I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2020
,
p
p
.
978
~
986
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
0
i
1
.
pp
9
7
8
-
986
978
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m/in
d
ex
.
p
h
p
/I
JE
C
E
Perf
o
r
m
a
nce
ana
ly
sis
o
f
bina
ry
and
m
ulticlas
s
m
o
del
s
using
a
z
ure
m
a
chine
le
a
rning
S
m
it
ha
Ra
j
a
g
o
pa
l
,
K
a
t
ig
a
ne
re
Sid
da
ra
m
a
pp
a
H
a
re
esh
a
,
P
o
o
rni
m
a
P
a
nd
ura
ng
a
K
un
da
pu
r
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter A
p
p
li
c
a
ti
o
n
s,
M
a
n
i
p
a
l
In
st
it
u
te
o
f
T
e
c
h
n
o
lo
g
y
,
M
a
n
ip
a
l
A
c
a
d
e
m
y
o
f
H
ig
h
e
r
Ed
u
c
a
ti
o
n
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
A
p
r
1
7
,
2
0
1
9
R
ev
i
s
ed
Sep
2
6
,
2
0
1
9
A
cc
ep
ted
Oct
3
,
2
0
1
9
Ne
tw
o
rk
d
a
ta
is
e
x
p
a
n
d
in
g
a
n
d
t
h
a
t
to
o
a
t
a
n
a
larm
in
g
ra
te.
Be
sid
e
s,
th
e
so
p
h
isti
c
a
ted
a
tt
a
c
k
to
o
ls
u
se
d
b
y
h
a
c
k
e
r
s
lea
d
to
c
a
p
ricio
u
s
c
y
b
e
r
th
re
a
t
lan
d
sc
a
p
e
.
T
ra
d
it
io
n
a
l
m
o
d
e
ls
p
ro
p
o
se
d
i
n
th
e
f
ield
o
f
n
e
tw
o
rk
in
tr
u
sio
n
d
e
tec
ti
o
n
u
sin
g
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
e
m
p
h
a
siz
e
m
o
re
o
n
im
p
ro
v
in
g
a
tt
a
c
k
d
e
tec
ti
o
n
ra
te
a
n
d
re
d
u
c
in
g
f
a
lse
a
lar
m
s
b
u
t
ti
m
e
e
ff
icie
n
c
y
is
o
f
ten
o
v
e
rlo
o
k
e
d
.
T
h
e
re
f
o
re
,
in
o
r
d
e
r
t
o
a
d
d
re
ss
t
h
is
li
m
it
a
ti
o
n
,
a
m
o
d
e
rn
so
l
u
ti
o
n
h
a
s
b
e
e
n
p
re
se
n
ted
u
sin
g
M
a
c
h
in
e
L
e
a
rn
in
g
-
as
-
a
-
S
e
r
v
ice
p
latf
o
r
m
.
T
h
e
p
ro
p
o
se
d
w
o
rk
a
n
a
ly
se
s
th
e
p
e
rf
o
rm
a
n
c
e
o
f
e
ig
h
t
tw
o
-
c
la
ss
a
n
d
t
h
re
e
m
u
lt
icla
ss
a
lg
o
rit
h
m
s
u
sin
g
UN
S
W
NB
-
1
5
,
a
m
o
d
e
rn
in
tru
si
o
n
d
e
tec
ti
o
n
d
a
tas
e
t.
8
2
,
3
3
2
tes
ti
n
g
sa
m
p
les
we
re
c
o
n
sid
e
re
d
to
e
v
a
lu
a
te
th
e
p
e
rf
o
r
m
a
n
c
e
o
f
a
lg
o
rit
h
m
s.
T
h
e
p
ro
p
o
se
d
tw
o
c
las
s
d
e
c
isio
n
f
o
re
st
m
o
d
e
l
e
x
h
ib
i
ted
9
9
.
2
%
a
c
c
u
ra
c
y
a
n
d
to
o
k
6
se
c
o
n
d
s
to
lea
rn
1
,
7
5
,
3
4
1
n
e
tw
o
rk
in
sta
n
c
e
s.
M
u
lt
icla
ss
c
las
sif
ic
a
ti
o
n
tas
k
wa
s
a
lso
u
n
d
e
rtak
e
n
w
h
e
re
in
a
tt
a
c
k
t
y
p
e
s
li
k
e
g
e
n
e
ric,
e
x
p
lo
it
s,
sh
e
ll
c
o
d
e
a
n
d
w
o
rm
s
we
re
c
las
si
f
ied
w
it
h
a
re
c
a
ll
p
e
rc
e
n
tag
e
o
f
9
9
%
,
9
4
.
4
9
%
,
9
1
.
7
9
%
a
n
d
9
0
.
9
%
re
sp
e
c
ti
v
e
l
y
b
y
th
e
m
u
lt
icla
ss
d
e
c
isio
n
f
o
re
st
m
o
d
e
l
th
a
t
a
lso
lea
p
f
ro
g
g
e
d
o
th
e
rs
in
term
s
o
f
train
in
g
a
n
d
e
x
e
c
u
ti
o
n
ti
m
e
.
K
ey
w
o
r
d
s
:
A
z
u
r
e
m
ac
h
i
n
e
lear
n
i
n
g
Dec
is
io
n
f
o
r
est
I
n
tr
u
s
io
n
d
etec
tio
n
L
o
ca
ll
y
d
ee
p
SVM
Mu
t
u
al
in
f
o
r
m
a
tio
n
UNSW
NB
-
15
Co
p
y
rig
h
t
©
2
0
2
0
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
P
o
o
r
n
i
m
a
P
an
d
u
r
a
n
g
a
K
u
n
d
ap
u
r
,
Dep
ar
t
m
en
t o
f
C
o
m
p
u
ter
A
p
p
licatio
n
s
,
Ma
n
ip
al
I
n
s
tit
u
te
o
f
T
ec
h
n
o
lo
g
y
,
Ma
n
ip
al
A
ca
d
e
m
y
o
f
Hi
g
h
er
E
d
u
ca
tio
n
,
Ma
n
ip
al,
I
n
d
ia
.
E
m
ail:
p
o
o
r
n
i
m
a.
g
ir
is
h
@
m
an
i
p
al.
ed
u
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
in
cr
ea
s
ed
u
s
e
o
f
d
ev
ice
s
ass
o
ciate
d
w
it
h
i
n
ter
n
et
g
e
n
e
r
ate
h
u
g
e
v
o
lu
m
es
o
f
n
et
w
o
r
k
d
ata
[
1
]
.
T
h
is
is
al
s
o
ac
co
m
p
a
n
ied
b
y
ad
v
a
n
ce
d
le
v
el
o
f
c
y
b
er
-
a
ttack
s
th
a
t
s
e
v
er
el
y
h
a
m
p
er
th
e
co
n
f
id
e
n
tialit
y
,
in
te
g
r
it
y
a
n
d
av
ailab
ilit
y
o
f
co
m
p
u
ter
r
eso
u
r
ce
s
[
2
,
3
]
.
R
o
b
u
s
t
n
e
t
w
o
r
k
in
tr
u
s
io
n
d
etec
ti
o
n
s
y
s
te
m
s
ar
e
th
e
n
ee
d
o
f
th
e
h
o
u
r
to
s
af
e
g
u
ar
d
co
n
f
id
en
t
ial
i
n
f
o
r
m
atio
n
a
g
ain
s
t
m
al
icio
u
s
ac
ti
v
itie
s
[
4
]
.
Ma
ch
i
n
e
lear
n
i
n
g
alg
o
r
ith
m
s
ar
e
co
m
m
o
n
l
y
u
s
e
d
to
ad
d
r
ess
th
e
p
r
o
b
lem
o
f
n
e
t
w
o
r
k
i
n
tr
u
s
io
n
d
etec
tio
n
[
5
]
.
W
h
en
e
v
er
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
ith
m
s
ar
e
e
m
p
l
o
y
ed
i
n
t
h
e
f
ie
ld
o
f
n
et
w
o
r
k
in
tr
u
s
io
n
d
etec
tio
n
,
t
w
o
r
ec
u
r
r
in
g
p
r
o
b
lem
s
ar
e
co
m
m
o
n
l
y
e
n
co
u
n
ter
ed
b
y
s
e
cu
r
it
y
e
x
p
er
ts
,
i.e
.
,
p
r
o
lo
n
g
ed
tr
ain
i
n
g
a
n
d
p
r
ed
ictio
n
ti
m
e.
T
h
e
tr
ain
in
g
ti
m
e
o
f
alg
o
r
ith
m
s
s
p
an
f
r
o
m
s
ec
o
n
d
s
to
h
o
u
r
s
[
6
,
7
]
.
T
h
e
lo
n
g
er
tr
ain
in
g
t
i
m
e
tak
e
n
b
y
t
h
e
I
n
tr
u
s
io
n
Dete
ct
io
n
S
y
s
te
m
s
to
an
al
y
s
e
t
h
e
d
ata
lead
s
to
s
u
b
s
tan
t
ial
d
ela
y
s
in
g
e
n
er
ati
n
g
aler
ts
[
8
,
9
]
,
o
b
v
io
u
s
l
y
co
n
s
id
er
ed
u
n
f
a
v
o
u
r
ab
le
in
th
e
f
ield
o
f
i
n
tr
u
s
io
n
d
etec
tio
n
r
esear
ch
.
T
h
e
p
r
o
b
lem
,
h
o
w
e
v
er
,
p
er
s
is
ts
b
ec
au
s
e
n
et
w
o
r
k
in
tr
u
s
io
n
d
etec
tio
n
i
n
v
o
lv
e
s
b
ig
d
ata
i
n
v
e
s
ti
g
atio
n
to
o
g
i
v
en
its
m
a
m
m
o
th
co
m
p
le
x
it
y
[
9
,
10]
.
A
cc
o
r
d
in
g
to
[
11
]
,
1
Gig
ab
its
p
er
s
ec
o
n
d
(
GB
P
s
)
o
f
n
et
w
o
r
k
tr
af
f
ic
alo
n
e
ca
n
i
n
tr
o
d
u
ce
b
ig
d
ata
ch
al
len
g
e
s
.
T
r
ad
itio
n
al
d
ata
m
i
n
i
n
g
to
o
ls
li
k
e
W
e
k
a,
Scik
it
lear
n
a
n
d
co
n
v
e
n
tio
n
al
n
u
m
er
ical
en
v
ir
o
n
m
e
n
ts
lik
e
Ma
tlab
m
a
y
n
o
t
b
e
ab
le
to
ad
d
r
ess
th
e
e
v
er
in
cr
e
asin
g
i
s
s
u
es
o
f
d
is
tr
ib
u
ted
d
ata
s
etti
n
g
s
[
1
2
]
.
P
er
f
o
r
m
a
n
ce
an
d
s
ca
lab
ilit
y
ar
e
th
e
t
w
o
m
aj
o
r
co
n
s
id
er
atio
n
s
f
o
r
co
n
d
u
ctin
g
n
et
w
o
r
k
in
tr
u
s
i
o
n
d
etec
tio
n
s
tu
d
y
.
B
ig
d
ata
p
r
o
ce
s
s
in
g
p
latf
o
r
m
s
lik
e
P
i
g
[
1
3
]
,
Sp
ar
k
m
ac
h
i
n
e
lear
n
i
n
g
[
1
4
]
an
d
A
z
u
r
e
m
a
ch
in
e
lear
n
in
g
[
1
5
]
ar
e
th
e
p
r
ef
er
r
ed
ch
o
ices
i
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
erfo
r
ma
n
ce
a
n
a
lysi
s
o
f b
in
a
r
y
a
n
d
mu
lticla
s
s
mo
d
els u
s
in
g
a
z
u
r
e
ma
ch
in
e
lea
r
n
in
g
(
S
mith
a
R
a
ja
g
o
p
a
l
)
979
th
e
m
o
d
er
n
s
ce
n
ar
io
g
iv
e
n
th
eir
ab
ilit
y
to
u
p
h
o
ld
m
e
m
o
r
y
r
eq
u
ir
e
m
en
ts
an
d
i
m
p
le
m
e
n
ta
tio
n
ess
e
n
tial
s
[
16]
.
Go
in
g
b
y
th
e
s
e
co
n
s
id
er
ati
o
n
s
,
it
i
s
i
m
p
er
ati
v
e
to
i
n
tr
o
d
u
ce
r
ad
ical
ad
v
an
ce
m
e
n
ts
to
in
tr
u
s
io
n
d
etec
tio
n
in
f
r
astru
ct
u
r
e.
A
z
u
r
e
Ma
ch
i
n
e
L
ea
r
n
i
n
g
i
s
o
n
e
s
u
c
h
Ma
ch
in
e
lear
n
i
n
g
as
-
a
Ser
v
ice
in
it
i
ativ
e
b
y
Mic
r
o
s
o
f
t
th
at
ca
n
b
e
e
m
p
lo
y
ed
to
d
e
v
elo
p
p
r
ed
ictiv
e
m
o
d
els.
T
h
e
p
r
o
p
o
s
ed
w
o
r
k
is
an
ill
u
s
t
r
atio
n
to
h
ig
h
lig
h
t
th
e
ad
v
an
ta
g
e
s
o
f
t
h
is
i
n
itiati
v
e
b
y
co
n
s
id
er
in
g
a
n
e
t
w
o
r
k
i
n
tr
u
s
io
n
d
etec
tio
n
u
s
e
ca
s
e
i
m
p
le
m
e
n
ted
th
r
o
u
g
h
s
u
p
er
v
i
s
ed
m
ac
h
i
n
e
lear
n
i
n
g
tech
n
iq
u
e
s
.
Gi
v
en
th
e
ex
i
s
t
en
ce
o
f
d
iv
er
s
e
a
lg
o
r
it
h
m
s
i
n
m
ac
h
in
e
lear
n
i
n
g
s
tu
d
y
,
it
is
o
f
te
n
ad
v
is
ab
le
to
i
n
v
e
s
ti
g
ate
t
h
e
p
er
f
o
r
m
an
ce
o
f
i
n
d
iv
id
u
al
a
lg
o
r
it
h
m
s
s
o
t
h
at
o
p
ti
m
al
p
r
ed
ictio
n
s
ca
n
b
e
ac
h
iev
ed
.
I
t
is
w
o
r
th
w
h
ile
to
m
e
n
tio
n
th
at
alg
o
r
it
h
m
s
p
er
f
o
r
m
d
if
f
er
en
tl
y
f
o
r
a
g
iv
e
n
d
ataset.
T
h
er
ef
o
r
e,
s
u
c
h
a
co
m
p
ar
at
iv
e
s
t
u
d
y
a
s
p
r
o
p
o
s
ed
,
b
ec
o
m
e
s
i
n
d
is
p
e
n
s
ab
le
i
n
t
h
e
f
ield
o
f
m
ac
h
in
e
lear
n
in
g
r
esear
ch
.
