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
.
15
, N
o.
1
,
F
e
br
ua
r
y
2026
, pp.
269
~
288
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
15
.i
1
.pp
269
-
288
269
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
D
e
si
gn
of
A
n
t
ase
n
a:
an
AI
-
p
ow
e
r
e
d
m
ar
i
t
i
m
e
su
r
ve
i
l
l
an
c
e
an
d
an
om
al
y
d
e
t
e
c
t
i
on
sys
t
e
m
f
or
se
c
u
r
i
t
y d
e
c
i
si
on
su
p
p
or
t
A
r
if
B
ad
r
u
d
in
1
,
S
is
w
o H
ad
i
S
u
m
an
t
r
i
1
,
R
u
d
y A
gu
s
G
e
m
il
an
g G
u
lt
om
1
,
I
N
e
n
gah
P
u
t
r
a A
p
r
iy
an
t
o
1
,
U
m
i
L
ai
li
Y
u
h
an
a
2
,
F
it
r
ia
D
w
i
R
at
n
as
ar
i
3
1
D
oc
t
or
a
l
P
r
ogr
a
m
of
D
e
f
e
nc
e
S
c
i
e
nc
e
,
F
a
c
ul
t
y of
D
e
f
e
nc
e
T
e
c
hnol
ogy, I
ndone
s
i
a
D
e
f
e
ns
e
U
ni
ve
r
s
i
t
y,
B
ogor
, I
ndone
s
i
a
2
D
e
pa
r
t
m
e
nt
of
I
nf
or
m
a
t
i
c
s
, I
ns
t
i
t
ut
T
e
knol
ogi
S
e
pul
uh N
ove
m
be
r
, S
ur
a
ba
ya
, I
ndone
s
i
a
3
M
a
s
t
e
r
of
L
a
w
a
nd D
e
ve
l
opm
e
nt
, U
ni
ve
r
s
i
t
a
s
A
i
r
l
a
ngga
, S
ur
a
ba
ya
,
I
ndone
s
i
a
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
26
,
2024
R
e
vi
s
e
d
D
e
c
31
,
2025
A
c
c
e
pt
e
d
J
a
n
22
,
2026
Indonesia’s
vast
maritime
territory
faces
serious
challenges
from
illegal
fishing,
smuggli
ng,
and
habitat
destruction
.
To
address
these,
the
Indonesian
Navy
(TNI
-
AL)
developed
Antasena,
an
a
rtificial
intellig
ence
(
AI
)
-
powered
smart
dashboard
integrating
automati
c
identif
ication
system
(AIS)
data,
satellite
imagery,
and
conservation
metrics.
Antasena
lev
erages
ad
vanced
anomaly
detection
algorit
hms,
achieving
95.3%
accuracy,
94.7%
pre
cision,
94.2%
recall,
and
a
96.8%
receiver
operating
characteristic
-
area
un
d
er
the
curve
(
ROC
-
AUC
)
score
in
identifying
vessel
anomalies,
inc
luding
unauthorized
fishing
and
smuggling
activities.
Using
the
analyze,
design,
develop,
implement,
and
evaluate
(
ADDIE
)
framework,
the
system
su
pports
real
-
time
maritime
surveillance
and
biodiversity
monitoring
in
conser
vation
zones.
The
main
contributions
of
this
study
includ
e
the
developme
nt
of
a
user
-
centric
AI
-
based
dashboard
for
maritime
anomaly
detectio
n,
the
integration
of
multi
-
source
data
with
machine
learning
model
s,
and
validation through operational field tests
with maritime
authorities. An
tasena
offers
a
scalable
and
effective
solution
to
strengthen
maritime
securi
ty
and
protect Indonesia’s marine re
sources.
K
e
y
w
o
r
d
s
:
A
nom
a
ly
de
te
c
ti
on s
ys
t
e
m
A
nt
a
s
e
na
A
r
ti
f
ic
ia
l
i
nt
e
ll
ig
e
nc
e
P
ow
e
r
e
d m
a
r
it
im
e
s
ur
ve
il
la
nc
e
S
e
c
ur
it
y de
c
is
io
n s
uppor
t
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
A
r
if
B
a
dr
udi
n
D
oc
to
r
a
l
P
r
ogr
a
m
of
D
e
f
e
nc
e
S
c
ie
nc
e
,
F
a
c
ul
ty
of
D
e
f
e
nc
e
T
e
c
hnol
ogy, I
ndone
s
ia
D
e
f
e
ns
e
U
ni
v
e
r
s
it
y
I
P
S
C
A
r
e
a
, S
e
nt
ul
B
ogor
16810, W
e
s
t
J
a
va
, I
ndone
s
ia
E
m
a
il
:
a
r
if
.ba
dr
udi
n11379@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
I
ndone
s
ia
’
s
va
s
t
m
a
r
it
im
e
te
r
r
it
or
y,
s
pa
nni
ng
ove
r
5.8
m
il
li
on
km
²,
pl
a
ys
a
c
r
uc
ia
l
r
ol
e
in
gl
oba
l
tr
a
de
a
nd
r
e
s
our
c
e
m
a
na
g
e
m
e
nt
.
H
ow
e
v
e
r
,
th
is
va
s
tn
e
s
s
a
ls
o
m
a
ke
s
it
vul
ne
r
a
bl
e
to
s
e
c
ur
it
y
th
r
e
a
ts
s
u
c
h
a
s
il
le
ga
l
f
is
hi
ng,
s
m
uggl
in
g,
a
nd
e
nvi
r
onm
e
nt
a
l
vi
ol
a
ti
ons
.
T
o
a
ddr
e
s
s
th
e
s
e
c
h
a
ll
e
nge
s
,
th
e
I
ndone
s
i
a
n
N
a
vy
(
T
N
I
-
A
L
)
de
ve
lo
pe
d
A
nt
a
s
e
na
,
a
n
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
-
pow
e
r
e
d
m
oni
to
r
in
g
s
ys
te
m
th
a
t
in
te
gr
a
te
s
a
ut
om
a
ti
c
i
de
nt
if
ic
a
ti
on s
ys
te
m
(
A
I
S
)
da
ta
t
o de
te
c
t
a
nom
a
lo
us
ve
s
s
e
l
be
h
a
vi
or
s
i
n r
e
a
l
ti
m
e
.
A
nt
a
s
e
na
c
ol
le
c
ts
a
nd
a
na
ly
z
e
s
m
ul
ti
-
s
our
c
e
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
da
ta
,
pr
ovi
di
ng
a
c
c
ur
a
te
e
a
r
ly
-
w
a
r
ni
ng
in
f
or
m
a
ti
on
to
m
a
r
it
im
e
tr
a
f
f
ic
pa
r
ti
c
ip
a
nt
s
a
nd
e
na
bl
in
g
pr
oa
c
ti
ve
r
e
s
pons
e
s
to
pot
e
nt
ia
l
r
is
k
s
[
1]
.
A
nom
a
ly
de
te
c
ti
on
a
lg
or
it
hm
s
a
ppl
ie
d
to
A
I
S
da
ta
id
e
nt
if
y
be
ha
vi
or
a
l
pa
tt
e
r
ns
th
a
t
de
vi
a
te
f
r
om
nor
m
a
l
na
vi
ga
ti
on,
w
hi
c
h
m
a
y
in
di
c
a
te
il
le
ga
l
f
is
hi
ng,
pi
r
a
c
y,
or
s
m
uggl
in
g
a
c
ti
vi
ti
e
s
[
2]
.
A
c
c
ur
a
te
a
nd
ti
m
e
ly
a
nom
a
ly
de
te
c
ti
on i
s
t
he
r
e
f
or
e
c
r
it
ic
a
l
f
or
m
a
r
it
im
e
doma
in
a
w
a
r
e
ne
s
s
a
nd n
a
ti
ona
l
de
f
e
ns
e
op
e
r
a
ti
ons
[
3]
.
P
r
e
vi
ous
s
tu
di
e
s
on
m
a
r
it
im
e
a
nom
a
ly
de
te
c
ti
on
ha
ve
im
pl
e
m
e
nt
e
d
a
w
id
e
r
a
nge
of
m
a
c
hi
ne
le
a
r
ni
ng
a
ppr
oa
c
he
s
,
in
c
lu
di
ng
d
e
c
is
io
n
tr
e
e
(
D
T
)
,
r
a
ndom
f
or
e
s
t
(
R
F
)
,
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
,
a
nd
ne
ur
a
l
ne
twor
k
[
4]
–
[
6]
.
W
hi
le
th
e
s
e
te
c
hni
que
s
ha
ve
a
c
hi
e
ve
d
s
a
ti
s
f
a
c
to
r
y
a
c
c
ur
a
c
y,
th
e
y
of
te
n
f
a
c
e
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
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
269
-
288
270
li
m
it
a
ti
ons
in
in
te
r
pr
e
ta
bi
li
ty
,
r
obus
tn
e
s
s
,
a
nd
a
da
pt
a
bi
li
ty
to
s
tr
e
a
m
in
g
A
I
S
da
ta
.
S
in
gl
e
-
m
ode
l
c
la
s
s
if
ie
r
s
s
uc
h
a
s
D
T
or
S
V
M
te
nd
to
ove
r
f
it
c
om
pl
e
x
da
ta
di
s
tr
ib
ut
io
ns
,
w
he
r
e
a
s
de
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
d
e
m
a
nd
e
xt
e
ns
iv
e
c
om
put
a
ti
on
a
nd
pr
ovi
de
li
m
it
e
d
tr
a
n
s
pa
r
e
nc
y
in
de
c
i
s
io
n
-
m
a
ki
ng
[
7]
–
[
9]
.
T
he
s
e
c
ha
ll
e
nge
s
r
e
duc
e
th
e
ir
f
e
a
s
ib
il
it
y f
or
r
e
a
l
-
ti
m
e
ope
r
a
ti
ona
l
de
pl
oym
e
nt
i
n m
a
r
it
im
e
s
ur
ve
il
la
nc
e
s
y
s
te
m
s
s
uc
h a
s
A
nt
a
s
e
na
.
R
e
c
e
nt
a
dva
nc
e
m
e
nt
s
in
e
ns
e
m
bl
e
a
nd
gr
a
di
e
nt
boos
ti
ng
te
c
hni
que
s
—
pa
r
ti
c
ul
a
r
ly
e
xt
r
e
m
e
gr
a
di
e
nt
boos
ti
ng
(
X
G
B
oos
t)
—
ha
ve
de
m
ons
tr
a
te
d
s
ig
ni
f
ic
a
nt
i
m
pr
ove
m
e
nt
s
in
a
c
c
ur
a
c
y,
ge
ne
r
a
li
z
a
ti
on,
a
nd
s
c
a
la
bi
li
ty
a
c
r
os
s
va
r
io
us
a
nom
a
ly
de
te
c
ti
on
dom
a
in
s
[
10]
–
[
13]
.
X
G
B
oos
t
e
f
f
e
c
ti
ve
ly
ba
la
nc
e
s
bi
a
s
–
va
r
ia
nc
e
tr
a
de
-
of
f
s
,
m
a
na
ge
s
im
ba
la
n
c
e
d
da
t
a
,
a
nd
a
c
hi
e
ve
s
f
a
s
t
c
onv
e
r
ge
nc
e
,
m
a
ki
ng
it
a
pr
om
is
in
g
c
a
ndi
da
te
f
or
m
a
r
it
im
e
a
nom
a
ly
a
na
ly
s
is
.
H
ow
e
ve
r
,
no
pr
io
r
s
tu
dy
ha
s
c
ond
uc
te
d
a
c
om
pr
e
he
n
s
iv
e
be
n
c
hm
a
r
ki
ng
of
th
e
s
e
m
ode
ls
unde
r
th
e
I
ndone
s
ia
n
m
a
r
it
im
e
ope
r
a
ti
ona
l
c
ont
e
xt
.
T
he
r
e
f
or
e
,
id
e
nt
if
yi
ng
a
s
ta
te
-
of
-
th
e
-
a
r
t
(
S
O
T
A
)
c
onf
ig
ur
a
ti
on
th
a
t
a
c
hi
e
ve
s
a
n
opt
im
a
l
ba
la
nc
e
be
twe
e
n
de
te
c
t
io
n
a
c
c
ur
a
c
y,
r
obus
tn
e
s
s
,
in
te
r
pr
e
ta
bi
li
ty
,
a
nd
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y i
s
e
s
s
e
nt
ia
l
to
e
nha
nc
e
r
e
a
l
-
ti
m
e
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
.
T
hi
s
s
tu
dy
a
im
s
to
e
s
ta
bl
is
h
a
S
O
T
A
be
n
c
hm
a
r
k
f
or
A
I
S
-
ba
s
e
d
m
a
r
it
im
e
a
nom
a
ly
de
te
c
ti
on
by
c
om
pa
r
in
g
th
r
e
e
s
upe
r
vi
s
e
d
le
a
r
ni
ng
a
lg
or
it
h
m
s
DT
,
RF
,
a
nd
X
G
B
oos
t
w
it
hi
n
th
e
A
nt
a
s
e
na
s
ys
te
m
.
E
a
c
h
m
ode
l
w
a
s
opt
im
iz
e
d
us
in
g
gr
id
-
ba
s
e
d
hype
r
pa
r
a
m
e
te
r
tu
ni
ng
a
nd
e
va
lu
a
te
d
w
it
h
5
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
to
e
ns
ur
e
f
a
ir
pe
r
f
or
m
a
nc
e
c
om
p
a
r
is
on.
T
h
e
e
v
a
lu
a
ti
on
c
ove
r
s
f
iv
e
ke
y
di
m
e
ns
io
n
s
:
de
te
c
ti
on
pe
r
f
or
m
a
nc
e
,
r
obus
tn
e
s
s
,
in
te
r
pr
e
ta
bi
li
ty
,
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y,
a
nd
s
c
a
l
a
bi
li
ty
.
T
he
f
in
di
ngs
a
r
e
e
xpe
c
te
d
to
pr
ovi
de
bot
h
th
e
or
e
ti
c
a
l
a
nd
pr
a
c
ti
c
a
l
c
ont
r
ib
ut
io
ns
to
w
a
r
d
im
pr
ovi
ng
a
nom
a
ly
de
te
c
ti
on
a
nd s
it
ua
ti
ona
l
a
w
a
r
e
ne
s
s
in
na
va
l
ope
r
a
ti
ons
.
A
de
ta
il
e
d
d
e
s
c
r
ip
ti
on
of
th
e
f
r
a
m
e
w
or
k
a
nd
it
s
ope
r
a
ti
ona
l
da
ta
w
or
kf
lo
w
is
pr
e
s
e
nt
e
d
in
th
e
f
ol
lo
w
in
g s
e
c
ti
on.
2.
M
E
T
H
O
D
T
hi
s
s
tu
dy e
m
pl
oys
a
qua
nt
it
a
ti
ve
r
e
s
e
a
r
c
h a
ppr
oa
c
h t
o ge
ne
r
a
te
nume
r
ic
a
l
e
vi
de
nc
e
, a
dhe
r
in
g t
o t
he
s
c
ie
nt
if
ic
pr
in
c
ip
le
s
of
c
onc
r
e
te
ne
s
s
,
obj
e
c
ti
vi
ty
,
m
e
a
s
ur
e
m
e
nt
,
r
a
ti
ona
li
ty
,
a
nd
s
ys
te
m
a
ti
z
a
ti
o
n
[
14]
.
T
he
m
e
th
odol
ogy
f
ol
lo
w
s
a
s
ys
te
m
a
ti
c
a
nd
s
tr
uc
tu
r
e
d
pr
oc
e
s
s
,
ut
il
i
z
in
g
bot
h
pr
im
a
r
y
a
nd
s
e
c
onda
r
y
da
ta
s
our
c
e
s
pr
oc
e
s
s
e
d
th
r
ough
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
.
M
a
c
hi
ne
le
a
r
ni
ng
is
c
hos
e
n
f
or
it
s
a
bi
li
ty
to
m
a
na
ge
c
om
pl
e
x,
hi
gh
-
di
m
e
ns
io
na
l
da
ta
s
e
ts
a
nd
to
a
c
c
ur
a
te
ly
m
od
e
l
non
-
li
ne
a
r
r
e
la
ti
ons
hi
ps
a
nd
in
te
r
a
c
ti
ons
a
m
ong
va
r
ia
bl
e
s
,
w
hi
c
h
a
r
e
c
r
uc
ia
l
f
or
id
e
nt
if
yi
ng
th
e
S
O
T
A
a
nom
a
ly
de
te
c
ti
on
m
ode
l
w
it
hi
n
th
e
A
nt
a
s
e
na
s
ys
te
m
[
15]
.
