.
T
he
id
e
a
w
a
s
pr
opos
e
d
to
im
pr
ove
th
e
de
te
c
ti
on
pr
oc
e
s
s
of
phi
s
hi
ng
da
ta
.
F
ig
ur
e
1
r
e
pr
e
s
e
nt
s
th
e
f
lo
w
of
pr
opos
e
d
id
e
a
,
w
he
r
e
a
s
F
ig
ur
e
2
r
e
pr
e
s
e
nt
s
th
e
c
om
pone
nt
s
of
th
e
pr
opos
e
d
a
lg
or
it
hm
.
T
he
s
te
ps
of
th
e
pr
oc
e
s
s
a
r
e
s
um
m
a
r
iz
e
d
as
in
A
lg
or
it
hm
1.
A
lg
or
it
hm
1. P
r
opos
e
d m
e
th
odol
ogy
F
or
e
a
c
h D
S
i
I
nput
D
S
i
C
le
a
n D
S
i
F
or
e
a
c
h F
S
i
A
ppl
y F
S
i
A
ppl
y M
L
m
e
th
ods
O
ut
put
r
e
s
ul
ts
//
r
e
s
ul
ts
a
f
te
r
f
e
a
tu
r
e
s
e
le
c
ti
on us
in
g D
L
a
nd D
T
A
ppl
y K
-
M
e
a
n
A
ppl
y M
L
m
e
th
ods
O
ut
put
r
e
s
ul
ts
//
r
e
s
ul
ts
a
f
te
r
c
lu
s
te
r
in
g
//
lo
op unti
l
a
ppl
yi
ng a
ll
F
S
i
E
nd f
or
//
lo
op unti
l
in
put
a
ll
D
S
i
E
nd f
or
F
ig
ur
e
1.
A
lg
or
it
hm
of
th
e
pr
opos
e
d
id
e
a
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4526
-
4536
4532
F
ig
ur
e
2
.
T
he
c
om
pone
nt
s
of
th
e
pr
opos
e
d
a
lg
or
it
hm
4.3.
S
im
u
la
t
io
n
r
e
s
u
lt
s
W
h
e
n
a
p
p
l
yi
n
g
f
e
a
t
ur
e
s
e
l
e
c
ti
o
n
m
e
t
h
od
s
(
R
S
,
S
c
C
,
a
n
d
P
C
A
)
to
t
h
e
or
i
gi
n
a
l
d
a
t
a
s
e
t
,
t
h
e
nu
m
b
e
r
of
t
h
e
b
e
s
t
i
n
f
o
r
m
a
t
i
v
e
f
e
a
t
ur
e
s
is
r
e
pr
e
s
e
n
t
e
d
in
T
a
b
l
e
4.
T
h
e
p
e
r
f
o
r
m
a
n
c
e
of
t
h
e
m
o
d
e
l
s
g
e
n
e
r
a
t
e
d
in
S
t
e
p
2
is
r
e
pr
e
s
e
n
t
e
d
in
T
a
b
l
e
s
5
to
7
. W
h
i
l
e
t
h
e
p
e
r
f
or
m
a
n
c
e
of
t
h
e
m
o
d
e
l
s
g
e
n
e
r
a
t
e
d
in
S
t
e
p
6
is
r
e
p
r
e
s
e
n
t
e
d
in
T
a
bl
e
s
8
to
10.
T
a
bl
e
4.
N
um
be
r
of
r
e
duc
ti
ons
in
D
S
s
D
a
t
a
s
e
t
R
a
w
DS
RS
S
c
C
P
C
A
D
S
1
31
23
4
21
D
S
2
49
7
8
3
T
a
bl
e
5.
M
e
a
s
ur
e
m
e
nt
s
a
f
te
r
a
ppl
yi
ng
RS
to
D
S
s
D
a
t
a
S
e
t
ML
A
c
c
ur
a
c
y
(%)
AUC
(%)
P
(%)
R
(%)
F
-
m
(
%
)
D
S
1
DL
94.33
98.91
93.9
96.08
94.97
DT
92.82
94.44
90.45
97.39
93.79
D
S
2
DL
86.91
95.67
94.09
78.78
85.75
DT
70.56
92.55
95.45
43.14
59.41
T
a
bl
e
6
. M
e
a
s
ur
e
m
e
nt
s
a
f
te
r
a
ppl
yi
ng S
c
C
t
o D
S
s
D
a
t
a
S
e
t
ML
A
c
c
ur
a
c
y (
%
)
A
U
C
(
%
)
P
(
%
)
R
(
%
)
F
-
m
(
%
)
D
S
1
DL
92.15
95.42
91.77
94.38
93.05
DT
91.61
95.20
92.21
92.78
92.49
D
S
2
DL
93.56
97.44
96.44
90.48
93.35
DT
78.16
61.59
97.11
58.05
72.62
T
a
bl
e
7
. M
e
a
s
ur
e
m
e
nt
s
a
f
te
r
a
ppl
yi
ng P
C
A
t
o D
S
s
D
a
t
a
S
e
t
ML
A
c
c
ur
a
c
y (
%
)
A
U
C
(
%
)
P
(
%
)
R
(
%
)
F
-
m
(
%
)
D
S
1
DL
93.86
98.60
94.10
94.94
94.51
DT
86.36
92.82
90.76
84.10
87.28
D
S
2
DL
69.10
75.39
67.93
72.27
70.03
DT
52.62
53.89
88.16
6.15
11.47
T
a
bl
e
8
.