T
h
e
o
b
j
ec
tiv
e
o
f
th
e
p
r
o
p
o
s
ed
w
o
r
k
is
to
an
al
y
s
e
t
h
e
p
er
f
o
r
m
an
ce
o
f
v
ar
io
u
s
alg
o
r
i
th
m
s
a
n
d
in
v
e
s
ti
g
ate
t
h
eir
tr
ai
n
i
n
g
t
i
m
e,
p
r
ed
ictio
n
ti
m
e,
attac
k
d
et
ec
tio
n
r
ate
an
d
f
alse
alar
m
r
ate
b
y
co
n
s
id
er
in
g
n
et
w
o
r
k
i
n
s
tan
ce
s
o
f
U
NSW
NB
-
1
5
d
ataset
o
n
a
s
o
p
h
is
t
icate
d
M
ac
h
i
n
e
lear
n
i
n
g
as
a
s
er
v
ice
(
M
L
aa
S)
p
latf
o
r
m
ca
lled
M
icr
o
s
o
f
t
A
z
u
r
e
Ma
ch
in
e
L
ea
r
n
i
n
g
St
u
d
io
(
MA
M
L
S).
A
m
o
d
er
n
a
n
d
a
co
m
p
r
e
h
e
n
s
i
v
e
d
ataset
is
ess
e
n
tial
to
ev
al
u
ate
t
h
e
ef
f
ec
tiv
e
n
e
s
s
o
f
t
h
e
p
r
o
p
o
s
e
d
ap
p
r
o
ac
h
an
d
UNSW
NB
-
1
5
d
ataset
s
er
v
e
s
th
e
p
u
r
p
o
s
e
[
1
7
-
1
9
]
.
A
s
ig
n
i
f
ica
n
t
ad
v
an
tag
e
o
f
a
n
y
M
L
a
aS
o
f
f
er
i
n
g
i
s
its
ab
ilit
y
to
s
av
e
co
m
p
u
tat
io
n
a
l
r
eso
u
r
ce
s
th
at
i
n
v
o
l
v
e
ex
ce
e
s
i
v
e
co
s
ts
[
20
,
2
1
]
.
T
h
e
n
o
v
elt
y
o
f
th
e
p
r
o
p
o
s
ed
a
p
p
r
o
ac
h
is
t
h
at
th
e
f
alse
alar
m
r
at
e
g
e
n
er
ated
b
y
t
w
o
clas
s
d
ec
is
io
n
f
o
r
est
m
o
d
el
is
q
u
ite
n
eg
l
ig
ib
le
an
d
th
e
attac
k
d
ete
ctio
n
ca
p
ab
ilit
y
o
f
m
u
lticla
s
s
d
ec
is
io
n
f
o
r
est
m
o
d
el
is
d
ef
in
ite
l
y
d
esira
b
le.
I
t
is
w
o
r
t
h
w
h
ile
to
m
e
n
tio
n
th
at
t
h
e
r
esu
lts
o
f
class
i
f
icatio
n
task
s
ar
e
q
u
ite
s
u
p
er
io
r
th
an
ex
is
tin
g
s
tate
o
f
th
e
ar
t
tech
n
iq
u
e
s
.
So
m
e
ex
is
ti
n
g
s
tu
d
ies
i
n
th
e
liter
at
u
r
e
h
av
e
e
x
p
lo
r
ed
t
h
e
p
er
f
o
r
m
an
ce
o
f
d
if
f
er
en
t
m
ac
h
in
e
lear
n
in
g
al
g
o
r
ith
m
s
o
n
UNSW
NB
-
15
d
ataset
as e
lu
cid
ated
b
elo
w
.
As
d
escr
ib
ed
in
[
1
8
]
,
s
ix
d
if
f
er
en
t
tec
h
n
iq
u
es
w
er
e
ap
p
lied
to
class
if
y
t
h
e
n
et
w
o
r
k
in
s
tan
ce
s
o
f
UNSW
NB
-
1
5
d
ataset.
T
h
e
h
ig
h
e
s
t
ac
c
u
r
ac
y
o
b
tain
ed
w
as
8
5
.
5
6
%
u
s
in
g
d
ec
is
io
n
tr
ee
th
at
also
g
e
n
er
ated
a
f
alse
alar
m
r
ate
o
f
1
5
.
7
8
%.
A
s
d
i
s
cu
s
s
ed
in
[
2
2
]
,
ex
p
er
i
m
e
n
tatio
n
w
as
co
n
d
u
cted
o
n
A
p
ac
h
e
Sp
ar
k
to
i
m
p
r
o
v
e
t
h
e
ac
c
u
r
a
c
y
a
n
d
it
c
an
b
e
n
o
ted
t
h
at
R
E
P
tr
ee
m
o
d
el
ac
h
ie
v
ed
a
n
ac
c
u
r
ac
y
o
f
9
3
.
5
6
%.
T
h
e
tr
ain
i
n
g
ti
m
e
tak
e
n
w
a
s
7
.
9
2
s
ec
o
n
d
s
t
o
lear
n
4
7
,
3
4
2
in
s
tan
ce
s
.
A
wr
ap
p
er
ap
p
r
o
ac
h
w
as
i
m
p
le
m
e
n
ted
i
n
[
2
3
]
u
s
i
n
g
g
en
et
ic
alg
o
r
it
h
m
a
n
d
v
ar
io
u
s
tr
ee
b
ased
class
if
ier
s
b
y
s
elec
tin
g
d
if
f
er
en
t
s
u
b
s
et
s
o
f
f
ea
t
u
r
es.
An
ac
cu
r
ac
y
o
f
8
1
.
4
2
%
an
d
a
f
alse
alar
m
r
ate
o
f
6
.
3
9
%
w
as
o
b
tai
n
ed
u
s
in
g
t
h
i
s
ap
p
r
o
ac
h
b
u
t
w
r
ap
p
er
ap
p
r
o
ac
h
es
ar
e
co
n
s
id
er
ed
to
b
e
co
m
p
u
tatio
n
all
y
e
x
h
a
u
s
tiv
e
[
2
4
]
.
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
f
o
u
r
cla
s
s
i
f
icatio
n
al
g
o
r
ith
m
s
l
ik
e
:
Dec
is
io
n
T
r
ee
,
R
an
d
o
m
Fo
r
est,
SVM
an
d
Naiv
e
B
a
y
e
s
w
er
e
co
m
p
ar
ed
an
d
A
p
ac
h
e
Sp
ar
k
w
a
s
u
s
ed
as
a
p
r
o
ce
s
s
in
g
p
ar
ad
ig
m
[
2
5
]
.
I
t
w
as
n
o
ticed
th
at
R
an
d
o
m
Fo
r
est
w
as
t
h
e
b
est
p
er
f
o
r
m
i
n
g
clas
s
i
f
ier
w
ith
9
7
.
4
9
%
ac
cu
r
ac
y
an
d
th
e
tr
ain
in
g
ti
m
e
w
a
s
r
ep
o
r
ted
as
5
.
6
9
s
ec
o
n
d
s
.
An
o
th
er
in
s
i
g
h
t
f
u
l
s
t
u
d
y
w
as
p
r
ese
n
ted
in
[
2
6
]
th
a
t
f
o
cu
s
s
ed
o
n
t
h
e
i
m
p
le
m
e
n
tat
io
n
o
f
s
u
p
er
v
is
ed
m
ac
h
in
e
lear
n
i
n
g
tec
h
n
iq
u
e
s
o
n
UN
SW
NB
-
1
5
d
ataset
to
test
th
eir
r
o
b
u
s
tn
e
s
s
.
E
m
p
ir
ical
r
es
u
lts
r
e
v
ea
led
th
at
lo
g
is
tic
r
eg
r
e
s
s
io
n
p
er
f
o
r
m
ed
b
etter
th
an
o
th
e
r
alg
o
r
ith
m
s
l
ik
e
T
r
ee
-
J
4
8
,
SV
M
an
d
Naiv
e
B
a
y
e
s
.
An
o
v
er
all
ac
cu
r
ac
y
o
f
8
9
.
2
6
%
w
a
s
r
ep
o
r
ted
b
y
lo
g
is
ti
c
r
eg
r
ess
io
n
m
o
d
el.
2.
M
E
T
H
O
DO
L
O
G
Y
T
h
is
s
ec
tio
n
d
escr
ib
es
in
d
etai
l
th
e
v
ar
io
u
s
asp
ec
ts
o
f
e
x
p
er
im
en
tatio
n
.
T
h
is
ar
ticle
f
o
c
u
s
e
s
o
n
eig
h
t
t
w
o
-
clas
s
a
n
d
t
h
r
ee
m
u
lticla
s
s
class
if
icatio
n
alg
o
r
it
h
m
s
.
C
la
s
s
i
f
icatio
n
m
o
d
els
w
er
e
d
esi
g
n
ed
i
n
f
o
u
r
d
i
f
f
er
en
t
s
tag
e
s
:
p
r
ep
r
o
ce
s
s
in
g
,
f
ea
t
u
r
e
s
elec
tio
n
u
s
i
n
g
m
u
t
u
al
i
n
f
o
r
m
atio
n
,
tu
n
i
n
g
o
f
h
y
p
er
-
p
ar
a
m
eter
s
an
d
d
esi
g
n
i
n
g
p
r
ed
ictiv
e
w
o
r
k
f
lo
w
s
.
B
asica
ll
y
,
UNSW
NB
-
1
5
h
as
4
7
f
ea
tu
r
es
an
d
t
w
o
clas
s
lab
els
.
T
h
e
d
ataset
h
as
co
n
tin
u
o
u
s
,
d
is
cr
ete
an
d
s
y
m
b
o
lic
f
ea
t
u
r
es
in
v
ar
ied
r
an
g
e
s
th
u
s
s
u
b
j
ec
ted
to
p
r
e
-
p
r
o
ce
s
s
in
g
.
D
u
r
in
g
th
e
e
x
p
er
i
m
e
n
tatio
n
,
all
n
o
m
i
n
al
f
ea
t
u
r
es
w
er
e
co
n
v
er
ted
i
n
to
in
te
g
er
s
.
N
u
m
er
ical
f
ea
tu
r
e
s
w
it
h
a
w
id
e
r
a
n
g
e
ar
e
d
if
f
ic
u
lt
to
h
an
d
le.
He
n
c
e
lo
g
ar
ith
m
ic
s
ca
li
n
g
w
as
ap
p
lied
to
d
ec
r
ea
s
e
th
eir
r
an
g
e
o
f
v
alu
e
s
.
B
o
o
lean
f
ea
t
u
r
es
d
id
n
o
t
n
ee
d
an
y
s
ca
l
in
g
.
Mi
n
-
m
ax
n
o
r
m
aliza
t
io
n
w
a
s
ap
p
lied
to
d
eter
m
in
e
t
h
e
s
m
al
lest
a
n
d
lar
g
est
v
alu
e
o
f
ea
ch
f
ea
t
u
r
e
in
t
h
e
r
an
g
e
[
0
,
1
]
.
(
1
)
I
n
(
1
)
,
m
i
n
an
d
m
a
x
r
e
f
er
to
th
e
m
i
n
i
m
u
m
a
n
d
m
ax
i
m
u
m
v
alu
e
s
o
f
ea
ch
f
ea
tu
r
e
“
i”.
E
a
ch
f
ea
tu
r
e
v
alu
e
V
is
s
ca
led
to
V
‟
.
Featu
r
e
s
co
r
in
g
w
a
s
u
s
ed
to
p
r
io
r
itize
th
e
f
ea
tu
r
es
f
o
llo
w
e
d
b
y
th
e
d
esig
n
o
f
w
o
r
k
f
lo
w
s
to
p
er
f
o
r
m
clas
s
i
f
i
ca
tio
n
tas
k
s
.
Up
o
n
ex
p
er
i
m
en
tatio
n
,
m
u
t
u
al
i
n
f
o
r
m
atio
n
y
ie
ld
ed
co
m
p
ar
ativ
e
l
y
b
etter
r
esu
lts
th
a
n
o
th
er
f
ilter
b
ased
f
ea
tu
r
e
s
elec
tio
n
m
eth
o
d
s
.
Mu
tu
al
i
n
f
o
r
m
atio
n
,
as
th
e
n
a
m
e
s
u
g
g
es
ts
is
a
m
ea
s
u
r
e
o
f
i
n
f
o
r
m
atio
n
b
et
w
ee
n
a
r
an
d
o
m
f
ea
t
u
r
e
„
x
‟
a
n
d
tar
g
et
v
ar
iab
le
„
y
‟
o
r
th
e
lab
el
[
2
7
]
.
T
h
e
m
u
t
u
al
in
f
o
r
m
atio
n
b
et
w
ee
n
t
w
o
v
ar
i
ab
les is
g
i
v
e
n
b
y
(
2
)
as e
x
p
lai
n
ed
in
(
2
)
an
d
(
3
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
978
-
986
980
∬
(
2
)
|
|
(
3)
In
(
2
)
in
clu
d
es
p
(
x
,
y
)
wh
ich
d
e
n
o
tes
t
h
e
j
o
in
t
p
r
o
b
ab
ilit
y
d
en
s
it
y
f
u
n
c
tio
n
,
p
(
x
)
,
p
(
y
)
ar
e
th
e
m
ar
g
i
n
al
d
en
s
i
t
y
f
u
n
ctio
n
s
.
I
n
t
h
e
co
n
te
x
t
o
f
f
ea
t
u
r
e
s
elec
tio
n
,
„
n
‟
r
ef
er
s
to
t
h
e
n
u
m
b
er
o
f
s
elec
ted
f
ea
t
u
r
es
an
d
i
s
k
n
o
w
n
a
s
j
o
in
t
m
u
t
u
al
i
n
f
o
r
m
atio
n
.
T
h
e
s
u
b
s
et
o
f
s
elec
ted
f
ea
t
u
r
es
i
s
r
ef
er
r
ed
to
as
X
S
a
s
g
iv
e
n
in
(
3
)
.
T
h
e
d
is
tr
ib
u
tio
n
o
f
tr
ain
in
g
a
n
d
tes
t
d
ataset
s
is
s
h
o
w
n
i
n
T
ab
le
1
.
I
t
ca
n
b
e
n
o
ted
th
at
t
h
er
e
w
a
s
n
o
r
ed
u
n
d
a
n
c
y
f
o
u
n
d
i
n
t
r
ain
in
g
a
n
d
test
in
g
d
is
tr
ib
u
tio
n
s
u
n
l
ik
e
b
e
n
ch
m
ar
k
d
atasets
[
1
7
,
1
8
]
.