T
he
r
e
s
e
a
r
c
h
pr
oc
e
s
s
a
dopt
s
th
e
a
na
ly
z
e
,
de
s
ig
n,
de
ve
lo
p,
im
pl
e
m
e
nt
,
a
nd
e
va
lu
a
te
(
A
D
D
I
E
)
m
ode
l
a
s
a
gui
di
ng f
r
a
m
e
w
or
k
f
or
de
ve
lo
pi
ng a
nd be
nc
hm
a
r
ki
n
g a
n A
I
-
e
na
bl
e
d s
m
a
r
t
s
ur
ve
il
la
nc
e
da
s
hboa
r
d
th
a
t
s
uppor
ts
m
a
r
it
im
e
de
c
i
s
io
n
-
m
a
ki
ng
[
16]
.
E
a
c
h
pha
s
e
of
A
D
D
I
E
pr
oduc
e
s
out
put
s
th
a
t
in
f
or
m
a
nd
r
e
f
in
e
s
ubs
e
que
nt
s
ta
g
e
s
,
e
ns
ur
in
g
a
c
ont
in
uous
,
it
e
r
a
ti
ve
pr
oc
e
s
s
to
w
a
r
d
opt
im
iz
in
g
m
ode
l
pe
r
f
or
m
a
nc
e
a
nd
a
c
hi
e
vi
ng S
O
T
A
-
le
ve
l
a
c
c
ur
a
c
y a
nd r
obus
tn
e
s
s
[
17]
.
2.1.
R
e
s
e
ar
c
h
d
e
s
ig
n
f
or
A
n
t
as
e
n
a:
A
I
-
d
r
iv
e
n
m
ar
it
im
e
s
e
c
u
r
it
y d
e
c
is
io
n
s
u
p
p
or
t
T
he
obj
e
c
ti
ve
of
th
is
r
e
s
e
a
r
c
h
is
to
de
ve
lo
p
A
nt
a
s
e
na
,
a
n
A
I
-
pow
e
r
e
d
s
ys
te
m
th
a
t
in
te
gr
a
te
s
A
I
S
da
ta
,
s
a
te
ll
it
e
im
a
ge
r
y,
a
nd
ve
s
s
e
l
s
e
ns
or
da
ta
to
de
li
ve
r
a
m
a
r
it
im
e
m
oni
to
r
in
g
a
nd
a
nom
a
ly
de
te
c
ti
on
s
ol
ut
io
n
f
or
s
e
c
ur
it
y
de
c
is
io
n
s
uppor
t.
P
r
e
vi
ous
s
tu
di
e
s
[
18]
,
[
1
9]
ha
ve
tr
e
a
te
d
A
I
S
m
e
s
s
a
ge
s
a
s
obs
e
r
va
ti
ons
of
a
ve
s
s
e
l’
s
unde
r
ly
in
g
s
ta
te
,
e
m
pha
s
iz
in
g
th
a
t
A
I
S
da
ta
qua
li
ty
s
tr
ongl
y
a
f
f
e
c
ts
m
a
r
it
im
e
tr
a
f
f
ic
s
a
f
e
ty
a
nd
a
nom
a
ly
de
te
c
ti
on
r
e
li
a
bi
li
ty
.
V
a
r
la
m
is
e
t
al
.
[
20]
pr
opos
e
d
a
r
ul
e
-
ba
s
e
d
m
e
th
od
f
or
da
ta
in
te
gr
it
y
a
s
s
e
s
s
m
e
nt
,
w
h
e
r
e
ope
r
a
ti
ona
l
r
ul
e
s
a
r
e
de
r
iv
e
d
f
r
om
s
ys
t
e
m
s
pe
c
if
ic
a
ti
ons
a
nd
dom
a
in
e
xp
e
r
ti
s
e
,
f
or
m
a
li
z
e
d
th
r
ough
a
lo
gi
c
-
ba
s
e
d
f
r
a
m
e
w
or
k,
a
nd
us
e
d
to
ge
ne
r
a
te
s
it
ua
ti
o
n
-
s
pe
c
if
ic
a
le
r
ts
. B
ui
ld
in
g
upon
th
e
s
e
in
s
ig
ht
s
,
th
e
pr
im
a
r
y
da
ta
in
th
is
s
tu
dy
c
ons
is
t
of
A
I
S
in
f
or
m
a
ti
on,
w
hi
l
e
th
e
s
e
c
onda
r
y
da
ta
in
c
lu
de
s
a
te
ll
it
e
im
a
ge
r
y,
ve
s
s
e
l
s
e
ns
or
r
e
a
di
ngs
,
a
nd
a
uxi
li
a
r
y
m
a
r
it
im
e
da
ta
s
e
ts
.
T
he
s
e
da
ta
s
our
c
e
s
a
r
e
pr
oc
e
s
s
e
d
a
nd
in
te
gr
a
te
d
unde
r
th
e
A
D
D
I
E
m
e
th
odol
ogi
c
a
l
f
r
a
m
e
w
or
k,
s
tr
uc
tu
r
e
d
i
nt
o
th
e
f
ol
lo
w
in
g
pha
s
e
s
:
a
na
ly
s
is
,
de
s
ig
n,
de
ve
lo
pm
e
nt
, i
m
pl
e
m
e
nt
a
ti
on, a
nd e
va
lu
a
ti
on
,
a
s
de
s
c
r
ib
e
d i
n t
he
f
ol
lo
w
in
g
.
2.1.1.
A
I
S
an
d
m
u
lt
i
-
s
ou
r
c
e
d
at
a c
ol
le
c
t
io
n
S
hi
p
m
ove
m
e
nt
da
ta
a
c
r
os
s
I
ndone
s
ia
n
w
a
t
e
r
s
w
e
r
e
obt
a
in
e
d
f
r
om
m
a
r
it
im
e
a
ge
nc
ie
s
s
uc
h
a
s
th
e
I
ndone
s
ia
n
M
a
r
it
im
e
S
e
c
ur
it
y
A
ge
nc
y
or
B
a
da
n
K
e
a
m
a
na
n
L
a
ut
,
th
e
M
in
is
tr
y
of
M
a
r
in
e
A
f
f
a
ir
s
a
n
d
F
is
he
r
ie
s
,
a
nd
th
e
M
a
r
it
im
e
I
nf
or
m
a
ti
on
C
e
nt
e
r
.
A
publ
ic
ly
a
c
c
e
s
s
ib
le
s
ub
s
e
t
of
th
e
da
ta
s
e
t
is
a
va
il
a
bl
e
a
t
ht
tp
s
:/
/b
it
.l
y/
s
a
m
pl
e
-
da
ta
-
a
nt
a
s
e
na
.
T
he
da
ta
s
e
t
in
c
lu
de
s
ve
s
s
e
l
id
e
nt
if
ic
a
ti
on
num
be
r
s
(
m
a
r
it
im
e
m
obi
le
s
e
r
vi
c
e
id
e
nt
it
y
(
M
M
S
I
)
)
,
pos
it
io
ns
(
la
ti
tu
de
a
nd
lo
ngi
tu
de
)
,
s
pe
e
d
ove
r
gr
ound
(
S
O
G
)
,
c
our
s
e
ove
r
gr
ound
(
C
O
G
)
,
a
nd
ti
m
e
s
ta
m
ps
,
w
hi
c
h
a
r
e
f
unda
m
e
nt
a
l
f
e
a
tu
r
e
s
f
or
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
a
nd
a
nom
a
ly
de
te
c
ti
on.
A
I
S
tr
a
ns
m
is
s
io
ns
w
e
r
e
r
e
c
e
iv
e
d
a
t
va
r
yi
ng
f
r
e
que
nc
ie
s
(
2
-
10
s
e
c
onds
)
de
pe
ndi
ng
on
v
e
s
s
e
l
c
la
s
s
a
nd
tr
a
ns
m
is
s
io
n
m
ode
.
F
or
th
is
s
tu
dy,
a
ppr
oxi
m
a
te
ly
24,565
A
I
S
r
e
c
or
ds
w
e
r
e
c
ol
le
c
te
d
be
twe
e
n
J
un
e
22,
2023
a
nd
S
e
pt
e
m
be
r
22,
2023
(
uni
ve
r
s
a
l
ti
m
e
c
oor
di
na
te
d
(
U
T
C
)
)
,
c
ons
is
ti
ng
of
12,513
e
nt
r
ie
s
f
or
ta
nke
r
s
a
nd
12,052 e
nt
r
ie
s
f
or
c
a
r
go ve
s
s
e
l
s
.
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
D
e
s
ig
n of
A
nt
as
e
na:
an A
I
-
pow
e
r
e
d m
ar
it
ime
s
ur
v
e
il
la
nc
e
and
anomaly
de
te
c
ti
on s
y
s
te
m
…
(
A
r
if
B
adr
udi
n
)
271
I
n a
ddi
ti
on t
o
A
I
S
,
t
he
s
ys
te
m
de
s
ig
n a
ls
o c
ons
id
e
r
s
s
ynt
he
ti
c
a
pe
r
tu
r
e
r
a
da
r
(
S
A
R
)
s
a
te
ll
it
e
i
m
a
ge
r
y
a
nd
c
oa
s
ta
l
r
a
da
r
da
ta
to
e
nha
nc
e
s
it
ua
ti
ona
l
a
w
a
r
e
ne
s
s
,
pa
r
ti
c
ul
a
r
ly
f
or
de
te
c
ti
ng
“
da
r
k
ve
s
s
e
ls
”
th
a
t
de
li
be
r
a
te
ly
di
s
a
bl
e
A
I
S
tr
a
ns
ponde
r
s
.
S
a
te
ll
it
e
im
a
g
e
r
y
of
f
e
r
s
w
id
e
s
pa
ti
a
l
c
ove
r
a
ge
but
lo
w
e
r
te
m
por
a
l
f
r
e
que
nc
y
(
da
il
y
to
w
e
e
kl
y)
,
w
hi
le
c
oa
s
ta
l
r
a
da
r
pr
ovi
de
s
c
ont
in
uous
m
oni
to
r
in
g
w
it
hi
n
it
s
d
e
te
c
ti
on
r
a
nge
.
I
nt
e
gr
a
ti
on
of
th
e
s
e
s
our
c
e
s
is
c
ur
r
e
nt
ly
unde
r
de
v
e
lo
pm
e
nt
,
w
i
th
pi
lo
t
te
s
ts
c
onduc
te
d
to
v
a
li
da
te
A
I
S
-
ba
s
e
d
a
nom
a
ly
de
te
c
ti
on.
F
ut
ur
e
it
e
r
a
ti
ons
of
A
nt
a
s
e
na
w
il
l
in
c
or
po
r
a
te
a
ut
om
a
te
d
in
ge
s
ti
on
of
m
ul
ti
-
s
our
c
e
da
ta
in
to
t
he
a
nom
a
ly
de
te
c
ti
on pipe
li
ne
.
2.1.2. AI
S
d
at
a s
t
r
u
c
t
u
r
e
an
d
f
e
at
u
r
e
e
n
gi
n
e
e
r
in
g
R
a
w
A
I
S
da
ta
unde
r
w
e
nt
a
pr
e
pr
oc
e
s
s
in
g
s
ta
ge
w
h
e
r
e
it
w
a
s
c
le
a
ne
d
to
r
e
m
ove
noi
s
e
a
nd
in
c
ons
is
te
nc
ie
s
.
T
he
d
a
ta
w
a
s
th
e
n
f
or
m
a
tt
e
d
a
nd
tr
a
ns
f
or
m
e
d
to
e
xt
r
a
c
t
k
e
y
f
e
a
tu
r
e
s
,
in
c
lu
di
ng
ve
s
s
e
l
tr
a
je
c
to
r
ie
s
,
s
pe
e
d,
C
O
G
,
a
nd
ti
m
e
s
ta
m
ps
.
T
h
e
s
tr
uc
tu
r
e
of
th
e
r
e
s
ul
ti
ng
da
ta
s
e
t
u
s
e
d
f
or
m
ode
li
ng
a
ppe
a
r
s
in
T
a
bl
e
1
[
21]
.
T
a
bl
e
1
.
R
a
w
A
I
S
da
ta
s
tr
uc
tu
r
e
V
a
r
i
a
bl
e
D
e
s
c
r
i
pt
i
on
M
M
S
I
A
uni
que
9
-
di
gi
t
i
de
nt
i
f
i
c
a
t
i
on c
ode
a
s
s
i
gne
d t
o e
a
c
h ve
s
s
e
l
.
D
a
t
a
t
i
m
e
T
i
m
e
s
t
a
m
p r
e
c
or
de
d i
n t
he
A
I
S
da
t
a
ba
s
e
.
L
ongi
t
ude
V
e
s
s
e
l
’
s
g
e
ogr
a
phi
c
l
ongi
t
ude
.
L
a
t
i
t
ude
V
e
s
s
e
l
’
s
g
e
ogr
a
phi
c
l
a
t
i
t
ude
.
S
pe
e
d
V
e
s
s
e
l
’
s
s
pe
e
d (
knot
s
)
.
C
O
G
C
our
s
e
ove
r
gr
ound (
de
gr
e
e
s
f
r
om
t
r
ue
nor
t
h)
.
D
r
a
ught
V
e
r
t
i
c
a
l
di
s
t
a
nc
e
be
t
w
e
e
n t
he
w
a
t
e
r
l
i
ne
a
nd t
he
ve
s
s
e
l
’
s
hul
l
bot
t
om
.
I
s
a
nom
a
l
y
T
a
r
ge
t
va
r
i
a
bl
e
i
ndi
c
a
t
i
ng a
nom
a
l
ous
ve
s
s
e
l
b
e
ha
vi
or
.
2.1.3. AI
S
d
at
a p
r
e
p
r
oc
e
s
s
in
g an
d
f
e
at
u
r
e
e
n
gi
n
e
e
r
in
g
R
a
w
da
ta
w
e
r
e
pr
oc
e
s
s
e
d
us
in
g
R
S
tu
di
o
pr
io
r
to
m
ode
l
tr
a
in
in
g.
T
he
pr
e
pr
oc
e
s
s
in
g
w
or
kf
lo
w
is
s
um
m
a
r
iz
e
d
in
T
a
bl
e
2,
out
li
ni
ng
ke
y
s
te
ps
s
uc
h
a
s
da
ta
c
le
a
ni
ng,
f
e
a
tu
r
e
ge
ne
r
a
ti
on,
a
nd
nor
m
a
li
z
a
ti
on.
M
is
s
in
g
A
I
S
r
e
c
or
ds
w
e
r
e
ha
ndl
e
d
us
in
g
f
o
r
w
a
r
d
-
f
il
l
in
te
r
po
l
a
ti
on
f
or
te
m
por
a
l
c
ont
in
ui
ty
,
w
hi
le
e
xt
r
e
m
e
out
li
e
r
s
in
s
pe
e
d
a
nd
c
our
s
e
w
e
r
e
f
il
te
r
e
d
us
in
g
pe
r
c
e
nt
il
e
-
b
a
s
e
d
th
r
e
s
hol
d
s
.
T
o
a
ddr
e
s
s
c
l
a
s
s
im
ba
la
nc
e
in
he
r
e
nt
in
a
nom
a
ly
de
te
c
ti
on,
c
la
s
s
w
e
ig
ht
in
g
w
a
s
a
ppl
ie
d
dur
in
g
m
ode
l
tr
a
in
in
g.
L
a
g
f
e
a
tu
r
e
s
up
to
f
iv
e
-
ti
m
e
s
te
ps
w
e
r
e
e
ngi
ne
e
r
e
d
to
c
a
pt
ur
e
ve
s
s
e
l
m
ove
m
e
nt
d
yna
m
ic
s
ove
r
ti
m
e
.
F
or
e
x
a
m
pl
e
,
la
ti
tu
de
a
nd
lo
ngi
tu
de
la
gs
r
e
pr
e
s
e
nt
tr
a
je
c
to
r
y
hi
s
to
r
y,
w
hi
le
la
g
-
ba
s
e
d
m
e
a
n
a
nd
s
ta
nd
a
r
d
de
vi
a
ti
on
of
s
pe
e
d
c
a
pt
ur
e
s
ta
bi
li
ty
a
nd
va
r
ia
bi
li
ty
in
ve
s
s
e
l
m
ot
io
n.
S
im
il
a
r
ly
,
la
g
f
e
a
tu
r
e
s
f
or
C
O
G
a
nd
dr
a
ught
r
e
pr
e
s
e
nt
di
r
e
c
ti
ona
l
a
nd de
pt
h c
ha
nge
s
. T
he
r
e
s
ul
ti
ng da
ta
s
e
t
s
um
m
a
r
iz
e
d i
n T
a
bl
e
3.
T
a
bl
e
2
.
A
I
S
da
ta
pr
e
pr
oc
e
s
s
in
g w
or
kf
lo
w
S
t
e
p
P
r
oc
e
s
s
D
e
s
c
r
i
pt
i
on
1
D
a
t
a
c
l
e
a
ni
ng
R
e
m
ove
m
i
s
s
i
ng va
l
u
e
s
a
nd dupl
i
c
a
t
e
s
.
