I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
11
,
No
.
3
,
J
u
n
e
2021
,
p
p
.
2500
~
2
5
0
7
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
1
i
3
.
pp
2
5
0
0
-
2
5
0
7
2500
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Co
m
pa
ra
tive
a
na
ly
sis
of
m
ultiple
c
la
ss
ificatio
n
m
o
d
e
ls to
i
m
pro
v
e
PM1
0
p
r
ediction
p
erfor
ma
nce
Yo
ng
-
J
in J
un
g
,
K
y
o
un
g
-
Wo
o
Cho
,
J
o
ng
-
Su
ng
L
ee
,
Cha
n
g
-
H
eo
n O
h
De
p
a
rt
m
e
n
t
o
f
El
e
c
tri
c
a
l
,
El
e
c
tro
n
ics
a
n
d
C
o
m
m
u
n
ica
ti
o
n
E
n
g
in
e
e
rin
g
,
Ko
re
a
Un
iv
e
rsity
o
f
T
e
c
h
n
o
lo
g
y
a
n
d
Ed
u
c
a
ti
o
n
(KO
REAT
ECH),
Ko
re
a
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 3
1
,
2
0
2
0
R
ev
i
s
ed
Sep
2
2
,
2
0
2
0
A
cc
ep
ted
Oct
1
4
,
2
0
2
0
W
it
h
th
e
in
c
re
a
sin
g
re
q
u
irem
e
n
t
o
f
h
ig
h
a
c
c
u
ra
c
y
f
o
r
p
a
rti
c
u
late
m
a
tt
e
r
p
re
d
ictio
n
,
v
a
rio
u
s
a
tt
e
m
p
ts
h
a
v
e
b
e
e
n
m
a
d
e
to
im
p
ro
v
e
p
re
d
icti
o
n
a
c
c
u
ra
c
y
b
y
a
p
p
ly
in
g
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s.
Ho
w
e
v
e
r,
th
e
c
h
a
ra
c
teristics
o
f
p
a
rti
c
u
late
m
a
tt
e
r
a
n
d
t
h
e
p
r
o
b
lem
o
f
th
e
o
c
c
u
rre
n
c
e
ra
te
b
y
c
o
n
c
e
n
tratio
n
m
a
k
e
it
d
iff
icu
lt
to
train
p
re
d
icti
o
n
m
o
d
e
ls,
re
su
lt
i
n
g
in
p
o
o
r
p
re
d
ictio
n
.
I
n
o
rd
e
r
to
s
o
lv
e
th
is
p
r
o
b
lem
,
in
th
i
s
p
a
p
e
r,
w
e
p
ro
p
o
se
d
m
u
lt
ip
le
c
la
ss
if
ic
a
ti
o
n
m
o
d
e
ls
f
o
r
p
re
d
ictin
g
p
a
rti
c
u
late
m
a
tt
e
r
c
o
n
c
e
n
tratio
n
s
re
q
u
ired
f
o
r
p
re
d
ictio
n
b
y
d
iv
id
in
g
th
e
m
in
to
A
QI
-
b
a
se
d
c
las
s
e
s.
We
d
e
sig
n
e
d
m
u
lt
ip
le
c
las
si
f
ica
ti
o
n
m
o
d
e
ls
u
sin
g
l
o
g
isti
c
re
g
re
ss
io
n
,
d
e
c
isio
n
tree
,
S
V
M
a
n
d
e
n
se
m
b
le
a
m
o
n
g
th
e
v
a
rio
u
s
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
ri
th
m
s.
T
h
e
c
o
m
p
a
riso
n
re
su
lt
s
o
f
th
e
p
e
rf
o
r
m
a
n
c
e
o
f
th
e
f
o
u
r
c
las
sif
ic
a
ti
o
n
m
o
d
e
ls
th
ro
u
g
h
e
rro
r
m
a
tri
c
e
s
c
o
n
f
ir
m
e
d
th
e
f
-
sc
o
re
o
f
0
.
8
2
o
r
h
ig
h
e
r
f
o
r
a
ll
th
e
m
o
d
e
ls
o
th
e
r
t
h
a
n
th
e
lo
g
isti
c
re
g
re
ss
io
n
m
o
d
e
l.
K
ey
w
o
r
d
s
:
C
las
s
i
f
icatio
n
D
ec
is
io
n
tr
ee
E
n
s
e
m
b
le
L
o
g
i
s
tic
r
eg
r
e
s
s
io
n
Ma
ch
i
n
e
lear
n
i
n
g
P
ar
ticu
late
m
atter
S
u
p
p
o
r
t v
ec
to
r
m
ac
h
i
n
e
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
:
C
h
a
n
g
-
Heo
n
O
h
Dep
ar
t
m
en
t o
f
E
lectr
ical,
E
lectr
o
n
ics an
d
C
o
m
m
u
n
icat
io
n
E
n
g
i
n
ee
r
i
n
g
Ko
r
ea
Un
iv
er
s
i
t
y
o
f
T
ec
h
n
o
lo
g
y
a
n
d
E
d
u
ca
tio
n
(
KO
R
E
A
T
E
C
H)
1
6
0
0
,
C
h
u
n
g
j
eo
l
-
r
o
,
B
y
eo
n
g
c
h
eo
n
-
m
y
eo
n
,
Do
n
g
n
a
m
-
g
u
,
C
h
eo
n
a
n
-
s
i,
C
h
u
n
g
c
h
eo
n
g
n
a
m
-
d
o
,
3
1
2
5
3
R
ep
u
b
lic
o
f
Ko
r
ea
E
m
ail:
c
h
o
h
@
k
o
r
ea
tec
h
.
ac
.
k
r
1.
I
NT
RO
D
UCT
I
O
N
P
ar
ticu
late
m
atter
is
a
s
u
b
s
t
an
ce
m
ad
e
u
p
o
f
v
ar
io
u
s
s
iz
es,
s
h
ap
es,
an
d
i
n
g
r
ed
ie
n
ts
.
P
ar
ticu
late
m
atter
,
d
iv
id
ed
in
to
10
,
2
.
5
ac
co
r
d
i
n
g
to
th
e
s
ize
o
f
1
0
,
2
.
5
o
r
less
,
af
f
ec
ts
o
u
r
h
ea
lth
b
y
ca
u
s
in
g
s
o
m
e
d
is
ea
s
es
s
u
c
h
a
s
ca
r
d
io
v
ascu
lar
,
r
esp
ir
ato
r
y
,
a
n
d
c
er
eb
r
o
v
ascu
lar
d
is
ea
s
e
s
.
A
cc
o
r
d
in
g
l
y
,
p
ar
tic
u
late
m
atter
w
as
c
lass
if
ied
as
a
d
a
n
g
er
o
u
s
s
u
b
s
ta
n
ce
,
an
d
it
i
s
a
n
al
y
ze
d
a
s
th
e
ca
u
s
e
o
f
d
ec
r
ea
s
in
g
t
h
e
v
i
talit
y
o
f
s
o
ciet
y
m
e
m
b
er
s
[
1
-
7
]
.
I
n
o
r
d
er
to
av
o
id
s
u
ch
h
ar
m
f
u
l
ef
f
ec
ts
o
f
p
ar
ticu
late
m
a
tter
as
m
u
c
h
as
p
o
s
s
ib
le,
it
h
as
b
ec
o
m
e
a
r
o
u
tin
e
p
r
ac
tice
to
ch
ec
k
t
h
e
in
f
o
r
m
atio
n
p
r
o
v
id
ed
b
ased
o
n
th
e
air
q
u
alit
y
in
d
ex
(
A
QI
)
,
w
h
ic
h
is
d
iv
id
ed
in
to
f
o
u
r
ca
te
g
o
r
ies:
'g
o
o
d
'
,
'm
o
d
er
ate'
,
'
b
ad
'
,
a
n
d
'v
er
y
b
ad
'
.
Ko
r
ea
'
s
p
ar
ticu
late
m
atter
p
r
ed
i
ctio
n
ac
cu
r
ac
y
w
a
s
ap
p
r
o
x
i
m
atel
y
6
0
%
in
2
0
1
5
,
an
d
th
e
Ko
r
ea
Me
teo
r
o
lo
g
ical
A
d
m
in
is
tr
atio
n
'
s
p
r
ed
ictio
n
p
r
o
ce
s
s
h
as
t
h
e
a
n
n
u
al
p
r
ed
icto
n
ac
c
u
r
ac
y
o
f
ap
p
r
o
x
i
m
atel
y
8
0
%.
Ho
w
e
v
er
,
th
i
s
i
s
i
n
f
o
r
m
atio
n
r
ef
lec
tin
g
o
n
th
e
w
ea
t
h
er
f
o
r
ec
aster
'
s
e
x
p
er
ien
ce
,
an
d
t
h
e
a
ct
u
al
p
ar
tic
u
late
m
atter
p
r
ed
ictio
n
m
o
d
el
s
h
o
w
s
t
h
e
ac
c
u
r
ac
y
o
f
ap
p
r
o
x
im
a
tel
y
5
0
% [
8
,
9
]
.
T
h
er
ef
o
r
e,
v
ar
io
u
s
attem
p
t
s
h
a
v
e
b
ee
n
m
ad
e
to
im
p
r
o
v
e
th
e
p
r
ed
ictio
n
ac
cu
r
ac
y
o
f
p
ar
ticu
l
ate
m
at
ter
b
y
ap
p
l
y
i
n
g
m
ac
h
i
n
e
lear
n
in
g
p
r
ed
ictio
n
alg
o
r
it
h
m
s
alo
n
g
w
it
h
co
n
v
e
n
tio
n
al
s
tatis
tical
t
ec
h
n
iq
u
es
[
1
0
,
1
1
]
.
Ho
w
e
v
er
,
th
e
ch
ar
ac
ter
is
tic
s
o
f
p
ar
ticu
late
m
at
ter
ar
is
in
g
f
r
o
m
v
ar
io
u
s
ex
ter
n
a
l
f
ac
to
r
s
an
d
th
e
p
r
o
b
lem
o
f
th
e
th
e
o
cc
u
r
r
en
ce
r
ate
b
y
co
n
ce
n
tr
atio
n
m
a
k
e
it
d
if
f
ic
u
lt
to
ef
f
ec
tiv
e
l
y
tr
ai
n
p
r
ed
ictio
n
m
o
d
el
s
.
I
n
o
r
d
er
t
o
i
m
p
r
o
v
e
t
h
e
p
r
ed
ictio
n
ac
c
u
r
ac
y
o
f
p
ar
tic
u
late
m
atter
co
n
ce
n
tr
at
io
n
s
,
K.
W
.
C
h
o
e
t
a
l.,
in
th
e
ir
s
t
u
d
y
,
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
C
o
mp
a
r
a
tive
a
n
a
lysi
s
o
f m
u
ltip
le
cla
s
s
ifica
tio
n
mo
d
els to
…
(
Yo
n
g
-
Jin
Ju
n
g
)
2501
p
r
o
p
o
s
ed
a
p
r
ed
ictio
n
m
o
d
el
t
h
at
s
ep
ar
ated
a
n
d
p
r
ed
icted
th
e
m
b
ased
o
n
a
s
p
ec
i
f
ic
co
n
ce
n
tr
atio
n
.
B
y
d
i
v
id
in
g
th
e
lo
w
an
d
h
ig
h
co
n
ce
n
tr
ati
o
n
s
b
ased
o
n
th
e
p
ar
tic
u
late
m
atter
co
n
ce
n
tr
atio
n
o
f
8
1
,
th
e
y
co
m
p
ar
ed
th
e
p
r
ed
ictio
n
p
er
f
o
r
m
an
ce
th
r
o
u
g
h
a
d
ee
p
n
e
u
r
al
n
et
w
o
r
k
-
b
ased
p
r
ed
ictio
n
m
o
d
el.