As
m
e
n
tio
n
ed
ab
o
v
e,
m
u
tu
al
i
n
f
o
r
m
atio
n
w
a
s
u
s
ed
as
f
ea
tu
r
e
s
co
r
in
g
m
et
h
o
d
av
ailab
le
as
a
m
o
d
u
le
o
n
A
z
u
r
e
Ma
ch
i
n
e
lear
n
i
n
g
s
tu
d
io
.
T
h
e
s
ali
en
t
f
ea
tu
r
e
s
,
as
lis
ted
i
n
T
ab
le
2
,
w
er
e
g
i
v
e
n
as
in
p
u
t
to
th
e
v
ar
io
u
s
class
i
f
ier
s
to
o
b
tain
th
e
b
est
p
o
s
s
ib
le
p
r
ed
ictio
n
s
f
r
o
m
t
h
e
m
.
I
n
o
r
d
e
r
to
a
p
tly
as
s
es
s
th
e
p
er
f
o
r
m
a
n
ce
o
f
all
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
co
n
s
id
er
ed
in
th
e
s
t
u
d
y
,
1
0
-
f
o
ld
cr
o
s
s
v
alid
at
io
n
w
a
s
ap
p
lied
an
d
a
s
ep
ar
ate
test
i
n
g
s
et
w
as
co
n
s
id
er
ed
f
o
r
ev
al
u
atio
n
.
C
r
o
s
s
v
a
lid
atio
n
b
ec
o
m
e
s
i
m
p
o
r
tan
t
i
n
m
ac
h
i
n
e
lear
n
i
n
g
r
esear
ch
to
co
n
tr
o
l
o
v
er
f
itti
n
g
a
n
d
co
r
r
o
b
o
r
ate
th
e
ca
p
ab
ilit
y
o
f
alg
o
r
ith
m
s
to
g
e
n
er
alize
o
n
i
n
d
ep
en
d
en
t
d
ata
(
test
in
g
s
et)
[
2
8
]
.
T
ab
le
1
.
Data
s
et
d
is
tr
ib
u
tio
n
C
l
a
ss
T
r
a
i
n
i
n
g
sam
p
l
e
s
T
e
st
i
n
g
sam
p
l
e
s
N
o
r
mal
5
6
0
0
0
3
7
0
0
0
A
n
a
l
y
si
s
2
0
0
0
6
7
7
B
a
c
k
d
o
o
r
1
7
4
6
5
8
3
R
e
c
o
n
n
a
i
ss
a
n
c
e
1
0
4
9
1
3
4
9
6
S
h
e
l
l
c
o
d
e
1
1
3
3
3
7
8
W
o
r
ms
1
3
0
44
DOS
1
2
2
6
4
4
0
8
9
F
u
z
z
e
r
s
1
8
1
8
4
6
0
6
2
G
e
n
e
r
i
c
4
0
0
0
0
1
8
8
7
1
Ex
p
l
o
i
t
s
3
3
3
9
3
1
1
1
3
2
T
o
t
a
l
1
,
7
5
,
3
4
1
8
2
,
3
3
2
T
ab
le
2
.
L
is
t o
f
s
alie
n
t
f
ea
t
u
r
es
S
l
.
N
o
N
a
me
o
f
t
h
e
f
e
a
t
u
r
e
F
e
a
t
u
r
e
sco
r
e
1
c
t
_
st
a
t
e
_
t
t
l
0
.
6
8
6
2
d
t
t
l
0
.
5
6
3
S
t
t
l
0
.
2
7
4
d
i
n
p
k
t
0
.
2
3
5
sme
a
n
0
.
2
0
6
r
a
t
e
0
.
1
9
9
7
c
t
_
d
s
t
_
s
p
o
r
t
_
l
t
m
0
.
1
9
6
8
sl
o
a
d
0
.
1
9
0
9
st
a
te
0
.
1
8
7
5
10
d
l
o
a
d
0
.
1
8
7
2
11
sb
y
t
e
s
0
.
1
8
5
12
d
p
k
t
s
0
.
1
7
5
13
d
b
y
t
e
s
0
.
1
7
1
14
dur
0
.
1
5
8
15
a
c
k
d
a
t
0
.
1
5
6
16
d
me
a
n
0
.
1
4
7
17
sy
n
a
c
k
0
.
1
3
8
18
t
c
p
r
t
t
0
.
1
3
1
2
.
1
.
Av
er
a
g
ed
perc
ept
ro
n
Av
er
ag
ed
P
er
ce
p
tr
o
n
is
a
s
i
m
p
lif
ied
f
o
r
m
o
f
n
eu
r
al
n
et
w
o
r
k
th
at
u
s
e
s
a
lin
ea
r
f
u
n
ctio
n
t
o
class
if
y
th
e
s
a
m
p
le
s
.
M
A
M
L
S
o
f
f
er
s
a
n
o
p
tio
n
o
f
s
etti
n
g
a
s
in
g
le
v
al
u
e
o
r
m
u
l
tip
le
v
alu
e
s
a
s
lear
n
i
n
g
r
ates
in
o
r
d
er
to
test
th
e
p
r
o
f
icie
n
c
y
o
f
t
w
o
class
A
v
er
a
g
ed
P
er
ce
p
tr
o
n
m
o
d
el.
Dif
f
er
en
t
p
ar
a
m
eter
v
a
lu
e
s
l
ik
e
0
.
1
,
0
.
5
an
d
1
.
0
w
er
e
s
et
a
s
lear
n
i
n
g
r
ate
to
d
eter
m
i
n
e
th
e
o
p
ti
m
al
co
n
f
i
g
u
r
a
tio
n
o
f
t
h
e
s
to
c
h
asti
c
g
r
ad
ien
t
d
escen
t
o
p
ti
m
izer
.
T
h
e
ad
v
an
tag
e
o
f
u
s
i
n
g
a
p
ar
a
m
eter
r
an
g
e
i
s
th
at
t
h
e
m
o
d
el
r
ep
r
is
es
o
v
er
s
ev
er
al
co
m
b
in
atio
n
s
ev
e
n
t
u
all
y
p
r
o
d
u
cin
g
t
h
e
o
p
ti
m
al
m
o
d
el.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
erfo
r
ma
n
ce
a
n
a
lysi
s
o
f b
in
a
r
y
a
n
d
mu
lticla
s
s
mo
d
els u
s
in
g
a
z
u
r
e
ma
ch
in
e
lea
r
n
in
g
(
S
mith
a
R
a
ja
g
o
p
a
l
)
981
2
.
2
.
B
a
y
es
po
int
m
a
chine
B
ay
e
s
p
o
in
t
m
ac
h
in
e
i
s
b
ase
d
o
n
B
ay
e
s
ia
n
p
r
in
cip
le
to
e
f
f
icien
tl
y
c
lass
if
y
n
et
w
o
r
k
i
n
s
tan
ce
s
b
y
ch
o
o
s
in
g
a
B
a
y
e
s
p
o
in
t
(
av
er
ag
e)
.
T
y
p
icall
y
,
iter
atio
n
s
ar
e
s
et
i
n
t
h
e
r
an
g
e
o
f
5
to
1
0
0
.
T
h
is
v
al
u
e
i
n
d
icate
s
th
e
n
u
m
b
er
o
f
ti
m
e
s
th
e
al
g
o
r
ith
m
iter
at
es
o
v
er
th
e
tr
ai
n
i
n
g
d
ata.
Nu
m
er
o
u
s
tr
ials
w
er
e
co
n
d
u
cted
b
y
v
ar
y
i
n
g
th
e
n
u
m
b
er
o
f
iter
atio
n
s
w
it
h
in
t
h
e
g
i
v
en
r
an
g
e
b
u
t
t
h
e
r
esu
lt
s
w
er
e
n
o
t
co
n
v
i
n
ci
n
g
en
o
u
g
h
an
d
lo
n
g
er
tr
ain
i
n
g
ti
m
e
w
as
o
b
s
er
v
ed
d
u
r
in
g
t
h
o
s
e
tr
ials
.
Ho
w
e
v
er
,
o
n
s
etti
n
g
t
h
e
n
u
m
b
er
o
f
tr
a
in
i
n
g
iter
atio
n
s
as
3
0
f
o
r
t
w
o
-
clas
s
B
a
y
e
s
P
o
in
t
Ma
c
h
i
n
e,
th
e
r
es
u
lt
s
o
b
tain
ed
w
er
e
s
a
tis
f
ac
to
r
y
a
n
d
th
i
s
w
a
s
t
h
e
b
as
is
f
o
r
r
etain
i
n
g
3
0
as th
e
n
u
m
b
er
o
f
tr
ain
i
n
g
iter
a
tio
n
s
f
o
r
th
e
ex
p
er
i
m
en
t.
2
.
3
.
B
o
o
s
t
ed
decisi
o
n t
re
e
B
o
o
s
ted
d
ec
is
io
n
tr
ee
is
an
en
s
e
m
b
le
m
o
d
el
p
r
i
m
ar
il
y
ai
m
ed
at
r
ec
tify
i
n
g
th
e
er
r
o
r
s
o
f
p
r
ev
io
u
s
l
y
b
u
ilt
tr
ee
s
.
T
h
e
f
o
u
r
cr
itic
al
h
y
p
er
-
p
ar
a
m
eter
v
al
u
es
as
s
h
o
w
n
i
n
T
ab
le
3
w
er
e
u
s
ed
to
ex
a
m
in
e
th
e
co
m
p
ete
n
ce
o
f
T
w
o
-
cl
ass
B
o
o
s
ted
Dec
is
io
n
T
r
ee
.
Her
e,
m
ax
i
m
u
m
n
u
m
b
er
o
f
leav
e
s
i
n
d
icate
th
e
m
ax
i
m
u
m
lea
v
es
t
h
at
ca
n
b
e
cr
ea
ted
in
an
y
tr
ee
.
T
h
e
s
ize
o
f
th
e
tr
ee
ca
n
b
e
i
n
cr
ea
s
ed
b
y
v
ar
y
i
n
g
t
h
is
v
alu
e
b
u
t
o
v
er
f
i
tti
n
g
an
d
p
r
o
lo
n
g
ed
tr
ai
n
i
n
g
ti
m
e
w
er
e
e
n
co
u
n
ter
ed
b
y
in
cr
ea
s
in
g
t
h
e
n
u
m
b
er
o
f
leav
e
s
.
Min
i
m
u
m
n
u
m
b
er
o
f
s
a
m
p
les
p
er
leaf
n
o
d
e
r
ef
er
s
to
th
e
n
u
m
b
er
o
f
ca
s
e
s
co
n
s
id
er
ed
to
cr
ea
te
a
leaf
n
o
d
e.
T
h
e
v
alu
e
1
0
s
i
g
n
if
ie
s
t
h
at
t
h
e
tr
ai
n
i
n
g
d
ata
co
n
tai
n
s
1
0
ca
s
es
m
ee
t
in
g
t
h
e
s
a
m
e
co
n
d
itio
n
a
s
t
h
e
r
u
les
f
o
r
m
u
lated
.
T
h
e
in
itial lea
r
n
i
n
g
r
ate
w
as a
s
s
i
g
n
ed
a
v
al
u
e
0
.
2
w
h
ich
b
asica
ll
y
h
i
n
ts
at
t
h
e
r
ate
o
f
co
n
v
er
g
en
ce
.
Fu
r
t
h
er
,
1
0
0
d
ec
is
io
n
tr
ee
s
wer
e
cr
ea
ted
in
t
h
e
e
n
s
e
m
b
le.
T
h
er
e
is
also
a
p
r
o
v
i
s
io
n
to
c
r
ea
te
m
o
r
e
t
h
an
1
0
0
tr
ee
s
b
u
t a
g
ai
n
,
th
e
tr
ai
n
i
n
g
t
i
m
e
b
ec
o
m
es c
o
n
s
id
er
ab
l
y
lo
n
g
er
,
h
en
ce
co
n
s
id
er
ed
in
ad
v
i
s
a
b
le.
T
ab
le
3
.
C
r
itical
p
ar
am
eter
s
u
s
ed
f
o
r
co
n
f
i
g
u
r
in
g
b
o
o
s
ted
d
e
cisi
o
n
tr
ee
s
2
.
4
.
Dec
is
io
n f
o
re
s
t
T
w
o
-
clas
s
Dec
is
io
n
f
o
r
est,
a
s
r
ec
o
m
m
e
n
d
ed
b
y
T
ea
m
A
zu
r
e
[
1
5
]
is
o
n
e
o
f
th
e
m
o
s
t
p
r
ef
er
r
ed
m
o
d
el
s
to
p
er
f
o
r
m
b
in
ar
y
cla
s
s
i
f
icatio
n
.
T
h
er
e
ar
e
t
w
o
r
es
a
m
p
lin
g
m
et
h
o
d
s
n
a
m
el
y
r
ep
licate
an
d
b
a
g
g
in
g
av
ailab
le
to
d
e
s
ig
n
a
t
w
o
cla
s
s
d
ec
is
io
n
f
o
r
est
m
o
d
el.
R
ep
li
ca
te
m
e
th
o
d
tr
ai
n
s
ea
ch
tr
ee
o
n
t
h
e
s
a
m
e
tr
ai
n
i
n
g
d
ata
w
h
er
ea
s
b
o
o
ts
tr
ap
ag
g
r
e
g
a
tin
g
o
r
b
ag
g
in
g
allo
w
s
ea
ch
t
r
ee
to
b
e
g
r
o
w
n
o
n
a
n
e
w
s
a
m
p
le.
I
t
ca
n
b
e
n
o
ted
th
at
t
h
e
v
al
u
es a
s
s
h
o
w
n
i
n
T
a
b
le
4
,
w
h
e
n
ass
i
g
n
ed
b
esto
w
e
d
th
e
o
p
ti
m
al
r
esu
lts
.
T
ab
le
4
.
C
r
itical
p
ar
am
eter
s
u
s
ed
f
o
r
co
n
f
i
g
u
r
in
g
d
ec
is
io
n
f
o
r
est
R
aisi
n
g
m
a
x
i
m
u
m
d
ep
th
led
to
a
m
a
x
i
m
u
m
p
r
ec
is
io
n
o
f
1
b
u
t
o
v
er
f
itti
n
g
w
as
n
o
ted
w
h
ic
h
also
r
esu
lted
i
n
a
lo
n
g
er
tr
ain
i
n
g
ti
m
e
(
n
o
t
d
esira
b
le)
.
T
h
e
n
u
m
b
er
o
f
r
an
d
o
m
s
p
lit
s
s
i
g
n
if
ies
t
h
e
n
u
m
b
er
o
f
s
p
lits
g
en
er
ated
p
er
n
o
d
e
f
r
o
m
w
h
ic
h
th
e
o
p
ti
m
al
s
p
lit
co
u
ld
b
e
ch
o
s
en
.
