T
E
L
K
O
M
N
I
K
A
T
elec
o
m
m
un
ica
t
io
n Co
m
pu
t
ing
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
2
4
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
,
p
p
.
1
62
~1
74
I
SS
N:
1
6
9
3
-
6
9
3
0
,
DOI
: 1
0
.
1
2
9
2
8
/TE
L
KOM
NI
KA.
v
2
4
i
1
.
2
7
4
5
1
162
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
Cha
lleng
es in rad
a
r
-
ba
sed no
n
-
sup
ercell
torna
do
det
ection
using
ma
chine lea
rning
appro
a
ches
K
ik
i
1,
3
,
Yo
nn
y
K
o
esm
a
ry
o
no
1
,
Ra
hm
a
t
H
ida
y
a
t
1
,
Do
na
ldi
Su
k
m
a
P
er
m
a
na
2
,
P
er
din
a
n
1
,
Abdu
lla
h Al
i
3
1
A
p
p
l
i
e
d
C
l
i
m
a
t
o
l
o
g
y
S
t
u
d
y
P
r
o
g
r
a
m
,
D
e
p
a
r
t
m
e
n
t
o
f
G
e
o
p
h
y
si
c
s
a
n
d
M
e
t
e
o
r
o
l
o
g
y
,
F
a
c
u
l
t
y
o
f
M
a
t
h
e
m
a
t
i
c
s
a
n
d
N
a
t
u
r
a
l
S
c
i
e
n
c
e
s,
I
P
B
U
n
i
v
e
r
si
t
y
,
B
o
g
o
r
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
C
l
i
m
a
t
o
l
o
g
y
,
I
n
d
o
n
e
si
a
n
A
g
e
n
c
y
o
f
M
e
t
e
o
r
o
l
o
g
y
C
l
i
ma
t
o
l
o
g
y
a
n
d
G
e
o
p
h
y
si
c
s (B
M
K
G
)
,
J
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
3
D
e
p
a
r
t
me
n
t
o
f
P
u
b
l
i
c
W
e
a
t
h
e
r
S
e
r
v
i
c
e
s,
I
n
d
o
n
e
s
i
a
n
A
g
e
n
c
y
o
f
M
e
t
e
o
r
o
l
o
g
y
C
l
i
m
a
t
o
l
o
g
y
a
n
d
G
e
o
p
h
y
si
c
s (B
M
K
G
)
,
J
a
k
a
r
t
a
,
I
n
d
o
n
e
si
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Au
g
20
,
2
0
2
5
R
ev
is
ed
No
v
28
2
0
2
5
Acc
ep
ted
Dec
8
,
2
0
2
5
To
rn
a
d
o
d
e
tec
ti
o
n
i
n
In
d
o
n
e
sia
re
m
a
in
s
c
h
a
ll
e
n
g
in
g
a
s
m
o
st
a
re
a
s
a
r
e
m
o
n
it
o
re
d
b
y
sin
g
le
-
p
o
lariz
a
ti
o
n
we
a
th
e
r
ra
d
a
r,
wh
il
e
d
u
a
l
-
p
o
lariz
a
ti
o
n
sy
ste
m
s
o
ffe
r
su
p
e
ri
o
r
d
e
tec
ti
o
n
c
a
p
a
b
il
it
ies
.
Th
is
stu
d
y
p
re
se
n
t
s
a
n
o
v
e
l
a
p
p
ro
a
c
h
b
y
a
p
p
ly
i
n
g
ra
n
d
o
m
f
o
re
st
(RF
)
a
n
d
XG
Bo
o
st
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
to
d
e
tec
t
to
r
n
a
d
o
e
s
u
si
n
g
sin
g
le
-
p
o
lariz
a
ti
o
n
ra
d
a
r
d
a
ta,
a
d
d
re
ss
in
g
a
c
rit
ica
l
g
a
p
in
tr
o
p
ica
l
to
r
n
a
d
o
m
o
n
it
o
r
in
g
wh
e
re
d
u
a
l
-
p
o
l
in
fra
s
tru
c
tu
re
is
li
m
it
e
d
.
F
o
u
r
to
r
n
a
d
o
c
a
se
s
in
S
u
ra
b
a
y
a
d
u
ri
n
g
2
0
2
4
we
re
a
n
a
ly
z
e
d
.
R
a
d
a
r
fe
a
tu
re
s
in
c
lu
d
in
g
re
flec
ti
v
it
y
,
ra
d
ial
v
e
l
o
c
it
y
,
v
o
rti
c
it
y
,
a
n
d
a
n
g
u
lar
m
o
m
e
n
tu
m
we
re
e
x
trac
ted
th
r
o
u
g
h
a
m
u
lt
i
-
e
lev
a
ti
o
n
sli
d
in
g
win
d
o
w
tec
h
n
iq
u
e
.
S
p
a
ti
a
l
lab
e
ls
we
re
a
ss
ig
n
e
d
b
a
se
d
o
n
re
p
o
rts
fro
m
t
h
e
I
n
d
o
n
e
sia
n
Ag
e
n
c
y
f
o
r
M
e
te
o
ro
l
o
g
y
,
Cl
ima
to
lo
g
y
a
n
d
G
e
o
p
h
y
sic
s
(BM
KG
)
with
a
7
.
5
k
m
ra
d
iu
s
fr
o
m
th
e
e
v
e
n
t
c
e
n
ter.
Th
e
d
a
tas
e
t
wa
s
b
a
lan
c
e
d
u
sin
g
sy
n
t
h
e
ti
c
m
in
o
rit
y
o
v
e
r
-
sa
m
p
li
n
g
tec
h
n
iq
u
e
(S
M
OTE)
.
Ev
a
l
u
a
ti
o
n
wa
s
p
e
rfo
rm
e
d
u
sin
g
th
e
lea
v
e
-
o
n
e
-
c
a
se
-
o
u
t
(
LOCO)
sc
h
e
m
e
.
Wi
th
in
-
c
a
se
e
v
a
lu
a
ti
o
n
sh
o
we
d
stro
n
g
p
e
rfo
rm
a
n
c
e
wit
h
a
re
a
u
n
d
e
r
th
e
c
u
rv
e
(AU
C)
>
0
.
9
4
fo
r
b
o
t
h
m
o
d
e
ls.
XG
Bo
o
st
a
c
h
iev
e
d
h
ig
h
e
r
p
ro
b
a
b
il
it
y
o
f
d
e
tec
ti
o
n
(P
OD
0
.
6
7
-
0
.
7
2
)
b
u
t
wi
th
e
lev
a
te
d
fa
lse
a
larm
ra
tes
(F
AR
u
p
to
7
0
%
)
.
RF
d
e
m
o
n
str
a
ted
m
o
re
b
a
lan
c
e
d
p
e
rfo
rm
a
n
c
e
(P
OD
0
.
6
1
-
0
.
6
5
,
F
AR
0
.
3
4
-
0
.
3
5
).
LOCO
e
v
a
lu
a
t
io
n
re
v
e
a
led
sig
n
ifi
c
a
n
t
P
OD
re
d
u
c
ti
o
n
a
n
d
F
AR
i
n
c
re
a
se
wh
e
n
tes
ted
o
n
n
e
w
c
a
se
s.
Th
is
in
d
ica
tes
g
e
n
e
ra
li
z
a
ti
o
n
c
h
a
ll
e
n
g
e
s
d
u
e
to
v
a
riab
il
it
y
in
to
rn
a
d
o
c
h
a
ra
c
teristics
.
Th
is
st
u
d
y
d
e
m
o
n
stra
tes
th
e
p
o
te
n
ti
a
l
o
f
m
a
c
h
in
e
le
a
rn
in
g
f
o
r
tro
p
ica
l
t
o
rn
a
d
o
e
a
rly
d
e
tec
ti
o
n
u
sin
g
re
a
d
il
y
a
v
a
il
a
b
le
sin
g
le
-
p
o
lariz
a
ti
o
n
ra
d
a
r.
K
ey
w
o
r
d
s
:
Ma
ch
in
e
lear
n
in
g
R
an
d
o
m
f
o
r
est
Sin
g
le
-
p
o
lar
ized
r
ad
ar
T
o
r
n
ad
o
es
XGBo
o
s
t
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Yo
n
n
y
Ko
esm
ar
y
o
n
o
Ap
p
lied
C
lim
ato
lo
g
y
Stu
d
y
Pr
o
g
r
am
,
De
p
ar
tm
en
t
o
f
Geo
p
h
y
s
ics an
d
Me
teo
r
o
lo
g
y
Facu
lty
o
f
Ma
th
em
atics a
n
d
N
atu
r
al
Scien
ce
s
,
I
PB
Un
iv
er
s
ity
B
o
g
o
r
,
I
n
d
o
n
esia
E
m
ail: y
o
n
n
y
@
ap
p
s
.
ip
b
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
Sev
er
e
co
n
v
ec
ti
v
e
s
to
r
m
s
p
r
esen
t
a
s
ig
n
if
ican
t
ch
allen
g
e
to
m
eteo
r
o
lo
g
ical
m
o
n
it
o
r
in
g
a
n
d
f
o
r
ec
asti
n
g
,
p
a
r
ticu
lar
ly
in
tr
o
p
ical
r
eg
io
n
s
.
T
h
ese
s
y
s
tem
s
f
r
eq
u
en
tly
p
r
o
d
u
ce
s
ig
n
if
ican
t
wea
th
er
p
h
en
o
m
en
a,
in
clu
d
in
g
h
ea
v
y
r
ain
,
th
u
n
d
er
s
to
r
m
s
,
h
ail,
an
d
to
r
n
ad
o
e
s
.
A
m
o
n
g
t
h
ese,
“
p
u
tin
g
b
eliu
n
g
”
a
lo
ca
l
ter
m
u
s
ed
in
I
n
d
o
n
esia
f
o
r
s
m
all
-
s
ca
le,
s
h
o
r
t
-
liv
ed
to
r
n
ad
o
es
is
a
f
r
eq
u
en
t
o
cc
u
r
r
en
ce
,
p
ar
ticu
lar
l
y
d
u
r
in
g
tr
an
s
itio
n
al
s
ea
s
o
n
s
[
1
]
-
[
3
]
.
T
o
r
n
ad
o
es,
th
o
u
g
h
ty
p
ically
s
m
all
an
d
s
h
o
r
t
-
liv
ed
in
I
n
d
o
n
esia,
ca
n
s
till
ca
u
s
e
s
ev
er
e
d
am
ag
e
to
in
f
r
astru
ctu
r
e
an
d
p
o
s
e
r
is
k
s
to
h
u
m
an
s
af
ety
[
4
]
.
Me
an
w
h
ile,
in
ter
m
s
o
f
ec
o
n
o
m
ic
lo
s
s
es,
f
o
r
ex
am
p
le
in
Sem
ar
an
g
,
C
en
tr
al
J
av
a,
wh
ic
h
ca
r
r
ies
a
to
r
n
ad
o
r
is
k
h
az
ar
d
lev
el
as
h
ig
h
as
2
8
.
5
0
2
%,
t
o
r
n
ad
o
es
ca
u
s
ed
an
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
C
h
a
llen
g
es in
r
a
d
a
r
-
b
a
s
ed
n
o
n
-
s
u
p
ercell
to
r
n
a
d
o
d
etec
tio
n
u
s
in
g
ma
ch
in
e
le
a
r
n
in
g
a
p
p
r
o
a
ch
es (
K
iki)
163
esti
m
ated
ec
o
n
o
m
ic
lo
s
s
o
f
u
p
to
USD
5
1
,
0
0
0
d
u
r
i
n
g
th
e
p
er
io
d
f
r
o
m
J
an
u
ar
y
2
0
1
4
to
Dec
em
b
er
2
0
1
8
[
5
]
.
T
h
is
ex
tr
em
e
p
h
en
o
m
en
o
n
ac
co
u
n
ts
f
o
r
ap
p
r
o
x
im
ately
2
1
%
o
f
all
n
atu
r
al
d
is
aster
s
th
at
o
c
cu
r
r
ed
in
I
n
d
o
n
esia
b
etwe
en
1
8
1
5
an
d
2
0
1
4
[
6
]
.
Giv
en
th
ese
r
is
k
s
,
u
n
d
er
s
tan
d
in
g
th
e
ch
ar
ac
ter
is
tics
o
f
I
n
d
o
n
esian
to
r
n
ad
o
es
is
cr
itical.
Sev
er
al
d
o
m
esti
c
s
tu
d
ies
h
av
e
attem
p
t
ed
to
a
n
aly
ze
t
h
ese
ev
en
ts
u
s
in
g
av
aila
b
le
m
eteo
r
o
lo
g
ical
d
ata.
Fo
r
e
x
am
p
le,
a
s
tu
d
y
b
y
Sis
wan
to
an
d
Su
p
ar
i
[
7
]
u
tili
ze
d
s
atellite
an
d
s
u
r
f
ac
e
o
b
s
er
v
atio
n
d
ata
to
id
e
n
tify
t
o
r
n
ad
o
p
r
ec
u
r
s
o
r
s
,
ch
ar
ac
ter
ized
b
y
a
r
ap
id
in
c
r
e
ase
in
r
elativ
e
h
u
m
id
ity
ap
p
r
o
x
im
ately
o
n
e
h
o
u
r
in
ad
v
an
ce
,
ac
co
m
p
an
ied
b
y
a
s
h
ar
p
tem
p
er
atu
r
e
d
r
o
p
a
n
d
t
h
e
p
r
esen
ce
o
f
a
“h
o
r
n
-
lik
e”
p
r
ess
u
r
e
an
o
m
aly
.