M
e
a
s
ur
e
m
e
nt
s
a
f
te
r
a
ppl
yi
ng
RS
-
K
-
m
e
a
n
to
D
S
s
D
a
t
a
S
e
t
ML
A
c
c
ur
a
c
y
(%)
AUC
(%)
P
(%)
R
(%)
F
-
m
(
%
)
D
S
1
DL
99.53
100
99.56
99.78
99.67
DT
98.16
98.08
99.16
98.29
98.72
D
S
2
DL
92.22
100
100
89.46
94.44
DT
92.16
99.31
98.42
90.89
94.50
T
a
bl
e
9
.
M
e
a
s
ur
e
m
e
nt
s
a
f
te
r
a
ppl
yi
ng
S
c
C
-
K
-
m
e
a
n
to
D
S
s
D
a
t
a
S
e
t
ML
A
c
c
ur
a
c
y
(%)
AUC
(%)
P
(%)
R
(%)
F
-
m
(
%
)
D
S
1
DL
99.37
100
98.90
100
99.45
DT
100
100
100
100
100
D
S
2
DL
96.88
100
100
96.31
98.12
DT
97.20
100
100
96.70
98.32
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
P
hi
s
hi
ng de
te
c
ti
on us
in
g c
lu
s
te
r
in
g and mac
hi
ne
l
e
a
r
ni
ng
(
L
uai
A
l
-
Shal
abi
)
4533
T
a
bl
e
10.
M
e
a
s
ur
e
m
e
nt
s
a
f
te
r
a
ppl
yi
ng
P
C
A
-
K
-
m
e
a
n
to
D
S
s
D
a
t
a
S
e
t
ML
A
c
c
ur
a
c
y
(%)
AUC
(%)
P
(%)
R
(%)
F
-
m
(
%
)
D
S
1
DL
99.53
100
99.47
100
99.71
DT
97.75
97.93
97.64
99.70
98.66
D
S
2
DL
96.01
99.87
100
94.84
97.34
DT
96.75
99.83
100
95.78
97.85
F
or
th
e
two
r
e
duc
e
d
da
ta
s
e
ts
,
th
e
a
c
c
ur
a
c
y
of
th
e
pr
opos
e
d
m
e
th
od
gi
ve
n
by
DT
a
nd
DL
w
a
s
th
e
hi
ghe
s
t
a
m
ong
a
ll
ot
he
r
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
c
om
pa
r
e
d
to
S
c
C
,
R
S
,
a
nd
P
C
A
.
T
h
e
de
te
c
ti
on
a
c
c
ur
a
c
ie
s
of
S
c
C
-
K
-
m
e
a
n,
RS
-
K
-
m
e
a
n,
a
nd
P
C
A
-
K
-
m
e
a
n
w
e
r
e
s
ig
ni
f
ic
a
nt
ly
im
pr
ove
d
by
a
ppl
yi
ng
th
e
k
-
m
e
a
n
c
lu
s
te
r
in
g
m
e
th
od
to
th
e
f
i
lt
e
r
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
(
S
c
C
,
R
S
,
a
nd
P
C
A
)
us
in
g
D
S
1
f
r
om
92.15
%
,
94.33
%
,
a
nd
93.86%
to
99.37
%
,
99.53
%
,
a
nd
99.53%
r
e
s
pe
c
ti
ve
ly
w
he
n
DL
c
la
s
s
if
ie
r
w
a
s
us
e
d
a
nd
f
r
om
91.61
%
,
92.82
%
,
a
nd
86.36%
to
100
%
,
98.16
%
,
a
nd
97.75%
r
e
s
pe
c
ti
ve
ly
w
he
n
DT
c
la
s
s
if
ie
r
w
a
s
us
e
d.
A
c
om
pa
r
is
on
w
a
s
a
ls
o
m
a
d
e
f
or
D
S
2,
a
nd
th
e
im
pr
ove
m
e
nt
w
a
s
ve
r
y
s
ig
ni
f
ic
a
nt
.
C
om
pa
r
in
g
th
e
th
r
e
e
s
ugge
s
te
d
hybr
id
m
e
th
ods
(
S
c
C
-
K
-
m
e
a
n,
RS
-
K
-
m
e
a
n,
a
nd
P
C
A
-
K
-
m
e
a
n)
,
th
e
a
c
c
ur
a
c
y
of
th
e
S
c
C
-
K
-
m
e
a
n
us
in
g
DT
w
a
s
th
e
be
s
t
a
m
ong
th
e
m
w
it
h
100%
a
c
c
ur
a
c
y
f
or
D
S
1
a
nd
97.2
%
f
o
r
D
S
2.
F
or
AUC
a
nd
pr
e
c
is
io
n
,
th
e
S
c
C
-
K
-
m
e
a
n
m
e
th
od
s
how
s
hi
ghe
r
pe
r
f
or
m
a
nc
e
(
m
os
tl
y
100
%
)
f
o
r
m
os
t
of
th
e
r
e
a
di
ngs
c
om
pa
r
e
d
to
RS
-
K
-
m
e
a
n,
a
nd
P
C
A
-
K
-
m
e
a
n
a
nd
v
e
r
y
hi
gh
va
lu
e
s
f
or
r
e
c
a
ll
a
nd
F
-
m
e
a
s
ur
e
.
It
is
c
le
a
r
f
r
om
th
e
r
e
s
ul
ts
th
a
t
a
ppl
yi
ng
K
-
m
e
a
n
c
lu
s
te
r
in
g
a
f
te
r
th
e
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
s
(
S
c
C
-
K
-
m
e
a
n,
RS
-
K
-
m
e
a
n,
a
nd
P
C
A
-
K
-
m
e
a
n)
a
c
hi
e
ve
s
hi
ghe
r
pe
r
f
or
m
a
nc
e
in
te
r
m
s
of
a
c
c
ur
a
c
y
,
A
U
C
,
pr
e
c
is
io
n
,
r
e
c
a
ll
,
a
nd
F
-
m
e
a
s
ur
e
f
or
th
e
s
e
le
c
te
d
D
S
s
.