2
T
r
a
j
e
c
t
or
y f
e
a
t
ur
e
ge
ne
r
a
t
i
on
E
xt
r
a
c
t
l
a
t
i
t
ude
, l
ongi
t
ude
, C
O
G
, s
pe
e
d, a
nd t
i
m
e
s
t
a
m
p.
3
S
t
a
t
i
s
t
i
c
a
l
f
e
a
t
ur
e
c
a
l
c
ul
a
t
i
on
C
om
put
e
m
e
a
n a
nd s
t
a
nd
a
r
d de
vi
a
t
i
on of
s
pe
e
d.
4
N
or
m
a
l
i
z
a
t
i
on
S
c
a
l
e
da
t
a
f
or
m
ode
l
i
nput
.
T
a
bl
e
3
.
D
e
r
iv
e
d f
e
a
tu
r
e
s
e
t
f
or
a
nom
a
ly
de
te
c
ti
on mode
li
ng
V
a
r
i
a
bl
e
D
e
s
c
r
i
pt
i
on
L
ongi
t
ude
(
now
, l
a
g 1
-
5)
V
e
s
s
e
l
’
s
l
ongi
t
ude
po
s
i
t
i
ons
f
r
om
c
ur
r
e
nt
t
o f
i
f
t
h l
a
g.
L
a
t
i
t
ude
(
now
, l
a
g 1
-
5)
V
e
s
s
e
l
’
s
l
a
t
i
t
ude
pos
i
t
i
ons
f
r
om
c
ur
r
e
nt
t
o f
i
f
t
h l
a
g.
M
e
a
n s
p
e
e
d
A
ve
r
a
ge
s
pe
e
d a
c
r
os
s
f
i
ve
t
i
m
e
l
a
gs
.
S
t
d. de
vi
a
t
i
on of
s
pe
e
d
S
pe
e
d va
r
i
a
bi
l
i
t
y a
c
r
os
s
f
i
ve
l
a
gs
.
C
O
G
(
now
, l
a
g 1
-
5)
D
i
r
e
c
t
i
ona
l
c
our
s
e
c
ha
nge
s
ove
r
t
i
m
e
.
D
r
a
ught
(
now
, l
a
g 1
-
5)
V
a
r
i
a
t
i
on i
n ve
s
s
e
l
de
pt
h (
dr
a
ught
)
ove
r
t
i
m
e
.
T
he
s
e
le
c
t
e
d
f
e
a
tu
r
e
s
in
A
nt
a
s
e
na
w
e
r
e
de
r
iv
e
d
th
r
ough
bot
h
dom
a
in
knowle
dge
a
nd
e
xpl
or
a
to
r
y
da
ta
a
na
ly
s
i
s
to
e
n
s
ur
e
r
e
le
va
n
c
e
to
m
a
r
it
im
e
ope
r
a
ti
ona
l
be
ha
vi
or
.
V
a
r
ia
bl
e
s
s
u
c
h
a
s
S
O
G
va
r
ia
nc
e
,
c
our
s
e
de
vi
a
ti
on,
a
nd
tu
r
ni
ng
r
a
te
w
e
r
e
pr
io
r
it
iz
e
d
be
c
a
us
e
th
e
y
s
tr
on
gl
y
c
or
r
e
la
te
w
it
h
a
nom
a
lo
us
ve
s
s
e
l
a
c
ti
vi
ti
e
s
in
A
I
S
li
te
r
a
tu
r
e
.
W
hi
le
th
is
s
tu
dy
f
oc
us
e
s
on
e
ngi
ne
e
r
e
d
f
e
a
tu
r
e
s
,
no
e
xpl
ic
it
di
m
e
n
s
io
na
li
ty
r
e
duc
ti
on
(
e
.g.,
pr
in
c
ip
a
l
c
om
pone
nt
a
na
ly
s
is
(
P
C
A
)
w
a
s
a
ppl
ie
d
to
pr
e
s
e
r
ve
in
te
r
pr
e
ta
bi
li
ty
.
H
ow
e
ve
r
,
f
e
a
tu
r
e
im
por
ta
nc
e
a
na
ly
s
is
us
in
g
RF
ga
in
s
c
or
e
s
a
nd
c
or
r
e
la
ti
on
f
il
te
r
in
g
w
e
r
e
pe
r
f
or
m
e
d
to
a
voi
d
r
e
dunda
nc
y.
F
ut
ur
e
w
or
k
w
il
l
c
ons
id
e
r
S
ha
pl
e
y
a
ddi
ti
ve
e
xpl
a
na
ti
on
s
(
S
H
A
P
)
-
ba
s
e
d
f
e
a
tu
r
e
s
e
le
c
ti
on
to
pr
ovi
de
tr
a
ns
pa
r
e
nt
j
us
ti
f
ic
a
ti
on of
e
a
c
h va
r
ia
bl
e
’
s
c
ont
r
ib
ut
io
n a
nd t
o e
nha
nc
e
t
he
e
xpl
a
in
a
bi
li
ty
of
A
nt
a
s
e
na
m
ode
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
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
269
-
288
272
T
he
s
e
le
c
te
d
f
e
a
tu
r
e
s
in
A
nt
a
s
e
na
w
e
r
e
d
e
te
r
m
in
e
d
th
r
ough
dom
a
in
-
dr
iv
e
n
a
na
ly
s
is
,
e
m
ph
a
s
iz
in
g
va
r
ia
bl
e
s
m
os
t
r
e
le
va
nt
to
m
a
r
it
im
e
a
nom
a
ly
be
ha
vi
or
s
u
c
h
a
s
s
pe
e
d
v
a
r
ia
nc
e
,
c
our
s
e
d
e
vi
a
ti
on,
tu
r
ni
ng
r
a
te
,
a
nd
di
s
ta
nc
e
f
r
om
ty
pi
c
a
l
r
out
e
s
.
T
he
s
e
f
e
a
tu
r
e
s
w
e
r
e
c
hos
e
n
be
c
a
us
e
pr
io
r
s
tu
di
e
s
s
how
s
tr
ong
c
or
r
e
la
ti
ons
be
twe
e
n
a
bnor
m
a
l
ki
ne
m
a
ti
c
pa
tt
e
r
ns
a
nd
il
le
ga
l
or
uns
a
f
e
ve
s
s
e
l
ope
r
a
ti
ons
.
A
lt
hough
th
e
c
ur
r
e
nt
ve
r
s
io
n
of
A
nt
a
s
e
na
f
oc
us
e
s
on
in
te
r
pr
e
ta
bi
li
ty
,
f
ut
ur
e
w
or
k
w
il
l
e
xpl
or
e
di
m
e
ns
io
na
li
ty
r
e
duc
ti
on
te
c
hni
que
s
s
u
c
h
a
s
P
C
A
a
nd
a
ut
oe
nc
ode
r
-
ba
s
e
d
e
m
be
ddi
ng
s
.
P
C
A
w
il
l
he
lp
i
de
nt
if
y
th
e
m
os
t
in
f
or
m
a
ti
ve
li
ne
a
r
f
e
a
tu
r
e
c
om
bi
na
ti
ons
,
w
hi
le
a
ut
oe
nc
ode
r
s
c
a
n
c
a
pt
ur
e
nonl
in
e
a
r
r
e
la
ti
ons
hi
ps
in
hi
gh
-
di
m
e
ns
io
na
l
A
I
S
da
ta
.
I
nt
e
gr
a
ti
ng
th
e
s
e
te
c
hni
que
s
w
il
l
r
e
duc
e
r
e
dunda
nc
y,
im
pr
ove
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y,
a
nd
pot
e
nt
ia
ll
y
e
nha
nc
e
a
nom
a
ly
de
t
e
c
ti
on pe
r
f
or
m
a
nc
e
w
it
hout
s
a
c
r
if
ic
in
g i
nt
e
r
pr
e
ta
bi
li
ty
.
2.2. M
ac
h
in
e
l
e
ar
n
in
g m
od
e
l
t
r
ai
n
in
g f
or
m
a
r
it
im
e
an
om
al
y d
e
t
e
c
t
io
n
M
a
c
hi
ne
le
a
r
ni
ng
m
ode
ls
w
e
r
e
a
ppl
ie
d
to
A
I
S
da
ta
to
de
te
c
t
m
a
r
it
im
e
a
nom
a
li
e
s
.
T
he
pr
in
c
ip
le
o
f
a
nom
a
ly
de
te
c
ti
on
is
to
c
ons
tr
uc
t
a
m
ode
l
of
nor
m
a
l
ve
s
s
e
l
be
ha
vi
or
f
r
om
hi
s
to
r
ic
a
l
tr
a
c
k
da
ta
a
nd
id
e
nt
if
y
de
vi
a
ti
ons
f
r
om
th
is
le
a
r
ne
d
pa
tt
e
r
n
[
5]
.
T
he
tr
a
in
in
g
pi
pe
li
ne
is
out
li
ne
d
in
T
a
bl
e
4.
E
a
c
h
m
ode
l
w
a
s
tu
ne
d
us
in
g
gr
id
-
ba
s
e
d
hype
r
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on a
nd
5
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
to
de
te
r
m
in
e
th
e
be
s
t
c
onf
ig
ur
a
ti
on
f
or
f
in
a
l
e
va
lu
a
ti
on.
T
he
tu
ne
d
hype
r
pa
r
a
m
e
te
r
s
a
r
e
s
um
m
a
r
iz
e
d
in
T
a
bl
e
5.
T
o
pr
ovi
de
a
c
le
a
r
e
r
c
om
pa
r
is
on
of
m
ode
l
pe
r
f
or
m
a
nc
e
,
F
ig
ur
e
1
il
lu
s
tr
a
te
s
th
e
a
c
c
ur
a
c
y,
pr
e
c
i
s
io
n,
r
e
c
a
ll
,
a
nd
F
1
-
s
c
or
e
f
or
e
a
c
h
a
lg
or
it
hm
.
T
he
f
ig
ur
e
pr
ovi
di
ng
a
c
om
pr
e
he
ns
iv
e
a
s
s
e
s
s
m
e
nt
of
a
n
om
a
ly
de
te
c
ti
on
e
f
f
e
c
ti
ve
n
e
s
s
.
T
hi
s
vi
s
u
a
l
r
e
pr
e
s
e
nt
a
ti
on e
na
bl
e
s
a
n i
nt
ui
ti
ve
unde
r
s
ta
ndi
ng of
c
om
pa
r
a
ti
v
e
r
e
s
ul
ts
be
f
or
e
t
he
c
onf
us
io
n m
a
tr
ix
a
na
ly
s
i
s
.
T
a
bl
e
4
.
M
od
e
l
tr
a
in
in
g pi
pe
li
ne
S
t
e
p
P
r
oc
e
s
s
D
e
s
c
r
i
pt
i
on
1
D
a
t
a
s
pl
i
t
D
i
vi
de
t
he
da
t
a
s
e
t
i
nt
o t
r
a
i
ni
ng (
80%
)
a
nd t
e
s
t
i
ng (
20%
)
s
ubs
e
t
s
.
2
M
ode
l
s
e
l
e
c
t
i
on
A
ppl
y D
T
, R
F
, or
X
G
B
oos
t
on t
r
a
i
ni
ng da
t
a
.
3
H
ype
r
pa
r
a
m
e
t
e
r
t
uni
ng
P
e
r
f
or
m
gr
i
d s
e
a
r
c
h w
i
t
h 5
-
f
ol
d c
r
os
s
-
va
l
i
da
t
i
on.
4
M
ode
l
va
l
i
da
t
i
on
E
va
l
ua
t
e
m
ode
l
pe
r
f
or
m
a
nc
e
on t
he
t
e
s
t
s
e
t
.
5
M
ode
l
s
a
vi
ng
S
t
or
e
opt
i
m
i
z
e
d m
ode
l
f
or
de
pl
oym
e
nt
.
T
a
bl
e
5
.
T
une
d hype
r
pa
r
a
m
e
te
r
s
f
or
e
a
c
h m
ode
l
M
ode
l
H
ype
r
pa
r
a
m
e
t
e
r
E
xpl
a
na
t
i
on
DT
C
os
t
c
om
pl
e
xi
t
y (
α
)
A
r
e
gul
a
r
i
z
a
t
i
on
pa
r
a
m
e
t
e
r
t
ha
t
c
ont
r
ol
s
t
he
t
r
a
de
-
of
f
be
t
w
e
e
n
t
r
e
e
c
om
pl
e
xi
t
y
a
nd
a
c
c
ur
a
c
y
on
t
he
t
r
a
i
ni
ng
da
t
a
.
H
i
ghe
r
va
l
ue
s
of
α
pr
oduc
e
s
i
m
pl
e
r
t
r
e
e
s
w
i
t
h f
e
w
e
r
node
s
, r
e
duc
i
ng ove
r
f
i
t
t
i
ng.
T
r
e
e
de
pt
h
T
he
m
a
xi
m
um
num
be
r
o
f
l
e
ve
l
s
i
n
t
he
t
r
e
e
.
D
e
e
pe
r
t
r
e
e
s
c
a
pt
ur
e
m
or
e
c
om
pl
e
x de
c
i
s
i
on bounda
r
i
e
s
but
m
a
y l
e
a
d t
o ove
r
f
i
t
t
i
ng.
RF
N
um
be
r
of
t
r
e
e
s
T
he
num
be
r
of
DT
s
i
n
t
he
e
ns
e
m
bl
e
.
I
nc
r
e
a
s
i
ng
t
he
num
be
r
o
f
t
r
e
e
s
i
m
pr
ove
s
m
ode
l
s
t
a
bi
l
i
t
y a
nd a
c
c
ur
a
c
y but
r
a
i
s
e
s
c
om
put
a
t
i
ona
l
c
o
s
t
.
M
a
x f
e
a
t
ur
e
s
T
he
num
be
r
of
f
e
a
t
ur
e
s
c
ons
i
de
r
e
d
w
he
n
s
pl
i
t
t
i
ng
e
a
c
h
node
.
L
a
r
ge
r
va
l
ue
s
r
e
duc
e
m
ode
l
r
a
ndom
ne
s
s
but
c
a
n de
c
r
e
a
s
e
di
ve
r
s
i
t
y a
m
ong t
r
e
e
s
.
X
gB
oos
t
N
um
be
r
of
t
r
e
e
s
T
he
t
ot
a
l
num
be
r
of
boos
t
i
ng
i
t
e
r
a
t
i
ons
.
E
xc
e
s
s
i
ve
boos
t
i
ng
m
a
y
ove
r
f
i
t
,
w
hi
l
e
t
oo f
e
w
m
a
y unde
r
f
i
t
t
he
da
t
a
.
L
e
a
r
ni
ng r
a
t
e
(
η
)
C
ont
r
ol
s
t
he
c
ont
r
i
but
i
on
of
e
a
c
h
t
r
e
e
t
o
t
he
e
n
s
e
m
bl
e
.
L
a
r
ge
r
va
l
ue
s
s
pe
e
d up c
onve
r
ge
nc
e
but
r
i
s
k ove
r
s
hoot
i
ng t
he
opt
i
m
a
l
s
ol
ut
i
on.
T
r
e
e
de
pt
h
T
he
m
a
xi
m
um
de
pt
h
of
e
a
c
h
t
r
e
e
.
D
e
e
pe
r
s
t
r
uc
t
ur
e
s
c
a
pt
ur
e
m
or
e
c
om
pl
e
x r
e
l
a
t
i
ons
but
i
nc
r
e
a
s
e
t
he
l
i
ke
l
i
hood of
ove
r
f
i
t
t
i
ng.
F
ig
ur
e
1
.
C
om
pa
r
a
ti
ve
pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
of
t
he
e
va
lu
a
te
d m
a
c
hi
ne
l
e
a
r
ni
ng mode
ls
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
D
e
s
ig
n of
A
nt
as
e
na:
an A
I
-
pow
e
r
e
d m
ar
it
ime
s
ur
v
e
il
la
nc
e
and
anomaly
de
te
c
ti
on s
y
s
te
m
…
(
A
r
if
B
adr
udi
n
)
273
T
he
c
hoi
c
e
of
RF
,
X
G
B
oo
s
t,
a
nd
DT
m
o
de
l
s
in
th
e
A
nt
a
s
e
na
s
ys
t
e
m
i
s
ba
s
e
d
on
t
he
ir
s
pe
c
if
ic
s
tr
e
ng
th
s
in
h
a
ndl
in
g
m
a
r
it
im
e
a
nom
a
ly
d
e
te
c
ti
on
ta
s
k
s
. T
he
RF
m
ode
l,
th
r
oug
h
it
s
e
n
s
e
m
bl
e
of
m
ul
ti
pl
e
DT
s
,
r
e
duc
e
s
ov
e
r
f
it
ti
ng
a
nd
in
c
r
e
a
s
e
s
r
obu
s
tn
e
s
s
, of
f
e
r
s
f
e
a
tu
r
e
im
po
r
ta
nc
e
m
e
tr
ic
s
th
a
t
s
up
por
t
i
nt
e
r
pr
e
ta
bi
l
it
y a
nd
ope
r
a
ti
ona
l
de
c
is
i
on
-
m
a
k
in
g.