R
u
s
m
ala
et
a
l.
[
8
]
a
n
aly
ze
d
a
to
r
n
ad
o
ev
en
t
i
n
J
ak
ar
ta
u
s
in
g
C
-
b
an
d
wea
th
er
r
ad
ar
d
ata
f
r
o
m
co
lu
m
n
m
ax
im
u
m
r
ef
lectiv
i
ty
(
C
MA
X)
,
v
er
tical
cu
t
(
VC
UT
)
,
an
d
c
o
n
s
tan
t
altitu
d
e
p
lan
p
o
s
itio
n
in
d
icato
r
(
v
elo
city
)
(
C
APPI
(
V)
)
at
0
.
5
k
m
,
1
.
0
k
m
,
a
n
d
1
.
5
k
m
,
c
o
m
b
in
e
d
with
h
o
r
izo
n
tal
win
d
(
HW
I
ND)
d
ata.
T
h
e
r
esu
lts
s
h
o
wed
th
at
th
e
to
r
n
ad
o
-
g
e
n
er
atin
g
c
o
n
v
ec
tiv
e
clo
u
d
d
e
v
elo
p
e
d
r
a
p
id
ly
,
with
r
ef
lectiv
ity
b
etwe
en
3
5
-
4
5
d
B
Z
an
d
win
d
s
p
ee
d
s
u
p
to
3
5
k
n
o
ts
.
An
o
th
er
s
tu
d
y
b
y
Kik
i
et
a
l.
[
3
]
a
n
aly
ze
d
th
e
s
p
atio
tem
p
o
r
al
d
is
tr
ib
u
tio
n
an
d
tr
en
d
s
o
f
to
r
n
ad
o
o
cc
u
r
r
en
ce
s
in
I
n
d
o
n
esia
o
v
e
r
th
e
p
ast
d
ec
ad
e
an
d
r
ep
o
r
ted
t
h
at
th
e
p
r
im
ar
y
to
r
n
ad
o
h
o
ts
p
o
ts
ar
e
lo
ca
ted
o
n
th
e
is
lan
d
o
f
J
av
a,
with
a
p
o
s
itiv
e
tr
en
d
in
to
r
n
ad
o
f
r
eq
u
e
n
cy
o
f
a
p
p
r
o
x
im
ately
1
2
e
v
en
ts
p
er
y
e
ar
.
Yu
d
is
tira
et
a
l
.
[
9
]
an
al
y
ze
d
u
p
p
e
r
-
air
d
ata
an
d
f
o
u
n
d
th
at
n
eg
ativ
e
lifte
d
in
d
ex
(
L
I
)
v
alu
es,
elev
ate
d
K
in
d
ex
(
K
I
)
,
to
tal
to
tals
in
d
ex
(
T
T
)
,
a
n
d
c
o
n
v
ec
tiv
e
av
ailab
le
p
o
ten
tial e
n
er
g
y
(
C
APE)
in
d
ices,
alo
n
g
with
h
ig
h
er
s
ev
er
e
wea
th
er
th
r
ea
t in
d
e
x
(
SW
E
AT
)
r
ea
d
in
g
s
,
in
d
icate
d
u
n
s
tab
le
atm
o
s
p
h
er
i
c
co
n
d
itio
n
s
co
n
d
u
civ
e
to
co
n
v
ec
tio
n
a
n
d
t
h
u
n
d
er
s
to
r
m
d
ev
elo
p
m
en
t.
Ov
er
all,
th
e
r
esu
lts
d
em
o
n
s
tr
ated
in
c
r
e
asin
g
atm
o
s
p
h
er
ic
in
s
tab
ilit
y
a
n
d
a
r
is
in
g
p
o
ten
tial
f
o
r
co
n
v
e
ctiv
e
an
d
lig
h
tn
in
g
ac
tiv
ity
co
m
p
ar
ed
to
th
e
p
r
ec
e
d
in
g
d
ay
s
.
Oth
er
s
tu
d
y
h
as
attem
p
ted
to
ev
alu
ate
co
m
m
u
n
ity
r
esil
ien
ce
to
to
r
n
ad
o
ev
en
ts
.
Fo
r
in
s
tan
ce
,
Hid
a
y
at
e
t
a
l.
[
1
0
]
co
n
d
u
cte
d
r
esear
c
h
i
n
Do
n
o
h
u
d
an
Villag
e,
C
en
tr
al
J
av
a,
an
d
id
en
tifie
d
a
m
o
d
e
r
ate
lev
el
o
f
r
esil
ien
ce
am
o
n
g
r
esid
en
ts
.
T
h
is
co
n
d
itio
n
was
i
n
f
lu
en
ce
d
b
y
f
ac
to
r
s
s
u
ch
as
ad
a
p
tab
ilit
y
to
ch
an
g
i
n
g
cir
c
u
m
s
tan
ce
s
an
d
ch
allen
g
es,
f
am
ilial
an
d
s
o
ci
al
s
u
p
p
o
r
t n
etwo
r
k
s
,
s
p
ir
itu
al
v
alu
es,
an
d
a
s
tr
o
n
g
s
en
s
e
o
f
p
u
r
p
o
s
e.
Desp
ite
th
e
s
e
an
aly
tical
ef
f
o
r
ts
,
th
e
d
ete
ctio
n
an
d
m
o
n
ito
r
in
g
o
f
s
u
ch
s
h
o
r
t
-
liv
ed
e
v
en
ts
r
em
ain
d
if
f
ic
u
lt,
n
ec
ess
itatin
g
r
o
b
u
s
t e
ar
ly
war
n
in
g
s
y
s
tem
s
.
T
h
e
p
r
i
m
ar
y
tech
n
o
l
o
g
ical
co
n
s
tr
ain
t
in
I
n
d
o
n
esia
lies
in
t
h
e
r
ad
a
r
in
f
r
astru
ctu
r
e,
wh
er
e
as
wea
th
er
r
ad
ar
is
r
ec
o
g
n
ized
as
th
e
p
r
im
ar
y
to
o
l
f
o
r
to
r
n
a
d
o
m
o
n
ito
r
in
g
an
d
f
o
r
ec
asti
n
g
g
lo
b
ally
[
1
1
]
.
W
h
ile
I
n
d
o
n
esia
n
Ag
en
cy
f
o
r
Me
teo
r
o
lo
g
y
,
C
lim
ato
lo
g
y
an
d
Geo
p
h
y
s
ics
(
B
MK
G)
h
as
b
eg
u
n
d
ep
l
o
y
in
g
d
u
al
-
p
o
lar
izatio
n
wea
th
er
r
ad
ar
s
,
t
h
e
n
etwo
r
k
r
em
ain
s
lim
ited
.
As
o
f
2
0
2
5
,
B
MK
G
o
p
er
ates
4
4
wea
th
e
r
r
a
d
ar
s
,
co
m
p
r
is
in
g
3
3
s
in
g
le
-
p
o
lar
izatio
n
C
-
b
a
n
d
r
a
d
ar
s
an
d
1
1
d
u
al
-
p
o
lar
izatio
n
s
y
s
tem
s
(
C
-
b
an
d
an
d
X
-
b
an
d
)
.
C
o
n
s
eq
u
en
tly
,
m
o
s
t
r
eg
io
n
s
,
in
clu
d
i
n
g
th
o
s
e
f
r
e
q
u
e
n
tly
ex
p
er
ien
ci
n
g
to
r
n
ad
o
e
v
en
ts
,
ar
e
p
r
ed
o
m
in
a
n
tly
b
y
s
in
g
le
-
p
o
lar
izatio
n
r
a
d
ar
s
y
s
tem
s
.
T
h
is
lim
i
tatio
n
r
estricts
th
e
av
ailab
ilit
y
o
f
ad
v
an
ce
d
p
o
lar
im
etr
ic
v
ar
iab
les
o
f
te
n
u
s
ed
in
m
o
d
er
n
s
to
r
m
d
etec
tio
n
,
cr
ea
tin
g
a
n
ee
d
f
o
r
m
eth
o
d
s
th
at
ca
n
m
ax
im
ize
th
e
u
tili
ty
o
f
ex
is
tin
g
s
in
g
le
-
p
o
la
r
izatio
n
d
ata.
I
n
th
e
g
lo
b
al
co
n
tex
t,
a
d
v
an
ce
s
in
d
ata
s
cien
ce
an
d
ar
tif
icial
in
tellig
en
ce
(
AI
)
o
f
f
er
s
ig
n
if
ican
t
o
p
p
o
r
tu
n
ities
to
co
m
p
lem
en
t
n
u
m
er
ical
wea
th
er
p
r
e
d
ictio
n
an
d
en
h
an
ce
r
ea
l
-
tim
e
g
u
id
an
ce
[
4
]
.
R
ec
en
t
y
ea
r
s
h
av
e
s
ee
n
th
e
s
u
cc
ess
f
u
l
ap
p
li
ca
tio
n
o
f
m
ac
h
in
e
lea
r
n
in
g
to
im
p
r
o
v
e
to
r
n
a
d
o
d
etec
tio
n
.
Fo
r
d
u
al
-
p
o
lar
izatio
n
d
ata,
Z
en
g
et
a
l.
[
1
2
]
in
tr
o
d
u
ce
d
an
ex
tr
em
e
g
r
a
d
ien
t
b
o
o
s
tin
g
(
XGBo
o
s
t)
-
b
ased
alg
o
r
ith
m
th
at
en
h
an
ce
d
d
etec
tio
n
ac
cu
r
ac
y
.
Ho
wev
e
r
,
p
r
o
g
r
ess
h
as
also
b
ee
n
m
a
d
e
u
s
in
g
s
in
g
le
-
r
a
d
ar
d
ata,
San
d
m
æ
l
et
a
l.
[
1
3
]
d
ev
elo
p
e
d
th
e
t
o
r
n
a
d
o
p
r
o
b
a
b
ilit
y
alg
o
r
ith
m
(
T
OR
P),
a
p
r
o
b
ab
ilis
tic
m
ac
h
in
e
lear
n
in
g
ap
p
r
o
ac
h
u
tili
zin
g
s
in
g
le
-
r
ad
ar
d
ata
to
esti
m
ate
to
r
n
ad
o
o
cc
u
r
r
en
ce
p
r
o
b
ab
il
ities
.
Veillette
et
a
l.
[
1
1
]
p
r
e
s
en
ted
th
e
to
r
n
ad
o
n
etwo
r
k
(
T
o
r
Net)
b
en
c
h
m
ar
k
d
ataset,
co
m
p
r
is
in
g
f
u
ll
-
r
eso
l
u
tio
n
p
o
lar
im
etr
ic
wea
th
e
r
r
a
d
ar
d
ata
to
f
ac
ilit
ate
d
ev
elo
p
in
g
an
d
e
v
alu
atin
g
m
ac
h
in
e
lear
n
i
n
g
al
g
o
r
ith
m
s
f
o
r
to
r
n
a
d
o
d
etec
tio
n
a
n
d
p
r
ed
i
ctio
n
.
Var
io
u
s
d
ee
p
lear
n
in
g
ar
ch
itectu
r
es
wer
e
e
v
alu
ated
f
o
r
to
r
n
ad
o
d
etec
tio
n
u
s
in
g
wea
th
er
r
ad
ar
d
ata,
d
em
o
n
s
tr
atin
g
th
e
p
o
ten
tial
o
f
th
ese
m
o
d
els
in
o
p
er
atio
n
al
s
ettin
g
s
[
1
1
]
.
Fu
r
th
er
m
o
r
e,
d
ee
p
lear
n
in
g
ar
ch
ite
ctu
r
es
h
av
e
s
h
o
wn
p
o
ten
tial
in
o
p
er
atio
n
al
s
ettin
g
s
,
with
Z
h
o
u
[
1
4
]
p
r
o
p
o
s
ed
a
h
y
b
r
id
m
o
d
els
co
m
b
in
i
n
g
Kalm
an
f
ilter
in
g
,
co
n
v
o
l
u
tio
n
al
n
eu
r
al
n
etwo
r
k
s
(
C
NNs),
an
d
b
id
ir
ec
tio
n
al
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
n
etwo
r
k
s
with
m
u
lti
-
h
ea
d
atten
tio
n
m
ec
h
an
is
m
s
to
im
p
r
o
v
e
to
r
n
ad
o
p
r
e
d
ictio
n
in
th
e
Un
ited
States
.
Su
f
i
et
a
l.
[
1
5
]
b
u
ilt
a
to
r
n
a
d
o
co
m
p
en
d
iu
m
th
at
en
co
m
p
ass
es
b
o
th
c
u
r
r
en
t
an
d
h
is
to
r
ical
r
e
co
r
d
s
o
f
to
r
n
ad
o
es
in
B
an
g
lad
esh
,
in
co
n
ju
n
ctio
n
with
AI
-
b
ased
r
e
g
r
ess
io
n
an
al
y
s
is
an
d
th
e
n
ew
d
ash
b
o
a
r
d
s
y
s
tem
,
wh
ich
wo
u
ld
en
a
b
le
an
y
s
tr
ateg
ic
d
ec
is
io
n
m
ak
er
to
m
ak
e
e
v
id
en
ce
-
b
as
ed
p
o
licy
d
ec
is
io
n
s
r
eg
a
r
d
in
g
to
r
n
ad
o
ev
en
ts
in
B
an
g
la
d
esh
.