A
nd
a
m
ong
th
e
m
,
S
c
C
-
K
-
m
e
a
n
w
a
s
th
e
be
s
t.
K
-
m
e
a
n,
in
th
is
w
or
k,
pr
ove
s
its
pe
r
f
or
m
a
nc
e
.
K
-
m
e
a
n
c
lu
s
t
e
r
in
g
a
f
te
r
f
e
a
tu
r
e
s
e
le
c
ti
on
us
in
g
R
S
,
S
c
C
,
a
nd
P
C
A
pe
r
f
or
m
s
be
tt
e
r
th
a
n
onl
y
f
e
a
tu
r
e
s
e
le
c
ti
on
(
R
S
,
S
c
C
,
a
nd
P
C
A
)
,
but
it
de
pe
nd
s
on
th
e
s
pe
c
if
ic
is
s
ue
th
a
t
be
in
g
s
ol
ve
d
a
nd
th
e
na
tu
r
e
of
da
ta
.
E
a
c
h
of
th
e
s
e
te
c
hni
que
s
s
e
r
ve
s
di
f
f
e
r
e
nt
pur
pos
e
s
a
nd
e
xc
e
ls
in
di
s
ti
nc
t
s
c
e
na
r
io
s
.
T
he
pe
r
f
or
m
a
nc
e
of
th
e
pr
opos
e
d
a
ppr
oa
c
h
w
a
s
c
om
pa
r
e
d
w
it
h
ot
he
r
hybr
id
a
ppr
oa
c
he
s
in
th
e
pr
e
vi
ous
w
or
k
us
e
d
in
de
te
c
ti
ng
phi
s
hi
ng
w
e
bs
it
e
s
.
T
h
e
a
c
c
ur
a
c
y
of
th
e
phi
s
hi
ng
hybr
id
a
ppr
oa
c
h
us
in
g
ge
ne
ti
c
a
lg
or
it
hm
(
GA
)
w
a
s
91.13%
[
23]
,
w
hi
le
it
w
a
s
95.76%
us
in
g
f
e
a
tu
r
e
s
-
ba
s
e
d
ANN
a
nd
K
-
m
e
doi
d
s
c
lu
s
te
r
in
g
a
lg
or
it
hm
[
24]
.
T
he
s
tu
dy
of
S
ul
e
m
a
n
a
nd
A
w
a
n
[
25]
us
in
g
a
n
ot
he
r
ge
ne
r
a
ti
ng
ge
ne
ti
c
a
lg
or
it
hm
(
Y
A
G
G
A
)
ga
ve
an
a
c
c
ur
a
c
y
r
e
a
c
he
d
95%
.
M
e
a
nw
hi
le
,
th
e
s
tu
dy
of
V
r
ba
nč
ič
et
al
.
[
26
]
us
in
g
th
e
ba
t
a
lg
o
r
it
hm
a
nd
hyb
r
id
ba
t
a
lg
or
it
hm
ga
ve
an
a
c
c
ur
a
c
y
of
96.5
%
.
F
o
r
th
e
w
o
r
k
th
a
t
us
e
d
D
P
I
,
S
D
N
,
a
nd
ANN,
th
e
a
c
c
ur
a
c
y
w
a
s
98.39%
[
27]
.
T
he
a
c
c
ur
a
c
y
of
th
e
m
e
th
od
th
a
t
us
e
s
S
c
C
a
nd
f
or
w
a
r
d
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
w
a
s
92.56%
[
31
].
As
our
pr
opos
e
d
m
e
th
od'
s
hi
gh
e
s
t
a
c
c
ur
a
c
y
f
or
th
e
U
C
I
phi
s
hi
ng
w
e
bs
it
e
s
w
a
s
100%
us
in
g
th
e
DT
c
la
s
s
if
ie
r
a
nd
99.37%
u
s
in
g
th
e
DL
c
la
s
s
if
ie
r
,
we
pr
oudl
y
c
onc
l
ude
d
th
a
t
our
a
ppr
oa
c
h
is
th
e
pi
one
e
r
in
s
ol
vi
ng
phi
s
hi
ng
pr
obl
e
m
s
.
F
ig
ur
e
s
3
(a
)
to
3
(
b)
c
om
pa
r
e
s
b
e
twe
e
n
m
e
a
s
ur
e
m
e
nt
s
be
f
or
e
a
n
d
a
f
te
r
a
ppl
yi
ng
th
e
K
-
m
e
a
n
c
lu
s
te
r
in
g
f
or
R
S
.
F
ig
ur
e
s
4
(a
)
to
4
(
b)
c
om
pa
r
e
s
be
twe
e
n
m
e
a
s
ur
e
m
e
nt
s
b
e
f
or
e
a
nd
a
f
te
r
a
ppl
yi
ng
th
e
K
-
m
e
a
n
c
lu
s
te
r
in
g
f
or
S
c
C
.
W
hi
le
F
ig
ur
e
s
5
(a
)
to
5
(
b)
c
om
pa
r
e
s
be
tw
e
e
n
m
e
a
s
ur
e
m
e
nt
s
be
f
or
e
a
nd
a
f
te
r
a
ppl
yi
ng
th
e
K
-
m
e
a
n
c
lu
s
te
r
in
g f
or
P
C
A
.
(
a
)
(
b)
F
ig
ur
e
3.
C
om
pa
r
is
ons
be
twe
e
n
m
e
a
s
ur
e
m
e
nt
s
b
e
f
or
e
a
nd
a
f
te
r
K
-
m
e
a
n
c
lu
s
te
r
in
g
–
RS
f
or
(
a
)
D
S
1
a
nd
(
b)
D
S
2
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
,
V
ol
.