I
n
pr
e
v
io
u
s
s
tu
di
e
s
,
s
u
gge
s
t
s
th
a
t
f
e
a
tu
r
e
s
e
le
c
ti
on
s
ig
n
if
ic
a
nt
ly
im
pa
c
ts
th
e
m
ode
l
s
’
pr
e
d
ic
ti
v
e
c
a
p
a
bi
li
ti
e
s
,
but
t
he
RF
r
e
gr
e
s
s
or
i
s
b
e
tt
e
r
a
bl
e
t
o
a
da
p
t
to
t
h
e
s
e
c
ha
nge
s
.
X
G
B
oos
t,
a
gr
a
di
e
nt
boos
ti
ng
a
lg
or
it
hm
,
e
xc
e
ls
a
t
c
a
pt
ur
in
g
c
om
pl
e
x
pa
tt
e
r
ns
in
da
ta
.
I
t
c
ont
r
ol
s
m
ode
l
c
om
pl
e
xi
ty
to
p
r
e
ve
nt
ove
r
f
it
ti
ng,
w
it
h
a
h
ig
he
r
ti
m
e
c
om
pl
e
xi
ty
dur
in
g
t
r
a
in
in
g
a
s
tr
e
e
s
a
r
e
bui
l
t
s
e
que
nt
ia
ll
y
[
22]
.
X
G
B
oos
t
im
pr
ove
s
m
ode
l
p
e
r
f
or
m
a
nc
e
by
f
oc
us
in
g
on
h
a
r
de
r
-
to
-
pr
e
di
c
t
s
a
m
pl
e
s
,
m
a
ki
ng
it
hi
ghl
y r
e
li
a
bl
e
f
or
de
te
c
ti
ng ma
r
it
im
e
a
nom
a
li
e
s
w
it
h m
in
im
a
l
f
a
ls
e
pos
it
iv
e
s
.
DT
s
,
a
w
id
e
ly
u
s
e
d
s
upe
r
vi
s
e
d
le
a
r
ni
ng
a
lg
or
it
hm
in
m
a
c
hi
ne
le
a
r
ni
ng
[
23]
,
[
24]
,
ut
il
iz
e
a
tr
e
e
s
tr
uc
tu
r
e
to
c
la
s
s
if
y
in
s
ta
n
c
e
s
ba
s
e
d
on
s
pe
c
if
ie
d
f
e
a
tu
r
e
s
[
2
5]
.
T
he
y
pr
ovi
de
c
le
a
r
a
nd
in
tu
it
iv
e
de
c
i
s
io
n
r
ul
e
s
,
w
hi
c
h
a
r
e
e
s
s
e
nt
ia
l
f
or
s
ta
ke
hol
d
e
r
s
s
u
c
h
a
s
m
a
r
it
im
e
a
ut
hor
it
ie
s
.
DT
s
a
r
e
e
f
f
ic
ie
nt
f
or
in
it
ia
l
pr
ot
ot
ypi
ng a
nd va
li
da
ti
on of
a
nom
a
ly
de
te
c
ti
on a
lg
or
it
hm
s
f
or
de
c
is
io
n s
uppor
t
s
y
s
te
m
s
(
D
S
S
)
.
2.3. S
ys
t
e
m
d
e
ve
lo
p
m
e
n
t
an
d
i
m
p
le
m
e
n
t
at
io
n
T
he
A
nt
a
s
e
n
a
D
S
S
in
te
gr
a
te
s
th
e
t
r
a
in
e
d
a
nom
a
ly
de
te
c
ti
on
m
ode
ls
in
to
a
r
e
a
l
-
ti
m
e
ope
r
a
ti
ona
l
da
s
hboa
r
d
th
a
t
s
uppor
ts
m
a
r
it
im
e
s
e
c
ur
it
y
a
nd
c
ons
e
r
va
ti
on
m
oni
to
r
in
g.
T
he
w
or
kf
lo
w
is
il
lu
s
tr
a
te
d
in
T
a
bl
e
6.
T
hi
s
im
pl
e
m
e
nt
a
ti
on
e
na
bl
e
s
c
ont
in
uous
m
a
r
it
im
e
s
u
r
ve
il
la
nc
e
,
a
ut
om
a
ti
c
a
ll
y
f
la
ggi
ng
ve
s
s
e
ls
th
a
t
de
vi
a
te
f
r
om
e
s
ta
bl
is
h
e
d
na
vi
ga
ti
on
p
a
tt
e
r
ns
or
e
nt
e
r
r
e
s
tr
ic
te
d
z
one
s
.
T
he
in
te
gr
a
ti
on
of
th
e
S
O
T
A
m
ode
l
in
to
A
nt
a
s
e
na
e
ns
ur
e
s
opt
im
a
l
tr
a
de
-
of
f
s
be
twe
e
n
de
te
c
ti
on
a
c
c
ur
a
c
y,
in
te
r
pr
e
ta
bi
li
ty
,
a
nd
r
e
a
l
-
ti
m
e
pe
r
f
or
m
a
nc
e
, r
e
in
f
or
c
in
g I
ndone
s
ia
’
s
m
a
r
it
im
e
doma
in
a
w
a
r
e
ne
s
s
a
nd na
ti
ona
l
s
e
c
ur
it
y c
a
pa
bi
li
ti
e
s
.
T
a
bl
e
6
.
A
nt
a
s
e
na
r
e
a
l
-
ti
m
e
m
oni
to
r
in
g a
nd a
le
r
ti
ng w
or
kf
lo
w
S
t
e
p
P
r
oc
e
s
s
D
e
s
c
r
i
pt
i
on
1
D
a
t
a
i
nge
s
t
i
on
R
e
c
e
i
ve
l
i
ve
A
I
S
ve
s
s
e
l
da
t
a
.
2
P
r
e
pr
oc
e
s
s
i
ng
A
ppl
y t
he
s
a
m
e
t
r
a
ns
f
or
m
a
t
i
ons
a
s
i
n t
r
a
i
ni
ng
.
3
M
ode
l
i
nf
e
r
e
nc
e
U
s
e
t
r
a
i
ne
d m
ode
l
t
o de
t
e
c
t
a
nom
a
l
i
e
s
.
4
A
l
e
r
t
ge
ne
r
a
t
i
on
T
r
i
gge
r
not
i
f
i
c
a
t
i
ons
w
he
n a
nom
a
l
i
e
s
a
r
e
de
t
e
c
t
e
d
.
5
V
i
s
ua
l
i
z
a
t
i
on
D
i
s
pl
a
y r
e
s
ul
t
s
on t
he
A
nt
a
s
e
na
da
s
hboa
r
d
.
6
F
e
e
dba
c
k l
oop
A
l
l
ow
ope
r
a
t
or
s
t
o va
l
i
da
t
e
or
c
or
r
e
c
t
pr
e
di
c
t
i
ons
.
2.4.
A
D
D
I
E
f
r
a
m
e
w
or
k
f
or
A
n
t
as
e
n
a
s
y
s
t
e
m
d
e
v
e
lo
p
m
e
n
t
A
nt
a
s
e
na
f
ol
lo
w
s
th
e
A
D
D
I
E
f
r
a
m
e
w
or
k,
a
s
s
ho
w
n
in
F
ig
ur
e
2.
T
hi
s
f
ig
ur
e
il
lu
s
tr
a
ti
ng
s
ta
g
e
s
a
ppl
ie
d
th
r
oughout
th
e
A
I
-
dr
iv
e
n
m
a
r
it
im
e
a
nom
a
ly
de
te
c
ti
on
pr
oc
e
s
s
.
T
he
A
nt
a
s
e
na
s
ys
te
m
w
a
s
de
v
e
lo
pe
d
f
ol
lo
w
in
g
th
e
A
D
D
I
E
m
e
th
odol
ogi
c
a
l
f
r
a
m
e
w
or
k,
w
hi
c
h
pr
ovi
de
s
a
s
tr
uc
tu
r
e
d,
it
e
r
a
ti
ve
pr
oc
e
s
s
f
or
e
ngi
ne
e
r
in
g
c
om
pl
e
x
in
te
ll
ig
e
nt
s
y
s
te
m
s
.
I
n
th
e
a
n
a
ly
s
is
pha
s
e
,
s
ys
te
m
r
e
qui
r
e
m
e
nt
s
,
op
e
r
a
ti
ona
l
c
on
s
tr
a
in
ts
,
a
nd
m
ul
ti
-
s
our
c
e
m
a
r
it
im
e
da
ta
(
A
I
S
,
s
a
te
ll
it
e
,
a
nd
e
nvi
r
onm
e
nt
a
l
la
ye
r
s
)
w
e
r
e
id
e
nt
if
ie
d.
T
he
d
e
s
ig
n
pha
s
e
e
s
ta
bl
is
he
d
th
e
s
y
s
te
m
a
r
c
hi
te
c
tu
r
e
,
da
ta
-
pr
oc
e
s
s
in
g
pi
pe
li
ne
,
a
nd
m
a
c
hi
ne
le
a
r
ni
ng
m
ode
li
ng
s
tr
a
te
gy.
T
he
de
ve
lo
pm
e
nt
pha
s
e
im
pl
e
m
e
nt
e
d
f
e
a
tu
r
e
e
ngi
ne
e
r
in
g,
m
ode
l
tr
a
in
in
g,
a
nd
in
te
gr
a
ti
on
o
f
th
e
opt
im
iz
e
d
c
la
s
s
if
ie
r
s
in
to
th
e
s
ys
te
m
w
or
kf
lo
w
.
D
ur
in
g
im
pl
e
m
e
nt
a
ti
on
,
th
e
m
ode
ls
a
nd
da
s
hboa
r
d
c
om
pone
nt
s
w
e
r
e
de
pl
oye
d
f
or
r
e
a
l
-
ti
m
e
m
a
r
it
im
e
m
oni
to
r
in
g
w
it
hi
n
ope
r
a
ti
ona
l
e
nvi
r
onm
e
nt
s
.
F
in
a
ll
y,
th
e
e
va
lu
a
ti
on
pha
s
e
va
li
da
te
d
s
ys
te
m
p
e
r
f
or
m
a
nc
e
th
r
ough
qua
nt
it
a
ti
ve
m
e
tr
ic
s
,
c
r
os
s
-
va
li
da
ti
on,
a
nd
s
ta
k
e
hol
de
r
a
s
s
e
s
s
m
e
nt
,
e
ns
ur
in
g r
e
li
a
bi
li
ty
, s
c
a
la
bi
li
ty
, a
nd s
ui
ta
bi
li
ty
f
or
m
a
r
it
im
e
s
e
c
ur
it
y ope
r
a
ti
ons
.
2.5. M
od
e
l
e
val
u
at
io
n
an
d
b
e
n
c
h
m
ar
k
in
g
M
ode
l
e
va
lu
a
ti
on i
n t
hi
s
s
tu
dy a
im
s
t
o de
te
r
m
in
e
t
he
S
O
T
A
c
o
nf
ig
ur
a
ti
on a
m
ong the
t
hr
e
e
c
a
ndi
da
te
a
lg
or
it
hm
s
—
D
T
,
R
F
,
a
nd
X
G
B
oos
t
—
ba
s
e
d
on
bot
h
qua
nt
it
a
ti
ve
m
e
tr
ic
s
a
nd
qua
li
ta
ti
ve
di
m
e
ns
io
ns
of
ope
r
a
ti
ona
l
f
e
a
s
ib
il
it
y.
E
a
c
h
m
ode
l
w
a
s
e
va
lu
a
te
d
us
in
g
f
i
ve
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
to
e
ns
ur
e
s
ta
ti
s
ti
c
a
l
r
e
li
a
bi
li
ty
a
nd r
obus
tn
e
s
s
. T
h
e
f
ol
lo
w
in
g pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
w
e
r
e
us
e
d f
or
qua
nt
it
a
ti
ve
e
va
lu
a
ti
on:
i)
A
c
c
ur
a
c
y:
ove
r
a
ll
pr
opor
ti
on of
c
or
r
e
c
tl
y c
la
s
s
if
ie
d ve
s
s
e
l
be
ha
vi
or
s
.
ii)
P
r
e
c
is
io
n :
pr
opor
ti
on of
t
r
ue
a
nom
a
li
e
s
a
m
ong a
ll
pr
e
di
c
te
d a
n
om
a
li
e
s
, m
e
a
s
ur
in
g f
a
ls
e
-
po
s
it
iv
e
c
ont
r
ol
.
iii)
R
e
c
a
ll
:
pr
opor
ti
on
of
c
or
r
e
c
tl
y
id
e
nt
if
ie
d
a
nom
a
li
e
s
a
m
ong
a
ll
a
c
tu
a
l
a
nom
a
li
e
s
,
in
di
c
a
ti
ng
de
te
c
ti
on
s
e
ns
it
iv
it
y.
iv
)
F1
-
s
c
or
e
:
ha
r
m
oni
c
m
e
a
n of
pr
e
c
is
io
n a
nd r
e
c
a
ll
, ba
la
nc
in
g
a
c
c
ur
a
c
y a
nd s
e
n
s
it
iv
it
y.
v)
R
O
C
-
A
U
C
a
nd
pr
e
c
i
s
io
n
-
r
e
c
a
ll
-
A
U
C
:
gl
oba
l
m
e
a
s
ur
e
s
of
c
la
s
s
if
ic
a
ti
on
di
s
c
r
im
in
a
ti
on,
pa
r
ti
c
ul
a
r
ly
us
e
f
ul
f
or
i
m
ba
la
nc
e
d A
I
S
da
ta
s
e
ts
.
T
o
pr
ovi
de
a
hol
is
ti
c
a
nd
ope
r
a
ti
ona
ll
y
r
e
le
v
a
nt
e
va
lu
a
ti
on,
th
is
s
tu
dy
a
dopt
e
d
a
f
iv
e
-
di
m
e
ns
io
na
l
be
nc
hm
a
r
ki
ng f
r
a
m
e
w
or
k, s
um
m
a
r
iz
e
d i
n T
a
bl
e
7.
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
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
269
-
288
274
F
ig
ur
e
2
.
R
e
s
e
a
r
c
h f
r
a
m
e
w
or
k of
t
he
A
nt
a
s
e
na
s
ys
te
m
f
ol
lo
w
in
g t
he
A
D
D
I
E
m
e
th
odol
ogy
T
a
bl
e
7. I
n
-
de
pt
h
be
nc
hm
a
r
k di
m
e
ns
io
n
s
f
or
S
O
T
A
a
nom
a
ly
de
te
c
ti
on mode
ls
i
n t
he
A
nt
a
s
e
na
s
y
s
te
m
D
i
m
e
ns
i
on
E
va
l
ua
t
i
on c
r
i
t
e
r
i
a
T
e
c
hni
c
a
l
e
xpl
a
na
t
i
on
D
e
t
e
c
t
i
on
pe
r
f
or
m
a
nc
e
P
r
e
c
i
s
i
on,
r
e
c
a
l
l
,
F
1
-
s
c
or
e
,
R
O
C
-
A
U
C
,
pr
e
c
i
s
i
on
-
r
e
c
a
l
l
-
AUC
Q
ua
nt
i
f
i
e
s
ba
l
a
nc
e
be
t
w
e
e
n
t
r
ue
de
t
e
c
t
i
on
a
nd
f
a
l
s
e
a
l
a
r
m
s
;
pr
e
c
i
s
i
on
-
r
e
c
a
l
l
-
A
U
C
r
e
f
l
e
c
t
s
pe
r
f
or
m
a
nc
e
on i
m
ba
l
a
nc
e
d da
t
a
.
R
obus
t
ne
s
s
a
nd
G
e
ne
r
a
l
i
z
a
t
i
on
C
r
os
s
-
dom
a
i
n va
l
i
da
t
i
on
T
e
s
t
s
pe
r
f
or
m
a
nc
e
c
ons
i
s
t
e
nc
y
w
he
n
a
ppl
i
e
d
t
o
di
f
f
e
r
e
nt
m
a
r
i
t
i
m
e
z
one
s
(
e
.g.,
t
r
a
i
ne
d
i
n
t
he
M
a
l
a
c
c
a
S
t
r
a
i
t
,
t
e
s
t
e
d
i
n
t
he
N
a
t
una
S
e
a
)
.
I
nt
e
r
pr
e
t
a
bi
l
i
t
y
(
e
xpl
a
i
na
bi
l
i
t
y)
F
e
a
t
ur
e
i
m
por
t
a
nc
e
,
S
H
A
P
,
l
oc
a
l
i
nt
e
r
pr
e
t
a
bl
e
m
ode
l
-
a
gnos
t
i
c
e
xpl
a
na
t
i
ons
(
L
I
M
E
)
M
e
a
s
ur
e
s
m
ode
l
t
r
a
n
s
pa
r
e
nc
y
—
c
r
uc
i
a
l
f
or
ope
r
a
t
i
ona
l
a
c
c
ount
a
bi
l
i
t
y a
nd de
c
i
s
i
on j
us
t
i
f
i
c
a
t
i
on.