T
h
e
p
r
ed
ictio
n
r
esu
lt
s
co
n
f
ir
m
ed
th
at
t
h
e
p
r
ed
ictio
n
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
lo
w
a
n
d
h
i
g
h
co
n
ce
n
tr
atio
n
s
w
a
s
i
m
p
r
o
v
ed
,
an
d
esp
ec
iall
y
it
s
h
o
w
ed
th
e
p
er
f
o
r
m
a
n
ce
i
m
p
r
o
v
e
m
e
n
t
o
f
2
0
.
6
2
%
f
o
r
th
e
h
ig
h
co
n
ce
n
tr
atio
n
s
[
9
]
.
T
h
e
s
tu
d
y
b
y
K.
Ka
y
a
et
a
l.
p
r
o
p
o
s
ed
a
s
o
lu
tio
n
to
an
u
n
b
alan
ce
d
p
r
o
b
lem
i
n
o
r
d
er
to
ad
d
r
ess
th
e
p
r
ed
ictio
n
p
r
o
b
le
m
o
f
t
h
e
r
eg
r
e
s
s
io
n
m
o
d
el
d
u
e
to
th
e
v
ar
iatio
n
i
n
th
e
o
cc
u
r
r
en
ce
r
ate
b
y
p
ar
ti
cu
late
m
a
tter
co
n
ce
n
tr
atio
n
.
T
h
ey
co
n
f
ir
m
ed
th
e
ac
cu
r
ac
y
o
f
ap
p
r
o
x
i
m
ate
l
y
8
0
%
in
th
e
en
t
ir
e
d
ata
s
et
b
y
m
a
k
in
g
th
e
n
u
m
b
er
o
f
s
a
m
p
les
o
f
t
h
e
cla
s
s
t
h
e
s
a
m
e
f
o
r
u
n
b
ala
n
ce
d
d
ata
th
r
o
u
g
h
t
h
e
p
r
o
p
o
s
ed
u
p
p
er
s
am
p
li
n
g
a
n
d
d
o
w
n
s
a
m
p
lin
g
[
1
2
]
.
I
n
th
is
p
ap
er
,
w
e
p
r
o
p
o
s
e
d
ata
class
if
icatio
n
m
o
d
els
b
y
co
n
c
en
tr
atio
n
to
i
m
p
r
o
v
e
th
e
p
er
f
o
r
m
an
ce
o
f
a
p
ar
ticu
late
m
atter
co
n
ce
n
tr
atio
n
p
r
ed
ictio
n
m
o
d
el.
Of
th
e
m
ac
h
i
n
e
lear
n
i
n
g
class
i
f
icat
i
o
n
m
o
d
els,
w
e
u
s
e
th
e
lo
g
i
s
tic
r
eg
r
es
s
io
n
,
d
e
cisi
o
n
tr
ee
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
(
SVM)
,
an
d
en
s
e
m
b
le
m
o
d
el
s
.
B
ased
o
n
th
e
A
QI
,
w
e
co
n
f
i
g
u
r
e
m
u
l
tip
le
class
i
f
icatio
n
m
o
d
els
b
y
d
iv
id
in
g
p
ar
ticu
late
m
atter
co
n
ce
n
tr
at
io
n
s
i
n
to
4
class
es.
I
n
o
r
d
er
to
a
p
p
ly
t
h
e
o
p
ti
m
al
p
ar
a
m
eter
s
to
th
e
m
o
d
els,
w
e
d
esig
n
t
h
e
m
o
d
els b
y
p
er
f
o
r
m
i
n
g
p
ar
a
m
eter
s
ea
r
ch
th
r
o
u
g
h
g
r
id
s
ea
r
ch
cr
o
s
s
v
ali
d
atio
n
.
W
e
p
er
f
o
r
m
m
o
d
el
e
v
a
lu
atio
n
u
s
in
g
t
h
e
er
r
o
r
m
atr
i
x
.
2.
DATA CO
L
L
E
C
T
I
O
N
AND
CO
NF
I
G
U
RAT
I
O
N
2
.
1
.
Da
t
a
c
o
llect
io
n a
nd
prepro
ce
s
s
ing
P
ar
ticu
late
m
a
tter
t
i
s
a
f
f
ec
ted
b
y
v
ar
io
u
s
f
ac
to
r
s
.
Air
p
o
llu
tan
ts
a
n
d
m
eteo
r
o
lo
g
ical
ele
m
en
ts
ar
e
t
y
p
ical,
w
h
ich
ar
e
co
m
m
o
n
l
y
ap
p
lied
to
s
tu
d
ies
f
o
r
p
r
ed
i
ctin
g
p
ar
ticu
late
m
atter
co
n
c
en
tr
at
io
n
s
[
1
3
-
1
6
]
.
B
ased
o
n
th
e
s
t
u
d
ies,
w
e
s
elec
ted
th
e
m
aj
o
r
da
ta
as sh
o
w
n
i
n
T
a
b
le
1
.
T
ab
le
1
.
Ma
j
o
r
d
ata
d
ef
in
itio
n
Ty
p
e
N
a
me
D
e
scri
p
t
i
o
n
A
i
r
p
o
l
l
u
t
a
n
t
s
10
T
h
e
a
v
e
r
a
g
e
p
a
r
t
i
c
u
l
a
t
e
ma
t
t
e
r
(
<
10
)
p
e
r
h
o
u
r
10ℎ
T
h
e
a
v
e
r
a
g
e
p
a
r
t
i
c
u
l
a
t
e
ma
t
t
e
r
(
<
10
)
o
f
t
h
e
p
r
e
v
i
o
u
s 1
h
o
u
r
3
T
h
e
a
v
e
r
a
g
e
o
z
o
n
e
o
f
t
h
e
p
r
e
v
i
o
u
s
1
h
o
u
r
T
h
e
a
v
e
r
a
g
e
c
a
r
b
o
n
mo
n
o
x
i
d
e
o
f
t
h
e
p
r
e
v
i
o
u
s 1
h
o
u
r
2
T
h
e
a
v
e
r
a
g
e
n
i
t
r
o
g
e
n
d
i
o
x
i
d
e
o
f
t
h
e
p
r
e
v
i
o
u
s 1
h
o
u
r
2
T
h
e
a
v
e
r
a
g
e
su
l
f
u
r
d
i
o
x
i
d
e
o
f
t
h
e
p
r
e
v
i
o
u
s
1
h
o
u
r
M
e
t
e
o
r
o
l
o
g
i
c
a
l
e
l
e
me
n
t
s
T
e
mp
e
r
a
t
u
r
e
T
h
e
a
v
e
r
a
g
e
t
e
mp
e
r
a
t
u
r
e
o
f
t
h
e
p
r
e
v
i
o
u
s
1
h
o
u
r
H
u
mi
d
i
t
y
T
h
e
a
v
e
r
a
g
e
h
u
m
i
d
i
t
y
o
f
t
h
e
p
r
e
v
i
o
u
s
1
h
o
u
r
W
i
n
d
S
p
e
e
d
T
h
e
a
v
e
r
a
g
e
w
i
n
d
sp
e
e
d
o
f
t
h
e
p
r
e
v
i
o
u
s
1
h
o
u
r
W
i
n
d
D
i
r
e
c
t
i
o
n
T
h
e
mo
st
f
r
e
q
u
e
n
t
w
i
n
d
d
i
r
e
c
t
i
o
n
o
f
t
h
e
p
r
e
v
i
o
u
s
1
h
o
u
r
A
cc
o
r
d
in
g
to
t
h
e
s
elec
ted
d
at
a,
w
e
co
llected
t
h
e
f
in
a
l
co
n
f
i
r
m
ed
d
ata
m
ea
s
u
r
ed
at
an
i
n
t
er
v
al
o
f
an
h
o
u
r
f
o
r
1
0
y
ea
r
s
f
r
o
m
2
0
0
9
to
2
0
1
8
at
th
e
m
ea
s
u
r
e
m
e
n
t
s
ta
tio
n
ar
o
u
n
d
C
h
eo
n
an
in
Ko
r
ea
.
A
ir
p
o
llu
t
io
n
d
ata
is
co
m
p
o
s
ed
o
f
10
,
10ℎ
,
3
,
,
O
2
,
an
d
2
,
an
d
m
eteo
r
o
lo
g
ical
ele
m
e
n
ts
c
o
n
s
is
t
o
f
te
m
p
er
at
u
r
e,
h
u
m
id
it
y
,
w
in
d
s
p
ee
d
,
an
d
w
i
n
d
d
ir
ec
tio
n
.
Sin
ce
s
o
m
e
d
ata
w
er
e
m
is
s
in
g
d
u
e
to
th
e
p
o
w
er
o
u
tag
e
a
n
d
m
ai
n
ten
a
n
ce
o
f
m
ea
s
u
r
e
m
e
n
t
eq
u
ip
m
e
n
t,
w
e
r
e
m
o
v
ed
all
d
ata
o
f
th
e
s
a
m
e
ti
m
e
w
h
e
n
th
e
m
is
s
in
g
d
a
ta
w
a
s
p
r
esen
t.
O
f
th
e
m
eteo
r
o
lo
g
ica
l
ele
m
en
ts
,
th
e
lar
g
e
s
t
w
i
n
d
d
ir
ec
tio
n
ex
p
r
ess
ed
in
az
i
m
u
th
,
th
at
is
,
th
e
0
°
an
d
3
6
0
°,
w
h
ic
h
w
er
e
o
f
te
n
u
s
ed
w
it
h
m
i
x
ed
n
o
tatio
n
,
w
er
e
u
n
i
f
ied
to
3
6
0
°.
T
h
er
e
is
a
n
ee
d
f
o
r
d
ata
p
r
ep
r
o
ce
s
s
in
g
to
p
er
f
o
r
m
clas
s
i
f
icat
io
n
th
r
o
u
g
h
m
ac
h
in
e
lear
n
i
n
g
alg
o
r
ith
m
s
u
s
i
n
g
th
e
co
llected
d
ata.
Sin
ce
w
e
u
s
ed
t
h
e
class
i
f
icat
i
o
n
alg
o
r
ith
m
b
ased
o
n
s
u
p
e
r
v
is
ed
lear
n
i
n
g
,
w
e
p
er
f
o
r
m
ed
cla
s
s
i
f
icatio
n
b
y
s
e
p
ar
ately
d
i
v
id
i
n
g
t
h
e
d
ata
co
r
r
esp
o
n
d
in
g
to
in
d
ep
en
d
e
n
t
v
ar
i
ab
les
an
d
th
e
d
ata
co
r
r
esp
o
n
d
in
g
to
d
ep
en
d
en
t
v
ar
iab
les.
T
h
e
in
d
ep
en
d
en
t
v
ar
iab
le
d
ata,
w
h
ic
h
in
c
lu
d
es
10ℎ
,
3
,
,
O
2
,
2
,
te
m
p
er
atu
r
e,
h
u
m
id
it
y
,
w
i
n
d
s
p
ee
d
an
d
w
i
n
d
d
ir
ec
tio
n
,
is
u
s
ed
to
p
r
ed
ict
th
e
r
an
g
e
o
f
p
ar
ticu
late
m
atter
co
n
ce
n
tr
atio
n
s
b
ased
o
n
th
e
A
QI
.
A
s
f
o
r
th
e
w
i
n
d
d
ir
ec
tio
n
,
it
is
n
ec
ess
ar
y
to
co
n
v
er
t
it
to
a
v
ec
to
r
f
o
r
m
b
ec
au
s
e
it
co
r
r
esp
o
n
d
s
to
ca
t
eg
o
r
ical
d
ata
ex
p
r
ess
ed
i
n
1
6
d
ir
ec
tio
n
s
.
T
h
er
ef
o
r
e,
th
r
o
u
g
h
o
n
e
-
h
o
t
e
n
co
d
in
g
,
w
e
co
n
v
er
ted
th
e
ca
teg
o
r
ies
co
r
r
esp
o
n
d
in
g
to
1
6
d
ir
ec
tio
n
s
to
1
6
v
ec
to
r
s
ex
p
r
ess
ed
in
0
an
d
1
.