C
o
m
p
lem
en
tin
g
th
is
,
Xu
e
et
a
l.
[
1
6
]
in
tr
o
d
u
ce
d
th
e
m
u
lti
-
task
id
en
tific
atio
n
n
etwo
r
k
(
MT
I
-
Net)
,
a
d
etec
tio
n
m
o
d
el
th
at
u
tili
ze
s
a
n
o
v
el
b
ac
k
b
o
n
e
with
s
p
atial
an
d
ch
an
n
el
atten
tio
n
u
n
its
.
T
h
is
ap
p
r
o
ac
h
h
as
p
r
o
v
en
h
ig
h
ly
ef
f
ec
tiv
e,
r
ed
u
cin
g
f
alse a
lar
m
r
ates f
r
o
m
0
.
9
4
to
0
.
4
6
a
n
d
ac
h
iev
in
g
a
n
ea
r
ly
f
o
u
r
f
o
ld
i
n
cr
ea
s
e
in
th
e
h
it r
ate.
Desp
ite
th
ese
g
lo
b
al
ad
v
an
ce
m
en
ts
,
th
er
e
r
em
ain
s
a
n
o
tab
le
g
ap
i
n
th
e
liter
atu
r
e
c
o
n
ce
r
n
in
g
th
e
ap
p
licatio
n
o
f
m
ac
h
in
e
lea
r
n
in
g
tech
n
iq
u
es
f
o
r
d
etec
tin
g
to
r
n
ad
o
es
in
s
p
ec
if
ically
wit
h
in
th
e
I
n
d
o
n
esian
m
ar
itime
co
n
tin
en
t.
M
o
s
t
cr
itical
is
th
e
lack
o
f
r
esear
ch
u
tili
z
in
g
s
in
g
le
-
p
o
lar
izatio
n
r
ad
a
r
d
ata
f
o
r
th
is
p
u
r
p
o
s
e,
wh
ich
co
n
s
titu
tes
th
e
b
u
lk
o
f
I
n
d
o
n
esia’
s
o
b
s
er
v
atio
n
al
n
etwo
r
k
.
Ad
d
r
ess
in
g
th
is
g
ap
is
ess
en
tial
f
o
r
d
ev
elo
p
in
g
d
etec
tio
n
s
y
s
tem
s
th
at
ar
e
ef
f
ec
tiv
e
in
th
e
co
n
tex
t
o
f
I
n
d
o
n
esia’
s
u
n
iq
u
e
m
eteo
r
o
lo
g
ic
al
co
n
d
itio
n
s
a
n
d
in
f
r
astru
ctu
r
e
c
o
n
s
tr
ain
ts
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
: 1
62
-
1
74
164
T
h
er
ef
o
r
e,
th
is
s
tu
d
y
aim
s
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
,
s
p
ec
if
ically
r
an
d
o
m
f
o
r
est
(
R
F)
an
d
XGBo
o
s
t,
in
d
etec
tin
g
to
r
n
ad
o
ev
e
n
ts
in
I
n
d
o
n
esia
u
s
in
g
s
in
g
le
-
p
o
lar
izatio
n
wea
th
e
r
r
ad
ar
d
ata.
A
d
d
itio
n
ally
,
t
h
e
s
tu
d
y
ass
ess
es
th
e
im
p
ac
t
o
f
in
c
o
r
p
o
r
ati
n
g
m
u
lti
-
elev
atio
n
r
ad
ar
d
ata
o
n
d
etec
tio
n
ac
cu
r
ac
y
.
B
y
ad
d
r
ess
in
g
th
e
i
d
en
tifie
d
r
esear
c
h
g
a
p
,
t
h
is
s
tu
d
y
co
n
tr
ib
u
tes
to
d
e
v
elo
p
in
g
m
o
r
e
ef
f
ec
tiv
e
ea
r
l
y
war
n
in
g
s
y
s
tem
s
f
o
r
to
r
n
ad
o
es
in
th
e
r
eg
io
n
.
T
h
e
f
i
n
d
in
g
s
wil
l
p
r
o
v
id
e
v
alu
ab
le
i
n
s
ig
h
ts
in
t
o
th
e
ap
p
licatio
n
o
f
m
ac
h
in
e
lear
n
i
n
g
tech
n
iq
u
es
f
o
r
tr
o
p
ical
r
eg
i
o
n
s
an
d
ar
ch
i
p
elag
ic
co
u
n
tr
ies,
wh
er
e
c
o
n
v
en
tio
n
al
d
etec
tio
n
m
eth
o
d
s
f
ac
e
u
n
iq
u
e
ch
allen
g
es d
u
e
to
co
m
p
lex
to
p
o
g
r
ap
h
y
an
d
m
eteo
r
o
lo
g
ical
p
atter
n
s
.
2.
M
E
T
H
O
D
2
.
1
.
M
a
t
er
ia
l
T
h
e
a
n
a
l
y
s
i
s
f
o
c
u
s
e
d
o
n
t
o
r
n
a
d
o
e
v
e
n
t
s
w
it
h
i
n
t
h
e
c
o
v
e
r
a
g
e
a
r
e
a
o
f
t
h
e
s
i
n
g
l
e
-
p
o
l
a
r
i
z
at
i
o
n
wea
t
h
e
r
r
a
d
a
r
o
p
e
r
a
t
e
d
b
y
B
M
KG
i
n
S
u
r
a
b
a
y
a
(
S
B
Y
)
.
T
h
is
l
o
c
a
ti
o
n
w
as
c
h
o
s
e
n
b
e
c
a
u
s
e
i
t
i
s
l
o
c
a
t
e
d
i
n
E
a
s
t
J
a
v
a
,
w
h
ic
h
i
s
o
n
e
o
f
t
h
e
p
r
o
v
i
n
c
e
s
w
i
t
h
t
h
e
h
i
g
h
e
s
t
f
r
e
q
u
e
n
c
y
o
f
t
o
r
n
a
d
o
e
v
e
n
t
s
i
n
I
n
d
o
n
e
s
i
a
i
n
r
e
c
e
n
t
y
e
a
r
s
[
3
]
.
Fo
u
r
d
o
c
u
m
en
ted
to
r
n
ad
o
e
v
en
ts
o
n
1
,
9
,
1
7
,
a
n
d
2
6
J
an
u
ar
y
2
0
2
4
wer
e
s
elec
ted
b
ased
o
n
d
etailed
in
cid
en
t
r
ep
o
r
ts
f
r
o
m
th
e
B
MK
G
[
1
7
]
.
T
h
ese
r
ep
o
r
ts
p
r
o
v
id
ed
ac
cu
r
ate
g
eo
g
r
ap
h
ic
co
o
r
d
in
ates,
ev
en
t
tim
in
g
s
,
d
u
r
atio
n
,
an
d
im
p
ac
t
d
escr
ip
tio
n
s
,
ess
en
tial f
o
r
s
p
at
ial
lab
elin
g
o
f
th
e
r
ad
ar
d
ata.
T
h
ese
ca
s
es we
r
e
s
elec
ted
p
r
im
ar
ily
b
ased
o
n
d
ata
av
ailab
ilit
y
an
d
t
h
e
ab
s
en
ce
o
f
s
ig
n
if
ican
t
r
a
d
a
r
b
ea
m
b
lo
ck
ag
e
d
u
e
to
ter
r
ain
at
th
e
ev
e
n
t
lo
ca
tio
n
s
.
Deta
iled
s
u
m
m
ar
ies
o
f
th
e
s
elec
ted
to
r
n
ad
o
ev
e
n
ts
,
in
clu
d
in
g
co
o
r
d
in
ates,
tim
in
g
s
,
an
d
r
e
p
o
r
ted
im
p
ac
ts
,
ar
e
p
r
esen
ted
in
T
ab
le
1
.
T
h
e
Su
r
ab
a
y
a
r
a
d
a
r
o
p
e
r
ates in
5
.
6
GHz
f
r
eq
u
e
n
cy
with
a
Ny
q
u
is
t
v
elo
ci
ty
o
f
3
2
m
/s
,
em
p
lo
y
in
g
a
s
tag
g
er
ed
p
u
ls
e
r
ep
etitio
n
f
r
eq
u
en
cy
(
PR
F)
s
ch
em
e
o
f
1
0
0
0
Hz
with
a
s
tag
g
er
in
g
3
/4
r
atio
an
d
a
s
p
atial
r
an
g
e
r
eso
lu
tio
n
o
f
2
5
0
m
eter
s
.
R
ad
ar
b
ea
m
b
lo
ck
a
g
e
an
al
y
s
is
r
esu
lts
f
o
r
th
e
th
r
ee
l
o
west
r
ad
ar
ele
v
atio
n
an
g
les
(
0
.
5
°,
0
.
8
°,
a
n
d
1
.
8
°)
a
r
e
p
r
esen
ted
i
n
Fig
u
r
es
1
(
a)
an
d
(
b
)
,
co
n
f
ir
m
in
g
m
in
i
m
al
o
r
n
e
g
lig
ib
le
b
l
o
ck
ag
e
at
all
ev
e
n
t
s
ites
.
T
ab
le
1
.
Selecte
d
to
r
n
ad
o
ca
s
e
s
tu
d
ies
D
a
t
e
/
t
i
me
Lo
c
a
t
i
o
n
I
mp
a
c
t
s
2
0
2
4
-
01
-
0
1
/
1
3
.
0
0
L
T
K
e
m
l
a
n
g
i
,
M
o
j
o
k
e
r
t
o
,
E
a
st
Ja
v
a
(
7
.
4
0
8
5
7
9
0
7
S
,
1
1
2
.
3
6
5
4
7
6
1
7
E)
1
6
h
o
u
ses
d
a
ma
g
e
d
,
se
v
e
r
a
l
t
r
e
e
s
u
p
r
o
o
t
e
d
,
t
w
o
v
e
h
i
c
l
e
s
h
i
t
b
y
f
a
l
l
e
n
t
r
e
e
s
2
0
2
4
-
01
-
1
7
/
1
4
.
3
0
L
T
M
e
g
a
l
u
h
,
Jo
mb
a
n
g
,
Ea
s
t
J
a
v
a
(
7
.
5
0
1
5
4
6
3
3
S
,
1
1
2
.
1
8
6
1
4
2
3
7
E)
8
h
o
u
se
s
d
a
ma
g
e
2
0
2
4
-
01
-
2
6
/
1
7
.
0
0
L
T
S
o
k
o
,
T
u
b
a
n
,
E
a
st
Ja
v
a
(
7
.
0
9
4
1
2
3
7
5
S
,
1
1
1
.
9
3
2
2
6
5
4
4
E)
D
o
z
e
n
s
o
f
h
o
u
s
e
s
d
a
m
a
g
e
d
,
o
n
e
i
n
j
u
r
y
2
0
2
4
-
02
-
0
4
/
1
6
.
0
0
L
T
Ta
r
i
k
,
S
i
d
o
a
r
j
o
,
E
a
st
Ja
v
a
(
7
.
4
4
8
3
7
7
7
7
S
,
1
1
2
.
5
1
7
8
3
6
1
1
E)
2
0
0
h
o
u
s
e
s
d
a
m
a
g
e
d
,
o
n
e
f
a
t
a
l
i
t
y
,
t
w
o
i
n
j
u
r
i
e
s
(
a)
(
b
)
Fig
u
r
e
1
.
C
o
v
e
r
ag
e
an
al
y
s
is
o
f
th
e
Su
r
ab
ay
a
wea
th
er
r
a
d
ar
; (
a)
b
ea
m
s
p
r
ea
d
in
g
f
o
r
o
n
e
v
o
l
u
m
e
s
ca
n
an
d
(
b
)
b
ea
m
b
lo
c
k
in
g
a
n
aly
s
is
f
o
r
th
e
th
r
ee
lo
west e
lev
atio
n
s
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
C
h
a
llen
g
es in
r
a
d
a
r
-
b
a
s
ed
n
o
n
-
s
u
p
ercell
to
r
n
a
d
o
d
etec
tio
n
u
s
in
g
ma
ch
in
e
le
a
r
n
in
g
a
p
p
r
o
a
ch
es (
K
iki)
165
2
.
2
.
M
et
ho
ds
2
.
2
.
1
.
Ra
da
r
da
t
a
ex
t
r
a
ct
io
n
a
nd
prepro
ce
s
s
ing
T
h
e
in
itial
p
h
ase
o
f
th
e
m
eth
o
d
o
lo
g
y
in
v
o
l
v
es
p
r
ep
ar
in
g
th
e
r
aw
r
ad
ar
o
b
s
er
v
atio
n
s
f
o
r
co
m
p
u
tatio
n
al
an
aly
s
is
.
R
ad
ar
d
ata
wer
e
s
to
r
ed
in
R
ain
b
o
w
-
5
f
o
r
m
at
(
“.
v
o
l”)
.