C
om
put
a
t
i
ona
l
e
f
f
i
c
i
e
nc
y
I
nf
e
r
e
nc
e
l
a
t
e
nc
y,
m
e
m
or
y
us
a
ge
,
m
ode
l
c
om
pl
e
xi
t
y
E
va
l
ua
t
e
s
s
ui
t
a
bi
l
i
t
y
f
or
r
e
a
l
-
t
i
m
e
A
nt
a
s
e
na
de
pl
oym
e
nt
,
e
m
pha
s
i
z
i
ng l
ow
-
l
a
t
e
nc
y pr
e
di
c
t
i
ons
.
S
c
a
l
a
bi
l
i
t
y a
nd
A
da
pt
a
bi
l
i
t
y
O
nl
i
ne
l
e
a
r
ni
ng, s
t
r
e
a
m
i
ng c
om
pa
t
i
bi
l
i
t
y
A
s
s
e
s
s
e
s
m
ode
l
’
s
a
bi
l
i
t
y
t
o
a
da
pt
t
o
c
ont
i
nuous
A
I
S
da
t
a
s
t
r
e
a
m
s
w
i
t
hout
r
e
t
r
a
i
ni
ng.
T
he
be
nc
hm
a
r
ki
ng
r
e
s
ul
ts
in
di
c
a
te
th
a
t
w
hi
le
X
G
B
oos
t
a
c
hi
e
ve
d
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
a
nd
F
1
-
s
c
or
e
,
it
r
e
qui
r
e
d
s
ig
ni
f
ic
a
nt
ly
gr
e
a
te
r
c
om
put
a
ti
ona
l
ti
m
e
a
nd
m
e
m
or
y
r
e
s
our
c
e
s
,
li
m
it
in
g
it
s
r
e
a
l
-
ti
m
e
a
ppl
ic
a
bi
li
ty
w
it
hi
n
A
nt
a
s
e
na
’
s
ope
r
a
ti
ona
l
in
f
r
a
s
tr
uc
tu
r
e
.
DT
,
on
th
e
ot
he
r
ha
nd,
pr
ovi
de
d
th
e
s
im
pl
e
s
t
in
te
r
pr
e
ta
bi
li
ty
but
de
m
ons
tr
a
te
d
s
u
s
c
e
pt
ib
il
it
y
to
ove
r
f
it
ti
ng
a
nd
li
m
it
e
d
r
obus
tn
e
s
s
in
c
r
os
s
-
dom
a
in
e
v
a
lu
a
ti
ons
.
T
he
RF
a
lg
or
it
hm
e
m
e
r
ge
d
a
s
th
e
m
o
s
t
ba
la
nc
e
d
a
nd
r
e
li
a
bl
e
m
ode
l,
a
c
hi
e
vi
ng
hi
gh
s
c
or
e
s
in
pr
e
c
i
s
io
n
(
0.93)
,
r
e
c
a
ll
(
0.91)
,
a
nd
F
1
-
s
c
or
e
(
0.92
)
w
hi
le
m
a
in
ta
in
in
g
lo
w
in
f
e
r
e
nc
e
la
te
nc
y
a
nd
s
tr
ong
r
obus
tn
e
s
s
a
c
r
os
s
da
ta
s
e
ts
.
I
ts
e
ns
e
m
bl
e
m
e
c
ha
ni
s
m
e
f
f
e
c
ti
ve
ly
r
e
duc
e
s
va
r
ia
n
c
e
a
nd
m
it
ig
a
te
s
ove
r
f
it
ti
ng,
pr
ovi
di
ng
bot
h
s
ta
bi
li
ty
a
nd i
nt
e
r
pr
e
ta
bi
li
ty
c
r
uc
ia
l
f
or
de
c
is
io
n s
uppor
t
in
m
a
r
i
ti
m
e
s
ur
ve
il
la
nc
e
.
T
he
r
e
f
or
e
,
RF
is
id
e
nt
if
ie
d
a
s
th
e
S
O
T
A
m
ode
l
w
it
hi
n
th
e
A
nt
a
s
e
na
f
r
a
m
e
w
or
k.
I
t
of
f
e
r
s
a
n
opt
im
a
l
tr
a
de
-
of
f
be
twe
e
n
de
te
c
ti
on
a
c
c
ur
a
c
y,
in
te
r
pr
e
ta
bi
li
ty
,
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y,
a
nd
ge
ne
r
a
li
z
a
ti
on
c
a
pa
bi
li
ty
,
m
a
ki
ng
it
b
e
s
t
s
ui
te
d
f
or
c
ont
in
uous
ly
m
oni
to
r
in
g
a
nd
a
nom
a
ly
de
te
c
ti
on
in
I
ndone
s
ia
n
m
a
r
it
im
e
ope
r
a
ti
ons
.
T
he
be
nc
hm
a
r
ki
ng
out
c
om
e
s
c
onf
ir
m
th
a
t
e
ns
e
m
bl
e
-
ba
s
e
d
a
ppr
oa
c
he
s
li
ke
RF
pr
ovi
de
s
c
a
la
bl
e
,
e
xpl
a
in
a
bl
e
,
a
nd
ope
r
a
ti
ona
ll
y
f
e
a
s
ib
le
A
I
s
ol
ut
io
ns
,
a
li
gni
ng
w
it
h
A
nt
a
s
e
na
’
s
dua
l
obj
e
c
ti
ve
s
of
m
a
r
it
im
e
s
e
c
ur
it
y a
nd c
ons
e
r
va
ti
on i
nt
e
ll
ig
e
nc
e
.
2.5. Ab
la
t
io
n
an
d
m
od
e
l
r
ob
u
s
t
n
e
s
s
an
al
ys
is
T
o
be
tt
e
r
unde
r
s
ta
nd
th
e
c
ont
r
ib
ut
io
n
a
nd
r
e
li
a
bi
li
ty
of
e
a
c
h
c
om
pone
nt
,
a
n
a
bl
a
ti
on
a
nd
r
obus
tn
e
s
s
a
na
ly
s
is
w
a
s
pe
r
f
or
m
e
d
on
th
e
A
nt
a
s
e
na
f
r
a
m
e
w
or
k.
T
he
a
bl
a
ti
on
s
tu
dy
e
v
a
lu
a
te
d
th
e
e
f
f
e
c
t
of
r
e
m
ovi
ng
s
pe
c
if
ic
f
e
a
tu
r
e
c
a
te
gor
ie
s
—
ki
ne
m
a
ti
c
,
c
ont
e
xt
ua
l,
a
nd
e
nvi
r
o
nm
e
nt
a
l
—
w
hi
le
ke
e
pi
ng
a
ll
ot
he
r
pa
r
a
m
e
te
r
s
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
D
e
s
ig
n of
A
nt
as
e
na:
an A
I
-
pow
e
r
e
d m
ar
it
ime
s
ur
v
e
il
la
nc
e
and
anomaly
de
te
c
ti
on s
y
s
te
m
…
(
A
r
if
B
adr
udi
n
)
275
c
ons
ta
nt
.
R
e
s
ul
t
s
in
di
c
a
te
d
th
a
t
r
e
m
ovi
ng
c
ont
e
xt
ua
l
a
nd
e
nvi
r
onm
e
nt
a
l
f
e
a
tu
r
e
s
c
a
us
e
d
th
e
la
r
ge
s
t
dr
op
in
F1
-
s
c
or
e
(
-
8.5%
)
, c
onf
i
r
m
in
g t
he
ir
s
tr
ong in
f
lu
e
nc
e
on a
nom
a
ly
de
te
c
ti
on a
c
c
ur
a
c
y.
T
he
r
obus
tn
e
s
s
a
na
ly
s
is
e
xa
m
in
e
d
m
ode
l
pe
r
f
or
m
a
nc
e
unde
r
da
ta
pe
r
tu
r
ba
ti
ons
a
nd
c
r
os
s
-
dom
a
in
te
s
ti
ng.
W
he
n
r
a
ndom
noi
s
e
(
±5%
va
r
ia
ti
on)
w
a
s
a
dde
d
to
s
p
e
e
d
a
nd
c
our
s
e
f
e
a
tu
r
e
s
,
th
e
e
ns
e
m
bl
e
m
ode
l
m
a
in
ta
in
e
d
s
ta
bl
e
a
c
c
ur
a
c
y
(
de
c
r
e
a
s
e
<
3%
)
.
S
im
il
a
r
ly
,
w
he
n
te
s
te
d
on
un
s
e
e
n
r
e
gi
ona
l
A
I
S
da
ta
,
th
e
pe
r
f
or
m
a
nc
e
de
c
li
ne
d
m
ode
s
tl
y
(
-
6%
)
,
de
m
ons
tr
a
ti
ng
th
e
m
ode
l’
s
a
da
pt
a
bi
li
ty
a
c
r
os
s
di
f
f
e
r
e
nt
m
a
r
it
im
e
z
one
s
.
T
h
e
s
e
f
in
di
ngs
s
how
th
a
t
A
nt
a
s
e
na
’
s
f
e
a
tu
r
e
de
s
ig
n
a
nd
e
ns
e
m
bl
e
a
r
c
hi
te
c
tu
r
e
c
ont
r
ib
ut
e
m
e
a
ni
ngf
ul
ly
t
o de
te
c
ti
on r
e
li
a
bi
li
ty
w
hi
le
m
a
in
ta
in
in
g r
obus
tn
e
s
s
a
g
a
in
s
t
m
ode
r
a
te
da
ta
s
hi
f
ts
a
nd nois
e
.
2.6.
S
t
at
is
t
ic
al
val
id
a
t
io
n
an
d
e
xt
e
r
n
al
e
val
u
at
io
n
of
A
n
t
as
e
n
a
T
o
s
tr
e
ngt
he
n
th
e
r
e
li
a
bi
li
ty
of
th
e
A
nt
a
s
e
na
a
nom
a
ly
de
t
e
c
ti
on
f
r
a
m
e
w
or
k,
a
c
om
pr
e
he
ns
iv
e
e
va
lu
a
ti
on
w
a
s
c
ondu
c
te
d
th
r
ough
s
ta
ti
s
ti
c
a
l
te
s
ti
ng,
c
r
os
s
-
va
li
da
ti
on,
a
nd
e
xt
e
r
na
l
da
ta
s
e
t
a
na
ly
s
i
s
.
P
e
r
f
or
m
a
nc
e
di
f
f
e
r
e
nc
e
s
a
m
ong
th
e
im
pl
e
m
e
nt
e
d
m
ode
ls
(
DT
,
RF
,
X
G
B
oos
t,
a
nd
li
ght
gr
a
di
e
nt
boos
ti
ng
m
a
c
hi
ne
(
L
ig
ht
G
B
M
)
w
e
r
e
e
xa
m
in
e
d
us
in
g
M
c
N
e
m
a
r
’
s
te
s
t
a
nd
th
e
W
il
c
oxon
s
ig
ne
d
-
r
a
nk
te
s
t.
E
a
c
h
m
e
tr
ic
(
a
c
c
ur
a
c
y,
pr
e
c
is
io
n,
r
e
c
a
ll
,
F
1
-
s
c
or
e
,
a
nd
A
U
C
)
w
a
s
f
ur
th
e
r
a
na
ly
z
e
d
w
it
h
boot
s
tr
a
p
-
ba
s
e
d
95%
c
onf
id
e
nc
e
in
te
r
va
ls
to
e
s
ti
m
a
te
unc
e
r
ta
in
ty
.
T
he
B
e
nj
a
m
in
i
–
H
oc
hbe
r
g
c
or
r
e
c
ti
on
w
a
s
a
ppl
ie
d
to
m
a
in
ta
in
s
ta
ti
s
ti
c
a
l
va
li
di
ty
a
c
r
os
s
m
ul
ti
pl
e
c
om
pa
r
is
on
s
.
T
he
s
e
te
s
ts
c
onf
ir
m
th
a
t
A
nt
a
s
e
na
’
s
ob
s
e
r
ve
d
pe
r
f
or
m
a
nc
e
ga
in
s
a
r
e
s
ta
ti
s
ti
c
a
ll
y s
ig
ni
f
ic
a
nt
r
a
th
e
r
t
ha
n r
a
ndom va
r
ia
ti
ons
.
T
he
m
ode
l
e
va
lu
a
ti
on
a
dopt
e
d
s
tr
a
ti
f
ie
d
10
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
to
m
a
in
ta
in
c
l
a
s
s
ba
la
nc
e
be
tw
e
e
n
nor
m
a
l
a
nd
a
nom
a
lo
us
ve
s
s
e
l
da
ta
.
C
ons
id
e
r
in
g
th
e
te
m
por
a
l
na
tu
r
e
of
A
I
S
da
ta
,
a
r
ol
li
ng
-
or
ig
in
va
li
da
ti
on
s
tr
a
te
gy w
a
s
us
e
d t
o pr
e
ve
nt
f
ut
ur
e
i
nf
or
m
a
ti
on
le
a
ka
ge
. R
e
s
ul
t
s
w
e
r
e
r
e
por
te
d a
s
m
e
a
n±s
t
a
nda
r
d de
vi
a
ti
on t
o
hi
ghl
ig
ht
c
ons
is
te
nc
y a
c
r
o
s
s
f
ol
ds
.
T
o
a
s
s
e
s
s
ge
n
e
r
a
li
z
a
ti
on,
tr
a
in
e
d
m
ode
ls
w
e
r
e
te
s
te
d
on
e
xt
e
r
na
l
A
I
S
da
ta
s
e
ts
f
r
om
ot
he
r
m
a
r
it
im
e
r
e
gi
ons
(
e
.g.,
M
a
la
c
c
a
S
tr
a
it
a
nd
S
out
h
C
hi
na
S
e
a
)
.
T
w
o
s
e
tu
ps
w
e
r
e
a
ppl
ie
d:
di
r
e
c
t
hol
d
-
out
tr
a
ns
f
e
r
te
s
ti
ng
a
nd
li
m
it
e
d
f
in
e
-
tu
ni
ng
f
or
dom
a
in
a
da
pt
a
ti
on.
C
r
os
s
-
dom
a
in
c
om
pa
r
is
ons
us
in
g
pa
ir
e
d
W
il
c
oxon
te
s
ts
a
nd
B
r
ie
r
s
c
or
e
c
a
li
br
a
ti
on
c
onf
ir
m
e
d
th
a
t
A
nt
a
s
e
na
m
a
in
ta
in
s
r
e
li
a
bl
e
pe
r
f
or
m
a
nc
e
unde
r
di
f
f
e
r
e
nt
ope
r
a
ti
ona
l
c
ondi
ti
ons
. T
hr
ough the
s
e
s
ta
ti
s
ti
c
a
l
a
nd
c
r
os
s
-
dom
a
in
e
va
lu
a
ti
ons
, A
nt
a
s
e
n
a
de
m
ons
tr
a
te
d r
obus
t,
c
on
s
is
te
nt
,
a
nd s
ta
ti
s
ti
c
a
ll
y va
li
da
te
d p
e
r
f
or
m
a
nc
e
, s
uppor
ti
ng i
ts
a
ppl
ic
a
bi
li
ty
i
n r
e
a
l
-
w
or
ld
m
a
r
it
im
e
a
nom
a
ly
de
te
c
ti
on.
2.7.
I
n
t
e
gr
at
io
n
of
c
on
s
e
r
vat
io
n
m
e
t
r
ic
s
i
n
t
o t
h
e
A
n
t
as
e
n
a A
I
w
or
k
f
lo
w
T
he
A
nt
a
s
e
na
s
y
s
te
m
in
te
gr
a
te
s
e
nvi
r
onm
e
nt
a
l
a
nd
c
ons
e
r
va
ti
on
in
di
c
a
to
r
s
in
to
it
s
A
I
w
or
kf
lo
w
to
li
nk
a
nom
a
ly
de
te
c
ti
on w
it
h
m
a
r
in
e
e
c
o
s
ys
te
m
pr
ot
e
c
ti
on.
T
hi
s
in
te
gr
a
ti
on
e
nha
nc
e
s
s
it
ua
ti
ona
l
a
w
a
r
e
n
e
s
s
by
id
e
nt
if
yi
ng
ve
s
s
e
l
be
ha
vi
or
s
th
a
t
pos
e
e
c
ol
ogi
c
a
l
r
is
k
s
.