Fo
r
th
e
r
e
m
ain
in
g
i
n
p
u
t
s
a
m
p
le
d
at
a
o
th
er
th
an
t
h
e
w
i
n
d
d
ir
ec
tio
n
,
th
e
y
ar
e
n
u
m
er
ical
d
ata
w
it
h
d
if
f
er
e
n
t
ch
ar
ac
ter
is
tic
s
,
an
d
th
e
y
w
er
e
co
n
v
er
ted
to
a
v
al
u
e
b
et
w
ee
n
0
an
d
1
th
r
o
u
g
h
m
i
n
m
ax
s
ca
li
n
g
i
n
o
r
d
er
to
u
n
i
f
y
th
e
r
an
g
e
o
f
n
u
m
er
ical
v
al
u
e
s
ex
p
r
ess
ed
ac
co
r
d
in
g
to
th
e
d
ata.
T
h
e
d
ep
en
d
en
t
v
ar
iab
le
d
ata
co
r
r
esp
o
n
d
t
o
10
,
an
d
as
s
h
o
w
n
in
T
ab
le
2
,
b
ased
o
n
th
e
A
QI
u
s
ed
as
a
f
o
r
ec
ast
b
y
th
e
Mi
n
is
tr
y
o
f
E
n
v
ir
o
n
m
e
n
t,
w
e
d
iv
id
ed
10
in
to
th
e
s
eq
u
e
n
tial
ca
teg
o
r
ies
:
'g
o
o
d
'
,
'
m
o
d
er
ate
'
,
'
b
ad
'
,
an
d
'
v
er
y
b
ad
'
,
an
d
e
x
p
r
ess
ed
th
e
m
a
s
f
o
u
r
clas
s
es o
f
0
,
1
,
2
,
an
d
3
,
r
esp
ec
tiv
el
y
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
11
,
No
.
3
,
J
u
n
e
2
0
2
1
:
2
5
0
0
-
2507
2502
T
ab
le
2
.
T
h
e
r
an
g
e
o
f
p
ar
ticu
l
ate
m
atter
co
n
ce
n
tr
atio
n
s
b
ase
d
o
n
A
QI
G
r
a
d
e
G
o
o
d
M
o
d
e
r
a
t
e
B
a
d
V
e
r
y
B
a
d
10
(
μ
g
/
3
)
0
~
3
0
3
1
~
8
0
8
1
~
1
5
0
1
5
0
~
2
.
2
.
Da
t
a
c
o
nfig
ura
t
io
n
T
h
e
d
ata
u
s
ed
i
n
t
h
e
s
u
p
er
v
is
e
d
lear
n
in
g
m
o
d
el
i
s
m
ai
n
l
y
co
m
p
o
s
ed
o
f
a
tr
ain
in
g
s
et
f
o
r
le
ar
n
in
g
a
n
d
a
test
s
et
f
o
r
ev
alu
atin
g
th
e
tr
ain
ed
m
o
d
el.
T
h
e
tr
ain
in
g
s
et
u
s
ed
to
tr
ain
th
e
m
o
d
el
is
s
u
b
d
iv
id
ed
in
to
a
tr
ain
s
et
an
d
a
v
alid
atio
n
s
et
b
ec
au
s
e
o
f
th
e
n
ee
d
to
v
er
if
y
w
h
et
h
er
tr
ain
i
n
g
is
w
ell
co
m
p
lete
d
.
I
n
th
is
p
ap
er
,
w
e
co
n
f
i
g
u
r
e
a
tr
ain
in
g
s
et
o
f
7
5
% a
n
d
a
test
s
et
o
f
2
5
%
w
it
h
t
h
e
p
r
ep
r
o
ce
s
s
ed
d
ata.
T
h
e
tr
ain
in
g
s
et
i
s
co
m
p
o
s
ed
o
f
a
tr
ain
s
et
o
f
8
0
%
an
d
a
v
alid
atio
n
s
et
o
f
2
0
%.
Fi
g
u
r
e
1
s
h
o
w
s
t
h
e
s
tr
u
ct
u
r
e
o
f
t
h
e
f
i
n
al
d
ata
u
s
ed
in
t
h
e
m
o
d
el,
a
n
d
T
ab
le
3
s
h
o
w
s
th
e
co
n
f
i
g
u
r
atio
n
o
f
t
h
e
d
ata
s
et.
Fig
u
r
e
1
.
Stru
ct
u
r
e
o
f
d
ata
s
et
T
ab
le
3
.
Data
s
et
co
n
f
ig
u
r
atio
n
S
t
r
u
c
t
u
r
e
o
f
d
a
t
a
se
t
S
a
mp
l
e
s
T
r
a
i
n
i
n
g
se
t
T
r
a
i
n
se
t
5
2
,
3
1
5
V
a
l
i
d
a
t
i
o
n
se
t
1
3
,
0
7
9
T
e
st
se
t
2
1
,
7
9
9
T
o
t
a
l
8
7
,
1
9
3
3.
CL
AS
SI
F
I
CAT
I
O
N
M
O
DE
L
DE
SI
G
N
3
.
1
.
L
o
g
is
t
ic
re
g
re
s
s
io
n
m
o
d
el
des
ig
n
L
o
g
i
s
tic
r
e
g
r
ess
io
n
i
s
a
n
al
g
o
r
ith
m
u
s
ed
to
p
r
ed
ict
t
h
e
lik
el
ih
o
o
d
o
f
a
n
e
v
en
t
u
s
i
n
g
a
li
n
ea
r
co
m
b
i
n
atio
n
o
f
i
n
d
ep
en
d
en
t
v
ar
iab
les.
As in
g
e
n
er
al
r
eg
r
es
s
io
n
an
al
y
s
is
,
it i
s
u
s
ed
in
f
u
t
u
r
e
p
r
ed
ictio
n
m
o
d
els
b
y
d
er
i
v
in
g
a
s
p
ec
i
f
ic
f
u
n
cti
o
n
th
r
o
u
g
h
th
e
r
elatio
n
s
h
ip
b
et
w
ee
n
d
ep
en
d
en
t
a
n
d
in
d
e
p
en
d
en
t
v
ar
iab
le
s
.
Ho
w
e
v
er
,
u
n
li
k
e
li
n
ea
r
r
eg
r
es
s
io
n
,
s
i
n
ce
th
e
p
r
ed
ictio
n
r
esu
lt
is
clas
s
i
f
ied
as
a
s
p
ec
if
ic
ca
teg
o
r
y
w
h
e
n
th
e
d
ep
en
d
en
t
v
ar
iab
le
is
ca
te
g
o
r
ical
d
ata,
it
is
u
s
ed
as
a
cl
ass
i
f
icatio
n
tech
n
iq
u
e
r
at
h
er
th
an
a
r
e
g
r
ess
io
n
tech
n
iq
u
e.
I
t
i
s
d
iv
id
ed
in
to
b
in
o
m
ial
o
r
m
u
lti
n
o
m
ial
d
e
p
en
d
in
g
o
n
t
h
e
ca
te
g
o
r
y
c
h
ar
ac
ter
is
tics
o
f
t
h
e
d
ep
en
d
en
t
v
ar
iab
le.
T
h
e
d
ep
en
d
en
t
v
ar
iab
le
f
o
r
tr
ain
i
n
g
th
e
class
if
ica
tio
n
m
o
d
el
o
f
p
ar
ticu
late
m
atte
r
co
n
ce
n
tr
atio
n
s
h
a
s
f
o
u
r
ca
te
g
o
r
ies.
A
cc
o
r
d
in
g
l
y
,
w
e
b
u
il
t
a
m
o
d
el
b
y
ap
p
l
y
i
n
g
a
m
u
lti
n
o
m
ia
l
lo
g
i
s
tic
r
eg
r
ess
io
n
m
et
h
o
d
.
Fo
r
a
m
o
d
el
p
r
ed
ictin
g
a
ce
r
tain
r
esu
l
t,
o
v
er
f
itt
in
g
o
r
u
n
d
e
r
f
itti
n
g
i
s
co
n
ti
n
g
e
n
t
o
n
t
h
e
in
ten
s
it
y
o
f
tr
ain
i
n
g
.
I
t
is
d
i
f
f
icu
l
t
f
o
r
a
m
o
d
el
w
it
h
o
v
er
f
i
tti
n
g
to
p
r
ed
ict
n
e
w
d
ata
s
i
n
ce
it
o
n
l
y
f
o
cu
s
es
o
n
tr
ain
i
n
g
d
ata.
I
n
th
e
ca
s
e
o
f
u
n
d
er
f
itt
in
g
,
th
er
e
is
a
p
r
o
b
lem
th
at
t
h
e
m
o
d
el
d
o
es
n
o
t
p
r
e
d
ict
m
o
s
t
o
f
t
h
e
d
ata
s
in
ce
it
d
o
es
n
o
t
id
en
ti
f
y
t
h
e
ch
ar
ac
ter
is
t
ics
o
f
th
e
d
ata
d
u
e
to
s
im
p
le
tr
ai
n
in
g
.
T
o
s
o
lv
e
th
ese
p
r
o
b
lem
s
,
lo
g
is
tic
r
eg
r
ess
io
n
b
asicall
y
u
s
e
s
L
2
r
eg
u
lar
izatio
n
p
ar
a
m
eter
c
[
1
7
]
.
T
h
er
ef
o
r
e,
f
o
r
b
etter
p
r
ed
ictio
n
p
er
f
o
r
m
a
n
ce
,
w
e
p
er
f
o
r
m
ed
th
e
s
ea
r
c
h
f
o
r
th
e
o
p
ti
m
al
c
v
alu
e
u
s
i
n
g
g
r
id
s
ea
r
ch
cr
o
s
s
v
alid
atio
n
t
o
f
in
d
t
h
e
v
al
u
e
o
f
c
t
h
at
f
it
s
th
e
m
o
d
el.
W
e
s
et
t
h
e
r
an
g
e
o
f
c
v
al
u
es
to
b
e
s
ea
r
ch
e
d
to
0
.
0
0
0
1
,
0
.
0
0
1
,
0
.
0
1
,
0
.
1
,
1
,
1
0
,
1
0
0
,
an
d
1
0
0
0
.
I
n
o
r
d
er
to
s
elec
t
p
ar
a
m
eter
s
w
it
h
h
i
g
h
g
en
er
aliza
tio
n
p
er
f
o
r
m
a
n
ce
,
we
s
et
th
e
c
v
p
ar
a
m
eter
o
f
k
-
f
o
l
d
cr
o
s
s
v
alid
atio
n
to
5
.
A
cc
o
r
d
in
g
l
y
,
t
h
e
c
v
al
u
e
w
a
s
s
eq
u
e
n
tia
ll
y
ac
ce
s
s
ed
to
c
o
m
p
ar
e
s
co
r
es
u
s
in
g
t
h
e
tes
t
s
et
af
ter
5
r
ep
etitiv
e
tr
ain
i
n
g
r
u
n
s
a
n
d
v
al
id
atio
n
s
.
Fo
r
p
r
ep
r
o
ce
s
s
in
g
o
f
v
alid
ati
o
n
f
o
ld
d
u
r
i
n
g
cr
o
s
s
v
alid
ati
o
n
,
w
e
s
ea
r
ch
ed
t
h
e
c
v
al
u
e
s
b
y
b
u
ild
i
n
g
th
e
p
ip
elin
e
o
f
m
in
m
ax
s
ca
ler
an
d
th
e
m
o
d
el.
T
ab
le
4
s
h
o
w
s
th
e
m
ea
n
test
s
co
r
e
an
d
c
v
alu
es
o
f
t
h
e
to
p
3
r
an
k
i
n
g
s
i
n
t
h
e
cr
o
s
s
v
al
id
ati
o
n
r
esu
l
ts
.
T
h
e
cr
o
s
s
v
alid
ati
o
n
r
esu
lts
s
h
o
w
ed
t
h
at
t
h
e
m
ea
n
tes
t
s
co
r
e
w
as
h
ig
h
e
s
t
w
it
h
0
.
8
0
8
9
5
8
w
h
en
th
e
c
v
al
u
e
w
as
1
0
.
0
,
th
u
s
we
s
elec
ted
th
e
c
v
al
u
e
to
b
e
ap
p
lied
to
l
o
g
is
tic
r
eg
r
ess
io
n
as
1
0
.