13
, N
o.
4
,
D
e
c
e
m
be
r
2024
:
4526
-
4536
4534
(
a
)
(
b)
F
ig
ur
e
4.
C
om
pa
r
is
ons
be
twe
e
n
m
e
a
s
ur
e
m
e
nt
s
b
e
f
or
e
a
nd
a
f
te
r
c
lu
s
te
r
in
g
–
S
c
C
f
or
(
a
)
D
S
1
a
nd
(
b)
D
S
2
(
a
)
(
b)
F
ig
ur
e
5.
C
om
pa
r
is
ons
be
twe
e
n
m
e
a
s
ur
e
m
e
nt
s
b
e
f
or
e
a
nd
a
f
te
r
c
lu
s
te
r
in
g
–
P
C
A
f
or
(
a
)
D
S
1
a
nd
(
b)
D
S
2
5.
C
O
M
P
L
E
X
I
T
Y
OF
F
E
A
T
U
R
E
S
E
L
E
C
T
I
O
N
A
N
D
C
L
U
S
T
E
R
I
N
G
M
E
T
H
O
D
S
C
om
pl
e
xi
ty
(
B
ig
O
-
N
ot
a
ti
on)
p
r
ovi
de
s
a
w
a
y
to
c
om
pa
r
e
a
nd
a
n
a
ly
z
e
th
e
e
f
f
ic
ie
nc
y
of
a
lg
or
it
hm
s
a
nd
to
unde
r
s
ta
nd
how
th
e
y
w
il
l
pe
r
f
or
m
as
th
e
in
put
s
iz
e
in
c
r
e
a
s
e
s
.T
he
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
of
RS
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
d
e
pe
nds
on
th
e
s
pe
c
if
ic
a
lg
or
it
hm
a
nd
a
ppr
oa
c
h
be
in
g
us
e
d.
T
he
c
om
pl
e
xi
ty
is
ty
pi
c
a
ll
y
e
xpr
e
s
s
e
d
in
te
r
m
s
of
th
e
num
be
r
of
in
s
ta
nc
e
s
(
n)
a
nd
th
e
num
be
r
of
f
e
a
tu
r
e
s
(
m
)
in
th
e
da
ta
s
e
t.
T
he
c
om
pl
e
xi
ty
is
ty
pi
c
a
ll
y
O
(
n*m
^
2)
in
th
e
w
or
s
t
c
a
s
e
[3
6
].
W
hi
le
th
e
c
om
p
ut
a
ti
ona
l
c
om
pl
e
xi
ty
of
S
c
C
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
ods
[3
7
]
de
pe
nds
on
th
e
s
pe
c
if
ic
a
lg
or
it
hm
s
a
nd
m
e
a
s
ur
e
s
us
e
d
f
or
f
e
a
tu
r
e
s
e
le
c
ti
on
w
it
hi
n
th
e
s
e
f
r
a
m
e
w
or
ks
.
B
ot
h
S
c
C
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
s
m
a
y
in
vol
ve
c
om
put
in
g
c
or
r
e
la
ti
ons
a
nd
s
ta
bi
li
ty
m
e
a
s
ur
e
s
f
or
f
e
a
tu
r
e
s
.
A
nd
th
e
c
om
pl
e
xi
ty
of
S
c
C
f
e
a
tu
r
e
s
e
le
c
ti
on
ty
pi
c
a
ll
y
de
pe
nd
s
on
c
om
put
in
g
f
e
a
tu
r
e
s
ta
bi
li
ty
,
w
hi
c
h
of
te
n
in
vol
ve
s
c
a
lc
ul
a
ti
ng
th
e
J
a
c
c
a
r
d
in
de
x
or
a
s
im
il
a
r
m
e
a
s
ur
e
f
or
a
s
s
e
s
s
in
g
f
e
a
tu
r
e
s
ta
bi
li
ty
a
c
r
os
s
s
ubs
e
ts
of
th
e
da
t
a
.
T
he
c
om
pl
e
xi
ty
is
O
(
n)
,
w
he
r
e
n
is
th
e
num
be
r
of
in
s
ta
nc
e
s
.
C
a
lc
ul
a
ti
ng
t
he
c
or
r
e
la
ti
o
n
b
e
tw
e
e
n
f
e
a
tu
r
e
s
a
nd
th
e
t
a
r
ge
t
v
a
r
ia
bl
e
(
e
.g.,
u
s
in
g
th
e
P
e
a
r
s
on
c
or
r
e
l
a
ti
on
c
oe
f
f
ic
i
e
nt
)
.
T
he
c
om
pl
e
xi
ty
of
c
o
m
put
in
g
c
or
r
e
la
t
io
n
s
is
of
te
n
O
(
n*m
)
,
w
he
r
e
m
is
th
e
nu
m
be
r
of
f
e
a
tu
r
e
s
.
T
h
e
ove
r
a
ll
c
om
pl
e
xi
ty
of
s
t
a
bi
li
t
y
-
c
or
r
e
l
a
ti
on
f
e
a
tu
r
e
s
e
le
c
ti
o
n
is
ty
pi
c
a
ll
y
dom
in
a
te
d
by
th
e
c
or
r
e
la
t
io
n
c
om
put
a
ti
on,
w
hi
c
h
is
O
(
n*m
)
,
a
s
s
um
in
g
th
a
t
th
e
s
ta
bi
li
ty
m
e
a
s
ur
e
is
r
e
l
a
ti
ve
l
y
e
f
f
i
c
ie
nt
.