Min
i
m
u
m
n
u
m
b
er
o
f
s
a
m
p
les
p
er
leaf
n
o
d
e
r
ef
er
s
to
th
e
n
u
m
b
er
o
f
ca
s
e
s
n
ee
d
ed
to
cr
ea
te
th
e
leaf
.
A
tte
m
p
t
s
w
er
e
m
ad
e
to
ascer
tain
w
h
eth
er
b
etter
r
esu
lt
s
co
u
ld
b
e
o
b
tain
ed
b
y
v
ar
y
in
g
t
h
e
v
al
u
es o
f
cr
itical
p
ar
a
m
eter
s
b
u
t
w
er
e
n
o
t e
f
f
ec
tu
a
l.
2
.
5
.
Dec
is
io
n j
un
g
le
Un
li
k
e
d
ec
is
io
n
tr
ee
s
th
a
t
allo
w
o
n
l
y
o
n
e
p
at
h
to
e
v
er
y
n
o
d
e,
a
d
ec
is
io
n
j
u
n
g
le
allo
w
s
m
u
l
tip
le
p
ath
s
f
r
o
m
r
o
o
t
to
ea
ch
leaf
.
Un
l
ik
e
d
ec
is
io
n
f
o
r
est
w
h
ic
h
u
s
es
t
r
ee
as
th
e
b
ase
lear
n
er
,
d
ec
is
i
o
n
j
u
n
g
le
e
m
p
lo
y
s
Dir
ec
ted
A
c
y
clic
Gr
ap
h
(
D
AG)
as
th
e
b
ase
lear
n
er
.
S
h
o
tt
o
n
et
al.
[
2
9
]
in
tr
o
d
u
ce
d
th
e
co
n
ce
p
t
o
f
d
ec
is
io
n
j
u
n
g
le
to
co
n
s
er
v
e
m
e
m
o
r
y
a
n
d
i
m
p
r
o
v
e
g
en
er
aliza
tio
n
.
N
u
m
b
er
o
f
o
p
ti
m
izatio
n
s
tep
s
p
er
d
ec
is
io
n
DAG
la
y
er
in
d
icate
s
t
h
e
n
u
m
b
er
o
f
s
tep
s
to
b
e
u
s
ed
to
en
h
a
n
ce
ea
ch
lev
e
l o
f
t
h
e
D
A
G.
T
h
e
v
a
lu
es a
s
e
n
u
m
er
ated
in
T
ab
le
5
w
er
e
u
s
ed
to
b
u
ild
th
e
m
o
d
el
an
d
v
ar
iatio
n
s
in
tr
o
d
u
ce
d
r
esu
lted
in
u
n
s
atis
f
ac
to
r
y
p
r
ed
ictio
n
s
.
T
ab
l
e
5
.
C
r
itical
p
ar
am
eter
s
u
s
ed
f
o
r
co
n
f
i
g
u
r
in
g
d
ec
is
io
n
j
u
n
g
le
M
a
x
.
l
e
a
v
e
s p
e
r
t
r
e
e
M
i
n
i
m
u
m
n
u
m
b
e
r
o
f
samp
l
e
s p
e
r
l
e
a
f
n
o
d
e
L
e
a
r
n
i
n
g
r
a
t
e
N
u
mb
e
r
o
f
t
r
e
e
s c
o
n
st
r
u
c
t
e
d
20
10
0
.
2
1
0
0
N
u
mb
e
r
o
f
d
e
c
i
si
o
n
t
r
e
e
s
M
a
x
i
m
u
m
d
e
p
t
h
o
f
d
e
c
i
si
o
n
t
r
e
e
N
u
mb
e
r
o
f
r
a
n
d
o
m s
p
l
i
t
s
p
e
r
n
o
d
e
M
i
n
i
m
u
m
n
u
m
b
e
r
o
f
samp
l
e
s p
e
r
l
e
a
f
n
o
d
e
8
32
1
2
8
1
N
u
mb
e
r
o
f
D
A
G
M
a
x
i
m
u
m
d
e
p
t
h
o
f
D
A
G
M
a
x
i
m
u
m
w
i
d
t
h
o
f
D
A
G
N
u
mb
e
r
o
f
o
p
t
i
m
i
z
a
t
i
o
n
s
t
e
p
s
p
e
r
d
e
c
i
si
o
n
D
A
G
l
a
y
e
r
8
32
1
2
8
2
0
4
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
978
-
986
982
2
.
6
.
L
o
ca
lly
deep
SVM
(
s
up
po
rt
v
ec
t
o
r
m
a
chi
ne
)
L
o
ca
ll
y
d
ee
p
k
er
n
e
l
ca
n
b
e
b
en
ef
icial
i
n
p
r
o
d
u
cin
g
b
etter
class
i
f
icatio
n
ac
c
u
r
ac
y
t
h
a
n
R
ad
ial
B
asi
s
Fu
n
ctio
n
(
R
B
F)
k
er
n
els
d
u
e
to
th
eir
ca
p
ab
ilit
y
to
i
n
cr
ea
s
e
f
e
atu
r
e
e
m
b
ed
d
in
g
a
n
d
attai
n
co
n
s
i
s
te
n
t sp
ee
d
[
3
0
]
.
Dep
th
o
f
t
h
e
tr
ee
is
a
h
y
p
er
-
p
ar
a
m
eter
u
s
ed
to
co
n
f
i
g
u
r
e
t
w
o
clas
s
lo
ca
ll
y
d
ee
p
S
V
M
w
h
ich
in
d
icate
s
th
e
m
ax
i
m
u
m
tr
ee
d
ep
th
.
C
h
o
o
s
in
g
an
ap
p
r
o
p
r
iate
v
alu
e
o
f
tr
ee
d
ep
th
b
ec
o
m
e
s
i
m
p
o
r
ta
n
t
s
in
ce
th
e
tr
ain
i
n
g
co
s
t
in
cr
ea
s
es
s
eq
u
e
n
tiall
y
w
it
h
tr
ee
d
ep
th
.
T
h
u
s
,
th
r
ee
r
eg
u
lar
izatio
n
p
ar
am
e
ter
s
wer
e
u
s
ed
to
co
n
tr
o
l
o
v
er
f
itti
n
g
n
a
m
el
y
la
m
b
d
a
(
W
)
,
la
m
b
d
a
th
eta
an
d
la
m
b
d
a
th
eta
p
r
i
m
e
s
et
a
t
0
.
1
,
0
.
0
1
an
d
0
.
0
1
r
esp
ec
tiv
el
y
.
L
a
m
b
d
a
i
n
d
icate
s
t
h
e
w
ei
g
h
t
to
b
e
ass
ig
n
ed
to
th
e
r
e
g
u
l
ar
izatio
n
ter
m
.
L
a
m
b
d
a
t
h
eta
d
ef
in
e
s
th
e
s
p
ac
e
b
et
w
ee
n
a
r
eg
io
n
b
o
u
n
d
ar
y
a
n
d
th
e
n
ea
r
est
d
ata
p
o
in
t.
L
a
m
b
d
a
t
h
eta
p
r
i
m
e,
a
p
ar
am
ete
r
n
ee
d
ed
to
co
n
tr
o
l
th
e
cu
r
v
at
u
r
e
in
d
ec
i
s
io
n
b
o
u
n
d
ar
ies is
also
a
n
in
teg
r
al
co
m
p
o
n
en
t r
eq
u
ir
ed
to
b
u
ild
t
h
e
t
wo
class
lo
ca
ll
y
d
ee
p
SVM.
Us
u
all
y
,
la
m
b
d
a
t
h
eta
a
n
d
la
m
b
d
a
t
h
eta
p
r
i
m
e
w
i
ll
b
e
o
n
e
te
n
t
h
o
f
la
m
b
d
a,
i
f
c
h
o
s
e
n
o
th
er
w
i
s
e
ca
u
s
e
s
o
v
er
f
itti
n
g
.
Sig
m
o
id
s
h
ar
p
n
es
s
r
ef
er
s
to
th
e
s
ca
li
n
g
p
ar
a
m
e
ter
.
Sig
m
o
id
k
er
n
el
is
q
u
ite
f
av
o
r
ab
le
d
u
e
to
its
g
en
e
s
is
f
r
o
m
n
eu
r
al
n
et
w
o
r
k
s
.
Ho
w
ev
er
,
it
s
u
s
a
g
e
i
s
n
o
t
en
co
u
r
ag
ed
w
id
el
y
d
u
e
to
i
ts
n
o
n
-
p
o
s
iti
v
e
s
e
m
i
d
ef
in
i
te
p
r
o
p
er
ties
[
3
1
]
.
Sig
m
o
id
k
er
n
el
d
o
es
n
o
t
s
atis
f
y
Me
r
ce
r
‟
s
t
h
eo
r
e
m
.
T
h
er
ef
o
r
e,
lar
g
e
v
a
lu
e
s
ca
n
n
o
t
b
e
ass
i
g
n
ed
to
s
ig
m
o
id
s
h
ar
p
n
es
s
.
S
m
aller
v
al
u
es
li
k
e
1
w
h
en
u
s
ed
ca
n
co
n
tr
o
l
th
e
th
r
es
h
o
ld
.
T
a
b
le
6
illu
s
tr
ates
th
e
cr
itical
d
ef
a
u
lt p
ar
a
m
eter
s
tu
n
ed
to
m
o
d
el
t
h
e
t
w
o
cla
s
s
l
o
ca
ll
y
d
e
ep
SVM.
T
ab
le
6
.
C
r
itical
p
ar
am
eter
s
u
s
ed
f
o
r
co
n
f
i
g
u
r
in
g
lo
ca
ll
y
d
ee
p
SVM
D
e
p
t
h
o
f
t
h
e
t
r
e
e
L
a
mb
d
a
L
a
mb
d
a
t
h
e
t
a
L
a
mb
d
a
t
h
e
t
a
p
r
i
me
S
i
g
mo
i
d
s
h
a
r
p
n
e
ss
3
0
.
1
0
.
0
1
0
.
0
1
1
2
.
7
.
Su
pp
o
rt
v
ec
t
o
r
m
a
chine
T
w
o
clas
s
SV
M
u
s
es
L
1
(
L
a
s
s
o
)
r
eg
u
lar
izatio
n
to
co
n
tr
o
l
o
v
er
f
it
tin
g
.
T
h
e
d
ef
a
u
lt
v
al
u
e
o
f
L
a
m
b
d
a
W
=0
.
0
0
1
w
a
s
s
et
as
w
ei
g
h
t si
n
ce
it is
p
r
ef
er
ab
le
to
u
s
e
a
n
o
n
-
ze
r
o
v
al
u
e
to
co
n
tr
o
l th
e
d
e
g
r
ee
o
f
o
v
er
f
itti
n
g
.
2
.
8
.
L
o
g
is
t
ic
re
g
re
s
s
io
n
Op
ti
m
izatio
n
to
ler
a
n
ce
is
a
t
h
r
es
h
o
ld
th
at
is
n
o
r
m
all
y
s
p
e
cif
ied
w
h
ile
d
esi
g
n
i
n
g
t
w
o
c
l
ass
lo
g
is
t
ic
r
eg
r
ess
io
n
m
o
d
el
u
s
in
g
L
-
B
FGS
(
li
m
ited
m
e
m
o
r
y
B
r
o
y
d
en
-
Fletc
h
er
-
Go
ld
f
ar
b
-
S
h
an
n
o
)
o
p
ti
m
izatio
n
[
1
5
]
.
T
h
is
m
o
d
el
n
ec
es
s
itate
s
p
r
o
p
er
tu
n
i
n
g
o
f
L
1
an
d
L
2
v
al
u
es
s
et
as
1
a
n
d
1
r
esp
ec
ti
v
el
y
.
T
h
e
m
e
m
o
r
y
s
ize
in
m
eg
ab
y
tes
u
s
ed
b
y
L
-
B
FG
S
o
p
ti
m
izer
w
as
s
et
as
2
0
w
h
ic
h
i
n
d
icate
s
t
h
e
p
ast
g
r
ad
ien
t
s
s
to
r
ed
in
m
e
m
o
r
y
f
o
r
th
e
ex
ec
u
tio
n
o
f
s
u
cc
ess
i
v
e
s
tep
s
.
I
f
th
e
m
e
m
o
r
y
s
ize
is
h
ig
h
er
,
th
e
n
in
all
p
o
s
s
ib
il
it
ies,
it
s
lo
w
s
d
o
w
n
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
an
d
th
e
m
o
d
el
en
d
s
u
p
b
ein
g
f
la
w
ed
.
T
h
r
ee
s
ig
n
if
ic
a
n
t
p
ar
a
m
eter
s
w
er
e
u
s
ed
to
b
u
ild
an
d
test
t
h
e
ef
f
ec
ti
v
e
n
es
s
o
f
t
w
o
c
lass
lo
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el
as
m
e
n
tio
n
ed
in
T
ab
le
7
.
R
eg
u
lar
izatio
n
is
o
f
te
n
ap
p
lied
to
class
if
icatio
n
p
r
o
b
le
m
s
i
n
o
r
d
er
to
m
i
n
i
m
ize
o
v
er
f
itti
n
g
.
T
ab
le
7
.
C
r
itical
p
ar
am
eter
s
u
s
ed
f
o
r
co
n
f
i
g
u
r
in
g
lo
g
i
s
tic
r
e
g
r
ess
io
n
L
1
R
e
g
u
l
a
r
i
z
a
t
i
o
n
L
2
R
e
g
u
l
a
r
i
z
a
t
i
o
n
M
e
mo
r
y
si
z
e
u
se
d
b
y
L
-
B
F
G
S
1
1
20
3.
E
XP
E
R
I
M
E
NT
A
L
RE
SUL
T
S AN
D
D
I
SC
USS
I
O
N
T
h
is
s
ec
tio
n
p
r
esen
ts
t
h
e
r
esu
lt
s
o
b
tain
ed
th
r
o
u
g
h
ex
p
e
r
i
m
en
tatio
n
u
s
i
n
g
M
A
M
L
S.
T
h
e
f
o
u
r
class
i
f
icatio
n
p
o
s
s
ib
il
itie
s
o
f
an
y
i
n
tr
u
s
io
n
d
etec
tio
n
s
tu
d
y
ar
e:
T
r
u
e
p
o
s
itiv
es(T
P),
T
r
u
e
n
eg
ati
v
es
(
T
N)
,
Fals
e
Ne
g
ati
v
es
(
FN)
an
d
Fals
e
P
o
s
itiv
e
s
(
FP
)
th
at
d
eter
m
in
e
t
h
e
s
ig
n
i
f
ica
n
t
p
er
f
o
r
m
a
n
ce
m
etr
ics
n
a
m
el
y
A
cc
u
r
ac
y
(
A
)
,
P
r
ec
is
io
n
(
P
)
,
R
ec
all(
R
)
,
F1
-
s
co
r
e(
F1
)
,
A
r
e
a
u
n
d
er
t
h
e
c
u
r
v
e(
A
U
C
)
a
n
d
f
alse
alar
m
r
ate.