0
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
C
o
mp
a
r
a
tive
a
n
a
lysi
s
o
f m
u
ltip
le
cla
s
s
ifica
tio
n
mo
d
els to
…
(
Yo
n
g
-
Jin
Ju
n
g
)
2503
T
ab
le
4
.
Gr
id
s
ea
r
ch
cr
o
s
s
v
ali
d
atio
n
r
esu
lt
s
(
lo
g
is
tic
r
eg
r
es
s
i
o
n
)
R
an
k
Me
an
test
s
co
r
e
c
1
0
.
8
0
8
9
5
8
1
0
.
0
2
0
.
8
0
8
9
2
7
1
0
0
0
.
0
3
0
.
8
0
6
8
1
7
1
.
0
3
.
2
.
Dec
is
io
n t
re
e
m
o
de
l des
ig
n
D
ec
is
io
n
tr
ee
i
s
a
w
id
el
y
u
s
e
d
m
o
d
el
f
o
r
class
i
f
icatio
n
an
d
r
eg
r
ess
io
n
.
I
t
i
s
b
asicall
y
a
n
alg
o
r
ith
m
th
at
lear
n
s
b
y
co
n
tin
u
o
u
s
l
y
a
n
s
w
er
i
n
g
q
u
est
io
n
s
to
ap
p
r
o
ac
h
a
s
p
ec
if
ic
d
ec
is
io
n
.
W
it
h
th
e
i
n
cr
ea
s
e
in
th
e
n
u
m
b
er
o
f
leaf
n
o
d
es,
th
e
ac
c
u
r
ac
y
o
f
t
h
e
tr
ain
i
n
g
s
e
t
in
cr
e
ases
b
u
t
o
v
er
f
itti
n
g
m
a
y
o
cc
u
r
[
1
8
]
.
On
e
o
f
th
e
m
et
h
o
d
s
u
s
ed
to
p
r
ev
en
t
o
v
er
f
itti
n
g
is
to
s
to
p
th
e
g
r
o
w
t
h
o
f
th
e
tr
ee
w
h
e
n
t
h
e
d
ep
th
o
f
th
e
tr
ee
r
ea
ch
es
a
ce
r
tain
lev
el.
T
h
e
p
ar
am
e
ter
th
at
li
m
its
t
h
e
d
ep
th
o
f
a
d
ec
is
i
o
n
tr
ee
is
m
ax
_
d
ep
th
,
an
d
w
e
ar
e
ab
le
to
im
p
r
o
v
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
m
o
d
el
b
y
ad
j
u
s
ti
n
g
th
e
d
ep
th
.
T
h
er
ef
o
r
e,
w
e
p
er
f
o
r
m
ed
th
e
s
ea
r
ch
f
o
r
th
e
o
p
tim
al
m
a
x
_
d
ep
th
v
alu
e
u
s
i
n
g
g
r
id
s
ea
r
ch
cr
o
s
s
v
alid
atio
n
.
W
e
s
et
t
h
e
r
an
g
e
o
f
m
ax
_
d
e
p
th
v
al
u
es
to
b
e
s
ea
r
ch
ed
to
1
~2
4
,
an
d
p
er
f
o
r
m
ed
t
h
e
s
ea
r
c
h
b
y
s
etti
n
g
th
e
cv
p
ar
a
m
eter
o
f
k
-
f
o
ld
cr
o
s
s
v
alid
atio
n
to
5
.
A
d
d
itio
n
all
y
,
f
o
r
p
r
ep
r
o
ce
s
s
in
g
o
f
v
al
id
atio
n
f
o
ld
d
u
r
in
g
cr
o
s
s
v
alid
atio
n
,
w
e
s
ea
r
c
h
ed
th
e
m
a
x
_
d
ep
th
v
al
u
es
b
y
b
u
ild
in
g
t
h
e
p
ip
elin
e
o
f
m
i
n
m
ax
s
ca
ler
an
d
th
e
m
o
d
el.
T
ab
le
5
s
h
o
w
s
t
h
e
m
e
an
test
s
co
r
e
an
d
m
ax
_
d
ep
th
v
alu
es
o
f
t
h
e
to
p
3
r
an
k
i
n
g
s
i
n
th
e
cr
o
s
s
v
alid
atio
n
r
esu
lt
s
.
T
h
e
c
r
o
s
s
v
alid
atio
n
r
esu
lt
s
s
h
o
w
ed
th
at
th
e
m
ea
n
t
est
s
co
r
e
w
a
s
h
i
g
h
e
s
t
w
ith
0
.
8
5
9
3
6
0
1
3
w
h
e
n
th
e
m
ax
_
d
ep
th
v
al
u
e
w
as
4
,
th
u
s
w
e
s
e
lecte
d
th
e
m
ax
_
d
ep
th
v
al
u
e
to
b
e
ap
p
lied
to
d
ec
is
io
n
tr
ee
as 4
.
T
ab
le
5
.
Gr
id
s
ea
r
ch
cr
o
s
s
v
ali
d
atio
n
r
esu
lt
s
(
d
ec
is
io
n
tr
ee
)
R
a
n
k
M
e
a
n
t
e
st
s
c
o
r
e
M
a
x
_
d
e
p
t
h
1
0
.
8
5
9
3
6
0
1
3
4
2
0
.
8
5
8
9
6
2
5
5
5
3
0
.
8
5
8
5
6
4
9
6
3
3
.
3
.
SV
M
m
o
del de
s
ig
n
SVM
is
o
n
e
o
f
m
ac
h
i
n
e
lear
n
i
n
g
m
et
h
o
d
s
an
d
is
a
s
u
p
er
v
i
s
e
d
lear
n
in
g
m
o
d
el
f
o
r
p
atter
n
r
ec
o
g
n
itio
n
an
d
d
ata
an
al
y
s
is
.
I
t
i
s
m
ai
n
l
y
u
s
ed
f
o
r
clas
s
i
f
icatio
n
a
n
d
r
eg
r
ess
io
n
a
n
al
y
s
is
.
G
iv
e
n
a
s
et
o
f
d
ata
b
elo
n
g
i
n
g
to
o
n
e
o
f
t
w
o
ca
teg
o
r
ies,
it
g
e
n
er
ates
a
n
o
n
-
s
to
c
h
a
s
tic
b
in
ar
y
li
n
ea
r
cla
s
s
i
f
icatio
n
m
o
d
el
th
at
d
eter
m
in
e
s
t
h
e
ca
teg
o
r
y
to
w
h
ic
h
n
e
w
d
ata
b
elo
n
g
s
b
ased
o
n
a
g
iv
en
d
ata
s
et.
T
h
e
g
en
er
ated
c
las
s
if
ica
tio
n
m
o
d
el
is
ex
p
r
ess
ed
as
a
b
o
u
n
d
ar
y
in
t
h
e
s
p
ac
e
o
n
to
w
h
ich
th
e
d
ata
is
m
ap
p
ed
.
I
t
is
an
alg
o
r
it
h
m
to
f
i
n
d
t
h
e
b
o
u
n
d
ar
y
w
it
h
th
e
lar
g
est
w
id
th
[
1
9
-
2
1
]
.
T
h
er
ef
o
r
e,
SVM
is
a
m
o
d
el
th
at
d
ef
i
n
es
a
b
aselin
e
f
o
r
cla
s
s
i
f
icatio
n
b
et
w
ee
n
ca
teg
o
r
ies
,
w
h
ic
h
i
s
ex
p
r
ess
ed
as a
d
ec
is
io
n
b
o
u
n
d
ar
y
.
I
n
SVM,
th
e
d
if
f
er
e
n
ce
in
p
er
f
o
r
m
an
ce
is
d
eter
m
i
n
ed
d
ep
e
n
d
in
g
o
n
h
o
w
t
h
e
d
ec
is
io
n
b
o
u
n
d
ar
y
is
d
ef
in
ed
,
an
d
i
t
is
cr
u
cial
to
f
i
n
d
th
e
o
p
ti
m
al
d
ec
is
io
n
b
o
u
n
d
ar
y
.
T
h
e
p
ar
a
m
eter
s
ap
p
lied
to
f
in
d
t
h
e
o
p
ti
m
al
d
ec
is
io
n
b
o
u
n
d
ar
y
ar
e
c
an
d
g
a
m
m
a.
C
i
s
a
p
ar
am
eter
t
h
at
ad
j
u
s
ts
t
h
e
allo
w
ab
le
r
an
g
e
o
f
o
u
tlier
s
b
y
co
n
tr
o
l
lin
g
t
h
e
m
ar
g
i
n
o
f
t
h
e
d
ec
is
io
n
b
o
u
n
d
ar
y
,
a
n
d
g
a
m
m
a
i
s
a
p
ar
a
m
e
ter
th
a
t
p
r
ev
e
n
t
s
o
v
er
f
itti
n
g
o
f
th
e
m
o
d
el
b
y
co
n
tr
o
llin
g
t
h
e
f
le
x
i
b
ilit
y
o
f
t
h
e
d
ec
is
io
n
b
o
u
n
d
ar
y
.
W
e
p
er
f
o
r
m
ed
th
e
s
ea
r
ch
f
o
r
t
h
e
o
p
ti
m
al
c
a
n
d
g
a
m
m
a
v
a
lu
es
to
f
i
n
d
th
e
o
p
ti
m
al
d
ec
is
io
n
b
o
u
n
d
ar
y
u
s
i
n
g
g
r
id
s
ea
r
ch
cr
o
s
s
v
alid
a
tio
n
.
W
e
s
et
t
h
e
r
an
g
e
o
f
c
an
d
g
a
m
m
a
v
al
u
es
to
b
e
s
ea
r
ch
e
d
to
0
.
0
0
0
1
,
0
.
0
0
1
,
0
.
0
1
,
0
.
1
,
1
,
1
0
,
1
0
0
,
an
d
1
0
0
0
,
an
d
p
er
f
o
r
m
ed
th
e
s
ea
r
ch
b
y
s
etti
n
g
th
e
cv
p
ar
a
m
e
ter
o
f
k
-
f
o
ld
cr
o
s
s
v
alid
atio
n
to
5
.
A
d
d
itio
n
al
l
y
,
f
o
r
p
r
ep
r
o
ce
s
s
in
g
o
f
v
a
lid
atio
n
f
o
ld
d
u
r
in
g
cr
o
s
s
v
alid
atio
n
,
w
e
s
ea
r
ch
ed
t
h
e
c
an
d
g
a
m
m
a
v
alu
e
s
b
y
b
u
ild
i
n
g
th
e
p
ip
elin
e
o
f
m
i
n
m
a
x
s
c
aler
an
d
th
e
m
o
d
el.
T
a
b
le
6
s
h
o
w
s
th
e
m
ea
n
test
s
co
r
e
an
d
r
elev
an
t
c
a
n
d
g
a
m
m
a
v
al
u
es
o
f
t
h
e
to
p
3
r
a
n
k
i
n
g
s
i
n
t
h
e
cr
o
s
s
v
alid
atio
n
r
esu
lts
.
T
h
e
cr
o
s
s
v
alid
atio
n
r
e
s
u
l
ts
s
h
o
w
ed
t
h
at
th
e
m
ea
n
te
s
t
s
co
r
e
w
a
s
h
ig
h
est
w
it
h
0
.
8
5
9
2
3
8
w
h
en
th
e
c
an
d
g
a
m
m
a
v
al
u
e
s
w
er
e
1
0
0
0
an
d
0
.
0
1
,
r
esp
ec
tiv
el
y
,
th
u
s
w
e
s
elec
ted
th
e
c
v
alu
e
an
d
th
e
g
a
m
m
a
v
al
u
e
to
b
e
ap
p
lied
to
SVM
as
1
0
0
0
an
d
0
.
0
1
,
r
esp
ec
tiv
el
y
.
T
ab
le
6
.
Gr
id
s
ea
r
ch
cr
o
s
s
v
ali
d
atio
n
r
esu
lt
s
(
SVM)
R
a
n
k
M
e
a
n
t
e
st
s
c
o
r
e
c
g
a
mm
a
1
0
.