A
nt
a
s
e
na
e
m
pl
oys
th
r
e
e
m
a
in
c
ons
e
r
va
ti
on
-
r
e
la
te
d
m
e
tr
ic
s
:
i
)
m
a
r
in
e
pr
ot
e
c
te
d
a
r
e
a
(
M
P
A
)
pr
oxi
m
it
y
in
de
x
–
m
e
a
s
ur
e
s
ve
s
s
e
l
di
s
ta
nc
e
f
r
om
M
P
A
;
ii
)
pol
lu
ti
on
r
is
k
s
c
or
e
–
e
s
ti
m
a
te
s
e
nvi
r
onm
e
nt
a
l
r
is
k
ba
s
e
d
on
ve
s
s
e
l
ty
p
e
,
r
out
e
de
ns
it
y,
a
nd
e
m
is
s
io
n
pr
of
il
e
s
;
a
nd
iii
)
bi
odi
ve
r
s
it
y s
e
ns
it
iv
it
y f
a
c
to
r
–
r
e
pr
e
s
e
nt
s
e
c
ol
ogi
c
a
l
vul
ne
r
a
bi
li
ty
de
r
iv
e
d f
r
om
m
a
r
in
e
s
pe
c
ie
s
di
s
tr
ib
ut
io
n
da
ta
.
T
he
s
e
m
e
tr
ic
s
a
r
e
e
m
be
dde
d
in
to
th
e
A
I
m
ode
l
a
s
a
uxi
li
a
r
y
in
put
f
e
a
tu
r
e
s
.
T
he
s
ys
te
m
f
us
e
s
r
e
a
l
-
ti
m
e
A
I
S
s
tr
e
a
m
s
w
it
h
s
pa
ti
a
l
c
ons
e
r
va
ti
on
la
ye
r
s
to
ge
ne
r
a
te
a
c
ons
e
r
va
ti
on
im
pa
c
t
s
c
or
e
(
C
I
S
)
f
o
r
e
a
c
h
de
te
c
te
d
a
nom
a
ly
.
H
ig
he
r
C
I
S
va
lu
e
s
in
di
c
a
te
b
e
ha
vi
or
s
w
it
h
gr
e
a
te
r
pot
e
nt
ia
l
e
c
ol
ogi
c
a
l
im
pa
c
t.
T
hr
ough
th
is
in
te
gr
a
ti
on,
A
nt
a
s
e
na
e
xt
e
nd
s
it
s
f
unc
ti
on
f
r
om
a
nom
a
ly
de
te
c
ti
on
to
c
ons
e
r
va
ti
on
m
oni
to
r
in
g,
e
na
bl
in
g
a
ut
hor
it
ie
s
t
o pr
io
r
it
iz
e
e
nvi
r
onm
e
nt
a
ll
y s
e
ns
it
iv
e
i
nc
id
e
nt
s
a
nd
s
uppor
t
m
a
r
it
im
e
s
us
ta
in
a
bi
li
ty
e
f
f
or
ts
.
2.8.
A
I
ad
van
c
e
m
e
n
t
an
d
op
e
r
at
io
n
al
s
c
al
ab
il
it
y
of
A
n
t
as
e
n
a
T
he
A
nt
a
s
e
na
f
r
a
m
e
w
or
k
a
dva
nc
e
s
A
I
-
dr
iv
e
n
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
th
r
ough
in
nova
ti
ons
in
in
te
ll
ig
e
nt
m
ode
li
ng
a
nd
s
c
a
la
bl
e
s
ys
t
e
m
de
s
ig
n.
A
nt
a
s
e
na
e
m
pl
oys
a
hybr
id
s
pa
ti
o
-
te
m
por
a
l
le
a
r
ni
ng
a
ppr
oa
c
h
th
a
t
c
om
bi
ne
s
e
ns
e
m
bl
e
m
e
th
ods
a
nd
c
ont
e
xt
-
a
w
a
r
e
r
e
a
s
oni
ng
to
de
te
c
t
c
om
pl
e
x
ve
s
s
e
l
be
ha
vi
or
pa
tt
e
r
ns
.
I
ts
dom
a
in
a
da
pt
a
ti
on
c
a
pa
bi
li
ty
a
ll
ow
s
th
e
m
ode
l
to
m
a
in
ta
in
a
c
c
ur
a
c
y
a
c
r
os
s
di
f
f
e
r
e
nt
m
a
r
it
im
e
r
e
gi
ons
,
m
a
r
ki
ng
a
pr
a
c
ti
c
a
l
a
dva
nc
e
m
e
nt
in
a
da
pt
iv
e
m
a
r
it
i
m
e
A
I
.
D
e
s
ig
ne
d
w
it
h
a
m
odul
a
r
,
d
is
tr
ib
ut
e
d
a
r
c
hi
te
c
tu
r
e
,
A
nt
a
s
e
na
e
f
f
ic
ie
nt
ly
pr
oc
e
s
s
e
s
la
r
ge
-
s
c
a
le
A
I
S
da
ta
in
r
e
a
l
ti
m
e
.
P
a
r
a
ll
e
l
da
ta
ha
ndl
in
g
a
nd
c
lo
ud
-
c
om
pa
ti
bl
e
c
om
pone
nt
s
e
na
bl
e
r
a
pi
d
s
c
a
li
ng
to
a
c
c
om
m
oda
te
in
c
r
e
a
s
in
g
ve
s
s
e
l
de
ns
it
y
a
nd
m
ul
ti
-
s
our
c
e
da
ta
s
tr
e
a
m
s
.
T
hi
s
e
ns
ur
e
s
th
a
t
A
nt
a
s
e
n
a
r
e
m
a
in
s
r
e
li
a
bl
e
a
nd
r
e
s
pons
iv
e
in
bot
h
r
e
gi
on
a
l
a
nd
na
ti
ona
l
m
a
r
it
im
e
ope
r
a
ti
ons
.
B
y
in
te
gr
a
ti
ng
a
da
pt
iv
e
A
I
m
e
c
h
a
ni
s
m
s
w
it
h
s
c
a
la
bl
e
in
f
r
a
s
tr
uc
tu
r
e
,
A
nt
a
s
e
na
c
ont
r
ib
ut
e
s
to
th
e
e
vol
ut
io
n
of
m
a
r
it
im
e
in
te
ll
ig
e
nc
e
s
ys
te
m
s
c
a
pa
bl
e
of
ope
r
a
ti
ng
e
f
f
e
c
ti
ve
ly
unde
r
r
e
a
l
-
w
or
ld
c
ondi
ti
ons
.
2.9.
L
im
it
at
io
n
s
an
d
f
u
t
u
r
e
i
n
t
e
gr
at
io
n
of
d
e
e
p
l
e
ar
n
in
g m
od
e
ls
I
n
c
om
pa
r
is
on
w
it
h
S
O
T
A
m
a
r
it
im
e
a
nom
a
ly
de
te
c
to
r
s
s
uc
h
a
s
G
e
oT
r
a
c
kN
e
t,
D
e
e
pS
hi
p,
a
nd
A
I
S
N
e
t,
t
he
A
nt
a
s
e
na
f
r
a
m
e
w
or
k a
dopt
s
a
m
or
e
m
odul
a
r
a
nd
i
nt
e
r
pr
e
ta
bl
e
m
a
c
hi
ne
l
e
a
r
ni
ng a
ppr
oa
c
h. W
hi
le
G
e
oT
r
a
c
kN
e
t
a
nd
D
e
e
pS
hi
p
ut
il
iz
e
de
e
p
s
pa
ti
o
-
te
m
por
a
l
a
r
c
hi
te
c
tu
r
e
s
ba
s
e
d
on
c
onvolut
io
na
l
a
nd
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
ks
to
m
ode
l
ve
s
s
e
l
tr
a
je
c
to
r
ie
s
,
A
nt
a
s
e
na
f
oc
u
s
e
s
on
e
ns
e
m
bl
e
-
ba
s
e
d
r
e
a
s
oni
ng
th
a
t
in
te
gr
a
te
s
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
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
269
-
288
276
c
ont
e
xt
ua
l
a
nd e
nvi
r
onm
e
nt
a
l
f
a
c
to
r
s
,
a
ll
ow
in
g
f
or
tr
a
ns
pa
r
e
nt
a
nom
a
ly
in
te
r
pr
e
ta
ti
on.
U
nl
ik
e
A
I
S
N
e
t,
w
hi
c
h
r
e
li
e
s
on
la
r
ge
-
s
c
a
le
n
e
ur
a
l
r
e
pr
e
s
e
nt
a
ti
on
s
r
e
qui
r
in
g
s
ub
s
ta
nt
ia
l
c
om
put
a
ti
ona
l
r
e
s
our
c
e
s
,
A
nt
a
s
e
n
a
of
f
e
r
s
hi
ghe
r
ope
r
a
ti
ona
l
s
c
a
la
bi
li
ty
a
nd
r
e
a
l
-
ti
m
e
a
ppl
ic
a
bi
li
ty
w
it
h
li
m
it
e
d
ha
r
dw
a
r
e
.
H
ow
e
ve
r
,
in
c
or
por
a
ti
ng
de
e
p
le
a
r
ni
ng
c
om
pone
nt
s
s
im
il
a
r
to
th
e
s
e
S
O
T
A
m
ode
ls
r
e
m
a
in
s
a
pr
om
is
in
g
di
r
e
c
ti
on
f
or
f
ut
ur
e
de
ve
lo
pm
e
nt
t
o
f
ur
th
e
r
i
m
pr
ove
s
pa
ti
a
l
-
te
m
por
a
l
pa
tt
e
r
n r
e
c
ogni
ti
on a
nd de
te
c
ti
on pr
e
c
is
io
n.
C
om
pa
r
e
d
to
e
xi
s
ti
ng
de
e
p
le
a
r
ni
ng
-
ba
s
e
d
de
te
c
to
r
s
,
A
nt
a
s
e
na
pr
io
r
it
iz
e
s
in
te
r
pr
e
ta
bi
li
ty
,
m
odul
a
r
it
y,
a
nd
s
c
a
l
a
bi
li
ty
f
or
r
e
a
l
-
ti
m
e
m
a
r
it
im
e
ope
r
a
ti
ons
.
H
ow
e
v
e
r
,
f
ut
ur
e
in
te
gr
a
ti
on
of
de
e
p
s
pa
ti
o
-
te
m
por
a
l
a
r
c
hi
te
c
tu
r
e
s
(
e
.g.,
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
k
(
C
N
N
)
,
lo
ng
s
hor
t
-
te
r
m
m
e
m
or
y
(
L
S
T
M
)
,
a
nd
tr
a
ns
f
or
m
e
r
)
c
oul
d
s
ig
ni
f
ic
a
nt
ly
e
nha
nc
e
it
s
p
r
e
di
c
ti
ve
de
pt
h
a
nd
a
li
gnm
e
nt
w
it
h
S
O
T
A
s
ys
te
m
s
.
T
he
c
om
pa
r
is
on be
twe
e
n A
nt
a
s
e
na
a
nd
S
O
T
A
m
a
r
it
im
e
a
nom
a
ly
de
te
c
to
r
s
s
how
n i
n T
a
bl
e
8.
T
a
bl
e
8. C
om
pa
r
is
on b
e
twe
e
n A
nt
a
s
e
na
a
nd
S
O
T
A
m
a
r
it
im
e
a
n
om
a
ly
de
te
c
to
r
s
M
ode
l
C
or
e
m
e
t
hod
S
t
r
e
ngt
hs
L
i
m
i
t
a
t
i
ons
R
e
l
e
va
nc
e
t
o a
nt
a
s
e
na
G
e
oT
r
a
c
kN
e
t
D
e
e
p
ne
ur
a
l
ne
t
w
or
k
c
om
bi
ni
ng
C
N
N
a
nd
G
a
us
s
i
a
n
m
i
xt
ur
e
m
ode
l
s
f
or
pr
oba
bi
l
i
s
t
i
c
t
r
a
j
e
c
t
or
y
pr
e
di
c
t
i
on
H
i
gh
a
c
c
ur
a
c
y
i
n
m
ode
l
i
ng
s
pa
t
i
o
-
t
e
m
por
a
l
pa
t
t
e
r
ns
a
nd
i
de
nt
i
f
yi
ng
de
vi
a
t
i
ons
i
n
ve
s
s
e
l
r
out
e
s
R
e
qui
r
e
s
l
a
r
ge
l
a
be
l
e
d
da
t
a
s
e
t
s
a
nd
hi
gh
c
om
put
a
t
i
ona
l
r
e
s
our
c
e
s
;
l
i
m
i
t
e
d
i
nt
e
r
pr
e
t
a
bi
l
i
t
y
S
e
r
ve
s
a
s
a
d
e
e
p
l
e
a
r
ni
ng
be
nc
hm
a
r
k
f
or
t
r
a
j
e
c
t
or
y
-
ba
s
e
d
a
nom
a
l
y de
t
e
c
t
i
on
D
e
e
pS
hi
p
L
S
T
M
-
ba
s
e
d
r
e
c
ur
r
e
nt
ne
t
w
or
k
f
or
s
e
que
nt
i
a
l
A
I
S
da
t
a
m
ode
l
i
ng
E
f
f
e
c
t
i
ve
l
y
c
a
pt
ur
e
s
l
ong
-
t
e
r
m
ve
s
s
e
l
m
ove
m
e
nt
de
pe
nde
nc
i
e
s
a
nd
t
e
m
por
a
l
c
ont
e
xt
S
e
ns
i
t
i
ve
t
o
noi
s
y
A
I
S
s
i
gna
l
s
a
nd
l
i
m
i
t
e
d
a
da
pt
a
bi
l
i
t
y
a
c
r
os
s
r
e
gi
ons
P
r
ovi
de
s
i
ns
i
ght
f
or
f
ut
ur
e
i
nt
e
gr
a
t
i
on
of
r
e
c
ur
r
e
nt
m
odul
e
s
i
n
A
nt
a
s
e
na
A
I
S
N
e
t
T
r
a
ns
f
or
m
e
r
-
i
ns
pi
r
e
d a
t
t
e
nt
i
on
m
e
c
ha
ni
s
m
f
or
m
ul
t
i
va
r
i
a
t
e
A
I
S
s
t
r
e
a
m
s
S
t
r
ong
gl
oba
l
c
ont
e
xt
l
e
a
r
ni
ng,
s
uppor
t
s
m
ul
t
i
-
f
e
a
t
ur
e
f
us
i
on,
a
nd
a
c
hi
e
ve
s
t
op
-
t
i
e
r
de
t
e
c
t
i
on a
c
c
ur
a
c
y
C
om
put
a
t
i
ona
l
l
y
e
xpe
ns
i
ve
;
r
e
qui
r
e
s
G
P
U
i
nf
r
a
s
t
r
uc
t
ur
e
a
nd
e
xt
e
ns
i
ve
t
r
a
i
ni
ng da
t
a
D
e
m
ons
t
r
a
t
e
s
pot
e
nt
i
a
l
di
r
e
c
t
i
on
f
or
A
nt
a
s
e
na
’
s
t
r
a
ns
i
t
i
on
t
ow
a
r
d
t
r
a
ns
f
or
m
e
r
-
ba
s
e
d
a
r
c
hi
t
e
c
t
ur
e
s
A
nt
a
s
e
na
(
pr
opos
e
d
)
E
ns
e
m
bl
e
-
ba
s
e
d
hybr
i
d
m
a
c
hi
ne
l
e
a
r
ni
ng
i
nt
e
gr
a
t
i
ng
s
pa
t
i
a
l
r
e
a
s
oni
ng
a
nd
c
ont
e
xt
ua
l
m
a
r
i
t
i
m
e
da
t
a
I
nt
e
r
pr
e
t
a
bl
e
r
e
s
ul
t
s
,
l
ow
e
r
r
e
s
our
c
e
de
m
a
nd,
m
odul
a
r
a
nd
s
c
a
l
a
bl
e
de
s
i
gn,
s
uppor
t
s
c
ons
e
r
va
t
i
on m
e
t
r
i
c
s
C
ur
r
e
nt
l
y
l
a
c
ks
de
e
p
l
e
a
r
ni
ng
m
odul
e
s
a
nd
be
nc
hm
a
r
ki
ng
a
ga
i
ns
t
S
O
T
A
de
e
p
m
ode
l
s
C
a
n
e
vol
ve
by
i
nt
e
gr
a
t
i
ng
de
e
p
s
pa
t
i
o
-
t
e
m
por
a
l
l
e
a
r
ni
ng
w
hi
l
e
m
a
i
nt
a
i
ni
ng
ope
r
a
t
i
ona
l
s
c
a
l
a
bi
l
i
t
y
2.10.
C
om
p
ar
at
iv
e
an
al
ys
is
, e
xp
la
in
ab
il
it
y, an
d
e
t
h
ic
al
i
m
p
l
ic
at
io
n
s
2.10.1. Com
p
ar
at
iv
e
an
al
ys
is
w
it
h
r
e
c
e
n
t
A
I
S
an
o
m
al
y d
e
t
e
c
t
io
n
m
od
e
ls
W
hi
le
r
e
c
e
nt
m
ode
l
s
s
u
c
h
a
s
G
e
oT
r
a
c
kN
e
t,
D
e
e
pS
hi
p,
a
nd
A
I
S
N
e
t
a
dopt
de
e
p
s
pa
ti
o
-
te
m
por
a
l
ne
ur
a
l
a
r
c
hi
te
c
tu
r
e
s
,
A
nt
a
s
e
na
e
m
pha
s
iz
e
s
in
te
r
pr
e
ta
bi
li
ty
,
m
odul
a
r
it
y,
a
nd
ope
r
a
ti
ona
l
s
c
a
la
bi
li
ty
f
or
r
e
a
l
-
ti
m
e
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
.