N
e
xt
m
e
th
od
is
t
he
c
or
r
e
la
t
io
n
-
ba
s
e
d
f
e
a
tu
r
e
s
e
le
c
ti
on,
w
hi
c
h
f
oc
us
e
s
on
c
om
put
i
ng
th
e
c
or
r
e
la
ti
on
b
e
tw
e
e
n
i
ndi
vi
d
ua
l
f
e
a
tu
r
e
s
a
nd
th
e
t
a
r
ge
t
v
a
r
ia
bl
e
.
T
he
c
om
pl
e
xi
t
y
is
ty
pi
c
a
ll
y
O
(
n
*m
)
,
w
he
r
e
n
is
t
he
num
b
e
r
of
in
s
ta
nc
e
s
,
a
n
d
m
is
th
e
num
b
e
r
of
f
e
a
tu
r
e
s
.
T
he
la
s
t
f
e
a
t
ur
e
s
e
le
c
ti
on
is
P
C
A
[3
8
]
w
hi
c
h
r
e
du
c
e
s
th
e
di
m
e
n
s
io
n
a
li
ty
of
th
e
d
a
ta
by
c
r
e
a
ti
n
g
a
ne
w
s
e
t
of
or
th
ogo
na
l
f
e
a
tu
r
e
s
c
a
ll
e
d
pr
in
c
ip
a
l
c
om
pone
nt
s
.
W
hi
le
P
C
A
it
s
e
lf
doe
s
n'
t
h
a
v
e
a
tr
a
di
ti
on
a
l
c
om
put
a
ti
on
a
l
c
o
m
pl
e
x
it
y
in
t
e
r
m
s
of
bi
g
O
not
a
ti
on,
it
i
nvol
v
e
s
c
a
lc
ul
a
ti
ng
e
i
ge
nv
e
c
t
or
s
a
nd
e
ig
e
n
va
lu
e
s
.
T
he
c
om
put
a
ti
on
a
l
c
om
pl
e
xi
ty
of
P
C
A
m
a
in
ly
de
pe
nd
s
on
th
e
s
in
gul
a
r
v
a
lu
e
de
c
om
po
s
it
i
on
(
S
V
D
)
or
e
ig
e
n
de
c
om
po
s
it
io
n
of
th
e
d
a
ta
'
s
c
o
va
r
ia
nc
e
m
a
tr
ix
.
T
he
c
om
pl
e
xi
ty
can
be
e
xpr
e
s
s
e
d
as
O
(
m
^
2*n)
+
O
(
m
^
3)
,
w
he
r
e
(
m
)
is
th
e
num
b
e
r
of
f
e
a
tu
r
e
s
(
or
ig
i
na
l
di
m
e
n
s
io
ns
)
,
a
nd
(
n)
is
th
e
n
um
be
r
of
in
s
ta
n
c
e
s
(
d
a
ta
p
oi
nt
s
)
.
T
he
f
ir
s
t
te
r
m
,
O
(
m
^
2*n)
,
r
e
pr
e
s
e
nt
s
th
e
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
of
c
a
lc
ul
a
ti
ng
th
e
c
ova
r
ia
nc
e
m
a
tr
ix
,
a
nd
th
e
s
e
c
ond
te
r
m
,
O
(
m
^
3)
,
r
e
pr
e
s
e
nt
s
th
e
c
om
pl
e
xi
ty
of
f
in
di
ng
th
e
e
ig
e
nve
c
to
r
s
a
nd
e
ig
e
nva
lu
e
s
of
th
e
c
ova
r
ia
nc
e
m
a
tr
ix
.
K
e
e
p
in
m
in
d
th
a
t
P
C
A
is
ty
pi
c
a
ll
y
us
e
d
to
tr
a
ns
f
or
m
th
e
da
ta
in
to
a
ne
w
s
pa
c
e
w
he
r
e
th
e
m
os
t
im
por
ta
nt
in
f
or
m
a
ti
on
is
r
e
ta
in
e
d,
r
a
th
e
r
th
a
n
s
e
l
e
c
ti
ng
a
s
ubs
e
t
of
th
e
or
ig
in
a
l
f
e
a
tu
r
e
s
.
T
he
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
of
c
lu
s
te
r
in
g
a
lg
or
it
hm
s
in
DL
can
va
r
y
w
id
e
ly
de
pe
ndi
ng
on
th
e
s
pe
c
if
ic
c
lu
s
te
r
in
g
m
e
th
od,
da
ta
s
iz
e
,
a
nd
c
h
a
r
a
c
te
r
is
ti
c
s
.
In
th
e
c
a
s
e
of
K
-
M
e
a
n
s
c
l
us
te
r
in
g
th
a
t
in
vol
ve
s
it
e
r
a
ti
ng
ove
r
th
e
d
a
ta
s
e
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
P
hi
s
hi
ng de
te
c
ti
on us
in
g c
lu
s
te
r
in
g and mac
hi
ne
l
e
a
r
ni
ng
(
L
uai
A
l
-
Shal
abi
)
4535
to
a
s
s
ig
n
da
t
a
poi
nt
s
to
c
lu
s
te
r
s
a
nd
upd
a
te
c
lu
s
t
e
r
c
e
nt
r
oi
ds
.
T
he
time
c
om
pl
e
xi
ty
f
or
K
-
M
e
a
n
s
is
ty
pi
c
a
ll
y
O
(
n*k*I
*d)
,
w
he
r
e
:n
is
th
e
num
be
r
of
da
ta
poi
nt
s
(
in
s
ta
nc
e
s
)
,
k
is
th
e
num
be
r
of
c
lu
s
te
r
s
,
I
is
th
e
num
be
r
of
it
e
r
a
ti
ons
,
d
is
th
e
num
be
r
of
f
e
a
tu
r
e
s
(
di
m
e
ns
io
ns
)
.