A
d
d
itio
n
al
l
y
,
t
h
e
tr
ai
n
i
n
g
a
n
d
ex
ec
u
t
io
n
ti
m
e
o
f
ea
ch
m
o
d
el
is
also
r
ep
o
r
ted
.
E
x
ec
u
tio
n
t
i
m
e
r
ef
er
s
to
th
e
ti
m
e
tak
en
b
y
t
h
e
m
o
d
el
to
o
u
tp
u
t
t
h
e
p
r
ed
ictio
n
s
(
C
las
s
lab
els
with
r
esp
ec
t
to
b
in
ar
y
an
d
attac
k
t
y
p
e
w
it
h
r
esp
ec
t
t
o
m
u
lt
iclas
s
m
o
d
els).
T
r
u
e
Po
s
itiv
e
(
Se
n
s
i
tiv
i
t
y
)
d
ef
i
n
e
s
th
e
n
u
m
b
er
o
f
p
o
s
itiv
e
s
a
m
p
le
s
co
r
r
ec
tly
clas
s
i
f
ied
as
p
o
s
itiv
e.
Fals
e
Ne
g
ati
v
e
(
FN)
is
th
e
n
u
m
b
er
o
f
p
o
s
i
tiv
e
e
x
a
m
p
le
s
w
r
o
n
g
l
y
c
lass
if
ied
as
n
eg
a
tiv
e.
Fals
e
P
o
s
iti
v
e
(
FP
)
i
s
t
h
e
n
u
m
b
er
o
f
n
e
g
ati
v
e
e
x
a
m
p
les
w
r
o
n
g
l
y
cla
s
s
i
f
ied
a
s
p
o
s
iti
v
e
an
d
T
r
u
e
Ne
g
ati
v
e
(
Sp
ec
if
icit
y
)
(
T
N)
is
th
e
n
u
m
b
er
o
f
n
eg
ativ
e
e
x
a
m
p
le
s
co
r
r
ec
tl
y
clas
s
i
f
ied
as
n
eg
a
tiv
e.
T
h
e
(
4
)
to
(
1
0
)
d
ef
in
e
th
ese
v
ar
io
u
s
p
er
f
o
r
m
an
ce
m
e
tr
ics:
(
4)
(
5
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
erfo
r
ma
n
ce
a
n
a
lysi
s
o
f b
in
a
r
y
a
n
d
mu
lticla
s
s
mo
d
els u
s
in
g
a
z
u
r
e
ma
ch
in
e
lea
r
n
in
g
(
S
mith
a
R
a
ja
g
o
p
a
l
)
983
(
6
)
(
7
)
Fals
e
P
o
s
itiv
e
R
ate
(
F
P
R
)
=
(
8
)
Fals
e
Ne
g
ati
v
e
R
ate
(
F
N
R
)
=
(
9
)
Fals
e
A
lar
m
R
ate
(
F
A
R
)
=
(
1
0
)
I
n
th
e
p
r
o
p
o
s
ed
w
o
r
k
,
ei
g
h
t
t
w
o
-
cla
s
s
cla
s
s
i
f
icatio
n
alg
o
r
ith
m
s
w
er
e
co
n
s
id
er
ed
an
d
th
eir
p
er
f
o
r
m
a
n
ce
w
a
s
an
al
y
s
ed
.
R
e
s
u
lt
s
ar
e
en
u
m
er
ated
in
T
ab
le
8
.
T
h
e
C
o
n
f
u
s
io
n
m
atr
i
x
s
h
o
w
n
b
elo
w
r
ep
r
esen
t
s
th
e
r
esu
l
ts
o
f
clas
s
i
f
icatio
n
o
b
tain
ed
f
r
o
m
th
r
ee
class
i
f
ie
r
s
n
a
m
el
y
m
u
lt
iclas
s
d
ec
is
io
n
f
o
r
est,
m
u
ltic
las
s
d
ec
is
io
n
j
u
n
g
le
an
d
m
u
lticla
s
s
lo
g
is
t
ic
r
eg
r
es
s
io
n
.
T
h
e
ac
tu
al
(
A
)
v
er
s
u
s
p
r
ed
i
cted
(
P
)
class
if
icatio
n
s
p
r
esen
ted
in
th
e
co
n
f
u
s
io
n
m
atr
ix
p
er
tain
to
th
e
ten
class
e
s
w
h
er
ei
n
th
e
to
p
m
o
s
t
r
o
w
s
i
g
n
i
f
ie
s
th
e
n
a
m
e
o
f
th
e
class
:
A
(
An
al
y
s
is
)
,
B
(
B
ac
k
d
o
o
r
)
,
D
(
DOS)
,
E
(
E
x
p
lo
its
)
,
F
(
Fu
zz
er
s
)
,
G
(
Gen
er
ic)
,
N
(
N
o
r
m
al)
,
R
(
R
ec
o
n
n
ai
s
s
a
n
ce
)
,
S
(
S
h
ellco
d
e)
an
d
W
(
W
o
r
m
s
)
.
T
h
e
R
e
s
u
lt
s
o
b
tain
ed
u
s
i
n
g
M
u
lt
icla
s
s
Dec
i
s
io
n
Fo
r
est
s
h
o
w
n
in
T
ab
le
9
(
C
o
n
f
u
s
io
n
Ma
tr
ix
1
)
,
T
a
b
le
1
0
(
C
o
n
f
u
s
io
n
Ma
tr
ix
2
)
,
an
d
T
ab
le
1
1
(
C
o
n
f
u
s
io
n
Ma
tr
i
x
3
)
.
T
ab
le
8
.
C
lass
if
icatio
n
r
es
u
lt
s
o
b
tai
n
ed
b
y
ap
p
l
y
i
n
g
eig
h
t t
wo
-
class
al
g
o
r
ith
m
s
co
n
s
id
er
ed
in
th
e
w
o
r
k
A
l
g
o
r
i
t
h
m
T
e
st
i
n
g
A
c
c
u
r
a
c
y
(
%)
P
r
e
c
i
si
o
n
(
%)
R
e
c
a
l
l
(
%)
F1
-
sco
r
e
A
U
C
F
A
R
(
%)
T
r
a
i
n
i
n
g
T
i
me
(
se
c
o
n
d
s)
Ex
e
c
u
t
i
o
n
T
i
me
(
se
c
o
n
d
s)
T
r
a
i
n
i
n
g
a
c
c
u
r
a
c
y
(
%)
A
v
e
r
a
g
e
P
e
r
c
e
p
t
r
o
n
7
7
.
2
8
0
.
3
8
8
.
1
0
.
8
4
0.
8
8
5
2
8
.
9
9
2
.
5
7
3
.
5
B
a
y
e
s p
o
i
n
t
ma
c
h
i
n
e
91
90
97
0
.
9
3
6
0
.
9
4
8
1
2
.
5
8
2
.
3
8
9
.
4
B
o
o
st
e
d
D
e
c
i
si
o
n
T
r
e
e
9
5
.
9
9
6
.
6
9
7
.
4
0
.
9
7
0
.
9
9
4
4
.
9
7
2
.
7
9
2
.
2
D
e
c
i
si
o
n
F
o
r
e
st
(
U
si
n
g
B
a
g
g
i
n
g
a
s
R
e
samp
l
i
n
g
)
9
9
.
2
9
9
%
9
9
.
6
0
.
9
9
4
1
1%
6
2
97
D
e
c
i
si
o
n
F
o
r
e
st
(
U
si
n
g
R
e
p
l
i
c
a
t
e
a
s
R
e
samp
l
i
n
g
)
9
9
.
5
9
9
.
4
9
9
.
8
0
.
9
9
6
1
0
.
7
6
2
9
6
.
6
D
e
c
i
si
o
n
J
u
n
g
l
e
9
4
.
6
9
4
.
3
98
0
.
9
6
1
0
.
9
7
.
2
6
.
5
2
.
9
9
2
.
2
L
o
c
a
l
l
y
d
e
e
p
S
V
M
9
3
.
3
9
1
.
6
9
9
.
3
0
.
9
5
3
0
.
9
7
5
10
7
.
8
3
.
3
9
1
.
0
S
V
M
8
5
.
7
8
9
.
4
8
9
.
6
0
.
8
9
5
0
.
9
1
7
1
6
.
5
7
.
9
3
.
5
8
3
.
8
L
o
g
i
st
i
c
R
e
g
r
e
ssi
o
n
9
1
.
6
9
0
.
8
9
7
.
6
0
.
9
4
1
0
.
9
5
2
1
1
.
7
7
3
8
8
.
8
T
ab
le
9
.
R
esu
lts
o
b
tain
ed
u
s
i
n
g
m
u
ltic
lass
d
ec
i
s
io
n
f
o
r
est (
co
n
f
u
s
io
n
m
atr
i
x
1)
A
A
B
D
E
F
G
N
R
S
W
R
e
c
a
l
l
(
%)
A
n
a
l
y
si
s
1
9
7
13
74
3
8
5
0
3
3
2
0
0
29
B
a
c
k
d
o
o
r
9
1
2
7
76
3
6
5
3
0
0
2
1
0
2
1
.
7
8
DOS
0
0
1
3
0
0
2
6
7
0
3
0
0
1
1
3
3
0
3
1
.
7
9
Ex
p
l
o
i
t
s
0
0
5
5
8
1
0
5
1
9
27
0
3
22
0
3
9
4
.
4
9
F
u
z
z
e
r
s
0
0
79
4
6
7
5
3
8
3
3
1
1
5
6
6
3
8
8
.
7
9
Ge
n
e
r
i
c
0
0
57
1
1
3
0
1
8
7
0
1
0
0
0
0
99
N
o
r
mal
0
0
0
24
6
6
6
0
3
6
3
1
0
0
0
0
9
8
.
1
3
R
e
c
o
n
n
a
i
ss
a
n
c
e
0
0
1
1
2
5
2
1
7
0
0
2
8
5
6
0
0
8
1
.
6
9
S
h
e
l
l
c
o
d
e
0
0
0
13
7
0
2
9
3
4
7
0
9
1
.
7
9
W
o
r
ms
0
0
0
4
0
0
0
0
0
40
9
0
.
9
P
r
e
c
i
si
o
n
(
%)
9
5
.
6
3
9
0
.
7
5
7
.
6
2
6
9
.
7
5
8
8
.
3
9
9
.
9
6
9
9
.
6
6
94
.
8
8
9
7
.
1
9
8
6
.
9
5
T
h
e
ti
m
e
ta
k
en
b
y
m
u
lt
icla
s
s
m
o
d
el
s
to
lear
n
n
u
m
er
o
u
s
n
et
w
o
r
k
in
s
ta
n
ce
s
a
n
d
s
u
b
s
eq
u
en
tl
y
d
is
tin
g
u
is
h
b
et
w
ee
n
at
tack
c
ateg
o
r
ies
an
d
n
o
r
m
al
p
atter
n
s
r
an
g
ed
b
et
w
ee
n
1
6
to
2
0
s
ec
o
n
d
s
.
T
h
e
least
tr
ain
i
n
g
ti
m
e
o
f
1
6
s
ec
o
n
d
s
w
as
tak
e
n
b
y
m
u
lticla
s
s
d
ec
is
io
n
f
o
r
est,
f
o
llo
w
ed
b
y
m
u
ltic
la
s
s
d
ec
is
io
n
j
u
n
g
le
th
at
to
o
k
1
8
s
ec
o
n
d
s
to
r
ec
o
g
n
ize
t
h
e
p
atter
n
s
b
elo
n
g
i
n
g
to
d
if
f
er
en
t
clas
s
es.
T
h
e
m
a
x
i
m
u
m
tr
ai
n
in
g
ti
m
e
o
f
2
0
s
ec
o
n
d
s
w
a
s
ta
k
e
n
b
y
m
u
lt
iclass
lo
g
i
s
tic
r
e
g
r
es
s
io
n
m
o
d
el.
I
t
is
w
o
r
th
w
h
ile
to
m
e
n
tio
n
t
h
at
th
e
e
x
ec
u
tio
n
ti
m
e
o
f
M
u
lticla
s
s
d
ec
is
io
n
f
o
r
est
an
d
Mu
lt
iclas
s
d
ec
is
i
o
n
j
u
n
g
le
w
a
s
r
ep
o
r
ted
as
6
an
d
6
.
5
s
ec
o
n
d
s
r
esp
ec
tiv
el
y
w
h
er
ea
s
M
u
lt
icla
s
s
lo
g
i
s
tic
r
eg
r
e
s
s
io
n
to
o
k
7
s
e
co
n
d
s
to
o
u
tp
u
t t
h
e
class
-
w
i
s
e
p
r
ed
ictio
n
s
.
P
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
978
-
986
984
T
ab
le
1
0
.
R
esu
lt o
b
tain
ed
u
s
i
n
g
m
u
ltic
lass
d
ec
i
s
io
n
j
u
n
g
le
(
c
o
n
f
u
s
io
n
m
atr
i
x
2)
A
A
B
D
E
F
G
N
R
S
W
R
e
c
a
l
l
(
%)
A
n
a
l
y
si
s
15
0
2
4
9
8
7
2
1
4
1
11
0
1
2
.
2
B
a
c
k
d
o
o
r
0
37
0
5
0
4
16
0
4
19
3
0
6
.
3
4
DOS
0
0
1
0
7
3
6
6
7
1
3
7
0
22
1
4
2
10
4
2
.
6
Ex
p
l
o
i
t
s
0
1
22
1
0
3
5
2
3
6
6
2
1
1
1
2
6
7
11
0
9
2
.
9
F
u
z
z
e
r
s
10
1
1
7
2
3
4
4
5
5
0
7
5
7
1
0
3
12
0
7
3
.
4
9
G
e
n
e
r
i
c
0
0
28
3
3
9
47
1
8
4
5
5
0
0
1
1
9
7
.
7
9
N
o
r
mal
10
0
0
4
3
0
3
8
4
8
0
3
2
5
9
7
1
1
1
0
4
8
8
.
1
R
e
c
o
n
n
a
i
ss
a
n
c
e
0
4
10
1
0
1
0
36
0
49
2
3
8
7
0
0
6
8
.