Vo
lu
m
e
s
ca
n
s
at
ea
ch
el
ev
atio
n
an
g
le
wer
e
p
r
o
ce
s
s
ed
to
ex
t
r
ac
t r
ef
lectiv
it
y
,
r
ad
ial
v
elo
city
,
a
n
d
s
p
ec
tr
u
m
wid
th
.
E
x
t
r
ac
ted
r
a
d
ar
d
ata
wer
e
o
r
g
a
n
ized
in
to
p
o
lar
c
o
o
r
d
i
n
ates
o
f
az
im
u
th
an
g
les
an
d
r
an
g
e
b
i
n
s
.
Su
b
s
eq
u
en
tly
,
m
is
s
in
g
o
r
er
r
o
n
eo
u
s
d
ata
wer
e
id
e
n
tifie
d
an
d
r
em
o
v
ed
u
s
in
g
n
u
m
er
ical
m
ask
in
g
p
r
o
ce
d
u
r
es
to
en
s
u
r
e
d
ata
q
u
ality
f
o
r
s
u
b
s
eq
u
en
t
a
n
aly
s
is
.
T
h
e
W
r
ad
lib
p
y
th
o
n
lib
r
ar
y
is
d
o
m
in
a
n
tly
u
s
ed
in
th
is
s
tu
d
y
to
ex
t
r
ac
t r
ad
ar
d
ata
[
1
8
]
.
2
.
2
.
2
.
F
ea
t
ure
e
x
t
ra
ct
i
o
n v
ia
t
he
s
lid
i
ng
wind
o
w
t
ec
hn
i
qu
e
Fo
llo
win
g
d
ata
clea
n
in
g
,
s
p
atial
f
ea
tu
r
es
wer
e
d
er
iv
ed
t
o
ca
p
tu
r
e
th
e
m
icr
o
-
s
ca
le
s
tr
u
ctu
r
e
o
f
co
n
v
ec
tiv
e
s
to
r
m
s
.
R
ad
ar
-
d
er
i
v
ed
f
ea
tu
r
es
wer
e
co
m
p
u
ted
u
s
in
g
a
s
lid
in
g
win
d
o
w
a
p
p
r
o
ac
h
,
f
o
llo
win
g
s
im
ilar
m
eth
o
d
o
l
o
g
ies
in
p
r
ev
io
u
s
s
tu
d
ies
[
1
2
]
.
E
ac
h
r
a
d
ar
s
ca
n
was
d
iv
id
ed
in
to
o
v
e
r
lap
p
in
g
4
×4
p
ix
el
s
p
atial
b
lo
ck
s
(
~2
k
m
2
ea
ch
,
b
ased
o
n
2
5
0
m
eter
r
eso
lu
tio
n
)
.
W
ith
in
ea
c
h
b
lo
c
k
,
we
e
x
tr
ac
ted
r
e
f
lectiv
ity
f
ea
tu
r
es
(
m
ea
n
,
m
ax
im
u
m
,
m
in
im
u
m
)
to
ch
a
r
ac
ter
ize
p
r
ec
ip
itatio
n
in
ten
s
i
ty
an
d
h
eter
o
g
en
eity
,
an
d
r
a
d
ial
v
elo
city
-
b
ased
f
ea
tu
r
es
(
m
ea
n
v
elo
city
,
d
elt
a
-
V,
r
o
tatio
n
al
v
elo
city
,
a
n
g
u
lar
m
o
m
en
t
u
m
)
to
ca
p
tu
r
e
k
i
n
em
atic
p
r
o
p
er
ties
.
Sh
ea
r
an
d
v
o
r
ticity
we
r
e
d
er
i
v
ed
f
r
o
m
v
elo
city
g
r
a
d
ien
ts
to
q
u
an
tify
r
o
tatio
n
al
m
o
tio
n
,
w
h
ile
m
ea
n
s
p
ec
tr
u
m
wid
th
ass
ess
ed
t
u
r
b
u
len
ce
in
te
n
s
ity
.
T
o
en
h
an
ce
s
en
s
itiv
ity
to
s
m
al
l
-
s
c
a
l
e
v
o
r
t
i
ce
s
,
a
c
e
n
t
r
al
2
×
2
p
i
x
e
l
s
u
b
-
b
l
o
c
k
w
i
t
h
i
n
e
a
c
h
4
×
4
b
l
o
c
k
w
a
s
a
n
a
l
y
z
e
d
s
e
p
a
r
a
t
el
y
,
a
d
o
p
t
e
d
f
o
l
l
o
w
i
n
g
t
h
e
m
e
t
h
o
d
o
l
o
g
y
p
r
o
p
o
s
e
d
b
y
[
1
3
]
,
w
h
i
c
h
d
e
m
o
n
s
t
r
a
t
e
d
t
h
a
t
t
h
i
s
s
p
a
t
i
al
e
x
t
e
n
t
e
f
f
e
c
t
i
v
el
y
c
a
p
t
u
r
e
s
c
o
n
v
e
c
t
i
v
e
s
t
o
r
m
m
o
r
p
h
o
l
o
g
y
w
h
i
l
e
p
r
e
s
e
r
v
i
n
g
t
h
e
r
e
s
o
l
u
ti
o
n
o
f
f
i
n
e
-
s
c
a
l
e
v
o
r
t
e
x
s
i
g
n
a
t
u
r
es
.
L
o
c
a
l
i
z
e
d
f
e
at
u
r
e
s
m
a
x
i
m
u
m
r
e
f
l
e
c
t
i
v
it
y
(
c
4
_
z
_
m
a
x
)
,
m
a
x
i
m
u
m
r
a
d
i
a
l
v
e
l
o
c
i
ty
(
c
4
_
v
_
m
a
x
)
,
m
e
a
n
s
p
e
c
t
r
u
m
w
i
d
t
h
(
c
4
_
w
_
a
v
g
)
,
an
d
c
e
n
t
r
a
l
v
o
r
ti
c
i
t
y
(
c
4
_
v
o
r
t
i
c
i
ty
)
—
w
e
r
e
e
x
t
r
a
c
t
e
d
t
o
f
o
c
u
s
o
n
f
i
n
e
-
s
c
a
le
r
o
t
at
i
o
n
a
l
s
i
g
n
a
t
u
r
es
a
s
s
o
c
i
a
te
d
w
i
t
h
t
o
r
n
a
d
o
e
s
.
T
h
e
s
l
i
d
i
n
g
w
i
n
d
o
w
o
p
er
a
t
e
d
w
i
t
h
s
t
r
i
d
e
=
1
,
e
n
s
u
r
i
n
g
d
e
n
s
e
s
a
m
p
l
i
n
g
a
n
d
m
a
x
i
m
i
z
i
n
g
d
e
t
e
ct
i
o
n
p
r
o
b
a
b
i
li
t
y
o
f
l
o
c
a
l
iz
e
d
v
o
r
t
e
x
s
t
r
u
ct
u
r
e
s
.
Fi
g
u
r
e
2
(
a
)
i
l
l
u
s
t
r
a
t
es
t
h
e
b
l
o
c
k
o
r
g
a
n
i
z
a
t
i
o
n
,
w
h
i
l
e
Fi
g
u
r
e
2
(
b
)
d
e
m
o
n
s
t
r
a
t
es
t
h
e
s
eq
u
en
tial o
v
er
lap
p
in
g
m
o
v
em
en
t a
cr
o
s
s
th
e
r
ad
ar
f
ield
.
(
a)
(
b
)
Fig
u
r
e
2
.
Featu
r
e
ex
tr
ac
tio
n
v
i
a
s
lid
in
g
win
d
o
w
s
ch
em
e:
(
a
)
4
×4
s
lid
in
g
win
d
o
w
an
d
2
×2
c
en
ter
b
lo
ck
f
r
o
m
r
ad
ar
d
ata
a
n
d
(
b
)
illu
s
tr
atio
n
o
f
o
v
e
r
lap
p
in
g
s
lid
in
g
win
d
o
w
m
o
v
em
en
t w
ith
s
tr
id
e
=
1
p
ix
el
ac
r
o
s
s
th
e
r
ad
ar
s
ca
n
2
.
2
.
3
.
L
a
belin
g
a
nd
s
pa
t
ia
l r
ef
er
encing
T
o
en
a
b
le
s
u
p
er
v
is
ed
lear
n
in
g
,
th
e
ex
tr
ac
ted
f
ea
tu
r
es
r
eq
u
ir
ed
p
r
ec
is
e
g
eo
s
p
atial
alig
n
m
en
t
with
h
is
to
r
ical
g
r
o
u
n
d
tr
u
th
r
ec
o
r
d
s
.
R
ad
ar
b
lo
ck
s
wer
e
g
eo
s
p
atially
r
ef
er
en
ce
d
b
y
co
n
v
er
tin
g
r
ad
ar
p
o
lar
co
o
r
d
in
ates
(
r
an
g
e
an
d
az
im
u
t
h
)
in
to
g
eo
g
r
ap
h
ic
latitu
d
e
a
n
d
lo
n
g
itu
d
e
co
o
r
d
in
ates.
A
r
a
d
iu
s
-
b
ased
lab
elin
g
ap
p
r
o
ac
h
was
im
p
lem
en
ted
,
wh
er
e
b
lo
ck
s
with
in
a
7
.
5
k
m
r
ad
iu
s
f
r
o
m
th
e
d
o
cu
m
en
ted
to
r
n
ad
o
ce
n
te
r
wer
e
lab
eled
as
p
o
s
itiv
e
(
lab
el
=
1
)
.
I
n
co
n
tr
ast,
b
lo
ck
s
o
u
ts
id
e
th
is
r
ad
iu
s
wer
e
lab
eled
n
e
g
ati
v
e
(
lab
el
=
0
)
.
T
h
e
to
r
n
ad
o
ce
n
ter
was
d
eter
m
in
e
d
f
r
o
m
B
MK
G
g
r
o
u
n
d
-
tr
u
th
r
ep
o
r
ts
.
T
h
is
ap
p
r
o
ac
h
en
s
u
r
e
d
co
n
s
is
ten
t
s
p
atia
l
lab
elin
g
o
f
to
r
n
ad
o
an
d
n
o
n
-
to
r
n
ad
o
b
lo
ck
s
w
h
ile
r
ed
u
cin
g
u
n
ce
r
tain
ties
f
r
o
m
r
ad
a
r
co
v
er
a
g
e
lim
itatio
n
s
.
T
h
e
7
.
5
k
m
r
ad
i
u
s
was se
lecte
d
b
ased
o
n
a
p
r
elim
in
ar
y
s
en
s
itiv
ity
an
aly
s
is
.
W
e
ev
alu
ated
lab
el
in
g
r
ad
ii o
f
2
.
5
k
m
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
: 1
62
-
1
74
166
5
.
0
k
m
,
7
.
5
k
m
,
an
d
1
0
.
0
k
m
r
elativ
e
to
th
e
r
ep
o
r
ted
to
r
n
ad
o
ce
n
ter
.
T
h
e
7
.
5
k
m
th
r
esh
o
ld
y
ield
ed
th
e
h
ig
h
est
p
r
o
b
a
b
ilit
y
o
f
d
etec
tio
n
(
POD)
in
o
u
r
v
alid
atio
n
s
et,
o
f
f
er
in
g
th
e
o
p
tim
al
b
alan
ce
b
etwe
e
n
en
co
m
p
ass
in
g
th
e
to
r
n
ad
o
’
s
p
ar
en
t
cir
cu
latio
n
a
n
d
m
in
im
izin
g
th
e
in
clu
s
io
n
o
f
n
o
n
-
r
o
tatio
n
al
s
to
r
m
ar
ea
s
.
Fig
u
r
e
3
illu
s
tr
ates
th
e
s
p
atial
d
is
tr
ib
u
tio
n
o
f
p
o
s
itiv
e
an
d
n
e
g
ativ
e
r
ad
a
r
b
lo
c
k
s
r
elativ
e
to
th
e
to
r
n
ad
o
ce
n
ter
.
Fig
u
r
e
3
.
Sp
atial
d
is
tr
ib
u
tio
n
o
f
r
ad
ar
b
lo
ck
s
af
ter
lab
elin
g
,
r
ed
cr
o
s
s
es r
ep
r
esen
t b
lo
c
k
s
lab
eled
as to
r
n
a
d
o
(
1
)
,
g
r
ay
p
o
in
ts
r
ep
r
esen
t
n
o
n
-
to
r
n
ad
o
b
lo
ck
s
(
0
)
,
a
n
d
th
e
b
lu
e
cir
cle
m
ar
k
s
th
e
d
o
cu
m
e
n
ted
to
r
n
ad
o
ce
n
ter
2
.
2
.
4
.
M
a
chine le
a
rning
a
lg
o
rit
hm
a
nd
ba
la
ncing
t
ec
hn
iq
ues
Hav
in
g
estab
lis
h
ed
a
lab
eled
d
ataset,
th
e
s
tu
d
y
p
r
o
ce
e
d
e
d
to
im
p
lem
en
t
an
d
ev
alu
ate
s
p
ec
if
ic
class
if
icatio
n
alg
o
r
ith
m
.
T
h
is
s
tu
d
y
em
p
lo
y
ed
two
s
u
p
er
v
is
ed
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
to
class
if
y
r
ad
ar
-
d
er
iv
ed
f
ea
tu
r
es
f
o
r
to
r
n
a
d
o
d
etec
tio
n
:
R
F
an
d
XGBo
o
s
t.