T
he
num
be
r
of
it
e
r
a
ti
ons
(I)
can
va
r
y,
a
nd
ty
pi
c
a
ll
y
K
-
M
e
a
ns
c
o
nve
r
ge
s
r
e
la
ti
ve
ly
qui
c
kl
y,
but
it
'
s
not
gua
r
a
nt
e
e
d
to
c
onve
r
ge
to
a
g
lo
ba
l
opt
im
um
.
C
lu
s
te
r
in
g
it
s
e
lf
doe
s
not
di
r
e
c
tl
y
a
f
f
e
c
t
th
e
time
c
om
pl
e
xi
ty
of
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
s
.
H
ow
e
ve
r
,
th
e
r
e
c
a
n
be
in
di
r
e
c
t
r
e
la
ti
o
ns
hi
ps
be
tw
e
e
n
c
lu
s
te
r
in
g
a
nd
f
e
a
tu
r
e
s
e
le
c
ti
on
th
a
t
m
a
y
im
pa
c
t
th
e
ove
r
a
ll
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
of
a
ML
pi
pe
li
ne
,
s
uc
h
as
pr
e
pr
oc
e
s
s
in
g,
f
e
a
tu
r
e
im
por
ta
nc
e
,
da
ta
s
iz
e
,
a
nd
pa
r
a
ll
e
li
z
a
ti
on.
6.
C
O
N
C
L
U
S
I
O
N
T
he
c
hoi
c
e
of
us
in
g
R
S
,
S
c
C
,
P
C
A
,
K
-
m
e
a
n,
D
T
,
or
DL
de
pe
nds
on
th
e
s
pe
c
if
ic
pr
obl
e
m
,
da
ta
,
a
nd
goa
ls
.
E
a
c
h
of
th
e
s
e
te
c
hni
que
s
ha
s
it
s
s
tr
e
ngt
hs
a
nd
w
e
a
kn
e
s
s
e
s
,
a
nd
th
e
r
ig
ht
c
hoi
c
e
s
houl
d
be
b
a
s
e
d
on
th
e
c
ha
r
a
c
te
r
is
ti
c
s
of
th
e
a
na
ly
s
is
.
T
hi
s
r
e
s
e
a
r
c
h
pr
opos
e
s
a
hybr
id
m
e
th
od
of
tr
a
di
ti
ona
l
f
e
a
tu
r
e
s
e
le
c
ti
on
m
e
th
od
s
,
K
-
m
e
a
n
c
lu
s
te
r
in
g
in
a
ddi
ti
on
to
c
la
s
s
if
ic
a
ti
on
us
in
g
DL
a
nd
DT
(
S
c
C
-
K
-
m
e
a
n)
f
or
two
di
f
f
e
r
e
nt
D
S
s
th
a
t
in
c
lu
de
da
ta
a
bout
phi
s
hi
ng
de
te
c
ti
on.
S
im
ul
a
ti
on
r
e
s
ul
ts
s
ho
w
th
a
t
th
e
pr
opos
e
d
a
lg
or
it
hm
out
pe
r
f
or
m
s
th
e
tr
a
di
ti
ona
l
te
s
te
d
m
e
th
ods
th
a
t
us
e
DL
a
nd
DT
to
ge
th
e
r
w
it
h
S
c
C
,
R
S
,
a
nd
P
C
A
m
e
th
ods
.
A
ls
o
th
e
ot
he
r
pr
opos
e
d
hybr
id
m
e
th
ods
th
a
t
us
e
DL
a
nd
DT
to
ge
th
e
r
w
it
h
RS
-
K
-
m
e
a
n
a
nd
P
C
A
-
K
-
m
e
a
n,
a
nd
ot
he
r
hybr
i
d
m
e
th
ods
e
xpl
a
in
e
d
e
a
r
li
e
r
in
s
e
c
ti
on
4.
F
ut
ur
e
s
tu
di
e
s
w
il
l
in
c
lu
d
e
c
om
pa
r
in
g
th
e
pr
opos
e
d
a
lg
or
it
hm
w
it
h
ot
he
r
ML
a
lg
or
it
hm
s
a
nd
in
ve
s
ti
ga
ti
ng
pr
o
s
pe
c
ts
f
or
de
ve
lo
pi
ng
to
ol
s
to
im
pr
ove
th
e
pe
r
f
or
m
a
nc
e
of
th
e
pr
opos
e
d
a
lg
or
it
hm
.
T
he
pr
opos
e
d
m
e
th
od
c
oul
d
a
l
s
o
be
te
s
te
d
a
g
a
in
s
t
f
ur
th
e
r
hi
ghl
y
di
m
e
ns
io
ne
d
phi
s
hi
ng
da
ta
s
e
ts
,
s
e
m
i
-
s
tr
uc
tu
r
e
d
a
nd
uns
tr
uc
tu
r
e
d
phi
s
hi
ng
da
ta
s
e
ts
,
a
nd
ot
he
r
ty
pe
s
of
a
tt
a
c
ks
s
uc
h
as
s
pa
m
a
nd
m
a
lwa
r
e
.
R
E
F
E
R
E
N
C
E
S
[
1
]
I
.
G
uyon
a
nd
A
.
E
l
i
s
s
e
f
,
“
an
i
nt
r
oduc
t
i
on
to
va
r
i
a
bl
e
a
nd
f
e
a
t
ur
e
s
e
l
e
c
t
i
on
,”
J
ou
r
nal
of
m
ac
hi
ne
l
e
ar
ni
ng
r
e
s
e
ar
c
h
,
vol
.
3,
pp.
1157
–
1182,
2003.