2
7
S
h
e
l
l
c
o
d
e
0
0
0
1
2
1
89
0
9
1
0
6
53
0
14
W
o
r
ms
0
6
0
5
2
0
1
0
0
30
68
P
r
e
c
i
si
o
n
(
%)
4
2
.
8
5
7
5
.
5
6
2
.
9
4
5
8
.
6
5
4
9
.
4
8
9
9
.
9
7
9
6
.
7
5
7
5
.
8
7
5
8
.
8
8
75
T
ab
le
1
1
.
R
esu
lt o
b
tain
ed
u
s
i
n
g
m
u
ltic
lass
lo
g
is
tic
r
eg
r
es
s
io
n
(
co
n
f
u
s
io
n
m
atr
ix
3)
A
A
B
D
E
F
G
N
R
S
W
R
e
c
a
l
l
(
%)
A
n
a
l
y
si
s
30
0
11
3
3
8
52
0
1
3
4
23
39
50
4
.
4
3
B
a
c
k
d
o
o
r
0
90
8
3
0
0
86
0
38
61
0
0
1
5
.
4
DOS
0
0
1
0
1
2
9
0
0
4
6
6
29
2
5
8
3
3
5
0
0
2
.
4
7
Ex
p
l
o
i
t
s
0
0
67
8
5
3
9
9
5
7
22
8
1
2
7
3
5
0
0
7
6
.
7
F
u
z
z
e
r
s
0
0
36
1
6
6
0
3
4
6
9
97
6
2
4
1
7
6
0
0
5
7
.
2
2
G
e
n
e
r
i
c
5
0
8
3
0
2
38
1
8
4
5
5
20
38
0
5
9
7
.
7
9
N
o
r
mal
20
11
40
1
8
5
0
4
4
0
3
0
3
0
5
9
9
77
0
0
8
2
.
7
R
e
c
o
n
n
a
i
ss
a
n
c
e
0
0
14
1
3
0
4
6
9
6
10
77
1
3
9
5
0
0
3
9
.
9
S
h
e
l
l
c
o
d
e
0
0
0
30
98
0
10
2
1
0
30
0
7
.
9
3
W
o
r
ms
0
0
0
33
0
0
0
0
0
11
25
P
r
e
c
i
si
o
n
(
%)
5
4
.
5
4
89
3
5
.
4
3
4
9
.
4
8
3
3
.
7
9
9
9
.
1
5
9
3
.
9
4
5
.
7
3
4
3
.
4
7
1
6
.
6
6
T
y
p
icall
y
,
a
n
y
I
n
tr
u
s
io
n
Dete
ctio
n
S
y
s
te
m
(
I
DS)
ai
m
s
a
t
i
m
p
r
o
v
in
g
t
h
e
attac
k
d
etec
tio
n
r
ate
an
d
r
ed
u
cin
g
f
al
s
e
alar
m
s
.
T
ec
h
n
icall
y
,
it
is
v
er
y
c
h
alle
n
g
i
n
g
to
ac
h
iev
e
a
lo
w
er
f
alse
alar
m
r
ate
in
s
p
ite
o
f
a
s
atis
f
ac
to
r
y
r
ec
all
p
er
ce
n
tag
e.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
d
e
m
o
n
s
tr
ates
th
a
t
th
e
d
ec
is
io
n
f
o
r
es
t
m
o
d
els
ar
e
q
u
ite
r
o
b
u
s
t.
A
r
i
g
o
r
o
u
s
in
v
es
tig
a
ti
o
n
o
f
all
th
e
m
o
d
els
co
n
s
id
er
ed
in
th
e
s
t
u
d
y
m
ad
e
s
o
m
e
i
n
ter
esti
n
g
r
ev
elat
io
n
s
as
elab
o
r
ated
in
th
is
s
ec
tio
n
.
Fals
e
A
lar
m
R
ate
(
F
AR
)
h
a
s
b
ee
n
co
n
s
id
er
ab
l
y
lo
w
(
<=
1
0
%)
w
it
h
r
esp
ec
t
to
f
o
u
r
b
i
n
ar
y
c
lass
if
ier
s
n
a
m
e
l
y
b
o
o
s
ted
d
ec
is
i
o
n
tr
ee
,
d
ec
is
io
n
f
o
r
est,
d
ec
is
io
n
j
u
n
g
le
an
d
l
o
ca
ll
y
d
ee
p
SVM
as
m
en
tio
n
ed
i
n
T
ab
le
8
.
P
ar
tic
u
lar
l
y
,
t
w
o
cla
s
s
d
ec
is
io
n
f
o
r
est
s
u
r
p
ass
ed
o
th
er
cla
s
s
i
f
ier
s
w
it
h
h
i
g
h
e
s
t
r
ec
al
l
r
ate
o
f
9
9
.
8
%
an
d
lo
w
est
F
AR
o
f
1
%
w
ith
b
ag
g
i
n
g
an
d
0
.
7
%
w
it
h
r
ep
licate
as
r
e
-
sa
m
p
lin
g
tec
h
n
iq
u
es
r
esp
ec
tiv
el
y
as
m
e
n
tio
n
ed
in
T
ab
le
8
.
A
lth
o
u
g
h
o
t
h
er
p
er
f
o
r
m
an
ce
m
ea
s
u
r
es
h
av
e
b
ee
n
u
s
ed
to
v
alid
at
e
th
e
ef
f
ec
t
iv
e
n
es
s
o
f
t
h
e
p
r
o
p
o
s
ed
m
o
d
els,
R
ec
all
a
n
d
F
AR
ar
e
th
e
t
w
o
s
ta
n
d
ar
d
m
e
tr
ics
w
id
el
y
e
m
p
lo
y
ed
in
in
tr
u
s
io
n
d
etec
t
io
n
r
esear
c
h
a
n
d
th
e
r
e
m
ai
n
i
n
g
m
etr
ic
s
s
er
v
e
as
s
u
p
p
le
m
e
n
tar
y
.
T
w
o
-
clas
s
lo
ca
ll
y
d
ee
p
SVM
h
as
p
er
f
o
r
m
ed
w
e
ll
w
ith
s
ee
m
in
g
l
y
g
o
o
d
attac
k
d
etec
tio
n
r
a
te
o
f
9
9
.
3
%
an
d
f
alse
alar
m
r
ate
as
lo
w
as
1
0
%.
T
w
o
-
clas
s
B
a
y
e
s
p
o
in
t
m
ac
h
i
n
e
(
B
P
M)
an
d
B
o
o
s
ted
Dec
is
io
n
T
r
ee
(
B
D
T
)
m
o
d
els
h
av
e
b
ee
n
co
n
s
i
s
ten
t
i
n
th
eir
p
er
f
o
r
m
a
n
ce
w
ith
9
7
%
r
ec
all.
Ho
w
e
v
er
,
B
P
M
h
as
r
ec
o
r
d
ed
a
h
ig
h
er
(
1
2
.
5
%)
FAR
as
co
m
p
ar
ed
to
B
D
T
m
o
d
el
(
4
.
9
%
as
F
AR
)
.
T
h
e
F
AR
r
ep
o
r
ted
b
y
A
v
er
a
g
ed
p
er
ce
p
tr
o
n
is
s
ee
m
i
n
g
l
y
h
i
g
h
i.
e.
,
2
8
.
9
%.
T
h
er
e
is
a
s
u
b
s
tan
tial
d
i
f
f
er
en
ce
b
et
w
ee
n
t
h
e
r
ec
all
p
er
ce
n
ta
g
e
o
f
t
w
o
-
clas
s
S
VM
a
n
d
t
w
o
-
clas
s
lo
ca
ll
y
d
ee
p
SVM
(
8
9
.
6
%
an
d
9
9
.
3
%
r
esp
ec
tiv
ely
)
.
T
w
o
class
SVM
‟
s
ca
p
ab
ilit
y
to
d
etec
t
f
alse
alar
m
s
h
as
n
o
t
b
ee
n
im
p
r
es
s
iv
e
s
in
ce
its
F
AR
is
r
ep
o
r
ted
to
b
e
as
h
i
g
h
as
1
6
.
5
%.
O
n
th
e
o
t
h
er
h
a
n
d
,
lo
ca
ll
y
d
ee
p
SVM
h
a
s
b
ee
n
co
m
p
ar
ati
v
el
y
b
etter
in
r
ed
u
c
in
g
f
al
s
e
alar
m
s
d
u
e
to
th
e
a
p
p
licatio
n
o
f
s
i
g
m
o
id
k
er
n
el.
T
w
o
clas
s
L
o
g
is
t
ic
r
eg
r
ess
io
n
h
a
s
b
ee
n
m
ed
io
cr
e
in
it
s
p
er
f
o
r
m
a
n
ce
w
i
th
a
r
ea
s
o
n
ab
le
r
ec
all
p
er
ce
n
tag
e
o
f
9
7
.
6
%
an
d
ap
p
ar
en
tl
y
a
h
ig
h
er
F
AR
o
f
1
1
.
7
%.
Net
w
o
r
k
s
a
m
p
les
i
n
a
n
y
d
ataset
ar
e
n
o
t
u
n
if
o
r
m
l
y
d
i
s
tr
ib
u
ted
ac
r
o
s
s
v
ar
io
u
s
clas
s
es
a
n
d
m
ac
h
i
n
e
lear
n
in
g
p
r
ac
titi
o
n
er
s
o
f
te
n
en
co
u
n
ter
th
e
p
r
o
b
lem
o
f
i
m
b
a
lan
ce
d
d
ataset
s
i
n
r
ea
l
ti
m
e
[
3
2
]
.
B
in
ar
y
class
i
f
icatio
n
alo
n
e
m
a
y
n
o
t
b
e
in
s
i
g
h
t
f
u
l
b
ec
au
s
e
t
w
o
cl
as
s
al
g
o
r
ith
m
s
ca
n
n
o
t
clas
s
i
f
y
t
h
e
s
a
m
p
le
s
i
n
to
a
p
ar
ticu
lar
attac
k
t
y
p
e
o
r
ca
teg
o
r
y
.
I
n
v
ie
w
o
f
t
h
e
ab
o
v
e
m
en
tio
n
ed
li
m
itatio
n
o
f
b
i
n
ar
y
clas
s
i
f
icatio
n
,
t
h
r
ee
alg
o
r
ith
m
s
w
er
e
e
m
p
lo
y
ed
to
p
er
f
o
r
m
m
u
lticla
s
s
clas
s
i
f
ica
tio
n
tas
k
s
.
E
m
p
ir
ical
i
n
v
est
ig
atio
n
d
e
m
o
n
s
tr
ated
th
at
m
u
lticla
s
s
d
ec
i
s
io
n
f
o
r
est
o
u
tp
er
f
o
r
m
ed
o
t
h
er
s
i
n
id
e
n
ti
f
y
in
g
v
ar
io
u
s
attac
k
t
y
p
es.T
h
e
r
ec
all
p
er
ce
n
tag
e
o
f
s
ev
e
n
class
e
s
in
cl
u
d
i
n
g
n
o
r
m
al
ar
e
q
u
ite
ap
p
ea
lin
g
ex
ce
p
t
A
n
al
y
s
is
,
B
ac
k
d
o
o
r
an
d
Den
ial
o
f
Ser
v
ice(
DOS)
a
s
p
r
ed
icted
b
y
m
u
l
ticlas
s
d
ec
is
io
n
f
o
r
est
(
as
e
n
u
m
er
ated
i
n
co
n
f
u
s
io
n
m
atr
i
x
1
)
.
On
t
h
e
m
u
lticla
s
s
clas
s
i
f
icatio
n
f
r
o
n
t,
t
h
e
r
es
u
lt
s
o
b
tai
n
ed
f
r
o
m
b
o
th
d
ec
i
s
io
n
j
u
n
g
le
a
n
d
l
o
g
is
tic
r
eg
r
es
s
io
n
s
w
er
e
tr
iv
ial.
B
o
th
th
ese
cla
s
s
i
f
ier
s
r
ep
o
r
ted
a
g
o
o
d
r
ec
all
p
er
ce
n
tag
e,
i.e
.
,
ab
o
v
e
9
0
%
w
it
h
r
esp
ec
t
to
o
n
ly
t
w
o
attac
k
ca
teg
o
r
ies
li
k
e
g
e
n
er
i
c
an
d
ex
p
lo
its
.
T
h
is
ca
n
b
e
attr
ib
u
ted
to
th
e
p
r
esen
ce
o
f
lar
g
er
s
a
m
p
les
in
P
P
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
P
erfo
r
ma
n
ce
a
n
a
lysi
s
o
f b
in
a
r
y
a
n
d
mu
lticla
s
s
mo
d
els u
s
in
g
a
z
u
r
e
ma
ch
in
e
lea
r
n
in
g
(
S
mith
a
R
a
ja
g
o
p
a
l
)
985
th
e
tr
ai
n
i
n
g
s
e
t
w
i
th
r
esp
ec
t
to
g
e
n
er
ic
a
n
d
e
x
p
lo
its
as
o
b
s
er
v
ab
le
f
r
o
m
T
ab
le
1
.
I
t
is
d
is
ce
r
n
ib
le
th
a
t
th
e
s
eq
u
en
ce
o
f
e
x
p
er
i
m
e
n
tati
o
n
co
n
d
u
cted
o
n
A
z
u
r
e
Ma
ch
i
n
e
L
ea
r
n
i
n
g
St
u
d
io
s
u
p
p
o
r
ted
b
y
an
in
g
e
n
io
u
s
s
et
o
f
alg
o
r
it
h
m
s
s
tr
en
g
t
h
en
ed
t
h
e
i
m
p
le
m
en
tatio
n
a
s
p
ec
t
s
i
n
c
e
o
v
er
all
attac
k
d
etec
tio
n
r
ate
is
v
is
ib
l
y
h
ig
h
a
n
d
f
alse
alar
m
r
ate
i
s
ap
p
ar
en
tl
y
lo
w
.
T
h
e
c
u
r
r
en
t
s
t
u
d
y
co
n
s
id
er
s
s
u
b
s
ta
n
tial
s
a
m
p
le
s
f
o
r
ex
p
er
i
m
e
n
tatio
n
(
2
5
7
,
6
7
3
n
et
w
o
r
k
in
s
ta
n
ce
s
i
n
clu
s
i
v
e
o
f
b
o
th
tr
ai
n
i
n
g
a
n
d
t
esti
n
g
d
ataset
s
)
.