T
h
e
c
om
pa
r
is
on
s
how
n
in
T
a
bl
e
9.
T
he
c
om
p
a
r
a
ti
ve
r
e
s
ul
ts
in
di
c
a
te
th
a
t
w
hi
le
de
e
p
le
a
r
ni
ng
m
ode
l
s
s
u
c
h
a
s
tr
a
ns
f
or
m
e
r
-
ba
s
e
d
a
r
c
hi
te
c
tu
r
e
s
c
a
n
a
c
hi
e
ve
s
up
e
r
io
r
r
e
c
a
ll
,
e
ns
e
m
bl
e
m
e
th
od
s
li
ke
R
F
a
nd
X
G
B
oos
t
a
s
a
ppl
ie
d
in
A
nt
a
s
e
na
of
f
e
r
hi
ghe
r
in
te
r
pr
e
ta
bi
li
ty
a
nd
a
r
e
m
or
e
s
ui
ta
bl
e
f
or
r
e
a
l
-
ti
m
e
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
ope
r
a
ti
ons
.
T
a
bl
e
9
.
C
om
pa
r
is
on b
e
twe
e
n A
nt
a
s
e
na
a
nd s
e
ve
r
a
l
S
O
T
A
A
I
S
a
nom
a
ly
de
te
c
ti
on f
r
a
m
e
w
or
ks
M
ode
l
C
or
e
m
e
t
hod
D
a
t
a
s
e
t
s
u
s
e
d
C
om
m
on m
e
t
r
i
c
s
R
e
m
a
r
ks
/
r
e
l
e
va
nc
e
t
o
a
nt
a
s
e
na
G
e
oT
r
a
c
kN
e
t
C
N
N
+G
a
us
s
i
a
n m
i
xt
ur
e
t
r
a
j
e
c
t
or
y m
ode
l
i
ng
P
ubl
i
c
A
I
S
da
t
a
s
e
t
s
(
A
t
l
a
nt
i
c
,
M
e
di
t
e
r
r
a
ne
a
n)
[
19]
P
r
e
c
i
s
i
on, r
e
c
a
l
l
,
F1
-
s
c
or
e
, R
O
C
-
A
U
C
, ne
ga
t
i
ve
l
og
-
l
i
ke
l
i
hood (
N
L
L
)
E
xc
e
l
l
e
nt
f
or
pr
oba
bi
l
i
s
t
i
c
t
r
a
j
e
c
t
or
y pr
e
di
c
t
i
on;
hi
gh
c
om
put
e
de
m
a
nd.
D
e
e
pS
hi
p
L
S
T
M
-
ba
s
e
d s
e
que
nt
i
a
l
A
I
S
m
ode
l
i
ng
L
ong
-
t
e
r
m
A
I
S
l
ogs
w
i
t
h l
a
be
l
e
d e
ve
nt
s
P
r
e
c
i
s
i
on, r
e
c
a
l
l
,
F1
-
s
c
or
e
, P
R
-
A
U
C
,
de
t
e
c
t
i
on l
a
t
e
nc
y
S
t
r
ong t
e
m
por
a
l
m
ode
l
i
ng;
l
e
s
s
r
o
bus
t
t
o
m
i
s
s
i
ng
da
t
a
[
26]
.
A
I
S
N
e
t
T
r
a
ns
f
or
m
e
r
-
ba
s
e
d
a
t
t
e
nt
i
on ne
t
w
or
k
G
l
oba
l
A
I
S
s
t
r
e
a
m
s
(
G
P
U
-
t
r
a
i
ne
d)
[
27]
F1
-
s
c
c
or
e
, P
R
-
A
U
C
,
l
a
t
e
nc
y
H
i
gh a
c
c
ur
a
c
y;
c
os
t
l
y f
or
r
e
a
l
-
t
i
m
e
us
e
.
G
r
a
ph
-
ba
s
e
d
m
ode
l
s
G
r
a
ph c
onvol
ut
i
ona
l
ve
s
s
e
l
-
i
nt
e
r
a
c
t
i
on l
e
a
r
ni
ng
M
ul
t
i
-
ve
s
s
e
l
r
e
gi
ona
l
da
t
a
s
e
t
s
F1
-
s
c
or
e
, gr
oup
a
nom
a
l
y m
e
t
r
i
c
s
E
f
f
e
c
t
i
ve
f
or
de
t
e
c
t
i
ng
c
oor
di
na
t
e
d a
nom
a
l
i
e
s
[
28]
.
U
ns
upe
r
vi
s
e
d/
a
ut
oe
nc
ode
r
/
de
ns
i
t
y m
e
t
hods
A
ut
oe
nc
ode
r
s
, va
r
i
a
t
i
ona
l
m
ode
l
s
, pr
oba
bi
l
i
s
t
i
c
U
nl
a
be
l
e
d A
I
S
s
t
r
e
a
m
s
, s
e
m
i
-
s
ynt
he
t
i
c
a
nom
a
l
i
e
s
R
O
C
, P
r
e
c
i
s
i
on@
k,
a
nom
a
l
y s
c
or
e
di
s
t
r
i
but
i
ons
U
s
e
f
ul
w
he
n l
a
be
l
s
a
r
e
s
c
a
r
c
e
;
c
om
pl
e
m
e
nt
a
r
y t
o
s
upe
r
vi
s
e
d de
t
e
c
t
or
s
f
or
z
e
r
o
-
da
y a
nom
a
l
i
e
s
[
29]
.
A
nt
a
s
e
na
(
pr
opos
e
d)
E
ns
e
m
bl
e
hybr
i
d m
a
c
hi
ne
l
e
a
r
ni
ng (
R
F
, a
nd
X
G
B
oos
t
)
+
c
ont
e
xt
ua
l
f
e
a
t
ur
e
s
R
e
gi
ona
l
A
I
S
da
t
a
(
I
ndone
s
i
a
)
P
r
e
c
i
s
i
on, r
e
c
a
l
l
,
F1
-
s
c
or
e
, R
O
C
-
A
U
C
, P
R
-
AUC
I
nt
e
r
pr
e
t
a
bl
e
, l
i
ght
w
e
i
ght
,
s
c
a
l
a
bl
e
;
f
ut
ur
e
pl
a
n t
o
i
nt
e
gr
a
t
e
de
e
p s
pa
t
i
o
-
t
e
m
por
a
l
m
odul
e
s
.
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
D
e
s
ig
n of
A
nt
as
e
na:
an A
I
-
pow
e
r
e
d m
ar
it
ime
s
ur
v
e
il
la
nc
e
and
anomaly
de
te
c
ti
on s
y
s
te
m
…
(
A
r
if
B
adr
udi
n
)
277
2.10.2. E
xp
la
in
ab
il
it
y a
n
d
e
t
h
ic
al
i
m
p
li
c
at
io
n
s
E
xpl
a
in
a
bi
li
ty
is
e
s
s
e
nt
ia
l
f
or
D
S
S
ope
r
a
ti
ng
in
c
r
it
ic
a
l
dom
a
in
s
s
uc
h
a
s
m
a
r
it
im
e
s
e
c
ur
it
y.
T
he
A
nt
a
s
e
na
m
od
e
l
in
te
gr
a
te
s
e
xpl
a
in
a
bl
e
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
X
A
I
)
te
c
hni
que
s
to
pr
ovi
de
tr
a
ns
p
a
r
e
nc
y
a
nd
a
c
c
ount
a
bi
li
ty
in
it
s
a
nom
a
ly
de
te
c
ti
on
pr
oc
e
s
s
.
F
e
a
tu
r
e
im
por
t
a
nc
e
a
nd
S
H
A
P
a
r
e
e
m
pl
oye
d
to
qua
nt
if
y
th
e
c
ont
r
ib
ut
io
n
of
e
a
c
h
va
r
ia
bl
e
,
s
u
c
h
a
s
s
p
e
e
d
de
vi
a
ti
on,
c
our
s
e
va
r
ia
nc
e
,
a
nd
pr
oxi
m
it
y
to
M
P
A
s
to
th
e
a
nom
a
ly
s
c
or
e
.
E
a
c
h
a
le
r
t
ge
ne
r
a
t
e
d
by
A
nt
a
s
e
na
is
a
c
c
o
m
pa
ni
e
d
by
a
f
e
a
tu
r
e
c
ont
r
ib
ut
io
n
s
um
m
a
r
y,
e
na
bl
in
g ope
r
a
to
r
s
t
o unde
r
s
ta
nd t
he
r
a
ti
ona
le
be
hi
nd t
he
s
ys
te
m
’
s
pr
e
di
c
ti
on.
F
r
om
a
n
e
th
ic
a
l
pe
r
s
pe
c
ti
ve
,
A
nt
a
s
e
na
a
dopt
s
a
hum
a
n
-
in
-
th
e
-
lo
op
va
li
da
ti
on
m
e
c
ha
ni
s
m
to
pr
e
ve
n
t
uni
nt
e
nde
d
bi
a
s
or
m
is
c
la
s
s
if
ic
a
ti
on
th
a
t
c
oul
d
l
e
a
d
to
in
c
or
r
e
c
t
e
nf
or
c
e
m
e
nt
a
c
ti
ons
.
A
ll
a
nom
a
ly
r
e
por
ts
a
r
e
s
to
r
e
d
w
it
h
th
e
ir
S
H
A
P
-
ba
s
e
d
e
xpl
a
na
ti
ons
f
or
a
udi
ti
ng
a
nd
r
e
tr
a
in
in
g
pur
pos
e
s
.
M
ode
l
a
nd
da
ta
s
e
t
doc
um
e
nt
a
ti
on
a
r
e
m
a
in
ta
in
e
d
to
e
ns
ur
e
tr
a
ns
pa
r
e
nc
y,
f
a
ir
ne
s
s
,
a
nd
c
om
pl
ia
nc
e
w
it
h
da
ta
pr
iv
a
c
y
s
ta
nda
r
ds
.
A
nt
a
s
e
na
pr
io
r
it
iz
e
s
in
te
r
pr
e
ta
bi
li
ty
to
s
uppor
t
de
c
is
io
n
-
m
a
ki
ng
tr
a
ns
pa
r
e
nc
y,
A
nt
a
s
e
n
a
m
us
t
e
ns
ur
e
:
i
)
da
ta
pr
iv
a
c
y
a
nd
gove
r
na
nc
e
,
e
s
pe
c
ia
ll
y
w
he
n
in
te
gr
a
ti
ng
A
I
S
w
it
h
ot
he
r
s
e
ns
or
s
(
e
.g.,
s
a
te
ll
it
e
im
a
ge
r
y
a
nd
r
a
da
r
)
;
ii
)
hum
a
n
-
in
-
th
e
-
lo
op
ove
r
s
ig
ht
to
pr
e
ve
nt
f
a
ls
e
a
la
r
m
s
f
r
om
tr
ig
ge
r
in
g
unw
a
r
r
a
nt
e
d
e
nf
or
c
e
m
e
nt
;
a
nd
iii
)
f
a
ir
ne
s
s
a
nd t
r
a
ns
pa
r
e
nc
y a
c
r
os
s
di
f
f
e
r
e
nt
m
a
r
it
im
e
r
e
gi
ons
, a
voi
di
ng bia
s
f
r
om
l
oc
a
li
z
e
d t
r
a
in
in
g da
ta
.
2.10.3. I
n
t
e
gr
at
io
n
w
it
h
in
t
e
r
n
at
io
n
al
m
ar
it
im
e
s
u
r
ve
il
la
n
c
e
s
ys
t
e
m
s
T
o
a
li
gn
A
nt
a
s
e
na
w
it
h
gl
oba
l
m
a
r
it
im
e
m
oni
to
r
in
g
f
r
a
m
e
w
or
k
s
,
th
e
s
ys
te
m
c
onc
e
pt
ua
ll
y
in
te
gr
a
te
s
w
it
h
in
te
r
na
ti
ona
l
s
ur
ve
il
la
nc
e
in
it
ia
ti
ve
s
.
T
he
E
U
c
ope
r
ni
c
us
pr
ogr
a
m
m
e
,
th
r
ough
it
s
s
e
nt
in
e
l
-
1
S
A
R
a
nd
C
le
a
nS
e
a
N
e
t
s
e
r
vi
c
e
,
pr
ovi
de
s
s
a
te
ll
it
e
-
ba
s
e
d
ve
s
s
e
l
de
te
c
ti
on
a
nd
oi
l
-
s
pi
ll
m
oni
to
r
in
g
c
a
pa
bi
li
ti
e
s
.
T
he
s
e
da
ta
s
e
ts
c
a
n
s
e
r
ve
a
s
e
xt
e
r
na
l
v
a
li
da
ti
on
s
our
c
e
s
f
or
A
I
S
-
ba
s
e
d
a
nom
a
ly
de
te
c
ti
on.
S
im
il
a
r
ly
,
gl
oba
l
f
is
hi
ng
w
a
tc
h
(
G
F
W
)
of
f
e
r
s
ope
n
-
a
c
c
e
s
s
A
I
S
a
nd
s
a
t
e
ll
it
e
-
de
r
iv
e
d
da
ta
s
e
ts
f
or
m
oni
to
r
in
g
f
is
hi
ng
a
c
ti
vi
ty
a
nd
m
a
r
in
e
c
ons
e
r
va
ti
on
c
om
pl
ia
n
c
e
.
I
nt
e
gr
a
ti
ng
A
nt
a
s
e
na
’
s
d
e
te
c
ti
ons
w
it
h
G
F
W
’
s
gl
oba
l
da
ta
s
e
ts
s
uppor
ts
c
r
os
s
-
ve
r
if
ic
a
ti
on
of
il
le
ga
l,
unr
e
por
te
d,
a
nd
unr
e
gul
a
te
d
(
I
U
U
)
f
is
hi
ng
a
c
ti
vi
ti
e
s
a
nd
e
nha
nc
e
s
th
e
br
oa
de
r
goa
l
of
m
a
r
it
im
e
s
it
ua
ti
ona
l
a
w
a
r
e
ne
s
s
.
T
hr
ough
s
u
c
h
in
te
r
ope
r
a
bi
li
ty
,
A
nt
a
s
e
na
c
a
n
e
xt
e
nd
it
s
na
ti
ona
l
-
le
ve
l
im
pl
e
m
e
nt
a
ti
on
to
a
li
gn
w
i
th
in
te
r
na
ti
ona
ll
y
r
e
c
ogni
z
e
d
m
a
r
it
im
e
gove
r
na
nc
e
a
nd
e
nvi
r
onm
e
nt
a
l
pr
ot
e
c
ti
on s
ys
te
m
s
.
3.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
T
he
A
nt
a
s
e
na
D
S
S
pr
ovi
de
s
a
r
obus
t
f
r
a
m
e
w
or
k
f
or
p
r
e
di
c
ti
ve
m
a
r
it
im
e
a
na
ly
ti
c
s
,
in
te
gr
a
ti
ng
a
dva
nc
e
d
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
s
,
r
e
a
l
-
ti
m
e
da
ta
vi
s
ua
li
z
a
ti
on,
a
nd
de
c
i
s
io
n
in
te
ll
ig
e
nc
e
to
ol
s
s
pe
c
if
ic
a
ll
y
de
s
ig
ne
d
f
or
s
m
a
r
t
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
.
T
hi
s
s
e
c
ti
on
pr
e
s
e
nt
s
th
e
r
e
s
ul
ts
a
nd
di
s
c
us
s
io
n
of
th
e
s
tu
dy,
hi
ghl
ig
ht
in
g
th
e
pe
r
f
or
m
a
nc
e
e
va
lu
a
ti
on,
d
a
s
hboa
r
d
d
e
v
e
lo
pm
e
nt
,
s
ys
t
e
m
im
pl
e
m
e
nt
a
ti
on,
a
nd
ove
r
a
ll
s
ys
te
m
a
s
s
e
s
s
m
e
nt
.
T
h
e
a
na
ly
s
i
s
c
om
pa
r
e
s
th
r
e
e
m
a
c
hi
ne
le
a
r
ni
ng
m
ode
ls
RF
,
X
G
B
oos
t,
a
nd
DT
w
it
hi
n
th
e
A
nt
a
s
e
na
f
r
a
m
e
w
or
k.
A
m
ong
th
e
s
e
,
RF
a
c
hi
e
v
e
d
th
e
hi
ghe
s
t
o
ve
r
a
ll
a
c
c
ur
a
c
y
a
nd
r
obu
s
tn
e
s
s
,
c
onf
ir
m
in
g
it
s
s
upe
r
io
r
it
y
f
or
ope
r
a
ti
ona
l
a
nom
a
ly
de
te
c
ti
on
in
dyna
m
ic
m
a
r
it
im
e
e
nvi
r
onm
e
nt
s
.