[
2
]
A.
K.
D
a
s
,
S.
S
e
ngupt
a
,
a
nd
S.
B
ha
t
t
a
c
ha
r
yya
,
“A
gr
oup
i
nc
r
e
m
e
nt
a
l
f
e
a
t
ur
e
s
e
l
e
c
t
i
on
f
or
c
l
a
s
s
i
f
i
c
a
t
i
on
us
i
ng
r
ough
s
e
t
t
he
or
y
ba
s
e
d
ge
ne
t
i
c
a
l
gor
i
t
hm
,”
A
ppl
i
e
d
Sof
t
C
om
put
i
ng
J
our
nal
,
vol
.
65,
pp.
400
–
411
, 2018, doi
:
10.1016/
j
.a
s
oc
.2018.01.040.
[
3]
A.
K.
J
a
i
n,
S.
P
a
r
a
s
ha
r
,
P.
K
a
t
a
r
e
,
a
nd
I.
S
ha
r
m
a
,
“
P
hi
s
hS
K
a
P
e
:
A
c
ont
e
nt
ba
s
e
d
a
ppr
oa
c
h
t
o
e
s
c
a
pe
phi
s
hi
ng
a
t
t
a
c
k
s
,”
P
r
oc
e
di
a
C
om
put
e
r
Sc
i
e
n
c
e
,
vol
.
171,
pp.
1102
–
1109,
2020,
doi
:
10.1016/
j
.pr
oc
s
.2020.04.118.
[
4]
A
.
A
nha
r
i
,
“
A
l
e
xa
da
t
a
s
e
t
,”
K
aggl
e
,
2023.
A
c
c
e
s
s
e
d:
M
a
y
1,
2023.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
w
w
w
.ka
ggl
e
.c
om
/
da
t
a
s
e
t
s
/
a
a
nha
r
i
/
a
l
e
xa
-
da
t
a
s
e
t
[
5]
O
pe
nP
hi
s
h,
“
O
pe
nP
hi
s
h
-
P
hi
s
hi
ng
I
nt
e
l
l
i
ge
nc
e
,
”
O
pe
n
P
hi
s
h
,
2021
.
A
c
c
e
s
s
e
d:
M
a
y
1,
2023.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
ope
nphi
s
h.c
om
/
[
6]
“
P
h
i
s
h
T
a
n
k
|
J
o
i
n
t
h
e
f
i
g
h
t
a
g
a
i
n
s
t
p
h
i
s
h
i
ng
”
P
h
i
s
h
T
a
nk
.
A
c
c
e
s
s
e
d
:
O
c
t
ob
e
r
1
,
20
2
3
.
[
O
n
l
i
ne
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
p
h
i
s
h
t
a
n
k
.
c
o
m
/
i
n
de
x
.
ph
p
[
7]
G.
S
onow
a
l
a
nd
K.
S.
K
uppus
a
m
y,
“
P
hi
D
M
A
–
A
phi
s
hi
ng
de
t
e
c
t
i
on
m
ode
l
w
i
t
h
m
ul
t
i
-
f
i
l
t
e
r
a
ppr
oa
c
h,”
J
our
nal
of
K
i
ng
Saud
U
ni
v
e
r
s
i
t
y
-
C
om
put
e
r
and
I
nf
or
m
at
i
on
Sc
i
e
nc
e
s
,
vol
.
32,
no.
1,
pp.
99
–
112,
20
20,
doi
:
10.1016/
j
.j
ks
uc
i
.2017.07.005.
[
8]
“
P
hi
s
hl
oa
d
da
t
a
s
e
t
,
”
P
hi
s
h
L
oad,
2023
.
A
c
c
e
s
s
e
d
:
M
a
y
1,
2023
.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
w
w
w
.m
e
di
e
n.i
f
i
.l
m
u.de
/
t
e
a
m
/
m
a
x.m
a
ur
e
r
/
f
i
l
e
s
/
phi
s
hl
oa
d/
dow
nl
oa
d.ht
m
l
[
9]
R.
S.
R
a
o,
A.
R.
P
a
i
s
,
a
nd
P.
A
na
nd,
“A
he
ur
i
s
t
i
c
t
e
c
hni
que
to
de
t
e
c
t
phi
s
hi
ng
w
e
bs
i
t
e
s
us
i
ng
T
W
S
V
M
c
l
a
s
s
i
f
i
e
r
,”
N
e
ur
a
l
C
om
put
i
ng
and
A
ppl
i
c
at
i
ons
,
vol
.
33,
no.
11,
pp.
5733
–
5752,
2021,
doi
:
10.1007/
s
00521
-
020
-
05354
-
z.
[
10]
M.
B
a
ba
gol
i
,
M.
P.
A
gha
ba
ba
,
a
nd
V.
S
ol
ouk,
“
H
e
ur
i
s
t
i
c
nonl
i
ne
a
r
r
e
gr
e
s
s
i
on
s
t
r
a
t
e
gy
f
or
de
t
e
c
t
i
ng
phi
s
hi
ng
w
e
bs
i
t
e
s
,”
Sof
t
C
om
put
i
ng,
vol
.
23,
no.
12,
pp.
4315
–
4327,
2019,
doi
:
10.1007/
s
00500
-
018
-
30
84
-
2.
[
11]
R.
M.
M
oha
m
m
a
d,
F.
T
ha
bt
a
h,
a
nd
L.
M
c
c
l
us
ke
y,
“
P
hi
s
hi
ng
w
e
bs
i
t
e
s
f
e
a
t
ur
e
s
,”
Sc
hool
of
C
om
put
i
ng and
E
ngi
ne
e
r
i
ng, U
ni
v
e
r
s
i
t
y
of
H
udde
r
s
f
i
e
l
d
,
pp.