I
t
is
w
o
r
t
h
w
h
ile
to
m
en
tio
n
t
h
at
tr
ain
i
n
g
ti
m
e
o
f
all
th
e
eig
h
t
t
w
o
-
cla
s
s
p
r
ed
ictiv
e
m
o
d
els
w
as
f
o
u
n
d
to
b
e
q
u
ite
m
i
n
i
m
al
as
r
ep
o
r
ted
in
t
h
e
r
an
g
e
o
f
6
to
9
s
ec
o
n
d
s
w
h
er
ea
s
m
u
lticla
s
s
cla
s
s
i
f
icat
io
n
m
o
d
els
to
o
k
r
elat
iv
el
y
lo
n
g
er
to
g
et
f
a
m
i
liar
w
it
h
d
i
f
f
er
e
n
t a
ttac
k
ca
teg
o
r
ie
s
.
4.
CO
NCLU
SI
O
N
AND
P
RO
S
P
E
CT
S
I
n
th
i
s
s
t
u
d
y
,
ei
g
h
t
t
w
o
-
clas
s
an
d
t
h
r
ee
m
u
lticla
s
s
cla
s
s
i
f
icatio
n
m
o
d
els
w
er
e
d
ev
e
lo
p
ed
u
s
in
g
UNSW
NB
-
1
5
d
atase
t.
B
ased
o
n
e
m
p
ir
ical
in
v
es
tig
a
tio
n
,
i
t
ca
n
b
e
s
tated
th
a
t
d
ec
is
io
n
f
o
r
est
ac
co
m
p
lis
h
ed
th
e
b
est
p
er
f
o
r
m
a
n
ce
.
Sin
ce
i
t
is
ex
tr
e
m
el
y
t
i
m
e
co
n
s
u
m
in
g
to
ex
ec
u
te
t
h
e
ex
p
er
i
m
e
n
ts
o
n
lo
ca
l
s
y
s
te
m
s
,
Mic
r
o
s
o
f
t
A
z
u
r
e
Ma
c
h
in
e
L
e
ar
n
in
g
Stu
d
io
(
M
A
M
L
S)
w
as
ch
o
s
en
f
o
r
e
x
p
er
i
m
e
n
tatio
n
.
A
p
ar
t
f
r
o
m
s
ta
n
d
ar
d
p
er
f
o
r
m
a
n
ce
m
e
tr
ics
lik
e
ac
c
u
r
ac
y
,
p
r
ec
is
io
n
,
r
ec
all,
f
1
-
s
c
o
r
e
an
d
A
UC
,
th
e
p
r
o
p
o
s
ed
w
o
r
k
also
co
n
s
id
er
ed
tr
ain
i
n
g
ti
m
e
a
n
d
ex
ec
u
tio
n
t
i
m
e
to
ev
al
u
ate
t
h
e
ef
f
ec
ti
v
e
n
es
s
o
f
th
e
al
g
o
r
it
h
m
s
.
T
h
e
p
r
o
p
o
s
ed
s
tu
d
y
h
a
s
h
ig
h
li
g
h
ted
th
at
M
AM
L
S
ca
n
s
er
v
e
as
an
e
x
p
ed
ien
t
I
n
teg
r
at
ed
Dev
elo
p
m
en
t
E
n
v
ir
o
n
m
e
n
t
(
I
DE
)
f
o
r
h
an
d
lin
g
lar
g
e
d
ataset
s
.
A
s
a
p
ar
t
o
f
f
u
t
u
r
e
w
o
r
k
,
it
w
ill
b
e
i
n
ter
esti
n
g
to
e
m
p
lo
y
d
if
f
er
e
n
t
i
n
tr
u
s
io
n
d
etec
tio
n
d
ataset
s
,
s
u
b
s
eq
u
en
t
l
y
g
au
g
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
v
ar
io
u
s
cla
s
s
i
f
i
er
s
.
E
x
p
er
ts
h
a
v
e
al
w
a
y
s
u
r
g
ed
th
e
r
esear
c
h
co
m
m
u
n
it
y
to
e
x
p
er
i
m
e
n
t
with
d
if
f
er
en
t
d
atasets
an
d
i
n
t
r
o
d
u
ce
n
o
v
el
tech
n
iq
u
e
s
f
o
r
n
et
w
o
r
k
in
tr
u
s
io
n
d
etec
tio
n
[
3
3
,
3
4
]
.
A
n
o
t
h
er
av
en
u
e
w
h
ich
ca
n
b
e
ex
p
lo
r
ed
in
f
u
t
u
r
e
ca
n
p
o
s
s
ib
l
y
i
n
cl
u
d
e
th
e
d
ep
lo
y
m
en
t
o
f
p
r
ed
ictiv
e
m
o
d
els
a
s
s
ca
lab
le
w
eb
s
er
v
ice
s
t
h
er
eb
y
le
v
e
r
ag
in
g
t
h
e
ca
p
ab
ilit
ie
s
o
f
M
A
M
L
S.
I
t
w
ill
b
e
tech
n
icall
y
c
h
alle
n
g
in
g
to
i
m
p
le
m
e
n
t
a
w
r
ap
p
er
b
ased
ap
p
r
o
ac
h
o
n
M
AM
L
S.
Su
ch
w
r
ap
p
er
b
ased
ap
p
r
o
ac
h
es
m
a
y
b
e
h
elp
f
u
l
to
d
em
o
n
s
tr
at
e
t
h
e
ef
f
ec
ti
v
en
e
s
s
o
f
MA
M
L
S,
e
v
en
t
u
all
y
r
es
u
t
in
g
i
n
a
p
er
ce
p
tiv
e
ass
es
s
m
en
t o
f
its
co
m
p
u
tatio
n
al
p
er
f
o
r
m
a
n
ce
.
RE
F
E
R
E
NC
E
S
[1
]
Ha
k
i
m
i,
Zah
ra
,
K
a
ri
m
F
e
a
z
,
M
o
rtez
a
B
a
ra
ti
,
"
A
F
lo
w
-
b
a
se
d
Distrib
u
ted
In
tru
si
o
n
De
tec
ti
o
n
S
y
ste
m
Us
in
g
M
o
b
il
e
Ag
e
n
ts,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
C
o
mp
u
ter
En
g
i
n
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
3
,
n
o
.
6,
p
p
.
7
3
2
-
7
4
0
,
2
0
1
3
.
[2
]
Ja
n
g
-
Ja
c
c
a
rd
J,
Ne
p
a
l
S
.
,
"
A
su
rv
e
y
o
f
e
m
e
rg
in
g
th
re
a
ts
in
c
y
b
e
rs
e
c
u
rit
y
,
"
J
o
u
rn
a
l
o
f
Co
m
p
u
ter
a
n
d
S
y
ste
m
S
c
ien
c
e
s
,
v
o
l.
80
,
n
o
.
5
,
p
p
.
9
7
3
-
93
,
2
0
1
4
.
[3
]
Ya
n
F
,
Jia
n
-
W
e
n
Y,
L
in
C.
,
"
Co
m
p
u
ter
n
e
t
w
o
rk
s
e
c
u
rit
y
a
n
d
tec
h
n
o
l
o
g
y
re
s
e
a
rc
h
,
"
In
2
0
1
5
S
e
v
e
n
t
h
In
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
M
e
a
su
rin
g
T
e
c
h
n
o
lo
g
y
a
n
d
M
e
c
h
a
tro
n
ics
Au
to
ma
ti
o
n
,
IEE
E
,
p
p
.
2
9
3
-
2
9
6
,
2
0
1
5
.
[4
]
Ya
z
d
a
n
i,
Na
v
id
M
o
sh
tag
h
i,
M
a
so
u
d
S
h
a
riat
P
a
n
a
h
i,
a
n
d
Eh
sa
n
S
a
d
e
g
h
i
P
o
o
r,
"
In
tel
li
g
e
n
t
De
tec
ti
o
n
o
f
In
tr
u
sio
n
in
to
Da
tab
a
se
s
U
sin
g
Ex
ten
d
e
d
Clas
sif
ier
S
y
ste
m
,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l
.
3
,
n
o
.
5
,
p
p
.
2
0
8
8
-
8
7
0
8
,
2
0
1
3
.
[5
]
A
b
u
ro
m
m
a
n
,
A
b
d
u
ll
a
A
m
in
,
a
n
d
M
a
m
u
n
Bi
n
I
b
n
e
Re
a
z
,
"
Re
v
ie
w
o
f
ID
S
De
v
e
lo
p
m
e
n
t
M
e
t
h
o
d
s
i
n
M
a
c
h
in
e
L
e
a
rn
in
g
,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
trica
l
a
n
d
C
o
mp
u
ter
E
n
g
i
n
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
6
,
n
o
.
5
,
pp.
2
4
3
2
-
2
4
3
6
,
2
0
1
6
.
[6
]
B
u
c
z
a
k
,
A
n
n
a
L
.
,
a
n
d
Er
h
a
n
G
u
v
e
n
,
"
A
su
rv
e
y
o
f
d
a
ta
m
in
in
g
a
n
d
m
a
c
h
in
e
lea
rn
i
n
g
m
e
th
o
d
s
f
o
r
c
y
b
e
r
se
c
u
ri
t
y
in
tru
si
o
n
d
e
tec
ti
o
n
,
"
IEE
E
C
o
mm
u
n
ica
ti
o
n
s S
u
rv
e
y
s
&
T
u
to
ria
ls
,
v
o
l.
1
8
,
n
o
.
2
,
p
p
.
1
1
5
3
-
1
1
7
6
,
2
0
1
5
.
[7
]
Oth
m
a
n
,
S
u
a
d
M
o
h
a
m
m
e
d
,
F
a
d
l
M
u
tah
e
r
Ba
-
A
l
w
i,
Na
b
e
e
l
T
.
A
l
so
h
y
b
e
,
a
n
d
Am
a
l
Y.
A
l
-
Ha
s
h
id
a
,
"
In
tru
si
o
n
d
e
tec
ti
o
n
m
o
d
e
l
u
sin
g
m
a
c
h
in
e
lea
rn
in
g
a
l
g
o
rit
h
m
o
n
B
ig
Da
t
a
e
n
v
iro
n
m
e
n
t,
"
J
o
u
rn
a
l
o
f
Bi
g
Da
t
a
,
v
o
l.
5
,
n
o
.
1
,
p
p
.
34
,
2
0
1
8
.
[8
]
T
c
h
a
k
o
u
c
h
t,
T
a
h
a
A
I
T
,
a
n
d
M
o
sta
f
a
Ezz
i
y
y
a
n
i,
"
Bu
il
d
in
g
a
fa
st
in
tru
si
o
n
d
e
tec
ti
o
n
sy
ste
m
fo
r
h
ig
h
-
s
p
e
e
d
-
n
e
tw
o
rk
s:
p
ro
b
e
a
n
d
DO
S
a
tt
a
c
k
s d
e
tec
ti
o
n
,
"
Pro
c
e
d
ia
Co
mp
u
ter
S
c
ien
c
e
,
v
o
l.
1
2
7
,
p
p
.
5
2
1
-
5
3
0
,
2
0
1
8
.
[9
]
Zu
e
c
h
,
Rich
a
r
d
,
T
a
g
h
i
M
.
Kh
o
s
h
g
o
f
taa
r,
a
n
d
Ra
n
d
a
ll
W
a
ld
,
"
I
n
tru
si
o
n
d
e
tec
ti
o
n
a
n
d
b
ig
h
e
ter
o
g
e
n
e
o
u
s
d
a
ta:
a
su
rv
e
y
,
"
J
o
u
rn
a
l
o
f
Bi
g
Da
t
a
,
v
o
l.
2
,
n
o
.
1
,
p
p
.
3
,
2
0
1
5
.
[1
0
]
S
u
th
a
h
a
ra
n
,
S
h
a
n
,
"
Big
d
a
ta
c
las
sif
ica
ti
o
n
:
P
ro
b
lem
s
a
n
d
c
h
a
ll
e
n
g
e
s
i
n
n
e
tw
o
rk
in
tru
si
o
n
p
re
d
ictio
n
w
it
h
m
a
c
h
in
e
lea
rn
in
g
,
"
ACM
S
IGM
ET
RICS
Per
fo
rm
a
n
c
e
Eva
l
u
a
t
io
n
Rev
iew
,
v
o
l.
4
1
,
n
o
.
4
,
p
p
.
70
-
73
,
2
0
1
4
.
[1
1
]
Na
ss
a
r,
M
o
h
a
m
e
d
,
Be
c
h
a
ra
a
l
B
o
u
n
a
,
a
n
d
Q
u
taib
a
h
M
a
ll
u
h
i,
"
S
e
c
u
re
o
u
tso
u
rc
in
g
o
f
n
e
tw
o
rk
f
lo
w
d
a
ta
a
n
a
l
y
sis,
"
In
2
0
1
3
IEE
E
In
ter
n
a
ti
o
n
a
l
C
o
n
g
re
ss
o
n
Bi
g
Da
t
a
,
IE
EE
,
p
p
.
4
3
1
-
4
3
2
,
2
0
1
3
.
[1
2
]
S
p
a
rk
s,
Ev
a
n
R.
,
Am
e
e
t
Tal
w
a
l
k
a
r,
V
irg
in
ia
S
m
it
h
,
Je
y
Ko
tt
a
lam
,
X
in
g
h
a
o
P
a
n
,
J
o
se
p
h
G
o
n
z
a
lez
,
M
ich
a
e
l
J.
F
ra
n
k
li
n
,
M
i
c
h
a
e
l
I.
Jo
r
d
a
n
,
a
n
d
T
i
m
Kra
sk
a
,
"
M
L
I:
A
n
A
P
I
f
o
r
d
istri
b
u
te
d
m
a
c
h
in
e
lea
rn
in
g
,
"
In
2
0
1
3
IEE
E
1
3
t
h
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
D
a
ta
M
in
in
g
,
IEE
E,
p
p
.
1
1
8
7
-
1
1
9
2
,
2
0
1
3
.
[1
3
]
Ca
sa
d
o
,
Ru
b
e
n
,
a
n
d
M
u
h
a
m
m
a
d
Yo
u
n
a
s,
"
Em
e
r
g
in
g
tren
d
s
a
n
d
te
c
h
n
o
l
o
g
ies
in
b
ig
d
a
ta
p
r
o
c
e
ss
in
g
,
"
Co
n
c
u
rr
e
n
c
y
a
n
d
Co
m
p
u
t
a
ti
o
n
:
Pr
a
c
ti
c
e
a
n
d
E
x
p
e
rie
n
c
e
,
v
o
l.
2
7
,
n
o
.
8
,
pp
.
2
0
7
8
-
2
0
9
1
,
2
0
1
5
.
[1
4
]
Zah
a
ria,
M
a
tei,
Re
y
n
o
ld
S
.
X
in
,
P
a
tri
c
k
W
e
n
d
e
ll
,
T
a
th
a
g
a
ta
Da
s,
M
ich
a
e
l
A
r
m
b
ru
st,
A
n
k
u
r
Da
v
e
,
X
ian
g
ru
i
M
e
n
g
e
t
a
l.