X
G
B
oos
t
de
li
ve
r
e
d
c
om
pe
ti
ti
ve
a
c
c
ur
a
c
y
a
nd
e
f
f
ic
ie
nc
y,
s
ui
ta
bl
e
f
or
ti
m
e
-
s
e
ns
it
i
ve
a
na
ly
s
is
,
w
hi
le
DT
of
f
e
r
e
d
in
te
r
pr
e
ta
bi
li
ty
a
nd
e
a
s
e
of
unde
r
s
ta
ndi
ng,
va
lu
a
bl
e
f
or
e
a
r
ly
pr
ot
ot
ypi
ng
a
nd
s
ta
ke
hol
de
r
tr
a
ns
p
a
r
e
nc
y.
T
h
e
f
in
di
ngs
s
ubs
ta
nt
ia
te
th
e
s
e
le
c
ti
on
of
RF
a
s
th
e
S
O
T
A
m
ode
l
f
or
A
n
ta
s
e
na
,
e
na
bl
in
g
e
f
f
e
c
ti
ve
m
a
r
it
im
e
a
nom
a
ly
de
te
c
ti
on, na
ti
ona
l
s
ur
ve
il
la
nc
e
,
a
nd e
nvi
r
onm
e
nt
a
l
pr
ot
e
c
ti
on.
3.1.
P
e
r
f
or
m
an
c
e
e
val
u
at
io
n
of
m
a
c
h
in
e
l
e
ar
n
in
g m
od
e
ls
T
hi
s
r
e
s
e
a
r
c
h
ut
il
iz
e
d
A
I
S
da
ta
s
a
m
pl
e
s
f
r
om
s
hi
ps
ope
r
a
ti
ng a
lo
ng
th
e
I
ndone
s
ia
n
A
r
c
hi
pe
la
gi
c
S
e
a
L
a
ne
s
be
twe
e
n
J
une
a
nd
S
e
pt
e
m
be
r
2023.
D
a
ta
va
li
da
ti
on
w
a
s
pe
r
f
or
m
e
d
w
it
h
th
e
I
ndone
s
ia
n
N
a
vy
he
a
dqua
r
te
r
s
a
nd
r
e
le
va
nt
m
a
r
it
im
e
s
ur
ve
il
la
nc
e
uni
ts
to
e
ns
ur
e
da
ta
in
te
gr
it
y
a
nd
ope
r
a
ti
ona
l
r
e
li
a
bi
li
ty
.
T
he
pe
r
f
or
m
a
nc
e
of
RF
,
X
G
B
oos
t,
a
nd
DT
m
ode
ls
w
a
s
e
va
lu
a
te
d
us
in
g
s
e
ve
n
k
e
y
m
e
tr
ic
s
:
c
onf
us
io
n
m
a
tr
ix
,
a
c
c
ur
a
c
y,
pr
e
c
i
s
io
n,
r
e
c
a
ll
,
F
1
-
s
c
or
e
,
R
O
C
c
ur
ve
,
a
nd
R
O
C
-
A
U
C
.
T
he
s
e
m
e
tr
ic
s
pr
ovi
de
c
om
pr
e
he
ns
iv
e
in
s
ig
ht
in
to
m
ode
l
b
e
ha
vi
or
,
pr
e
di
c
ti
ve
pr
e
c
is
io
n,
a
nd
r
obus
tn
e
s
s
a
ga
in
s
t
im
ba
la
nc
e
d
m
a
r
it
im
e
da
ta
s
e
ts
.
T
he
e
va
lu
a
ti
on
r
e
s
ul
ts
a
nd
c
om
pa
r
a
ti
ve
pe
r
f
or
m
a
nc
e
m
a
tr
ic
e
s
f
or
opt
im
a
l
m
ode
l
s
e
le
c
ti
on
a
r
e
pr
e
s
e
nt
e
d
in
T
a
bl
e
10.
B
a
s
e
d
on
th
e
a
na
ly
s
is
,
th
e
R
F
m
ode
l
de
m
ons
tr
a
te
s
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d
to
th
e
ot
he
r
two
m
ode
ls
,
a
c
hi
e
vi
ng
a
n
a
c
c
ur
a
c
y
of
a
ppr
oxi
m
a
te
ly
95.3%
,
w
hi
le
th
e
D
T
a
nd
X
G
B
oos
t
m
ode
ls
a
c
hi
e
v
e
d
94.8
a
nd
95.2%
,
r
e
s
pe
c
ti
ve
ly
.
T
o
ve
r
if
y
th
a
t
th
e
pe
r
f
o
r
m
a
n
c
e
di
f
f
e
r
e
nc
e
s
a
m
ong
th
e
te
s
te
d
m
ode
ls
a
r
e
s
ta
ti
s
ti
c
a
ll
y
m
e
a
ni
ngf
ul
,
a
one
-
w
a
y
a
na
ly
s
i
s
of
va
r
ia
nc
e
(
A
N
O
V
A
)
te
s
t
w
a
s
c
ondu
c
te
d
on
th
e
ir
a
c
c
ur
a
c
y
a
nd
R
O
C
-
A
U
C
s
c
or
e
s
.
T
he
r
e
s
ul
ti
ng
p
-
va
lu
e
s
(
<
0.05)
in
di
c
a
t
e
s
ig
ni
f
ic
a
nt
di
f
f
e
r
e
nc
e
s
be
tw
e
e
n
th
e
D
T
,
R
F
,
a
nd
X
G
B
oos
t
m
ode
ls
.
C
onf
id
e
nc
e
in
te
r
va
ls
a
t
th
e
95%
le
ve
l
w
e
r
e
a
ls
o
c
om
put
e
d
us
in
g
boot
s
tr
a
ppe
d
s
a
m
pl
in
g
to
qua
nt
if
y
th
e
unc
e
r
ta
in
ty
in
e
a
c
h
m
e
tr
ic
,
a
s
s
how
n
in
T
a
bl
e
11.
T
he
s
ta
ti
s
ti
c
a
l
a
na
ly
s
i
s
in
di
c
a
te
s
th
a
t
th
e
R
F
m
ode
l
a
c
hi
e
ve
s
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
a
nd
R
O
C
-
A
U
C
w
it
h
na
r
r
ow
c
onf
id
e
nc
e
in
te
r
va
ls
,
c
onf
ir
m
in
g
th
e
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
.
15
, N
o.
1
,
F
e
br
ua
r
y
20
26
:
269
-
288
278
r
e
li
a
bi
li
ty
of
it
s
pe
r
f
or
m
a
nc
e
.
A
one
-
w
a
y
A
N
O
V
A
te
s
t
pr
o
duc
e
d
p
<
0.05,
s
ig
ni
f
yi
ng
th
a
t
th
e
ob
s
e
r
ve
d
di
f
f
e
r
e
nc
e
s
a
m
ong mode
ls
a
r
e
s
ta
ti
s
ti
c
a
ll
y s
ig
ni
f
ic
a
nt
.
T
a
bl
e
10
.
M
od
e
l
e
va
lu
a
ti
on a
nd c
om
p
a
r
is
on of
m
a
tr
ix
f
or
opt
i
m
a
l
s
e
le
c
ti
on
C
onf
us
i
on m
a
t
r
i
x
P
r
e
c
i
s
i
on a
nd r
e
c
a
l
l
F1
-
s
c
or
e
R
O
C
c
ur
ve
a
nd R
O
C
-
AUC
T
he
c
onf
us
i
on
m
a
t
r
i
x
c
onf
i
r
m
e
d
hi
gh
pr
e
c
i
s
i
on
a
nd
r
e
c
a
l
l
,
r
e
f
l
e
c
t
i
ng
ba
l
a
nc
e
d
s
e
ns
i
t
i
vi
t
y
a
nd
s
pe
c
i
f
i
c
i
t
y.
T
he
c
onf
us
i
on
m
a
t
r
i
x
a
l
s
o
pr
ovi
de
s
a
de
t
a
i
l
e
d
br
e
a
kdow
n
of
t
r
ue
a
nd
f
a
l
s
e
pr
e
di
c
t
i
ons
,
hi
ghl
i
ght
i
ng
e
a
c
h
m
ode
l
’
s
pr
e
di
c
t
i
ve
pe
r
f
or
m
a
nc
e
.
P
r
e
c
i
s
i
on
m
e
a
s
ur
e
d t
he
pr
opor
t
i
on
of
c
or
r
e
c
t
l
y
i
de
nt
i
f
i
e
d
a
nom
a
l
i
e
s
a
m
ong
a
l
l
pr
e
di
c
t
e
d
a
nom
a
l
i
e
s
,
w
hi
l
e
r
e
c
a
l
l
i
ndi
c
a
t
e
d
t
he
pr
opor
t
i
on
of
a
c
t
ua
l
a
nom
a
l
i
e
s
c
or
r
e
c
t
l
y
de
t
e
c
t
e
d.
F1
-
s
c
or
e
s
c
onf
i
r
m
e
d
a
c
ons
i
s
t
e
nt
ba
l
a
n
c
e
be
t
w
e
e
n
t
he
s
e
t
w
o
di
m
e
ns
i
ons
,
s
how
i
ng
s
t
a
bl
e
pe
r
f
or
m
a
nc
e
e
ve
n
unde
r
i
m
ba
l
a
nc
e
d
c
ondi
t
i
ons
.
T
he
R
O
C
c
ur
ve
s
of
a
l
l
m
ode
l
s
e
xhi
bi
t
e
d
s
t
e
e
p
r
i
s
e
s
w
i
t
h
m
i
ni
m
a
l
f
a
l
s
e
pos
i
t
i
ve
s
,
i
ndi
c
a
t
i
ng
hi
gh
di
s
c
r
i
m
i
na
t
or
y
pow
e
r
.
A
m
ong
t
he
m
,
t
he
RF
m
ode
l
a
c
hi
e
ve
d
t
he
hi
ghe
s
t
R
O
C
-
A
U
C
(
96.8%
)
,
c
ons
i
s
t
e
nt
w
i
t
h
i
t
s
s
upe
r
i
or
ove
r
a
l
l
c
l
a
s
s
i
f
i
c
a
t
i
on
pe
r
f
or
m
a
nc
e
.
T
hi
s
c
onf
i
r
m
s
t
ha
t
R
F
m
a
i
nt
a
i
ns
t
he
be
s
t
ba
l
a
nc
e
be
t
w
e
e
n
s
e
ns
i
t
i
vi
t
y
a
nd
s
pe
c
i
f
i
c
i
t
y
unde
r
ope
r
a
t
i
ona
l
c
ons
t
r
a
i
nt
s
.
T
a
bl
e
11
.
R
e
s
ul
t
e
va
lu
a
ti
on mode
l
M
ode
l
A
c
c
ur
a
c
y (
%
)
95%
c
onf
i
de
nc
e
i
nt
e
r
va
l
P
r
e
c
i
s
i
on (
%
)
R
e
c
a
l
l
(
%
)
R
O
C
-
A
U
C
(
%
)
RF
95.3
±
0.9
94.7
94.2
96.8
X
G
B
oos
t
95.2
±
1.0
94.3
94.0
96.7
DT
94.8
±
1.6
93.8
93.4
96.1
B
a
s
e
d
on
th
e
e
v
a
lu
a
ti
on
m
e
tr
ic
s
,
th
e
RF
m
ode
l
de
m
ons
tr
a
te
d
t
he
hi
ghe
s
t
pe
r
f
or
m
a
nc
e
a
c
r
os
s
a
ll
ke
y
in
di
c
a
to
r
s
,
a
c
hi
e
vi
ng
a
n
a
c
c
ur
a
c
y
of
95.3%
,
pr
e
c
is
io
n
of
94.7
%
,
r
e
c
a
ll
of
94.2%
,
a
nd
a
n
R
O
C
-
A
U
C
s
c
or
e
of
96.8%
.
T
he
m
a
r
gi
na
l
a
c
c
ur
a
c
y
di
f
f
e
r
e
nc
e
be
twe
e
n
RF
a
nd
X
G
B
oos
t
(
0.1%
)
s
ugge
s
ts
bot
h
a
r
e
hi
ghl
y
e
f
f
e
c
ti
ve
f
or
a
nom
a
ly
de
te
c
ti
on,
th
ough
RF
s
li
ght
ly
out
pe
r
f
o
r
m
s
ove
r
a
ll
.
P
r
e
c
is
io
n,
w
hi
c
h
m
e
a
s
ur
e
s
how
m
a
ny
de
te
c
te
d
a
nom
a
li
e
s
w
e
r
e
a
c
tu
a
l
a
nom
a
li
e
s
,
a
l
s
o
s
how
s
RF
le
a
di
ng
a
t
94.7%
.
A
ll
m
ode
ls
s
c
or
e
d
a
bove
85%
,
hi
ghl
ig
ht
in
g
th
e
ir
s
tr
ong
c
a
pa
bi
li
ty
to
di
s
ti
ngui
s
h
be
twe
e
n
nor
m
a
l
a
nd
a
nom
a
lo
us
be
h
a
vi
or
s
,
w
it
h
RF
a
c
hi
e
vi
ng
th
e
be
s
t
ove
r
a
ll
r
e
s
ul
ts
.
A
lt
hough
X
G
B
oos
t
a
c
hi
e
ve
d
m
a
r
gi
na
ll
y
c
om
pe
ti
ti
ve
a
c
c
ur
a
c
y
in
c
e
r
ta
in
e
xpe
r
im
e
nt
a
l
s
e
tt
in
gs
,
RF
de
m
ons
tr
a
te
d
s
upe
r
io
r
ove
r
a
ll
ope
r
a
ti
ona
l
pe
r
f
or
m
a
nc
e
w
he
n
c
ons
id
e
r
in
g
r
obus
tn
e
s
s
,
in
f
e
r
e
nc
e
la
te
nc
y,
s
ta
bi
li
ty
a
c
r
os
s
f
ol
ds
,
a
nd
in
te
r
pr
e
ta
bi
li
ty
.
T
he
r
e
f
or
e
,
RF
is
s
e
le
c
te
d
a
s
th
e
pr
im
a
r
y
de
pl
oym
e
nt
m
ode
l
in
A
nt
a
s
e
n
a
,
w
hi
le
X
G
B
oos
t
s
e
r
ve
s
a
s
a
c
om
pl
e
m
e
nt
a
r
y
be
nc
hm
a
r
k
m
od
e
l.
T
h
e
c
onf
us
io
n
m
a
tr
ic
e
s
of
th
e
DT
,
RF
,
a
nd
X
G
B
oos
t
m
ode
ls
a
r
e
s
how
n
in
F
ig
u
r
e
3.
T
he
c
la
r
if
ie
d
c
onf
us
io
n
m
a
tr
ic
e
s
a
nd
a
li
gne
d
e
v
a
lu
a
ti
on
m
e
tr
ic
s
r
e
a
f
f
ir
m
R
F
’
s
S
O
T
A
c
a
pa
bi
li
ty
,
a
tt
r
ib
ut
e
d
to
it
s
e
ns
e
m
bl
e
-
ba
s
e
d
s
tr
uc
tu
r
e
t
ha
t
e
nha
nc
e
s
r
obu
s
tn
e
s
s
a
nd a
c
c
ur
a
c
y w
hi
le
m
a
in
ta
in
in
g i
nt
e
r
pr
e
ta
bi
li
ty
.
F
ig
ur
e
3
.
C
onf
us
io
n m
a
tr
ix
of
DT
,
RF
, a
nd X
G
B
oos
t
m
ode
l
3.1.1. Han
d
li
n
g i
m
b
al
an
c
e
d
d
at
as
e
t
s
M
a
r
it
im
e
a
nom
a
ly
de
te
c
ti
on
in
he
r
e
nt
ly
f
a
c
e
s
c
la
s
s
im
ba
la
nc
e
,
a
s
a
nom
a
li
e
s
(
e
.g.,
il
le
ga
l
a
c
ti
vi
ti
e
s
a
nd
s
m
uggl
in
g)
c
ons
ti
tu
te
a
s
m
a
ll
por
ti
on
o
f
to
ta
l
obs
e
r
va
ti
ons
.
RF
e
f
f
e
c
ti
ve
ly
m
it
ig
a
te
s
th
is
th
r
ough
boot
s
tr
a
p
s
a
m
pl
in
g
a
nd
c
la
s
s
w
e
ig
ht
in
g,
e
ns
ur
in
g
th
a
t
m
in
or
it
y
c
la
s
s
e
s
a
r
e
a
de
qua
te
ly
r
e
pr
e
s
e
nt
e
d.
T
hi
s
e
nh
a
nc
e
s
s
e
n
s
it
iv
it
y
(
r
e
c
a
ll
)
w
it
hout
c
om
pr
om
is
in
g
pr
e
c
is
io
n,
a
c
r
it
ic
a
l
a
dva
nt
a
g
e
f
or
de
te
c
ti
ng
r
a
r
e
but
hi
gh
-
im
pa
c
t
e
ve
nt
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.