T
h
e
R
F
alg
o
r
ith
m
is
an
en
s
em
b
le
lear
n
in
g
m
eth
o
d
b
ased
o
n
ag
g
r
e
g
atin
g
m
u
ltip
le
d
ec
is
io
n
tr
ee
s
[
1
9
]
,
[
2
0
]
.
E
ac
h
tr
ee
is
tr
ain
ed
o
n
a
r
an
d
o
m
b
o
o
ts
tr
ap
s
am
p
le
o
f
th
e
tr
ain
in
g
d
ata,
an
d
at
ea
ch
s
p
lit,
a
r
an
d
o
m
s
u
b
s
et
o
f
f
ea
tu
r
es
is
co
n
s
id
er
ed
.
T
h
e
f
in
al
p
r
e
d
ictio
n
is
o
b
tain
ed
b
y
m
aj
o
r
ity
v
o
tin
g
am
o
n
g
th
e
o
u
tp
u
ts
o
f
in
d
iv
i
d
u
al
tr
ee
s
.
Ma
th
em
atica
lly
,
th
e
R
F
class
i
f
ier
ca
n
b
e
e
x
p
r
ess
ed
as in
(
1
)
.
̂
=
(
ℎ
1
(
)
,
ℎ
2
(
)
,
…
.
,
ℎ
(
)
)
(
1
)
wh
er
e
ℎ
(
)
d
en
o
tes
th
e
p
r
ed
ictio
n
o
f
th
e
t
-
th
d
ec
is
io
n
tr
ee
f
o
r
in
p
u
t
f
ea
tu
r
e
v
ec
to
r
,
an
d
is
th
e
to
tal
n
u
m
b
er
o
f
tr
ee
s
.
XGBo
o
s
t
[
2
1
]
is
a
s
c
alab
le
an
d
ef
f
icien
t
im
p
lem
en
t
atio
n
o
f
g
r
ad
ie
n
t
-
b
o
o
s
ted
d
ec
is
io
n
tr
ee
s
(
GB
DT
)
.
XGBo
o
s
t
b
u
ild
s
ad
d
itiv
e
m
o
d
els
in
a
f
o
r
war
d
s
tag
e
-
wis
e
f
ash
io
n
,
wh
e
r
e
n
ew
tr
ee
s
a
r
e
f
itted
to
co
r
r
ec
t
th
e
r
esid
u
al
er
r
o
r
s
o
f
p
r
i
o
r
tr
e
es.
T
h
e
o
b
jectiv
e
f
u
n
ctio
n
ℒ
m
in
im
iz
ed
d
u
r
in
g
tr
ain
in
g
co
n
s
is
ts
o
f
a
r
eg
u
la
r
ized
lo
s
s
,
ex
p
r
ess
ed
as in
(
2
)
.
ℒ
=
∑
(
,
̂
)
=
1
+
∑
Ω
(
)
=
1
(
2
)
wh
er
e
l
(
y
i
,
y
̂
i
)
d
en
o
tes
th
e
lo
s
s
f
u
n
ctio
n
(
e.
g
.
,
lo
g
is
tic
lo
s
s
f
o
r
b
in
a
r
y
class
if
icatio
n
)
b
etwe
en
th
e
tr
u
e
lab
el
y
an
d
th
e
p
r
ed
icted
lab
el
y
̂
i
,
f
k
is
th
e
k
-
th
tr
ee
,
a
n
d
Ω(
f
)
r
e
p
r
esen
ts
a
r
eg
u
lar
izatio
n
ter
m
p
e
n
alizin
g
th
e
co
m
p
lex
ity
o
f
ea
ch
tr
ee
to
a
v
o
i
d
o
v
er
f
itti
n
g
.
B
o
th
R
F
an
d
XG
B
o
o
s
t
wer
e
tr
ain
ed
u
s
in
g
th
e
r
a
d
ar
-
d
er
i
v
ed
f
ea
tu
r
e
s
et
d
escr
ib
ed
p
r
e
v
io
u
s
ly
.
H
y
p
er
p
ar
am
eter
s
wer
e
s
elec
ted
e
m
p
ir
ically
b
ased
o
n
p
r
elim
in
a
r
y
ex
p
er
im
en
ts
:
f
o
r
R
F,
th
e
n
u
m
b
er
o
f
tr
ee
s
was
s
e
t
to
1
0
0
;
f
o
r
XGBo
o
s
t,
th
e
n
u
m
b
er
o
f
esti
m
ato
r
s
was
s
et
to
2
0
0
0
with
a
lear
n
i
n
g
r
ate
o
f
0
.
0
1
,
m
ax
im
u
m
tr
ee
d
e
p
th
o
f
5
,
an
d
a
co
l
u
m
n
s
u
b
s
am
p
lin
g
r
atio
o
f
0
.
1
.
Du
e
to
th
e
s
ev
er
e
im
b
alan
ce
b
etwe
en
p
o
s
itiv
e
(
to
r
n
ad
o
)
an
d
n
eg
ativ
e
s
am
p
les,
a
s
y
n
th
eti
c
m
in
o
r
ity
o
v
er
s
am
p
lin
g
tec
h
n
iq
u
e
(
SMOT
E
)
[
2
2
]
was
ap
p
lied
p
r
io
r
to
m
o
d
el
tr
ain
in
g
.
SMOT
E
g
en
er
ates
s
y
n
th
etic
ex
am
p
les
o
f
th
e
m
in
o
r
ity
class
b
y
in
ter
p
o
latin
g
b
etwe
en
e
x
is
tin
g
m
in
o
r
ity
class
s
am
p
les
an
d
th
eir
n
ea
r
est
n
eig
h
b
o
r
s
.
Giv
en
a
m
in
o
r
ity
s
am
p
le
x
an
d
o
n
e
o
f
its
n
ea
r
est
n
eig
h
b
o
r
s
x
nn
,
a
s
y
n
th
etic
s
am
p
le
x
new
is
g
en
er
ated
as
in
(
3
)
.
T
h
e
in
itial
d
ataset
e
x
h
ib
ited
a
s
ev
er
e
class
im
b
ala
n
ce
in
h
er
en
t
to
th
e
r
ar
e
n
atu
r
e
o
f
to
r
n
ad
o
e
v
en
ts
.
T
h
e
r
ad
ar
d
o
m
ai
n
was
g
r
id
d
e
d
in
to
a
1
0
0
0
×
1
0
0
0
m
atr
ix
,
y
i
eld
in
g
a
to
tal
o
f
1
,
0
0
0
,
0
0
0
p
o
ten
tial
s
am
p
les
p
er
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
C
h
a
llen
g
es in
r
a
d
a
r
-
b
a
s
ed
n
o
n
-
s
u
p
ercell
to
r
n
a
d
o
d
etec
tio
n
u
s
in
g
ma
ch
in
e
le
a
r
n
in
g
a
p
p
r
o
a
ch
es (
K
iki)
167
s
ca
n
.
B
ased
o
n
th
e
7
.
5
k
m
r
ad
iu
s
lab
elin
g
cr
iter
ia,
t
h
e
m
in
o
r
ity
class
(
p
o
s
itiv
e/to
r
n
ad
o
)
c
o
m
p
r
is
ed
ap
p
r
o
x
im
ately
2
,
8
0
0
s
am
p
les,
wh
er
ea
s
th
e
m
ajo
r
ity
class
(
n
eg
ativ
e/n
o
n
-
t
o
r
n
a
d
o
)
ac
c
o
u
n
te
d
f
o
r
th
e
r
em
ain
in
g
~9
9
7
,
0
0
0
s
am
p
les.
SMOT
E
g
en
er
ated
s
y
n
th
etic
m
in
o
r
ity
s
am
p
les
to
ac
h
iev
e
p
ar
ity
with
th
e
m
ajo
r
ity
class
.
C
o
n
s
eq
u
en
tly
,
th
e
p
o
s
t
-
b
alan
ci
n
g
d
ataset
co
n
s
is
ted
o
f
a
p
p
r
o
x
i
m
ately
9
9
7
,
0
0
0
p
o
s
itiv
e
s
am
p
l
es
(
co
m
b
in
in
g
r
ea
l
an
d
s
y
n
th
etic
in
s
tan
ce
s
)
an
d
9
9
7
,
0
0
0
n
e
g
ativ
e
s
am
p
les,
r
esu
ltin
g
in
a
to
tal
tr
ain
in
g
d
ataset
o
f
ap
p
r
o
x
im
ately
2
,
0
0
0
,
0
0
0
s
am
p
les.
SMOT
E
i
m
p
r
o
v
es
ac
cu
r
ac
y
in
u
n
b
alan
ce
d
d
atasets
b
y
g
en
er
atin
g
s
y
n
th
etic
s
am
p
les
th
at
m
ain
tain
th
e
s
am
e
s
tatis
tical
d
is
tr
ib
u
tio
n
as th
e
to
r
n
a
d
o
d
ata
[
2
3
]
.
=
+
(
−
)
(
3
)
wh
er
e
is
a
r
an
d
o
m
n
u
m
b
er
in
[
0
,
1
]
.
T
h
is
b
alan
cin
g
p
r
o
ce
d
u
r
e
was
ap
p
lied
o
n
ly
to
th
e
tr
ain
in
g
s
et
to
av
o
i
d
in
f
o
r
m
atio
n
leak
ag
e
i
n
to
th
e
te
s
t set.
2
.
2
.
5
.
M
o
del t
ra
ini
ng
a
nd
v
a
lid
a
t
io
n
T
h
e
co
m
p
iled
d
ataset
was d
iv
id
ed
in
to
tr
ain
in
g
an
d
test
in
g
s
u
b
s
ets u
s
in
g
a
7
0
-
3
0
% sp
lit,
s
t
r
atif
ied
b
y
class
lab
els
to
en
s
u
r
e
p
r
o
p
o
r
ti
o
n
al
r
e
p
r
esen
tatio
n
o
f
b
o
t
h
to
r
n
ad
o
an
d
n
o
n
-
t
o
r
n
ad
o
s
am
p
l
es.
B
ef
o
r
e
tr
ain
i
n
g
,
f
ea
tu
r
e
s
ca
lin
g
was
p
er
f
o
r
m
ed
u
s
in
g
th
e
Stan
d
ar
d
Scaler
to
n
o
r
m
alize
f
ea
tu
r
e
d
is
tr
ib
u
tio
n
s
,
th
er
eb
y
im
p
r
o
v
in
g
th
e
co
n
v
e
r
g
en
ce
b
eh
a
v
io
r
a
n
d
s
tab
ilit
y
o
f
th
e
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
.
T
h
e
RF
a
n
d
XG
B
o
o
s
t
m
o
d
els
wer
e
tr
ain
ed
a
n
d
v
alid
ated
o
n
th
is
b
alan
ce
d
a
n
d
s
ca
led
d
ataset.
T
ab
le
2
s
u
m
m
ar
izes
t
h
e
f
ea
t
u
r
es
ex
tr
ac
ted
f
o
r
th
is
an
aly
s
is
.
Mo
d
el
p
er
f
o
r
m
an
ce
was
q
u
an
t
itativ
ely
ass
es
s
ed
u
s
in
g
m
u
ltip
le
ev
alu
atio
n
m
etr
ics.
T
h
e
ar
ea
u
n
d
er
th
e
cu
r
v
e
(
AUC)
[
2
4
]
a
n
d
t
h
e
r
ec
e
iv
er
o
p
er
atin
g
ch
ar
ac
te
r
is
tic
(
R
OC
)
[
2
5
]
cu
r
v
e
was
u
s
ed
to
m
ea
s
u
r
e
th
e
m
o
d
el
’
s
ab
ilit
y
to
d
is
cr
im
in
ate
b
etwe
e
n
p
o
s
itiv
e
a
n
d
n
eg
ativ
e
class
es
.
T
h
e
POD,
e
q
u
iv
alen
t
to
r
ec
al
l,
was c
alcu
lated
to
q
u
an
tify
th
e
p
r
o
p
o
r
tio
n
o
f
co
r
r
ec
tly
id
e
n
tifie
d
to
r
n
ad
o
ev
e
n
ts
.
I
n
a
d
d
itio
n
,
th
e
f
alse
alar
m
r
ate
(
FAR
)
was
ev
alu
ated
to
in
d
icate
th
e
f
r
eq
u
en
cy
o
f
in
co
r
r
ec
t
p
o
s
itiv
e
p
r
ed
ictio
n
s
.
T
h
e
F1
-
s
co
r
e
was
co
m
p
u
ted
as
a
h
ar
m
o
n
ic
m
ea
n
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all,
p
r
o
v
id
in
g
a
b
alan
ce
d
m
ea
s
u
r
e
o
f
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
L
astl
y
,
a
co
n
f
u
s
io
n
m
atr
ix
was
an
aly
ze
d
to
o
f
f
er
a
co
m
p
r
eh
en
s
iv
e
v
i
ew
o
f
tr
u
e
p
o
s
itiv
e,
tr
u
e
n
eg
a
tiv
e,
f
alse
p
o
s
itiv
e,
an
d
f
alse
n
eg
at
iv
e
r
ates.