8
5
9
2
3
8
1
0
0
0
0
.
0
1
2
0
.
8
5
9
1
7
7
1
1
3
0
.
8
5
9
1
4
6
1
0
.
1
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
11
,
No
.
3
,
J
u
n
e
2
0
2
1
:
2
5
0
0
-
2507
2504
3
.
4
.
E
ns
e
m
ble
m
o
del
des
ig
n
E
n
s
e
m
b
le
is
a
tec
h
n
iq
u
e
t
h
at
g
en
er
ate
s
a
p
o
w
er
f
u
l
m
o
d
el
b
y
co
m
b
i
n
i
n
g
m
u
lt
ip
le
m
o
d
els
to
ac
h
ie
v
e
b
etter
p
r
ed
ictio
n
p
e
r
f
o
r
m
a
n
ce
as
co
m
p
ar
ed
w
it
h
u
s
in
g
an
i
n
d
iv
id
u
al
m
ac
h
in
e
lear
n
in
g
m
o
d
el.
W
h
en
m
u
ltip
le
m
o
d
el
s
ar
e
co
m
b
i
n
ed
,
th
e
a
m
o
u
n
t
o
f
ca
lc
u
latio
n
is
g
e
n
e
r
all
y
in
cr
ea
s
ed
,
y
e
t
it
p
r
ev
en
ts
o
v
er
f
it
tin
g
m
o
r
e
ef
f
ec
tiv
e
l
y
t
h
an
u
s
in
g
a
n
in
d
i
v
id
u
al
m
o
d
el
an
d
it
h
as
th
e
a
d
v
an
ta
g
e
o
f
s
h
o
w
i
n
g
b
etter
p
er
f
o
r
m
an
ce
t
h
an
an
in
d
iv
id
u
al
m
o
d
el
i
f
t
h
e
p
er
f
o
r
m
an
ce
o
f
an
in
d
i
v
id
u
al
m
o
d
el
is
p
o
o
r
[
2
2
-
2
4
].
E
n
s
e
m
b
le
is
m
ai
n
l
y
d
iv
id
ed
i
n
to
a
co
llectio
n
m
et
h
o
d
o
lo
g
y
a
n
d
a
b
o
o
s
tin
g
m
et
h
o
d
o
lo
g
y
.
T
h
e
co
llectio
n
m
et
h
o
d
o
lo
g
y
h
as
th
e
p
r
ed
eter
m
in
ed
s
et
o
f
m
o
d
els
to
b
e
u
s
ed
,
b
u
t
th
e
b
o
o
s
tin
g
m
et
h
o
d
o
lo
g
y
g
r
ad
u
all
y
i
n
cr
ea
s
e
s
th
e
m
o
d
els
to
b
e
u
s
ed
.
I
n
th
is
s
tu
d
y
,
w
e
co
m
b
in
ed
th
e
lo
g
i
s
tic
r
eg
r
ess
io
n
,
d
ec
is
io
n
tr
ee
,
an
d
SVM
m
o
d
els
p
r
ev
io
u
s
l
y
d
esi
g
n
ed
,
w
h
ich
co
r
r
esp
o
n
d
s
to
th
e
co
llectio
n
m
et
h
o
d
o
lo
g
y
,
to
b
u
ild
an
en
e
m
b
le
m
o
d
el.
Fi
g
u
r
e
2
s
h
o
w
s
t
h
e
s
tr
u
ctu
r
e
o
f
t
h
e
en
s
e
m
b
le
m
o
d
el.
Fig
u
r
e
2
.
Stru
ct
u
r
e
o
f
en
s
e
m
b
l
e
m
o
d
el
T
h
e
tr
ain
in
g
s
e
t
d
ata
ar
e
u
s
ed
as
an
in
p
u
t
v
ar
iab
le
to
t
h
e
co
m
b
in
ed
lo
g
is
t
ic
r
eg
r
es
s
io
n
,
d
e
cisi
o
n
tr
ee
,
an
d
SVM
m
o
d
els,
an
d
th
e
p
r
ed
icted
r
esu
lts
ar
e
o
u
tp
u
tted
f
r
o
m
an
i
n
d
iv
id
u
al
m
o
d
el.
T
h
e
f
i
n
al
p
r
ed
ictio
n
r
esu
lt
s
ar
e
g
en
er
ated
b
y
v
o
tin
g
o
n
th
e
o
u
tp
u
tted
r
esu
lt
s
[
2
5
]
.
V
o
tin
g
is
d
iv
id
ed
in
to
h
ar
d
a
n
d
s
o
f
t
v
o
tin
g
.
Har
d
v
o
tin
g
s
i
m
p
l
y
s
elec
ts
th
e
f
in
al
p
r
ed
ictio
n
b
ased
o
n
t
h
e
p
r
ed
ictio
n
r
es
u
lts
o
f
a
n
i
n
d
iv
id
u
al
m
o
d
el.
T
h
e
v
o
tin
g
m
et
h
o
d
o
f
th
e
en
s
e
m
b
le
m
o
d
e
l
d
esig
n
ed
in
t
h
is
s
t
u
d
y
is
s
o
f
t
v
o
tin
g
,
w
h
ich
s
e
lects
th
e
f
in
a
l
p
r
ed
i
ctio
n
b
ased
o
n
th
e
s
u
m
o
f
co
n
d
itio
n
al
p
r
o
b
ab
ilit
ies o
f
an
i
n
d
iv
id
u
al
m
o
d
el.
4.
P
E
RF
O
RM
ANCE E
VA
L
U
AT
I
O
N
W
e
ev
alu
ated
class
if
icatio
n
p
er
f
o
r
m
an
ce
u
s
i
n
g
th
e
p
r
ev
io
u
s
l
y
co
n
f
i
g
u
r
ed
d
ata
s
et
an
d
d
e
s
ig
n
ed
th
e
class
i
f
icatio
n
m
o
d
els.
Fo
r
p
er
f
o
r
m
a
n
ce
ev
al
u
atio
n
,
w
e
u
s
ed
p
r
ec
is
io
n
,
r
ec
all,
an
d
f
-
s
co
r
e
b
ased
o
n
th
e
er
r
o
r
m
atr
i
x
.
Fig
u
r
e
3
s
h
o
w
s
th
e
er
r
o
r
m
atr
ices
cr
ea
ted
b
ased
o
n
th
e
class
i
f
icatio
n
r
esu
lt
s
o
f
th
e
tr
ain
ed
m
o
d
els
.
T
ab
le
7
s
h
o
w
s
th
e
p
er
f
o
r
m
a
n
ce
ev
alu
atio
n
o
f
th
e
clas
s
i
f
ic
atio
n
m
o
d
els
ca
lcu
lated
b
y
r
e
f
er
r
in
g
to
th
e
er
r
o
r
m
atr
ices
.
W
h
en
t
h
e
l
o
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el
p
r
ed
icted
'g
o
o
d
'
,
th
e
p
r
ec
is
io
n
w
a
s
h
i
g
h
e
s
t
w
i
th
0
.
8
6
8
5
.
W
h
en
th
e
p
r
ed
ictio
n
w
as
p
er
f
o
r
m
ed
b
ased
o
n
th
e
in
p
u
t
d
ata
o
f
'
m
o
d
er
ate
'
,
t
h
e
r
ec
all
w
a
s
h
i
g
h
est
w
it
h
0
.
9
3
4
1
.
On
th
e
o
th
er
h
a
n
d
,
th
e
cla
s
s
i
f
icat
i
o
n
d
id
n
o
t
w
o
r
k
w
e
ll
f
o
r
‘
b
ad
’
a
n
d
‘
v
er
y
b
ad
’
.
E
s
p
ec
iall
y
,
t
h
e
p
r
ed
ictio
n
w
as
n
o
t
m
ad
e
at
all
f
o
r
'v
er
y
b
ad
'
.
W
h
en
t
h
e
d
ec
is
io
n
tr
ee
m
o
d
el
p
r
ed
icted
'
m
o
d
er
ate
'
,
th
e
p
r
ec
is
io
n
w
as
h
i
g
h
e
s
t
w
it
h
0
.
8
9
7
7
.
W
h
en
th
e
p
r
ed
ict
io
n
w
as
p
er
f
o
r
m
ed
b
ased
o
n
t
h
e
i
n
p
u
t
d
ata
o
f
'
m
o
d
er
ate
'
,
th
e
r
ec
all
w
a
s
h
ig
h
e
s
t
w
it
h
0
.
9
023
.
On
t
h
e
o
t
h
er
h
a
n
d
,
th
e
p
r
ec
is
io
n
an
d
r
ec
all
f
o
r
'
b
ad
'
an
d
'v
er
y
b
ad
'
s
h
o
w
ed
r
e
lativ
el
y
lo
w
v
al
u
e
s
co
m
p
ar
ed
to
'
g
o
o
d
'
a
n
d
'
m
o
d
e
r
ate
’
.
T
h
e
SVM
m
o
d
el
s
h
o
w
e
d
th
e
h
ig
h
est
p
r
ec
is
io
n
a
n
d
r
ec
all
w
it
h
0
.
8
9
9
7
an
d
0
.
8
9
9
7
,
r
esp
ec
tiv
el
y
,
f
o
r
'
m
o
d
er
ate
’
.
As
in
th
e
d
ec
is
io
n
tr
ee
m
o
d
el
,
th
e
p
r
ec
is
io
n
an
d
r
ec
all
f
o
r
'
b
ad
'
an
d
'v
er
y
b
ad
'
s
h
o
w
ed
r
elativ
el
y
lo
w
v
alu
es
co
m
p
ar
ed
to
'
g
o
o
d
'
an
d
'
m
o
d
er
ate
’
.
T
h
e
en
s
e
m
b
le
m
o
d
el
s
h
o
w
ed
th
e
h
ig
h
e
s
t
p
r
ec
is
io
n
an
d
r
ec
all
w
it
h
0
.
8
9
9
7
an
d
0
.
8
9
9
7
f
o
r
'
m
o
d
er
ate
’
.
Ho
w
e
v
er
,
th
e
p
r
ec
is
io
n
a
n
d
r
ec
all
f
o
r
'
b
ad
'
a
n
d
'
v
er
y
b
ad
'
s
h
o
w
ed
r
elativ
el
y
lo
w
v
alu
e
s
co
m
p
ar
e
d
to
'
g
o
o
d
'
a
n
d
'm
o
d
er
ate’
,
r
e
s
u
lt
in
g
i
n
d
if
f
ic
u
lt
y
class
i
f
y
in
g
t
h
e
r
elev
a
n
t
class
es.
T
h
e
an
al
y
s
i
s
r
esu
l
ts
b
ase
d
o
n
th
e
p
r
ec
i
s
io
n
an
d
r
ec
all
s
h
o
w
ed
th
at
t
h
e
p
r
ec
is
io
n
an
d
r
ec
all
o
f
'g
o
o
d
'
an
d
'm
o
d
er
ate'
w
er
e
r
elati
v
el
y
h
ig
h
er
t
h
an
t
h
o
s
e
o
f
'
b
ad
'
an
d
'
v
er
y
b
ad
'
.
W
h
e
n
an
al
y
ze
d
t
h
r
o
u
g
h
t
h
e
er
r
o
r
m
atr
ix
es
i
n
Fi
g
u
r
e
3
,
it
w
a
s
co
n
f
ir
m
ed
t
h
at,
o
f
t
h
e
i
n
p
u
t
d
at
a,
th
e
p
r
o
p
o
r
tio
n
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
C
o
mp
a
r
a
tive
a
n
a
lysi
s
o
f m
u
ltip
le
cla
s
s
ifica
tio
n
mo
d
els to
…
(
Yo
n
g
-
Jin
Ju
n
g
)
2505
d
ata
co
r
r
esp
o
n
d
in
g
to
'
g
o
o
d
'
a
n
d
'm
o
d
er
ate'
u
s
ed
f
o
r
clas
s
if
i
ca
tio
n
w
as
h
i
g
h
.