1
–
7,
201
5
.
[
12]
R
. M
oha
m
m
a
d a
nd
L
. M
c
C
l
us
ke
y, “
P
hi
s
hi
ng w
e
bs
i
t
e
s
,”
U
C
I
M
ac
hi
ne
L
e
ar
ni
n
g R
e
pos
i
t
or
y
, 2012, doi
:
10.24432/
C
51W
2X
.
[
13]
K.
L.
C
hi
e
w
,
C.
L.
T
a
n,
K.
S.
W
ong,
K.
S.
C.
Y
ong,
a
nd
W.
K.
T
i
ong,
“A
n
e
w
hybr
i
d
e
ns
e
m
bl
e
f
e
a
t
ur
e
s
e
l
e
c
t
i
on
f
r
a
m
e
w
or
k
f
or
m
a
c
hi
ne
l
e
a
r
ni
ng
-
ba
s
e
d
phi
s
hi
ng
de
t
e
c
t
i
on
s
ys
t
e
m
,”
I
nf
or
m
at
i
on
Sc
i
e
nc
e
s
,
vol
.
484,
pp.
153
–
166,
2019,
doi
:
10.1016/
j
.i
ns
.2019.01.064.
[
14]
M.
M.
Y
a
dol
l
a
hi
,
F.
S
hoe
l
e
h,
E.
S
e
r
ka
ni
,
A.
M
a
da
ni
,
a
nd
H.
G
ha
r
a
e
e
,
“
A
n
a
da
pt
i
ve
m
a
c
hi
ne
l
e
a
r
ni
ng ba
s
e
d
a
ppr
oa
c
h f
or
phi
s
hi
ng
de
t
e
c
t
i
on
us
i
ng
hybr
i
d
f
e
a
t
ur
e
s
,”
2019
5t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
W
e
b
R
e
s
e
ar
c
h,
I
C
W
R
2019
,
pp.
281
–
286,
2019,
doi
:
10.1109/
I
C
W
R
.2019.8765265.
[
15]
S.
S
m
a
di
,
N.
A
s
l
a
m
,
a
nd
L.
Z
ha
ng,
“
D
e
t
e
c
t
i
on
of
onl
i
ne
phi
s
hi
ng
e
m
a
i
l
u
s
i
ng
dyna
m
i
c
e
vol
vi
ng
ne
ur
a
l
ne
t
w
or
k
ba
s
e
d
on
r
e
i
nf
or
c
e
m
e
nt
l
e
a
r
ni
ng,”
D
e
c
i
s
i
on
Suppor
t
Sy
s
t
e
m
s
,
vol
.
107,
pp.
88
–
102,
2018
,
doi
:
10.1016/
j
.ds
s
.2018.01.001.
[
16]
J
.
N
a
z
a
r
i
o,
“
I
nde
x
of
/
~
j
os
e
/
phi
s
hi
ng,”
M
onk
e
y
.
A
c
c
e
s
s
e
d:
M
a
y 1, 2023.
[
O
nl
i
n
e
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
m
onke
y.or
g/
~
j
os
e
/
phi
s
hi
ng/
[
17]
“
I
nde
x
of
/
ol
d/
publ
i
c
c
or
pus
,”
Spam
A
s
s
as
s
i
n
,
2019
.
A
c
c
e
s
s
e
d
:
M
a
y
1,
2023
.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
ps
:
/
/
s
pa
m
a
s
s
a
s
s
i
n.a
pa
c
he
.or
g/
ol
d/
publ
i
c
c
or
pus
/
[
18]
W.
W
e
i
,
Q.
K
e
,
J.
N
ow
a
k,
M.
K
or
yt
kow
s
ki
,
R.
S
c
he
r
e
r
,
a
nd
M.
W
o
ź
ni
a
k,
“
A
c
c
ur
a
t
e
a
nd
f
a
s
t
U
R
L
phi
s
hi
ng
d
e
t
e
c
t
or
:
A
c
onvol
ut
i
ona
l
ne
ur
a
l
ne
t
w
or
k
a
ppr
oa
c
h,”
C
om
put
e
r
N
e
t
w
or
k
s
,
vol
.
178,
2020,
d
oi
:
10.1016/
j
.c
om
ne
t
.2020.107275.
[
19]
“
C
om
m
on
c
r
a
w
l
-
ope
n
r
e
pos
i
t
or
y
of
w
e
b
c
r
a
w
l
da
t
a
,”
C
o
m
m
on
C
r
aw
l
.
A
c
c
e
s
s
e
d:
M
a
y
01,
2023.
[
O
nl
i
ne
]
.
A
va
i
l
a
bl
e
:
ht
t
p:
/
/
c
om
m
onc
r
a
w
l
.or
g/
[
20]
S.
S
m
a
di
,
N.
A
s
l
a
m
,
L.
Z
ha
ng,
R.
A
l
a
s
e
m
,
a
nd
M.
A.
H
os
s
a
i
n,
“
D
e
t
e
c
t
i
on
of
phi
s
hi
ng
e
m
a
i
l
s
us
i
ng
da
t
a
m
i
ni
ng
a
l
gor
i
t
h
m
s
,”
SK
I
M
A
2015
-
9t
h
I
nt
e
r
nat
i
onal
C
onf
e
r
e
nc
e
on
Sof
t
w
a
r
e
,
K
no
w
l
e
dge
,
I
nf
or
m
at
i
on
M
anage
m
e
nt
and
A
ppl
i
c
at
i
ons
,
2016,
doi
:
10.1109/
S
K
I
M
A
.2015.7399985.
Evaluation Warning : The document was created with Spire.PDF for Python.