T
o
r
ig
o
r
o
u
s
ly
ass
es
s
th
e
g
en
er
aliza
tio
n
ca
p
ab
ilit
y
o
f
th
e
m
o
d
els,
a
leav
e
-
one
-
ca
s
e
-
o
u
t
(
L
OC
O)
ev
alu
atio
n
was
co
n
d
u
cted
,
wh
er
e
th
e
m
o
d
el
was
tr
ain
ed
o
n
all
b
u
t
o
n
e
ca
s
e
an
d
test
ed
o
n
th
e
r
em
ain
i
n
g
u
n
s
ee
n
ev
en
t.
T
h
is
ap
p
r
o
ac
h
en
ab
led
th
e
ev
alu
atio
n
o
f
m
o
d
el
r
o
b
u
s
tn
ess
ac
r
o
s
s
d
if
f
er
en
t
to
r
n
a
d
o
ca
s
es,
em
p
h
asizin
g
th
e
ch
allen
g
es
o
f
g
en
er
alizin
g
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els
f
o
r
s
m
all
-
s
ca
le
tr
o
p
ical
v
o
r
tices
with
h
ig
h
ly
lo
ca
lized
ch
ar
ac
ter
is
tics
.
T
ab
le
2
.
Featu
r
es e
x
tr
ac
te
d
in
th
e
s
tu
d
y
F
e
a
t
u
r
e
s
D
e
f
i
n
i
t
i
o
n
P
h
y
s
i
c
a
l
i
n
t
e
r
p
r
e
t
a
t
i
o
n
R
e
f
l
e
c
t
i
v
i
t
y
Z_
ma
x
_
1
,
Z
_
m
a
x
_
2
,
Z_
m
a
x
_
3
,
Z_
a
v
g
_
1
,
Z_
a
v
g
_
2
,
Z
_
a
v
g
_
3
,
Z_
m
i
n
_
1
,
Z_
mi
n
_
2
,
Z_
m
i
n
_
3
,
c
4
_
z
_
m
a
x
_
1
,
c
4
_
z
_
m
a
x
_
2
,
c
4
_
z
_
ma
x
_
3
R
e
f
l
e
c
t
i
v
i
t
y
(
ma
x
/
a
v
g
/
m
i
n
)
f
r
o
m
h
o
r
i
z
o
n
t
a
l
p
o
l
a
r
i
z
a
t
i
o
n
f
r
o
m
t
i
l
t
1
,
2
,
a
n
d
3
;
c
4
i
s
c
e
n
t
r
a
l
-
b
l
o
c
k
r
e
f
l
e
c
t
i
v
i
t
y
o
n
t
h
e
l
o
c
a
l
i
z
e
d
c
o
r
e
o
f
t
h
e
st
o
r
m
V
e
l
o
c
i
t
y
V
_
a
v
g
_
1
,
V
_
a
v
g
_
2
,
V
_
a
v
g
_
3
,
r
o
t
a
t
i
o
n
a
l
_
v
e
l
o
c
i
t
y
_
a
v
g
_
1
,
r
o
t
a
t
i
o
n
a
l
_
v
e
l
o
c
i
t
y
_
a
v
g
_
2
,
r
o
t
a
t
i
o
n
a
l
_
v
e
l
o
c
i
t
y
_
a
v
g
_
3
,
r
o
t
a
t
i
o
n
a
l
_
v
e
l
o
c
i
t
y
_
m
a
x
_
1
,
r
o
t
a
t
i
o
n
a
l
_
v
e
l
o
c
i
t
y
_
m
a
x
_
2
,
r
o
t
a
t
i
o
n
a
l
_
v
e
l
o
c
i
t
y
_
m
a
x
_
3
,
a
n
g
u
l
a
r
_
m
o
me
n
t
u
m
_
ma
x
_
1
,
a
n
g
u
l
a
r
_
m
o
me
n
t
u
m
_
ma
x
_
2
,
a
n
g
u
l
a
r
_
m
o
me
n
t
u
m
_
ma
x
_
3
,
a
n
g
u
l
a
r
_
m
o
me
n
t
u
m
_
a
v
g
_
1
,
a
n
g
u
l
a
r
_
m
o
me
n
t
u
m
_
a
v
g
_
2
,
a
n
g
u
l
a
r
_
m
o
me
n
t
u
m
_
a
v
g
_
3
,
d
e
l
t
a
_
V
_
1
,
d
e
l
t
a
_
V
_
2
,
d
e
l
t
a
_
V
_
3
,
s
h
e
a
r
_
mi
n
_
1
,
sh
e
a
r
_
m
i
n
_
2
,
s
h
e
a
r
_
m
i
n
_
3
,
sh
e
a
r
_
ma
x
_
1
,
sh
e
a
r
_
m
a
x
_
2
,
s
h
e
a
r
_
m
a
x
_
3
,
c
4
_
v
o
r
t
i
c
i
t
y
_
1
,
c
4
_
v
o
r
t
i
c
i
t
y
_
2
,
c
4
_
v
o
r
t
i
c
i
t
y
_
3
R
a
d
i
a
l
v
e
l
o
c
i
t
y
(
ma
x
/
m
i
n
/
a
v
g
)
f
r
o
m
h
o
r
i
z
o
n
t
a
l
p
o
l
a
r
i
z
a
t
i
o
n
f
r
o
m
t
i
l
t
1
,
2
,
a
n
d
3
;
r
o
t
a
t
i
o
n
a
l
v
e
l
o
c
i
t
y
r
e
p
r
e
se
n
t
s
t
h
e
ma
g
n
i
t
u
d
e
o
f
t
h
e
i
n
b
o
u
n
d
a
n
d
o
u
t
b
o
u
n
d
w
i
n
d
c
o
u
p
l
e
t
;
a
n
g
u
l
a
r
mo
me
n
t
u
m
a
n
d
v
o
r
t
i
c
i
t
y
r
e
p
r
e
s
e
n
t
s
t
h
e
r
a
d
i
u
s
o
f
a
r
o
t
a
t
i
n
g
c
o
l
u
m
n
c
o
n
t
r
a
c
t
s
(
st
r
e
t
c
h
i
n
g
)
,
w
i
n
d
s
p
e
e
d
s
i
n
c
r
e
a
se
;
d
e
l
t
a
_
V
a
n
d
s
h
e
a
r
c
a
p
t
u
r
e
t
h
e
l
a
t
e
r
a
l
c
h
a
n
g
e
i
n
w
i
n
d
s
p
e
e
d
o
v
e
r
a
sh
o
r
t
d
i
s
t
a
n
c
e
S
p
e
c
t
r
a
l
w
i
d
t
h
W
_
a
v
g
_
1
,
W
_
a
v
g
_
2
,
W
_
a
v
g
_
3
,
W
_
max
_
1
,
W
_
ma
x
_
2
,
W
_
ma
x
_
3
,
W
_
mi
n
_
1
,
W
_
mi
n
_
2
,
W
_
mi
n
_
3
,
c
4
_
w
_
a
v
g
_
1
,
c
4
_
w
_
a
v
g
_
2
,
c
4
_
w
_
a
v
g
_
3
,
c
4
_
w
_
m
a
x
_
1
,
c
4
_
w
_
m
a
x
_
2
,
c
4
_
w
_
ma
x
_
3
S
p
e
c
t
r
a
l
w
i
d
t
h
(
m
a
x
,
a
v
g
,
mi
n
)
f
r
o
m
h
o
r
i
z
o
n
t
a
l
p
o
l
a
r
i
z
a
t
i
o
n
f
r
o
m
t
i
l
t
1
,
2
,
a
n
d
3
;
c
4
i
s
c
e
n
t
r
a
l
-
b
l
o
c
k
r
e
f
l
e
c
t
i
v
i
t
y
o
n
t
h
e
l
o
c
a
l
i
z
e
d
c
o
r
e
o
f
t
h
e
st
o
r
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
: 1
62
-
1
74
168
I
n
a
d
d
itio
n
to
t
h
e
L
OC
O
ev
al
u
atio
n
,
in
f
er
en
ce
was
also
p
er
f
o
r
m
ed
o
n
an
in
d
ep
e
n
d
en
t
to
r
n
a
d
o
e
v
en
t
th
at
was
n
o
t
u
s
ed
d
u
r
in
g
m
o
d
el
tr
ain
in
g
.
T
h
is
in
f
e
r
en
ce
s
tep
s
er
v
ed
as
a
r
ea
l
-
wo
r
ld
v
alid
atio
n
to
f
u
r
th
er
e
x
am
in
e
th
e
m
o
d
el’
s
o
p
er
atio
n
al
ap
p
lic
ab
ilit
y
an
d
its
ab
ilit
y
to
p
r
e
d
ic
t u
n
s
ee
n
ca
s
es b
ased
s
o
lely
o
n
th
e
ex
tr
ac
ted
r
a
d
ar
f
ea
tu
r
es.
T
h
e
o
v
er
all
wo
r
k
f
l
o
w
o
f
th
e
t
o
r
n
ad
o
d
etec
tio
n
s
y
s
tem
,
en
co
m
p
ass
in
g
d
ata
p
r
ep
ar
atio
n
,
f
ea
tu
r
e
en
g
in
ee
r
in
g
,
an
d
class
if
icatio
n
s
tag
es p
r
esen
ted
in
Fig
u
r
e
4
.
Fig
u
r
e
4
.
Flo
wch
ar
t
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
lo
g
y
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
3
.
1
.
Sp
a
t
ia
l
f
ea
t
ure
dis
t
ributio
n a
nd
f
ea
t
ure
im
po
rt
a
nce
E
ac
h
p
a
n
el
in
Fig
u
r
e
s
5
(
a
)
to
(
d
)
d
is
p
lay
s
f
o
u
r
s
p
atial
m
a
p
s
o
n
th
e
lef
t:
v
o
r
ticity
,
d
elta
-
v
,
an
g
u
lar
m
o
m
en
tu
m
,
an
d
z
_
m
ax
r
ef
lec
tiv
ity
,
all
d
e
r
iv
ed
f
r
o
m
th
e
lo
west
r
ad
ar
elev
atio
n
(
0
.
5
°).
T
h
e
b
lu
e
d
o
t
in
d
icate
s
th
e
ce
n
ter
o
f
th
e
r
ep
o
r
ted
t
o
r
n
ad
o
.
R
ig
h
t
-
s
id
e
b
ar
p
lo
ts
s
h
o
w
th
e
t
o
p
f
iv
e
m
o
s
t
im
p
o
r
tan
t
f
e
atu
r
es
ac
co
r
d
i
n
g
to
th
e
R
F
an
d
XGB
m
o
d
els.
f
e
atu
r
e
s
elec
tio
n
p
atter
n
s
g
e
n
e
r
ally
em
p
h
asize
z_
m
ax
an
d
r
o
tatio
n
al
f
ea
t
u
r
es,
co
n
s
is
ten
t w
ith
s
p
atial
in
ten
s
if
icatio
n
n
ea
r
th
e
ev
en
t c
e
n
ter
.
T
h
e
an
al
y
s
is
o
f
r
ad
ar
-
d
er
iv
e
d
f
ea
tu
r
es
r
ev
ea
ls
th
at
r
e
f
lectiv
ity
(
Z
)
r
e
m
ain
s
th
e
m
o
s
t
r
o
b
u
s
t
in
d
icato
r
f
o
r
id
e
n
tify
in
g
th
e
co
r
e
o
f
to
r
n
ad
o
-
p
r
o
d
u
ci
n
g
s
to
r
m
s
in
tr
o
p
ic
al
s
ettin
g
s
.
As
v
is
u
alize
d
in
t
h
e
s
p
atial
d
is
tr
ib
u
tio
n
m
ap
s
Fig
u
r
e
s
5
(
a)
to
(
d
)
,
m
ax
i
m
u
m
r
ef
lectiv
ity
(
Z
_
m
ax
)
c
o
n
s
is
ten
tly
ex
h
ib
its
co
n
ce
n
tr
ated
h
ig
h
v
alu
es
in
th
e
im
m
ed
iate
v
icin
ity
o
f
t
h
e
r
e
p
o
r
ted
to
r
n
ad
o
ce
n
ter
s
ac
r
o
s
s
all
f
o
u
r
ca
s
es.
T
h
is
alig
n
s
with
th
e
p
h
y
s
ical
u
n
d
er
s
tan
d
i
n
g
th
at
s
tr
o
n
g
co
n
v
ec
tiv
e
u
p
d
r
af
ts
,
m
a
n
if
est
ed
as
h
ig
h
r
ef
lectiv
ity
co
r
es,
ar
e
a
p
r
er
e
q
u
is
ite
f
o
r
s
u
s
tain
in
g
th
ese
s
to
r
m
s
y
s
tem
s
.
I
n
co
n
tr
ast,
d
y
n
am
ic
f
ea
t
u
r
es
s
u
ch
as
v
o
r
ticity
,
d
elta
-
V,
a
n
d
an
g
u
lar
m
o
m
en
tu
m
s
h
o
w
s
ig
n
if
ican
t
v
ar
iab
ilit
y
b
etwe
en
ev
en
ts
.