Fo
r
an
u
n
b
al
an
ce
d
m
o
d
el
w
it
h
th
e
h
ig
h
p
r
o
p
o
r
tio
n
o
f
a
s
p
ec
if
ic
cla
s
s
,
t
h
e
g
e
n
er
all
y
u
s
ed
ac
cu
r
ac
y
i
s
m
ea
n
i
n
g
less
.
T
h
e
r
ef
o
r
e,
w
e
s
h
o
u
ld
ev
alu
a
te
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
m
o
d
el
b
y
tak
i
n
g
in
to
ac
co
u
n
t
a
ll
cla
s
s
e
s
w
ith
t
h
e
s
a
m
e
p
r
o
p
o
r
tio
n
,
an
d
u
s
ed
th
e
m
ea
n
o
f
m
ac
r
o
f
-
s
co
r
e
.
Oth
er
th
a
n
th
e
lo
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el
w
i
th
an
f
-
s
co
r
e
o
f
0
.
4
9
3
0
,
t
h
e
o
th
er
m
o
d
el
s
s
h
o
w
ed
th
e
s
i
m
ilar
s
co
r
es
w
i
th
0
.
8
2
6
4
,
0
.
8
2
7
7
,
an
d
0
.
8
2
6
5
.
(
a)
(
b
)
(
c)
(
d
)
Fig
u
r
e
3
.
C
o
n
f
u
s
io
n
m
atr
ix
;
(
a)
L
o
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el
,
(
b
)
Dec
is
io
n
tr
ee
m
o
d
el
,
(
c)
SVM
m
o
d
el
,
(
d
)
E
n
s
e
m
b
le
m
o
d
el
T
ab
le
7
.
C
lass
if
icatio
n
p
er
f
o
r
m
an
ce
e
v
al
u
atio
n
b
y
m
o
d
els
L
o
g
i
st
i
c
r
e
g
r
e
ssi
o
n
D
e
c
i
si
o
n
t
r
e
e
S
V
M
En
se
mb
l
e
c
l
a
ss
p
r
e
c
i
si
o
n
r
e
c
a
l
l
p
r
e
c
i
si
o
n
r
e
c
a
l
l
p
r
e
c
i
si
o
n
r
e
c
a
l
l
p
r
e
c
i
si
o
n
r
e
c
a
l
l
0
(
g
o
o
d
)
0
.
8
6
8
5
0
.
8
3
7
8
0
.
8
6
1
6
0
.
8
5
2
8
0
.
8
5
7
7
0
.
8
5
7
8
0
.
8
5
7
9
0
.
8
5
7
7
1
(
mo
d
e
r
a
t
e
)
0
.
8
1
6
2
0
.
9
3
4
1
0
.
8
9
7
7
0
.
9
0
2
3
0
.
8
9
9
7
0
.
8
9
9
7
0
.
8
9
9
7
0
.
8
9
9
8
2
(
b
a
d
)
0
.
6
9
2
5
0
.
1
5
1
0
0
.
7
8
8
1
0
.
7
8
8
5
0
.
7
8
8
5
0
.
7
8
8
5
0
.
7
8
8
1
0
.
7
8
8
5
3
(
v
e
r
y
b
a
d
)
0
.
0
0
0
0
0
.
0
0
0
0
0
.
7
6
3
0
0
.
7
5
7
4
0
.
7
6
4
7
0
.
7
6
4
7
0
.
7
6
3
0
0
.
7
5
7
4
f
-
sco
r
e
(
mac
r
o
)
0
.
4
9
3
0
0
.
8
2
6
4
0
.
8
2
7
7
0
.
8
2
6
5
5.
CO
NCLU
SI
O
N
I
n
p
r
ed
ictin
g
p
ar
ticu
la
te
m
att
er
co
n
ce
n
tr
atio
n
s
,
t
h
er
e
is
a
p
r
o
b
lem
o
f
tr
ain
in
g
p
ar
ticu
la
te
m
atter
co
n
ce
n
tr
atio
n
p
r
ed
ictio
n
m
o
d
els
b
ec
au
s
e
o
f
th
e
c
h
ar
ac
ter
is
tics
o
f
p
ar
ticu
late
m
atter
.
I
n
o
r
d
er
to
s
o
lv
e
th
is
p
r
o
b
lem
,
v
ar
io
u
s
s
t
u
d
ies
h
a
v
e
b
ee
n
u
n
d
er
w
a
y
s
u
c
h
as
p
er
f
o
r
m
i
n
g
p
r
ed
ictio
n
b
y
d
iv
id
in
g
p
ar
ticu
late
m
atte
r
co
n
ce
n
tr
atio
n
s
b
ased
o
n
a
s
p
e
cif
ic
co
n
ce
n
tr
atio
n
.
I
n
th
i
s
p
a
p
er
,
to
im
p
r
o
v
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
p
ar
ticu
la
te
m
atter
co
n
ce
n
tr
atio
n
p
r
ed
ictio
n
m
o
d
el,
w
e
p
r
o
p
o
s
ed
m
u
lt
ip
le
class
i
f
icatio
n
m
o
d
els
t
h
at
p
r
o
v
id
ed
p
ar
ticu
late
m
atter
co
n
ce
n
tr
at
io
n
s
i
n
f
o
u
r
class
es
b
ased
o
n
th
e
AQI
.
T
o
th
is
e
n
d
,
w
e
co
n
f
i
g
u
r
ed
d
ata
s
ets
b
y
s
elec
tin
g
air
p
o
llu
tan
t
d
ata
an
d
m
eteo
r
o
lo
g
ical
ele
m
e
n
ts
co
llected
at
an
i
n
ter
v
al
o
f
a
n
h
o
u
r
f
o
r
1
0
y
ea
r
s
ar
o
u
n
d
C
h
eo
n
an
.
As
t
h
e
clas
s
i
f
icatio
n
m
o
d
el
s
i
n
th
i
s
s
t
u
d
y
,
w
e
u
s
e
d
th
e
lo
g
i
s
tic
r
eg
r
ess
io
n
,
d
ec
is
io
n
tr
ee
,
SVM,
en
s
e
m
b
le.
I
n
o
r
d
er
t
o
ap
p
ly
o
p
ti
m
a
l
p
ar
am
eter
s
to
ea
ch
m
o
d
el,
w
e
s
e
ar
ch
ed
th
e
p
ar
a
m
eter
s
th
r
o
u
g
h
g
r
id
s
ea
r
ch
cr
o
s
s
v
alid
atio
n
.
W
e
b
u
ilt
th
e
en
s
e
m
b
le
m
o
d
el
b
y
co
m
b
i
n
i
n
g
th
e
lo
g
is
tic
r
eg
r
ess
io
n
,
d
ec
is
i
o
n
tr
ee
,
a
n
d
SVM
m
o
d
el
s
in
to
o
n
e.
W
e
u
s
ed
e
r
r
o
r
m
atr
ix
e
s
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
f
o
u
r
m
u
l
tip
le
class
i
f
icatio
n
m
o
d
els.
L
o
g
i
s
tic
r
eg
r
es
s
io
n
s
h
o
w
ed
p
o
o
r
p
r
ec
is
io
n
,
r
e
ca
ll
,
an
d
f
-
s
co
r
e
co
m
p
ar
ed
to
o
th
er
cl
ass
i
f
icatio
n
m
o
d
els.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
11
,
No
.
3
,
J
u
n
e
2
0
2
1
:
2
5
0
0
-
2507
2506
Dec
is
io
n
tr
ee
,
SVM,
a
n
d
en
s
e
m
b
le
m
o
d
e
ls
all
s
h
o
w
ed
th
e
p
r
ec
is
io
n
an
d
r
ec
all
w
it
h
0
.
8
5
o
r
h
ig
h
er
f
o
r
'g
o
o
d
'
an
d
'm
o
d
er
ate'
b
ased
o
n
t
h
e
A
QI
,
w
h
er
ea
s
th
e
y
s
h
o
w
ed
0
.
7
5
~0
.
7
9
f
o
r
'
’
b
ad
'
a
n
d
'v
er
y
b
ad
'
.
W
e
co
n
f
ir
m
ed
th
at
th
i
s
w
a
s
b
ec
au
s
e
th
e
p
ar
ticu
late
m
atter
d
ata
u
s
ed
in
th
e
class
i
f
icatio
n
m
o
d
els
w
er
e
u
n
b
alan
ce
d
d
ata
w
it
h
th
e
h
i
g
h
p
r
o
p
o
r
tio
n
o
f
a
s
p
e
cif
ic
clas
s
.
A
cc
o
r
d
in
g
l
y
,
w
e
v
er
if
ied
t
h
e
s
co
r
es
o
f
t
h
e
m
o
d
els
b
y
ta
k
i
n
g
in
to
ac
co
u
n
t
all
clas
s
es
w
i
th
t
h
e
s
a
m
e
p
r
o
p
o
r
tio
n
,
an
d
f
o
u
n
d
th
at
th
e
m
o
d
els
o
th
er
t
h
an
t
h
e
lo
g
is
tic
r
e
g
r
ess
io
n
m
o
d
el
s
h
o
w
ed
a
s
co
r
e
o
f
0
.
8
2
o
r
h
ig
h
er
.
O
f
th
e
s
e
m
o
d
els,
th
e
SVM
m
o
d
el
s
h
o
w
ed
t
h
e
b
est
class
i
f
icatio
n
p
er
f
o
r
m
a
n
ce
w
it
h
0
.
8
2
7
7
.
I
n
f
u
tu
r
e,
in
o
r
d
er
to
ad
d
r
ess
th
e
p
r
o
b
lem
o
f
u
n
b
alan
ce
d
d
ata,
w
e
ar
e
g
o
in
g
to
co
m
p
ar
e
clas
s
i
f
icatio
n
p
er
f
o
r
m
an
ce
t
h
r
o
u
g
h
t
h
e
alg
o
r
it
h
m
ch
an
g
e
s
o
f
th
e
clas
s
i
f
icatio
n
m
o
d
el
s
an
d
d
esig
n
a
p
ar
ticu
late
m
at
ter
co
n
ce
n
tr
atio
n
p
r
ed
ictio
n
m
o
d
el
b
ased
o
n
th
e
i
m
p
r
o
v
ed
clas
s
i
f
icatio
n
m
o
d
el
s
.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
is
r
esear
ch
w
a
s
s
u
p
p
o
r
ted
b
y
B
asic
Scie
n
ce
R
e
s
ea
r
ch
P
r
o
g
r
a
m
th
r
o
u
g
h
th
e
Natio
n
al
R
esear
ch
Fo
u
n
d
atio
n
o
f
Ko
r
ea
(
NR
F)
f
u
n
d
ed
b
y
t
h
e
Mi
n
is
tr
y
o
f
E
d
u
ca
tio
n
(
2
0
1
9
R
1
I
1
A
3
A
0
1
0
5
9
0
3
8
)
RE
F
E
R
E
NC
E
S
[1
]
G
.
W
.
Ev
a
n
s,
"
A
ir
P
o
ll
u
ti
o
n
a
n
d
Hu
m
a
n
Be
h
a
v
io
r,
"
J
o
u
rn
a
l
o
f
S
o
c
ia
l,
v
o
l
.
3
7
,
n
o
.
1
,
p
p
.
9
5
-
1
2
5
,
A
p
r
.
2
0
1
0
.
[2
]
M
.
S
.
S
e
o
,
"
T
h
e
I
m
p
a
c
t
o
f
P
a
rti
c
u
late
M
a
tt
e
r
o
n
Eco
n
o
m
ic
A
c
ti
v
it
y
,
"
T
h
e
Ko
re
a
n
W
o
me
n
Eco
n
o
mi
sts
Asso
c
ia
ti
o
n
,
v
o
l.
1
2
,
p
p
.
7
5
-
1
0
0
,
Ju
n
.
2
0
1
5
.
[3
]
A
.
V
a
lav
a
n
id
is,
K
.
F
io
tak
is,
a
n
d
T
.
V
lac
h
o
g
ian
n
i
.