,
"
A
p
a
c
h
e
sp
a
rk
:
a
u
n
if
ied
e
n
g
in
e
f
o
r
b
ig
d
a
ta
p
ro
c
e
ss
in
g
,
"
Co
mm
u
n
ica
ti
o
n
s
o
f
t
h
e
ACM
,
v
o
l.
5
9
,
n
o
.
1
1
,
p
p
.
56
-
65
,
2
0
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
978
-
986
986
[1
5
]
T
e
a
m
,
Az
u
re
M
L
,
"
A
z
u
re
M
L
:
An
a
to
m
y
o
f
a
m
a
c
h
in
e
lea
rn
in
g
se
rv
ice
,
"
In
Co
n
fer
e
n
c
e
o
n
Pre
d
i
c
ti
v
e
AP
Is
a
n
d
Ap
p
s
,
p
p
.
1
-
1
3
.
2
0
1
6
.
[1
6
]
El
sh
a
w
i,
Ra
d
w
a
,
S
h
e
rif
S
a
k
r,
Do
m
e
n
ico
T
a
li
a
,
a
n
d
P
a
o
l
o
T
ru
n
f
io
,
"
Big
d
a
ta
sy
ste
m
s
m
e
e
t
m
a
c
h
in
e
lea
rn
in
g
c
h
a
ll
e
n
g
e
s: T
o
w
a
rd
s b
ig
d
a
ta
sc
ien
c
e
a
s a se
rv
ice
,
"
Bi
g
d
a
ta
re
se
a
r
c
h
,
v
o
l.
14
,
p
p
.
1
-
11
,
2
0
1
8
.
[1
7
]
M
o
u
sta
f
a
,
No
u
r,
a
n
d
Jill
S
lay
,
"
UN
S
W
-
NB1
5
:
a
c
o
m
p
re
h
e
n
siv
e
d
a
ta
se
t
f
o
r
n
e
tw
o
rk
in
tru
si
o
n
d
e
tec
ti
o
n
sy
ste
m
s
(UN
S
W
-
NB1
5
n
e
tw
o
rk
d
a
ta
se
t),
"
In
2
0
1
5
mil
it
a
ry
c
o
mm
u
n
ica
ti
o
n
s
a
n
d
in
f
o
rm
a
ti
o
n
sy
ste
ms
c
o
n
fer
e
n
c
e
(
M
il
CIS
)
,
IEE
E,
p
p
.
1
-
6,
2
0
1
5
.
[1
8
]
M
o
u
sta
f
a
,
No
u
r,
a
n
d
Jill
S
lay
,
"
Th
e
e
v
a
lu
a
ti
o
n
o
f
Ne
tw
o
rk
A
n
o
m
a
ly
De
t
e
c
ti
o
n
S
y
ste
m
s:
S
tatisti
c
a
l
a
n
a
ly
sis
o
f
th
e
UN
S
W
-
NB1
5
d
a
ta
se
t
a
n
d
th
e
c
o
m
p
a
ri
so
n
w
it
h
th
e
KD
D9
9
d
a
t
a
se
t,
"
In
fo
rm
a
ti
o
n
S
e
c
u
rity
J
o
u
rn
a
l:
A
Gl
o
b
a
l
Per
sp
e
c
ti
v
e
,
v
o
l.
2
5
,
n
o
.
1
-
3
,
p
p
.
18
-
31
,
2
0
1
6
.
[1
9
]
M
o
u
sta
f
a
,
No
u
r,
a
n
d
Jill
S
lay
,
"
Th
e
sig
n
if
ica
n
t
f
e
a
tu
re
s
o
f
th
e
UN
S
W
-
NB1
5
a
n
d
th
e
KD
D9
9
d
a
ta
s
e
ts
f
o
r
n
e
tw
o
rk
in
tru
si
o
n
d
e
tec
ti
o
n
sy
ste
m
s,
"
In
2
0
1
5
4
t
h
in
ter
n
a
ti
o
n
a
l
wo
rk
sh
o
p
o
n
b
u
il
d
in
g
a
n
a
lys
is
d
a
ta
se
t
s
a
n
d
g
a
t
h
e
rin
g
e
x
p
e
rie
n
c
e
re
tu
rn
s fo
r se
c
u
rity (
BA
DG
ER
S
)
,
IE
EE
,
p
p
.
2
5
-
31
,
2
0
1
5
.
[2
0
]
Rib
e
iro
,
M
a
u
ro
,
Ka
tarin
a
G
ro
li
n
g
e
r,
a
n
d
M
iri
a
m
A
M
Ca
p
re
tz,
"
M
laa
s
:
M
a
c
h
in
e
lea
rn
in
g
a
s
a
se
rv
ice
,
"
In
2
0
1
5
IEE
E
1
4
th
In
ter
n
a
ti
o
n
a
l
C
o
n
fer
e
n
c
e
o
n
M
a
c
h
in
e
L
e
a
rn
i
n
g
a
n
d
A
p
p
li
c
a
ti
o
n
s (
ICM
L
A)
,
IEE
E
,
p
p
.
8
9
6
-
9
0
2
,
2
0
1
5
.
[2
1
]
T
a
f
ti
,
A
h
m
a
d
P
.
,
Eri
c
L
a
Ro
se
,
Jo
n
a
th
a
n
C.
Ba
d
g
e
r,
R
o
ss
Kle
ima
n
,
a
n
d
P
e
g
g
y
P
e
issig
,
"
M
a
c
h
in
e
lea
rn
in
g
-
as
-
a
-
se
rv
ice
a
n
d
it
s
a
p
p
li
c
a
ti
o
n
to
m
e
d
ica
l
in
f
o
rm
a
ti
c
s,
"
In
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
M
a
c
h
in
e
L
e
a
r
n
in
g
a
n
d
Da
t
a
M
in
in
g
in
Pa
tt
e
r
n
Rec
o
g
n
it
i
o
n
,
S
p
rin
g
e
r,
Ch
a
m
,
p
p
.
2
0
6
-
2
1
9
,
2
0
1
7
.
[2
2
]
Da
h
iy
a
,
P
riy
a
n
k
a
,
a
n
d
De
v
e
sh
Ku
m
a
r
S
riv
a
sta
v
a
,
"
N
e
t
w
o
rk
in
tru
sio
n
d
e
tec
ti
o
n
in
b
ig
d
a
tas
e
t
u
sin
g
S
p
a
rk
,
"
Pro
c
e
d
ia
c
o
mp
u
ter
sc
ien
c
e
,
v
o
l.
132
,
p
p
.
2
5
3
-
2
6
2
,
2
0
1
8
.
[2
3
]
Kh
a
m
m
a
ss
i
,
Ch
a
o
u
k
i,
a
n
d
S
a
o
u
ss
e
n
Kric
h
e
n
,
"
A
GA
-
L
R
w
r
a
p
p
e
r
a
p
p
r
o
a
c
h
f
o
r
f
e
a
tu
re
se
lec
ti
o
n
i
n
n
e
tw
o
rk
in
tru
si
o
n
d
e
tec
ti
o
n
,
"
c
o
mp
u
ter
s
&
se
c
u
rity
,
v
o
l.
70
,
p
p
.
2
5
5
-
2
7
7
,
2
0
1
7
.
[2
4
]
Am
b
u
sa
id
i,
M
o
h
a
m
m
e
d
A
.
,
X
ian
g
ji
a
n
He
,
P
riy
a
d
a
rsi
Na
n
d
a
,
a
n
d
Zh
iy
u
a
n
T
a
n
,
"
Bu
il
d
i
n
g
a
n
in
t
ru
sio
n
d
e
tec
ti
o
n
s
y
st
e
m
u
sin
g
a
f
il
ter
-
ba
se
d
f
e
a
tu
re
se
lec
ti
o
n
a
lg
o
rit
h
m
,
"
IEE
E
tra
n
sa
c
ti
o
n
s
o
n
c
o
m
p
u
ter
s
,
v
o
l.
6
5
,
n
o
.
1
0
,
p
p
.
2
9
8
6
-
2
9
9
8
,
2
0
1
6
.
[2
5
]
Be
lo
u
c
h
,
M
u
sta
p
h
a
,
S
a
lah
El
H
a
d
a
j,
a
n
d
M
o
h
a
m
e
d
Id
h
a
m
m
a
d
,
"
P
e
rf
o
r
m
a
n
c
e
e
v
a
lu
a
ti
o
n
o
f
in
tr
u
sio
n
d
e
tec
ti
o
n
b
a
se
d
o
n
m
a
c
h
in
e
lea
rn
in
g
u
sin
g
A
p
a
c
h
e
S
p
a
rk
,
"
Pro
c
e
d
ia
Co
mp
u
t
e
r S
c
ien
c
e
,
v
o
l.
1
2
7
,
p
p
.
1
-
6
,
2
0
1
8
.
[2
6
]
Bh
a
m
a
r
e
,
De
v
a
l,
T
a
r
a
S
a
l
m
a
n
,
M
o
h
a
m
m
e
d
S
a
m
a
k
a
,
A
i
m
a
n
Erb
a
d
,
a
n
d
Ra
j
Ja
in
,
"
F
e
a
sib
il
it
y
o
f
su
p
e
rv
ise
d
m
a
c
h
in
e
lea
rn
in
g
f
o
r
c
lo
u
d
se
c
u
rit
y
,
"
In
2
0
1
6
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
I
n
fo
rm
a
ti
o
n
S
c
ien
c
e
a
n
d
S
e
c
u
rity
(
ICIS
S
)
,
p
p
.
1
-
5,
IEE
E,
2
0
1
6
.
[2
7
]
V
e
rg
a
ra
,
Jo
rg
e
R.
,
a
n
d
P
a
b
lo
A
.
Estev
e
z
,
"
A
re
v
ie
w
o
f
f
e
a
tu
re
s
e
lec
ti
o
n
m
e
th
o
d
s
b
a
se
d
o
n
m
u
tu
a
l
in
f
o
rm
a
ti
o
n
,
"
Ne
u
ra
l
c
o
mp
u
ti
n
g
a
n
d
a
p
p
li
c
a
t
io
n
s
,
v
o
l.
2
4
,
n
o
.
1
,
p
p
.
1
7
5
-
1
8
6
,
2
0
1
4
.
[2
8
]
S
m
it
h
,
T
o
n
y
C.
,
a
n
d
E
ib
e
F
ra
n
k
,
"
In
tro
d
u
c
in
g
m
a
c
h
in
e
lea
rn
in
g
c
o
n
c
e
p
ts
w
it
h
W
EKA
,
"
In
S
ta
ti
s
ti
c
a
l
g
e
n
o
mic
s
,
Hu
ma
n
a
Pre
ss
,
Ne
w
Y
o
rk
,
N
Y
,
p
p
.
3
5
3
-
3
7
8
,
2
0
1
6
.
[2
9
]
S
h
o
t
to
n
,
Ja
m
i
e
,
T
o
b
y
S
h
a
rp
,
P
u
s
h
m
e
e
t
Ko
h
li
,
S
e
b
a
stian
No
w
o
z
in
,
J
o
h
n
W
in
n
,
a
n
d
A
n
to
n
io
Crim
in
isi,
"
De
c
isio
n
ju
n
g
les
:
Co
m
p
a
c
t
a
n
d
r
ich
m
o
d
e
ls
f
o
r
c
las
si
f
ica
ti
o
n
,
"
In
A
d
v
a
n
c
e
s
in
Ne
u
ra
l
I
n
f
o
rm
a
ti
o
n
Pro
c
e
ss
in
g
S
y
ste
ms
,
p
p
.
2
3
4
-
242
,
2
0
1
3
.
[3
0
]
Jo
se
,
Cij
o
,
P
ra
so
o
n
G
o
y
a
l
,
P
a
rv
Ag
g
r
wa
l,
a
n
d
M
a
n
ik
V
a
rm
a
,
"
L
o
c
a
l
d
e
e
p
k
e
rn
e
l
lea
rn
in
g
f
o
r
e
ff
i
c
ien
t
n
o
n
-
li
n
e
a
r
sv
m
p
re
d
ictio
n
,
"
I
n
In
ter
n
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
m
a
c
h
i
n
e
lea
rn
i
n
g
,
p
p
.
4
8
6
-
4
9
4
.
2
0
1
3
.
[3
1
]
Ho
w
le
y
,
T
o
m
,
a
n
d
M
ic
h
a
e
l
G
.
M
a
d
d
e
n
,
"
T
h
e
g
e
n
e
ti
c
k
e
rn
e
l
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
:
De
sc
rip
ti
o
n
a
n
d
e
v
a
lu
a
ti
o
n
,
"
Arti
fi
c
ia
l
in
telli
g
e
n
c
e
re
v
iew
,
v
o
l.
2
4
,
n
o
.
3
-
4,
p
p
.
3
7
9
-
3
9
5
,
2
0
0
5
.
[3
2
]
Kra
w
c
z
y
k
,
B
a
rto
sz
,
"
L
e
a
rn
in
g
f
r
o
m
i
m
b
a
lan
c
e
d
d
a
ta:
o
p
e
n
c
h
a
ll
e
n
g
e
s
a
n
d
f
u
tu
re
d
irec
ti
o
n
s,
"
Pro
g
re
ss
in
Arti
fi
c
ia
l
In
telli
g
e
n
c
e
,
v
o
l.
5
,
n
o
.
4
,
p
p
.
2
2
1
-
2
3
2
,
2
0
1
6
.
[3
3
]
M
o
u
sta
f
a
,
No
u
r
,
Jia
n
k
u
n
Hu
,
a
n
d
Jill
S
lay
,
"
A
h
o
li
stic
re
v
iew
o
f
N
e
t
w
o
rk
A
n
o
m
a
l
y
De
te
c
ti
o
n
S
y
ste
m
s:
A
c
o
m
p
re
h
e
n
siv
e
su
rv
e
y
,
"
J
o
u
rn
a
l
o
f
Ne
two
rk
a
n
d
Co
mp
u
ter
A
p
p
l
ica
ti
o
n
s
,
v
o
l.
1
2
8
,
p
p
.
33
-
55
,
2
0
1
9
.
[3
4
]
Rin
g
,
M
a
rk
u
s,
S
a
ra
h
W
u
n
d
e
rli
c
h
,
De
n
iz
S
c
h
e
u
ri
n
g
,
D
iete
r
L
a
n
d
e
s,
a
n
d
A
n
d
re
a
s
Ho
th
o
,
"
A
su
rv
e
y
o
f
n
e
t
w
o
rk
-
b
a
se
d
in
tr
u
sio
n
d
e
tec
ti
o
n
d
a
ta se
ts,
"
Co
mp
u
ter
s
&
S
e
c
u
rity
,
2
0
1
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.