W
h
ile
lo
ca
lized
clu
s
ter
s
o
f
h
ig
h
v
o
r
ticity
wer
e
d
is
tin
ct
in
th
e
ca
s
es
o
f
J
an
u
ar
y
1
an
d
J
an
u
ar
y
1
7
,
th
e
s
ig
n
atu
r
es
we
r
e
m
u
c
h
m
o
r
e
d
is
p
er
s
ed
o
r
wea
k
er
in
th
e
o
th
e
r
two
ca
s
es.
T
h
is
in
co
n
s
is
ten
cy
h
ig
h
lig
h
ts
th
e
ch
allen
g
e
o
f
d
etec
tin
g
n
o
n
-
s
u
p
er
ce
ll
t
o
r
n
a
d
o
es
co
m
p
ar
e
d
to
s
u
p
er
ce
ll
c
o
u
n
ter
p
a
r
ts
,
wh
ich
ty
p
ically
p
o
s
s
ess
co
h
er
en
t a
n
d
p
er
s
is
ten
t m
eso
cy
clo
n
es.
T
h
e
f
ea
tu
r
e
im
p
o
r
ta
n
ce
ev
alu
a
tio
n
u
s
in
g
R
F
an
d
XGB
f
u
r
th
e
r
co
r
r
o
b
o
r
ates
th
ese
s
p
atial
o
b
s
er
v
atio
n
s
.
As
s
h
o
wn
in
th
e
b
ar
ch
ar
ts
in
F
ig
u
r
e
s
5
(
a)
to
(
d
)
,
b
o
th
m
o
d
els
h
ea
v
ily
p
r
io
r
itize
r
ef
lectiv
ity
-
b
ased
f
ea
tu
r
es
(
e.
g
.
,
Z
_
m
ax
,
Z
_
av
g
)
as
th
e
p
r
im
ar
y
p
r
e
d
icto
r
s
.
Ho
wev
er
,
th
e
m
o
d
els
d
iv
er
g
e
in
t
h
eir
s
ec
o
n
d
a
r
y
s
elec
tio
n
s
;
XGB
ten
d
s
to
ass
ig
n
h
ig
h
er
im
p
o
r
tan
ce
to
r
o
tatio
n
al
v
elo
city
a
n
d
a
n
g
u
lar
m
o
m
en
tu
m
d
er
iv
e
d
f
r
o
m
h
ig
h
er
elev
atio
n
s
,
wh
er
ea
s
R
F
f
o
cu
s
es
m
o
r
e
o
n
n
ea
r
-
s
u
r
f
ac
e
s
tatis
tics
.
T
h
is
d
iv
er
g
en
ce
s
u
g
g
ests
th
at
wh
ile
th
er
m
o
d
y
n
am
ic
in
ten
s
ity
(
r
ef
lectiv
ity
)
is
a
u
n
iv
er
s
al
p
r
ed
icto
r
,
th
e
k
i
n
em
atic
(
r
o
tatio
n
al)
s
ig
n
atu
r
es
o
f
I
n
d
o
n
esian
to
r
n
ad
o
es
a
r
e
to
o
v
ar
iab
le
t
o
b
e
ca
p
tu
r
ed
u
n
i
f
o
r
m
ly
b
y
d
i
f
f
er
en
t a
l
g
o
r
ith
m
s
.
3
.
2
.
M
o
del per
f
o
rma
nce
a
n
d o
pera
t
io
na
l t
ra
de
-
o
f
f
s
T
h
e
q
u
an
titativ
e
ev
al
u
atio
n
f
o
cu
s
ed
o
n
c
o
m
p
ar
i
n
g
th
e
p
r
e
d
ictiv
e
p
er
f
o
r
m
a
n
ce
o
f
r
an
d
o
m
f
o
r
est
(
T
DA
-
R
F)
an
d
XGBo
o
s
t
(
T
DA
-
XGB)
m
o
d
els
ac
r
o
s
s
f
o
u
r
t
o
r
n
ad
o
ca
s
es,
u
s
in
g
R
OC
cu
r
v
es
an
d
a
s
u
ite
o
f
ev
alu
atio
n
m
et
r
ics
in
clu
d
in
g
AUC,
F1
-
s
co
r
e,
POD,
FAR
,
an
d
co
n
f
u
s
io
n
m
atr
ices
Fig
u
r
es
6
(
a)
to
(
d
)
.
Acr
o
s
s
all
ca
s
es,
b
o
th
m
o
d
els
ac
h
iev
e
d
co
n
s
is
ten
tly
h
i
g
h
AUC
v
alu
es
ab
o
v
e
0
.
9
4
,
in
d
icatin
g
e
x
ce
llen
t
d
is
cr
im
in
atio
n
ca
p
ab
ilit
y
b
etwe
en
to
r
n
a
d
o
an
d
n
o
n
-
to
r
n
a
d
o
s
am
p
les.
No
tab
ly
,
T
DA
-
R
F
ten
d
ed
to
y
ield
h
ig
h
er
F1
-
s
co
r
es
in
th
r
e
e
o
u
t
o
f
f
o
u
r
ca
s
es,
s
u
g
g
esti
n
g
a
b
etter
b
alan
ce
b
etwe
en
p
r
ec
is
io
n
an
d
r
ec
all.
Fo
r
in
s
tan
ce
,
in
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
C
h
a
llen
g
es in
r
a
d
a
r
-
b
a
s
ed
n
o
n
-
s
u
p
ercell
to
r
n
a
d
o
d
etec
tio
n
u
s
in
g
ma
ch
in
e
le
a
r
n
in
g
a
p
p
r
o
a
ch
es (
K
iki)
169
SB
Y_
2
0
2
4
0
1
0
1
,
T
DA
-
R
F
ac
h
iev
ed
an
F1
-
s
co
r
e
o
f
0
.
6
2
9
wi
th
a
POD
o
f
0
.
6
1
1
an
d
FAR
o
f
0
.
3
5
3
,
co
m
p
ar
e
d
to
T
DA
-
XGB’s lo
wer
F1
-
s
co
r
e
o
f
0
.
5
9
5
d
esp
ite
a
h
ig
h
er
PO
D
o
f
0
.
7
2
2
.
Ho
wev
er
,
a
ty
p
ical
p
atter
n
e
m
er
g
ed
wh
er
e
T
DA
-
XGB
s
h
o
wed
s
u
p
er
io
r
POD,
ca
p
tu
r
i
n
g
m
o
r
e
to
r
n
ad
o
ca
s
es
(
tr
u
e
p
o
s
itiv
es),
alb
eit
at
h
ig
h
e
r
FAR
.
Fo
r
e
x
a
m
p
le,
o
n
2
0
2
4
-
01
-
1
7
,
T
DA
-
X
GB
r
ea
ch
ed
a
POD
o
f
0
.
7
2
3
b
u
t
with
a
FAR
o
f
0
.
6
4
7
,
wh
ile
T
DA
-
R
F
o
f
f
er
e
d
a
m
o
r
e
co
n
s
e
r
v
ativ
e
p
er
f
o
r
m
a
n
c
e
with
a
lo
wer
FA
R
o
f
0
.
3
4
1
b
u
t a
ls
o
s
lig
h
tly
lo
we
r
POD
(
0
.
6
5
1
)
.
T
h
is
tr
ad
e
-
o
f
f
h
ig
h
lig
h
ts
th
e
cr
itical
b
alan
ce
b
etwe
en
s
en
s
itiv
ity
an
d
s
p
ec
if
icity
in
o
p
e
r
atio
n
al
ap
p
licatio
n
s
.
I
n
ter
esti
n
g
ly
,
i
n
2
0
2
4
-
02
-
0
4
,
wh
ic
h
h
ad
a
m
u
c
h
lar
g
er
test
s
a
m
p
le
s
ize,
b
o
th
m
o
d
els
ac
h
iev
ed
r
elativ
ely
b
alan
ce
d
AUCs
(
0
.
9
4
6
f
o
r
T
DA
-
R
F
an
d
0
.
9
4
2
f
o
r
T
DA
-
XGB).
Ho
wev
er
,
T
DA
-
XGB’s
FA
R
s
o
ar
ed
to
0
.
6
9
3
,
s
ig
n
alin
g
p
o
ten
tial
o
v
e
r
-
s
en
s
itiv
ity
wh
en
e
x
p
o
s
ed
to
lar
g
e
im
b
alan
ce
d
d
atasets
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
5
.
Sp
atial
d
is
tr
ib
u
tio
n
o
f
r
ad
ar
-
d
er
iv
e
d
f
ea
tu
r
es a
n
d
f
e
atu
r
e
im
p
o
r
tan
ce
f
o
r
ea
ch
t
o
r
n
ad
o
ca
s
e:
(
a)
2
0
2
4
-
01
-
0
1
,
(
b
)
2
0
2
4
-
01
-
1
7
,
(
c)
2
0
2
4
-
01
-
2
6
,
an
d
(
d
)
2
0
2
4
-
02
-
02
T
h
ese
r
esu
lts
u
n
d
e
r
s
co
r
e
th
e
ch
allen
g
e
o
f
ac
h
iev
in
g
g
e
n
er
aliza
b
le
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
h
ig
h
ly
lo
ca
lized
an
d
s
m
all
-
s
ca
le
to
r
n
a
d
o
ev
e
n
ts
.
W
h
ile
en
s
em
b
le
m
e
th
o
d
s
lik
e
RF
o
f
f
er
s
tab
ilit
y
,
X
GB
o
o
s
t
m
ay
b
etter
ca
p
tu
r
e
r
ar
e
ev
e
n
ts
,
alb
eit
with
a
h
ig
h
er
r
is
k
o
f
f
alse
p
o
s
itiv
es.
T
h
e
v
ar
iab
ilit
y
in
p
er
f
o
r
m
a
n
c
e
ac
r
o
s
s
ca
s
es
a
ls
o
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N:
1
6
9
3
-
6
9
3
0
TEL
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
2
4
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
6
: 1
62
-
1
74
170
s
u
p
p
o
r
ts
th
e
n
ee
d
f
o
r
tailo
r
ed
m
o
d
el
ca
lib
r
atio
n
an
d
p
o
s
s
ib
ly
s
p
atio
tem
p
o
r
al
s
tr
atif
icatio
n
d
u
r
in
g
m
o
d
el
tr
ai
n
in
g
an
d
d
e
p
lo
y
m
e
n
t.
F
r
o
m
a
n
o
p
e
r
a
t
i
o
n
a
l
p
e
r
s
p
e
c
t
i
v
e
,
t
h
i
s
t
r
a
d
e
-
o
f
f
d
i
c
t
a
t
e
s
d
i
s
t
i
n
c
t
d
e
p
l
o
y
m
e
n
t
s
t
r
a
t
e
g
i
e
s
d
e
p
e
n
d
i
n
g
o
n
t
h
e
f
o
r
e
c
a
s
t
i
n
g
p
h
i
l
o
s
o
p
h
y
.
A
h
i
g
h
-
s
e
n
s
i
t
i
v
i
t
y
m
o
d
e
l
l
i
k
e
T
D
A
-
X
G
B
a
c
t
s
a
s
a
n
e
f
f
e
c
t
i
v
e
“
s
a
f
e
t
y
n
e
t
”
o
r
“
e
a
r
l
y
v
i
g
i
l
a
n
c
e
”
t
o
o
l
,
e
n
s
u
r
i
n
g
t
h
a
t
f
o
r
e
c
a
s
t
e
r
s
a
r
e
a
l
e
r
t
e
d
t
o
m
o
s
t
p
o
t
e
n
t
i
a
l
t
h
r
e
a
t
s
.
H
o
w
e
v
e
r
,
i
t
s
h
i
g
h
F
A
R
i
m
p
l
i
e
s
t
h
a
t
i
t
c
a
n
n
o
t
b
e
u
s
e
d
a
s
a
f
u
l
l
y
a
u
t
o
m
a
t
e
d
w
a
r
n
i
n
g
t
r
i
g
g
e
r
,
a
s
t
h
i
s
w
o
u
l
d
l
i
k
e
l
y
l
e
a
d
t
o
“
w
a
r
n
i
n
g
f
a
t
i
g
u
e
”
a
m
o
n
g
t
h
e
p
u
b
l
i
c
.
C
o
n
v
e
r
s
e
l
y
,
t
h
e
m
o
r
e
c
o
n
s
e
r
v
a
t
i
v
e
T
D
A
-
R
F
,
w
i
t
h
i
t
s
l
o
w
e
r
F
A
R
,
i
s
b
e
t
t
e
r
s
u
i
t
e
d
f
o
r
s
c
e
n
a
r
i
o
s
w
h
e
r
e
m
a
i
n
t
a
i
n
i
n
g
p
u
b
l
i
c
t
r
u
s
t
a
n
d
r
e
d
u
c
i
n
g
f
a
l
s
e
p
o
s
i
t
i
v
e
s
i
s
p
r
i
o
r
i
t
i
z
e
d
,
t
h
o
u
g
h
i
t
c
a
r
r
i
e
s
t
h
e
r
i
s
k
o
f
m
i
s
s
i
n
g
w
e
a
k
e
r
e
v
e
n
t
s
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
6
.