,
"
A
irb
o
rn
e
P
a
rti
c
u
late
M
a
tt
e
r
a
n
d
Hu
m
a
n
He
a
lt
h
:
T
o
x
ico
lo
g
ica
l
A
s
s
e
ss
m
e
n
t
a
n
d
Im
p
o
rtan
c
e
o
f
S
ize
a
n
d
Co
m
p
o
siti
o
n
o
f
P
a
rt
icle
s
f
o
r
Ox
id
a
ti
v
e
D
a
m
a
g
e
a
n
d
Ca
rc
in
o
g
e
n
ic
M
e
c
h
a
n
ism
s
,
"
J
o
u
rn
a
l
o
f
En
v
iro
n
me
n
ta
l
S
c
ien
c
e
a
n
d
He
a
lt
h
,
Pa
rt
C,
v
o
l.
2
6
,
n
o
.
4
,
p
p
.
3
3
9
-
3
6
2
,
2
0
0
8
.
[4
]
J.
O.
A
n
d
e
rso
n
,
Jo
se
f
G
.
T
h
u
n
d
iy
il
a
n
d
A
n
d
re
w
S
to
lb
a
c
h
,
"
Clea
rin
g
th
e
A
ir:
A
Re
v
iew
o
f
th
e
Ef
f
e
c
t
s
o
f
P
a
rti
c
u
late
M
a
tt
e
r
A
ir
P
o
ll
u
t
io
n
o
n
Hu
m
a
n
He
a
lt
h
,
"
J
o
u
rn
a
l
o
f
M
e
d
ica
l
T
o
x
ico
lo
g
y
,
v
o
l.
8
,
n
o
.
2
,
p
p
.
1
6
6
-
1
7
5
,
2
0
1
2
.
[5
]
K.
H.
Ki
m
,
E.
Ka
b
ir
a
n
d
S
.
Ka
b
ir
,
"
A
R
e
v
ie
w
o
n
th
e
Hu
m
a
n
He
a
lt
h
Im
p
a
c
t
o
f
A
irb
o
rn
e
P
a
rti
c
u
l
a
te
M
a
tt
e
r,
"
En
v
iro
n
me
n
t
In
ter
n
a
ti
o
n
a
l,
v
o
l.
7
4
,
p
p
.
1
3
6
-
1
4
3
,
2
0
1
5
.
[6
]
N.
J.
Him
e
,
e
t
a
l.
,
"
A
Co
m
p
a
ris
o
n
o
f
th
e
He
a
lt
h
Ef
f
e
c
ts
o
f
Am
b
ien
t
P
a
rti
c
u
late
M
a
tt
e
r
A
ir
P
o
ll
u
t
io
n
F
ro
m
F
iv
e
Em
issio
n
S
o
u
rc
e
s,"
In
ter
n
a
t
io
n
a
l
J
o
u
r
n
a
l
o
f
E
n
v
iro
n
me
n
t
a
l
Res
e
a
rc
h
a
n
d
Pu
b
li
c
He
a
l
th
,
v
o
l.
1
5
,
n
o
.
6
,
p
p
.
1
-
2
4
,
2
0
1
6
,
d
o
i:
1
0
.
3
3
9
0
/
ij
e
rp
h
1
5
0
6
1
2
0
6
.
[7
]
W
o
rld
He
a
lt
h
Org
a
n
iza
ti
o
n
(W
HO
),
"
He
a
lt
h
e
ffe
c
ts
o
f
p
a
rti
c
u
late
m
a
tt
e
r.
P
o
li
c
y
i
m
p
li
c
a
ti
o
n
s
f
o
r
c
o
u
n
tri
e
s
i
n
e
a
ste
rn
Eu
ro
p
e
,
Ca
u
c
a
su
s a
n
d
c
e
n
tral
A
sia
,
"
Reg
io
n
a
l
Offi
c
e
f
o
r E
u
ro
p
e
,
2
0
1
3
.
[8
]
Bo
a
rd
o
f
A
d
it
a
n
d
I
n
sp
e
c
ti
o
n
(BA
I),
"
W
e
a
th
e
r
f
o
re
c
a
st an
d
e
a
rth
q
u
a
k
e
n
o
ti
f
ica
ti
o
n
sy
st
e
m
o
p
e
ra
ti
o
n
,
"
In
ter
n
a
ti
o
n
a
l
T
HE
Bo
a
rd
o
f
A
u
d
it
a
n
d
In
s
p
e
c
ti
o
n
o
f
KORE
A,
2
0
1
7
.
[9
]
K.
W
.
Ch
o
,
e
t
a
l.
,
"
S
e
p
a
ra
ti
o
n
P
r
e
d
ictio
n
M
o
d
e
l
b
y
Co
n
c
e
n
tratio
n
b
a
se
d
o
n
De
e
p
Ne
u
ra
l
Ne
tw
o
rk
f
o
r
I
m
p
ro
v
in
g
P
M
1
0
F
o
re
c
a
st
A
c
c
u
ra
c
y
,
"
J
o
u
rn
a
l
o
f
t
h
e
Ko
re
a
In
st
it
u
te
o
f
In
f
o
r
ma
ti
o
n
a
n
d
Co
mm
u
n
ica
ti
o
n
E
n
g
i
n
e
e
rin
g
,
v
o
l.
2
4
,
n
o
.
1
,
p
p
.
8
-
1
4
,
2
0
2
0
.
[1
0
]
J.
W
.
Ch
a
a
n
d
J.
Y.
Kim
,
"
De
v
e
lo
p
m
e
n
t
o
f
Da
ta
M
i
n
in
g
A
lg
o
rit
h
m
f
o
r
Im
p
le
m
e
n
tatio
n
o
f
F
in
e
Du
st
Nu
m
e
rica
l
P
re
d
ictio
n
M
o
d
e
l,
"
J
o
u
rn
a
l
o
f
t
h
e
Ko
re
a
In
stit
u
te
o
f
I
n
f
o
rm
a
ti
o
n
a
n
d
Co
mm
u
n
ica
ti
o
n
En
g
in
e
e
rin
g
,
v
o
l.
2
2
,
n
o
.
4
,
p
p
.
5
9
5
-
6
0
1
,
2
0
1
8
.
[1
1
]
A
.
Ch
a
lo
u
lak
o
u
,
G
.
G
riv
a
s,
a
n
d
N
.
S
p
y
re
ll
is
,
"
N
e
u
ra
l
Ne
t
w
o
rk
a
n
d
M
u
l
ti
p
le
Re
g
re
ss
io
n
M
o
d
e
ls
f
o
r
P
M
1
0
P
re
d
ictio
n
in
A
th
e
n
s:
A
Co
m
p
a
ra
ti
v
e
A
ss
e
ss
m
e
n
t,
"
J
o
u
rn
a
l
o
f
t
h
e
Ai
r
&
W
a
ste
M
a
n
a
g
e
me
n
t
Asso
c
i
a
ti
o
n
,
v
o
l.
5
3
,
no.
1
0
,
p
p
.
1
1
8
3
-
1
1
9
0
,
2
0
0
3
.
[1
2
]
K.
Ka
y
a
a
n
d
S
.
G
.
Og
u
d
u
c
u
,
"
A
Bin
a
ry
Cla
ss
i
f
ica
ti
o
n
M
o
d
e
l
f
o
r
P
M
1
0
L
e
v
e
ls,"
2
0
1
8
3
r
d
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
C
o
mp
u
ter
S
c
ien
c
e
a
n
d
En
g
in
e
e
rin
g
(
UBM
K),
2
0
1
8
,
p
p
.
3
6
1
-
3
6
6
.
[1
3
]
J.
M
.
Ha
n
,
Ja
e
-
G
o
o
Kim
,
a
n
d
Ki
-
H
y
u
n
Ch
o
,
"
V
e
rif
y
a
Ca
u
sa
l
Re
l
a
ti
o
n
sh
i
p
b
e
tw
e
e
n
F
in
e
Du
st
a
n
d
A
ir
Co
n
d
it
i
o
n
-
W
e
a
th
e
r
Da
ta i
n
S
e
lec
ted
A
re
a
b
y
Co
n
tam
in
a
ti
o
n
F
a
c
to
rs,"
T
h
e
jo
u
rn
a
l
o
f
Bi
g
d
a
t
a
,
v
o
l
.
2
,
n
o
.
1
,
p
p
.
1
7
-
2
6
,
2
0
1
7
.
[1
4
]
X
.
Z
h
a
o
,
e
t
a
l.
,
"
A
D
e
e
p
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
f
o
r
A
ir
Qu
a
li
t
y
Clas
si
f
ica
ti
o
n
,
"
J
o
u
rn
a
l
o
f
I
n
fo
rm
a
ti
o
n
Hi
d
in
g
a
n
d
M
u
lt
ime
d
i
a
S
ig
n
a
l
Pro
c
e
ss
in
g
,
v
o
l.
9
,
p
p
.
3
4
6
-
3
5
4
,
2
0
1
8
.
[1
5
]
B.
T
.
On
g
,
K
.
S
u
g
iu
ra
,
a
n
d
K
.
Z
e
tt
su
.
,
"
Dy
n
a
m
ic
p
re
-
train
in
g
o
f
De
e
p
Re
c
u
rre
n
t
Ne
u
ra
l
Ne
tw
o
rk
s
f
o
r
p
re
d
icti
n
g
e
n
v
iro
n
m
e
n
tal
m
o
n
it
o
rin
g
d
a
ta,"
2
0
1
4
IE
EE
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
Bi
g
D
a
ta
(
Bi
g
Da
t
a
),
2
0
1
4
,
p
p
.
7
6
0
-
7
6
5
.
[1
6
]
X
.
L
i,
e
t
a
l.
,
"
L
o
n
g
sh
o
rt
-
term
m
e
m
o
r
y
n
e
u
ra
l
n
e
tw
o
rk
f
o
r
a
ir
p
o
l
lu
tan
t
c
o
n
c
e
n
trati
o
n
p
re
d
i
c
ti
o
n
s:
M
e
th
o
d
d
e
v
e
lo
p
m
e
n
t
a
n
d
e
v
a
lu
a
ti
o
n
,
"
En
v
iro
n
me
n
ta
l
Po
ll
u
t
io
n
,
v
o
l.
2
3
1
,
p
p
.
9
9
7
-
1
0
0
4
,
2
0
1
7
.
[1
7
]
S
.
H.
Je
o
n
a
n
d
Y.
S
.
S
o
n
,
"
P
re
d
i
c
ti
o
n
o
f
f
in
e
d
u
st
P
M
1
0
u
sin
g
a
d
e
e
p
n
e
u
ra
l
n
e
tw
o
rk
m
o
d
e
l,
"
T
h
e
Ko
re
a
n
jo
u
rn
a
l
o
f
a
p
p
li
e
d
st
a
ti
stic
s,
v
o
l
.
3
1
,
n
o
.
2
,
p
p
.
2
6
5
-
2
8
5
,
2
0
1
8
.
[1
8
]
R.
S
.
M
ic
h
a
lsk
i,
e
t
a
l
.
,
“
L
e
a
rn
in
g
Ef
f
icie
n
t
Clas
sif
i
c
a
ti
o
n
P
r
o
c
e
d
u
r
e
s
a
n
d
T
h
e
ir
A
p
p
li
c
a
ti
o
n
t
o
Ch
e
ss
En
d
G
a
m
e
s,
”
M
a
c
h
in
e
L
e
a
r
n
in
g
,
p
p
.
4
6
3
-
4
8
2
,
1
9
8
3
.
[1
9
]
C.
Co
rtes
a
n
d
V
.
V
a
p
n
ik
,
"
S
u
p
p
o
rt
-
v
e
c
to
r
n
e
tw
o
rk
s,"
M
a
c
h
in
e
L
e
a
rn
in
g
,
v
o
l.
2
0
,
n
o
.
3
,
p
p
.
2
7
3
-
2
9
7
,
1
9
9
5
.
[2
0
]
P
.
H.
Hu
y
n
h
a
n
d
T
h
a
n
h
-
Ng
h
i
Do
,
"
En
h
a
n
c
i
n
g
g
e
n
e
e
x
p
re
ss
io
n
c
las
si
f
ica
ti
o
n
o
f
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
s
w
it
h
g
e
n
e
ra
ti
v
e
a
d
v
e
rs
a
rial
n
e
tw
o
rk
s,
"
J
o
u
rn
a
l
o
f
in
f
o
rm
a
ti
o
n
a
n
d
c
o
mm
u
n
ica
ti
o
n
c
o
n
v
e
rg
e
n
c
e
e
n
g
i
n
e
e
rin
g
,
v
o
l
.
1
7
,
n
o
.
1
,
p
p
.
1
4
-
2
0
,
2
0
1
9
.
[2
1
]
G
.
W
a
n
g
a
n
d
S
.
Y.
S
h
i
n
,
"
A
n
Im
p
ro
v
e
d
Tex
t
Clas
si
f
ica
ti
o
n
M
e
th
o
d
f
o
r
S
e
n
ti
m
e
n
t
Clas
sif
ic
a
ti
o
n
,
"
J
o
u
r
n
a
l
o
f
in
fo
rm
a
ti
o
n
a
n
d
c
o
mm
u
n
ica
ti
o
n
c
o
n
v
e
rg
e
n
c
e
e
n
g
i
n
e
e
rin
g
,
v
o
l.
1
7
,
n
o
.
1
,
p
p
.
4
1
-
4
8
,
2
0
1
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
C
o
mp
a
r
a
tive
a
n
a
lysi
s
o
f m
u
ltip
le
cla
s
s
ifica
tio
n
mo
d
els to
…
(
Yo
n
g
-
Jin
Ju
n
g
)
2507
[2
2
]
L
.
Ro
k
a
c
h
,
"
En
se
m
b
le
-
b
a
se
d
c
las
sif
iers
,
"
Arti
fi
c
ia
l
I
n
telli
g
e
n
c
e
Rev
iew,
v
o
l.
3
3
,
n
o
.
1
-
2
,
p
p
.
1
-
3
9
,
2
0
1
0
.
[2
3
]
R.
P
o
li
k
a
r,
"
En
se
m
b
le
b
a
se
d
sy
s
tem
s
in
d
e
c
isio
n
m
a
k
in
g
,
"
IEE
E
Circ
u
it
s
a
n
d
S
y
ste
ms
M
a
g
a
zin
e
,
v
o
l.
6
,
n
o
.
3
,
p
p
.
2
1
-
4
5
,
2
0
0
6
.
[2
4
]
R.
E.
S
u
tan
t
o
a
n
d
S
.
H.
L
e
e
,
"
En
se
m
b
le
o
f
De
g
ra
d
e
d
A
rti
f
ici
a
l
In
t
e
ll
ig
e
n
c
e
M
o
d
u
les
A
g
a
in
st
A
d
v
e
rsa
rial
A
tt
a
c
k
s
o
n
Ne
u
ra
l
Ne
tw
o
rk
s,"
J
o
u
rn
a
l
o
f
i
n
fo
rm
a
ti
o
n
a
n
d
c
o
mm
u
n
ica
t
io
n
c
o
n
v
e
rg
e
n
c
e
e
n
g
in
e
e
rin
g
,
v
o
l.
1
6
,
n
o
.
3
,
p
p
.
1
4
8
-
1
5
2
,
2
0
1
8
.
[2
5
]
A
.
Hu
sin
a
n
d
K.
R.
Ku
-
M
a
h
a
m
u
d
,
"
A
n
t
S
y
ste
m
a
n
d
W
e
ig
h
ted
Vo
ti
n
g
M
e
th
o
d
f
o
r
M
u
lt
i
p
le
Clas
sif
ier
S
y
ste
m
s,
"
In
ter
n
a
t
io
n
a
l
J
o
u
rn
a
l
o
f
E
lec
trica
l
a
n
d
C
o
mp
u
ter
En
g
in
e
e
rin
g
(
IJ
ECE
)
,
v
o
l.
8
,
n
o
.
6
,
p
p
.
4
7
0
5
-
4
7
1
2
,
2
0
1
8
.
B
I
O
G
RAP
H
I
E
S
O
F
AUTH
O
RS
Yo
ng
-
J
in
J
un
g
,
r
ec
eiv
ed
h
is
B
.
S.
in
E
lectr
o
n
ics
E
n
g
i
n
ee
r
in
g
f
r
o
m
Ko
n
g
j
u
Natio
n
al
Un
i
v
er
s
it
y
,
C
h
eo
n
an
,
So
u
t
h
K
o
r
ea
,
in
2
0
1
4
,
an
d
h
is
M.
S.
in
E
lectr
ical,
E
lectr
o
n
ics
an
d
C
o
m
m
u
n
ica
tio
n
E
n
g
i
n
ee
r
in
g
f
r
o
m
Ko
r
ea
Un
iv
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
an
d
E
d
u
ca
tio
n
(
KORE
A
T
E
C
H)
in
2
0
1
6
.
He
is
cu
r
r
en
tl
y
p
u
r
s
u
i
n
g
a
P
h
.
D.
in
E
lectr
ical,
E
lectr
o
n
ics
an
d
C
o
m
m
u
n
icatio
n
E
n
g
i
n
ee
r
in
g
at
t
h
e
Ko
r
ea
Un
i
v
er
s
i
t
y
o
f
T
ec
h
n
o
lo
g
y
a
n
d
E
d
u
ca
t
io
n
(
KORE
A
T
E
C
H)
.
His
r
esear
ch
in
ter
es
ts
in
cl
u
d
e
m
ac
h
i
n
e
lear
n
in
g
,
d
ata
an
al
y
s
is
an
d
d
ee
p
lear
n
in
g
.
K
y
o
un
g
-
Wo
o
Cho
,
r
ec
eiv
ed
h
is
B
.
S.
in
E
lectr
o
n
ics
E
n
g
i
n
ee
r
i
n
g
f
r
o
m
Ko
n
g
j
u
Natio
n
al
Un
i
v
er
s
i
t
y
,
C
h
eo
n
a
n
,
So
u
th
Ko
r
ea
,
in
2
0
1
3
,
a
n
d
h
is
M.
S.
in
E
lectr
ical,
E
lectr
o
n
ics
a
n
d
C
o
m
m
u
n
icati
o
n
E
n
g
in
ee
r
i
n
g
f
r
o
m
Ko
r
ea
Un
i
v
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
an
d
E
d
u
ca
tio
n
(
KORE
A
T
E
C
H)
in
2
0
1
5
.
He
is
c
u
r
r
en
tl
y
p
u
r
s
u
i
n
g
a
P
h
.
D.
in
E
lectr
ical,
E
lectr
o
n
ics
an
d
C
o
m
m
u
n
icati
o
n
E
n
g
in
ee
r
i
n
g
at
th
e
Ko
r
ea
Un
i
v
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
an
d
E
d
u
ca
tio
n
(
KORE
A
T
E
C
H)
.
His
r
esear
ch
in
ter
e
s
ts
i
n
cl
u
d
e
m
ac
h
in
e
lear
n
i
n
g
,
d
ee
p
lear
n
in
g
a
n
d
p
ar
ticu
late
m
a
tter
s
p
r
ed
ictio
n
.
J
o
ng
-
Su
ng
L
ee
,
r
ec
eiv
ed
h
is
B
.
S in
I
n
f
o
r
m
atio
n
C
o
m
m
u
n
ic
atio
n
f
r
o
m
Ko
r
ea
Naz
ar
en
e
Un
i
v
er
s
it
y
,
C
h
eo
n
a
n
,
So
u
th
K
o
r
ea
,
in
2
0
1
1
,
an
d
h
is
M.
S
i
n
E
lectr
ical,
E
lectr
o
n
ics
a
n
d
C
o
m
m
u
n
ica
tio
n
E
n
g
i
n
ee
r
in
g
f
r
o
m
Ko
r
ea
Un
iv
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
an
d
E
d
u
ca
tio
n
(
KORE
A
T
E
C
H)
i
n
2
0
1
6
.
He
is
cu
r
r
en
tl
y
p
u
r
s
u
i
n
g
a
P
h
.
D.
in
E
lectr
ical,
E
lectr
o
n
ics
an
d
C
o
m
m
u
n
icatio
n
E
n
g
i
n
ee
r
in
g
at
t
h
e
Ko
r
ea
Un
i
v
er
s
i
t
y
o
f
T
ec
h
n
o
lo
g
y
a
n
d
E
d
u
ca
t
io
n
(
KORE
A
T
E
C
H)
.
His
r
esear
c
h
i
n
ter
est
s
i
n
clu
d
e
5
G
n
et
wo
r
k
,
m
ac
h
in
e
lear
n
i
n
g
an
d
d
ee
p
lear
n
in
g
.
Cha
ng
-
He
o
n
O
h
,
r
ec
eiv
ed
th
e
B
.
S.
a
n
d
M.
S.E
.
d
eg
r
ee
s
in
telec
o
m
m
u
n
icatio
n
a
n
d
in
f
o
r
m
atio
n
en
g
i
n
ee
r
in
g
f
r
o
m
Ko
r
ea
A
er
o
s
p
ac
e
Un
i
v
er
s
it
y
,
K
y
u
n
g
g
i
-
Do
,
Ko
r
ea
,
in
1
9
8
8
an
d
1
9
9
0
,
r
esp
ec
tiv
el
y
.
He
r
ec
eiv
ed
th
e
P
h
.
D.
d
eg
r
ee
in
av
io
n
ics
e
n
g
i
n
ee
r
i
n
g
f
r
o
m
Ko
r
ea
A
er
o
s
p
ac
e
Un
iv
e
r
s
it
y
,
K
y
u
n
g
g
i
-
Do
,
Ko
r
ea
,
in
1
9
9
6
.
Fro
m
Feb
r
u
ar
y
1
9
9
0
to
Au
g
u
s
t
1
9
9
3
,
h
e
w
as
w
it
h
Han
j
in
E
lectr
o
n
ic
s
C
o
.
,
w
h
e
r
e
h
e
w
a
s
i
n
v
o
l
v
ed
in
t
h
e
r
esear
ch
an
d
d
ev
e
lo
p
m
en
t
o
f
r
ad
io
co
m
m
u
n
icatio
n
&
m
o
n
ito
r
i
n
g
s
y
s
te
m
s
.
Fro
m
Octo
b
er
1
9
9
3
to
Feb
r
u
ar
y
1
9
9
9
,
h
e
w
a
s
w
it
h
th
e
C
DM
A
R
&
D
ce
n
ter
o
f
Sa
m
s
u
n
g
E
lectr
o
n
ics
C
o
.
,
w
h
er
e
h
e
w
as
in
v
o
l
v
ed
in
th
e
d
esig
n
an
d
d
ev
elo
p
m
en
t
o
f
C
DM
A
ce
llu
lar
s
y
s
te
m
s
a
n
d
C
D
M
A
P
C
S
s
y
s
te
m
s
f
o
r
s
u
cc
e
s
s
f
u
l
co
m
m
er
cial
C
DM
A
d
ep
lo
y
m
en
t
i
n
Ko
r
ea
.
Sin
ce
Ma
r
ch
1
9
9
9
,
h
e
h
as
b
ee
n
w
it
h
th
e
Sc
h
o
o
l
o
f
E
lectr
ical,
E
lectr
o
n
ics
an
d
C
o
m
m
u
n
icati
o
n
E
n
g
.
,
Ko
r
ea
Un
i
v
er
s
it
y
o
f
T
ec
h
n
o
lo
g
y
an
d
E
d
u
ca
tio
n
(
KORE
A
T
E
C
H)
,
C
h
eo
n
a
n
,
Ko
r
ea
,
w
h
er
e
h
e
i
s
cu
r
r
e
n
tl
y
a
p
r
o
f
ess
o
r
.
His
r
esear
ch
in
ter
est
is
in
th
e
ar
ea
s
o
f
w
i
r
eless
/
m
o
b
ile
co
m
m
u
n
ica
tio
n
,
w
ir
eless
lo
ca
lizatio
n
,
I
o
T
an
d
en
g
in
ee
r
i
n
g
ed
u
ca
tio
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.