R
OC
cu
r
v
es a
n
d
ev
a
lu
atio
n
m
etr
ics f
o
r
r
an
d
o
m
f
o
r
est (
T
DA
-
R
F)
an
d
XGBo
o
s
t (
T
DA
-
XGB)
m
o
d
els u
s
in
g
th
e
t
o
p
f
ea
tu
r
es o
f
ea
ch
ca
s
e:(
a)
2
0
2
4
-
01
-
0
1
,
(
b
)
2
0
2
4
-
01
-
1
7
,
(
c)
2
0
2
4
-
01
-
2
6
,
an
d
(
d
)
2
0
2
4
-
02
-
02
C
o
m
p
ar
ed
t
o
s
tu
d
ies
lik
e
Z
e
n
g
et
a
l.
[2
6
]
,
wh
ic
h
a
p
p
lied
m
ac
h
in
e
lear
n
in
g
to
d
etec
t
lar
g
e
-
s
ca
le
s
u
p
er
ce
ll
to
r
n
ad
o
es
u
s
in
g
d
u
al
-
p
o
lar
izatio
n
r
a
d
ar
an
d
e
n
v
ir
o
n
m
en
tal
d
ata,
o
u
r
m
o
d
els
tr
ain
ed
s
o
lely
o
n
s
in
g
le
-
p
o
lar
izatio
n
r
a
d
ar
f
ea
tu
r
es
d
e
m
o
n
s
tr
ated
co
m
p
ar
ab
le
AUC
b
u
t
g
en
er
ally
l
o
wer
F1
-
s
co
r
es
an
d
h
ig
h
e
r
FAR
s
.
T
h
is
d
is
cr
ep
an
cy
ar
is
es
f
r
o
m
th
e
in
h
er
e
n
tly
s
m
all
s
ca
le,
s
h
o
r
t
d
u
r
atio
n
,
an
d
d
is
o
r
g
an
ize
d
n
atu
r
e
o
f
to
r
n
ad
o
ev
en
ts
in
I
n
d
o
n
esia,
wh
ich
lack
th
e
c
o
h
er
e
n
t r
o
tatio
n
al
s
ig
n
a
tu
r
es f
o
u
n
d
i
n
tem
p
er
ate
s
u
p
e
r
ce
lls
.
Ad
d
itio
n
ally
,
lim
ited
s
am
p
le
s
ize
an
d
lab
el
u
n
ce
r
tain
ty
f
r
o
m
f
ield
-
b
a
s
ed
r
ep
o
r
ts
p
r
esen
t
r
o
b
u
s
t
m
o
d
el
lear
n
in
g
an
d
g
en
er
aliza
tio
n
ch
allen
g
es.
W
h
ile
RF
an
d
XGBo
o
s
t
s
h
o
we
d
p
r
o
m
is
in
g
r
esu
lts
,
p
ar
ticu
la
r
ly
in
r
ec
all,
th
eir
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
C
h
a
llen
g
es in
r
a
d
a
r
-
b
a
s
ed
n
o
n
-
s
u
p
ercell
to
r
n
a
d
o
d
etec
tio
n
u
s
in
g
ma
ch
in
e
le
a
r
n
in
g
a
p
p
r
o
a
ch
es (
K
iki)
171
p
er
f
o
r
m
an
ce
v
ar
ied
s
ig
n
if
ica
n
tly
ac
r
o
s
s
ca
s
es,
r
ef
lectin
g
t
h
e
d
if
f
icu
lty
o
f
b
u
ild
in
g
u
n
iv
er
s
al
d
etec
to
r
s
f
o
r
tr
o
p
ical
to
r
n
a
d
o
es with
o
u
t
b
r
o
ad
er
en
v
i
r
o
n
m
e
n
tal
co
n
tex
t
o
r
h
ig
h
-
r
eso
lu
tio
n
m
u
lti
-
s
o
u
r
ce
d
ata.
B
o
th
m
o
d
els
ac
h
iev
ed
h
ig
h
A
UC
v
alu
es
ac
r
o
s
s
all
ca
s
es.
T
DA
-
XGB
g
en
er
ally
o
f
f
er
ed
a
h
ig
h
er
POD
at
th
e
ex
p
en
s
e
o
f
an
in
cr
ea
s
e
d
FAR
,
wh
ile
T
DA
-
R
F
p
r
o
v
id
ed
m
o
r
e
b
alan
ce
d
p
e
r
f
o
r
m
a
n
ce
.
T
h
ese
r
esu
lts
h
ig
h
lig
h
t
th
e
tr
a
d
e
-
o
f
f
b
etwe
e
n
s
en
s
itiv
ity
an
d
s
p
ec
if
icity
wh
e
n
d
etec
tin
g
s
m
all
-
s
ca
le,
n
o
n
-
s
u
p
er
ce
ll
to
r
n
ad
o
es
f
r
o
m
r
a
d
ar
-
d
er
iv
ed
f
ea
tu
r
es.
T
o
f
u
r
th
er
e
v
alu
ate
th
e
g
en
er
aliza
tio
n
ab
ilit
y
o
f
th
e
m
o
d
el
s
,
a
L
OC
O
ev
alu
atio
n
was
co
n
d
u
cted
i
n
wh
ich
ea
ch
o
f
th
e
f
o
u
r
ca
s
es wa
s
h
eld
o
u
t a
s
th
e
test
s
et
wh
ile
th
e
r
em
ain
in
g
th
r
ee
wer
e
u
s
ed
f
o
r
tr
ain
in
g
.
T
h
e
r
esu
lts
o
f
th
is
r
ig
o
r
o
u
s
ev
alu
atio
n
ar
e
p
r
esen
t
ed
in
Fig
u
r
e
7
.
L
OC
O
r
esu
lt
s
r
ev
ea
l
a
n
o
tab
l
e
d
ec
lin
e
in
o
v
er
all
d
etec
tio
n
p
er
f
o
r
m
a
n
ce
co
m
p
a
r
ed
to
th
e
with
in
-
ca
s
e
ev
alu
ati
o
n
s
.
W
h
ile
th
e
ar
ea
u
n
d
e
r
th
e
R
OC
cu
r
v
e
(
AUC)
r
em
ain
ed
r
elativ
el
y
s
tab
le
ab
o
v
e
0
.
8
5
f
o
r
m
o
s
t
ca
s
es
b
o
th
r
an
d
o
m
f
o
r
est
(
T
DA
-
R
F)
an
d
XG
B
o
o
s
t
(
T
DA
-
XGB)
s
u
f
f
er
ed
s
ig
n
if
ican
t d
r
o
p
s
in
F1
-
s
co
r
es
an
d
POD.
T
DA
-
XGB
g
en
er
ally
ac
h
iev
e
d
h
ig
h
e
r
P
OD
ac
r
o
s
s
all
ca
s
es,
in
d
icatin
g
s
tr
o
n
g
er
s
en
s
itiv
ity
to
to
r
n
ad
o
o
cc
u
r
r
en
ce
s
;
h
o
w
ev
er
,
th
is
ca
m
e
at
th
e
co
s
t
o
f
s
u
b
s
tan
tially
h
ig
h
er
FAR
,
in
s
o
m
e
ca
s
es
ex
ce
e
d
in
g
7
0
%.
Fo
r
in
s
tan
ce
,
wh
en
SB
Y
_
2
0
2
4
0
1
1
7
was
u
s
ed
as
th
e
u
n
s
ee
n
test
s
et,
T
DA
-
XGB
r
ea
ch
ed
a
POD
o
f
0
.
6
6
7
b
u
t
with
a
FAR
o
f
0
.
7
2
3
,
h
ig
h
l
ig
h
tin
g
th
e
ten
d
e
n
cy
o
f
th
e
m
o
d
el
to
o
v
er
-
p
r
ed
ict
p
o
s
itiv
es.
Me
an
wh
ile,
T
DA
-
R
F
s
h
o
wed
m
o
r
e
co
n
s
er
v
ativ
e
b
eh
av
io
r
,
p
r
o
d
u
c
in
g
lo
wer
FAR
b
u
t
f
ailin
g
to
d
etec
t
a
s
ig
n
if
ican
t p
o
r
tio
n
o
f
to
r
n
a
d
o
in
s
tan
ce
s
.
T
h
ese
f
in
d
in
g
s
h
av
e
s
ig
n
if
ica
n
t
im
p
licatio
n
s
f
o
r
o
p
er
atio
n
a
l
im
p
lem
en
tatio
n
in
tr
o
p
ical
r
eg
io
n
s
lik
e
I
n
d
o
n
esia.
T
h
e
cu
r
r
en
t
p
er
f
o
r
m
an
ce
lev
els
s
u
g
g
est
th
at
th
ese
m
ac
h
in
e
lear
n
i
n
g
(
ML
)
m
o
d
els
ar
e
b
est
u
tili
ze
d
as
d
ec
is
io
n
-
s
u
p
p
o
r
t
to
o
ls
r
at
h
er
th
an
s
tan
d
alo
n
e
a
u
to
m
ated
war
n
in
g
s
y
s
tem
s
.
Giv
en
th
e
s
h
o
r
t
lead
tim
es
o
f
to
r
n
ad
o
e
v
en
ts
,
th
e
m
o
d
el
o
u
t
p
u
ts
ca
n
s
er
v
e
as
a
“f
ir
s
t
-
g
u
ess
”
g
u
id
an
ce
f
ield
,
d
r
awin
g
th
e
f
o
r
ec
aster
’
s
atten
tio
n
to
s
p
ec
if
ic
s
to
r
m
ce
lls
th
at
ex
h
ib
it
m
icr
o
-
s
ca
le
r
o
tatio
n
al
c
h
ar
ac
ter
is
tics
o
f
ten
i
n
v
is
ib
le
t
o
th
e
n
a
k
ed
e
y
e
o
n
s
tan
d
ar
d
r
a
d
ar
d
is
p
lay
s
.
B
y
in
t
eg
r
atin
g
t
h
e
ML
p
r
o
b
ab
ilit
y
m
ap
s
with
en
v
i
r
o
n
m
e
n
tal
an
aly
s
is
,
f
o
r
ec
aster
s
ca
n
f
ilter
o
u
t
th
e
f
alse
alar
m
s
g
e
n
er
ated
b
y
T
DA
-
XGB,
ef
f
ec
t
iv
ely
co
m
b
i
n
in
g
h
u
m
a
n
ex
p
e
r
tis
e
with
m
ac
h
in
e
s
en
s
itiv
ity
.
Fig
u
r
e
7
.
Mo
d
el
p
er
f
o
r
m
a
n
ce
u
s
in
g
L
OC
O
ev
alu
atio
n
f
o
r
b
o
th
T
DA
-
R
F a
n
d
T
DA
-
XGB
T
h
ese
r
esu
lts
d
em
o
n
s
tr
ate
th
e
co
m
p
le
x
ity
o
f
g
en
er
alizin
g
r
ad
ar
-
b
ased
m
ac
h
in
e
lear
n
i
n
g
m
o
d
els
to
d
etec
t
h
ig
h
ly
lo
ca
lized
an
d
s
h
o
r
t
-
liv
ed
tr
o
p
ical
v
o
r
tex
p
h
e
n
o
m
en
a.
Un
lik
e
s
ig
n
if
ican
t
s
u
p
er
ce
ll
to
r
n
ad
o
es,
wh
ich
ten
d
to
e
x
h
ib
it
co
n
s
is
t
en
t
s
p
atial
an
d
s
tr
u
ctu
r
al
r
a
d
ar
p
atter
n
s
,
to
r
n
ad
o
es
in
I
n
d
o
n
esia
s
h
o
w
ca
s
e
-
d
ep
en
d
e
n
t
v
a
r
iab
ilit
y
in
b
o
th
s
tr
u
ctu
r
e
a
n
d
r
ad
ar
s
ig
n
atu
r
e,
lim
itin
g
cr
o
s
s
-
ca
s
e
m
o
d
el
p
er
f
o
r
m
an
ce
.
T
h
e
L
OC
O
ev
alu
atio
n
h
ig
h
lig
h
ts
th
e
n
ee
d
f
o
r
f
u
tu
r
e
r
esear
ch
o
n
in
co
r
p
o
r
atin
g
d
o
m
ain
ad
a
p
tatio
n
,
tem
p
o
r
al
en
s
em
b
le
m
eth
o
d
s
,
o
r
ad
d
itio
n
al
atm
o
s
p
h
er
ic
p
r
e
d
icto
r
s
to
b
o
o
s
t g
en
er
aliza
tio
n
ac
r
o
s
s
o
p
er
atio
n
ally
d
iv
er
s
e
co
n
d
itio
n
s
.
T
h
e
L
OC
O
ev
alu
atio
n
r
esu
lt
s
f
u
r
th
er
h
i
g
h
lig
h
t
th
e
c
h
allen
g
es
in
b
u
ild
in
g
g
e
n
er
aliza
b
l
e
m
ac
h
in
e
lear
n
in
g
m
o
d
els
f
o
r
to
r
n
ad
o
e
s
in
tr
o
p
ical
e
n
v
ir
o
n
m
en
ts
li
k
e
I
n
d
o
n
esia.
Un
lik
e
th
e
wit
h
in
-
ca
s
e
ev
alu
atio
n
,
wh
er
e
th
e
m
o
d
els
p
er
f
o
r
m
e
d
well
o
n
d
ata
d
r
awn
f
r
o
m
th
e
s
am
e
ev
en
t,
L
OC
O
r
esu
lts
r
e
v
ea
led
p
e
r
f
o
r
m
an
ce
Evaluation Warning : The document was created with Spire.PDF for Python.