I
nte
rna
t
io
na
l J
o
urna
l o
f
E
v
a
lua
t
io
n a
nd
Resea
rc
h in E
du
ca
t
io
n (
I
J
E
RE
)
Vo
l.
15
,
No
.
5
,
Octo
b
er
20
26
,
p
p
.
4
2
0
7
~
4
2
1
9
I
SS
N:
2
2
5
2
-
8
8
2
2
,
DOI
: 1
0
.
1
1
5
9
1
/
ijer
e
.
v
15
i
5
.
3
9
4
2
7
4207
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ere.
ia
esco
r
e.
co
m
The impa
ct
o
f
v
o
ca
tiona
l t
ra
ining
on fema
le labo
r
pr
o
ductivity
a
n eva
lua
tion in
Vietnam
’s
co
nstr
uction sec
tor
T
ruc
T
ha
i P
hu
o
ng
,
L
a
m
Ng
uy
en
So
n
F
a
c
u
l
t
y
o
f
C
i
v
i
l
E
n
g
i
n
e
e
r
i
n
g
,
I
n
d
u
st
r
i
a
l
U
n
i
v
e
r
s
i
t
y
o
f
H
o
C
h
i
M
i
n
h
C
i
t
y
,
H
o
C
h
i
M
i
n
h
C
i
t
y
,
V
i
e
t
n
a
m
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Mar
10
,
2
0
2
6
R
ev
is
ed
J
u
l
2
,
2
0
2
6
Acc
ep
ted
Au
g
5
,
2
0
2
6
Vie
tn
a
m
’s
c
o
n
stru
c
ti
o
n
se
c
to
r
h
a
s
e
x
p
e
rien
c
e
d
str
o
n
g
g
ro
wt
h
t
h
r
o
u
g
h
ra
p
i
d
u
rb
a
n
iza
ti
o
n
a
n
d
in
fra
str
u
c
tu
re
in
v
e
stm
e
n
t.
Ho
we
v
e
r,
fe
m
a
le
wo
rk
e
rs
c
o
n
ti
n
u
e
t
o
fa
c
e
sig
n
ifi
c
a
n
t
b
a
rr
iers
,
in
c
lu
d
i
n
g
h
i
g
h
in
f
o
rm
a
l
e
m
p
lo
y
m
e
n
t
ra
tes
(6
0
–
6
5
%
)
,
1
5
–
2
0
%
lo
we
r
l
a
b
o
r
p
r
o
d
u
c
ti
v
it
y
t
h
a
n
m
a
les
,
a
n
d
li
m
it
e
d
a
c
c
e
s
s
to
v
o
c
a
ti
o
n
a
l
trai
n
in
g
(
o
n
ly
1
9
–
2
5
%
h
o
l
d
c
e
rti
fica
tes
).
Th
is
st
u
d
y
in
v
e
stig
a
tes
th
e
imp
a
c
t
o
f
t
h
e
sh
a
re
o
f
fe
m
a
les
a
g
e
d
1
5
+
with
v
o
c
a
ti
o
n
a
l
train
in
g
c
e
rti
fica
tes
(S
TF
W1
5
+
)
o
n
fe
m
a
le
c
o
n
stru
c
ti
o
n
lab
o
r
p
r
o
d
u
c
t
iv
it
y
(F
CLP
).
Us
in
g
a
sh
o
rt
ti
m
e
-
se
ries
d
a
tas
e
t
(2
0
1
0
–
2
0
2
4
,
n
=
1
5
)
fro
m
Ge
n
e
ra
l
S
tatisti
c
s
Offic
e
(
G
S
O
)
o
f
Vie
tn
a
m
,
In
tern
a
ti
o
n
a
l
Lab
o
r
Org
a
n
iza
ti
o
n
(
ILO
)
,
a
n
d
Wo
rl
d
Ba
n
k
,
th
e
stu
d
y
a
p
p
l
i
e
d
F
o
re
stDiffu
sio
n
sy
n
th
e
ti
c
d
a
ta
p
ip
e
li
n
e
-
ti
m
e
-
se
ries
to
g
e
n
e
ra
te
7
5
h
i
g
h
-
q
u
a
li
t
y
sy
n
th
e
ti
c
sa
m
p
les
(to
tal
9
0
o
b
se
rv
a
ti
o
n
s).
An
a
rti
ficia
l
n
e
u
r
a
l
n
e
two
r
k
(AN
N)
m
o
d
e
l
in
teg
r
a
ted
with
S
Ha
p
ley
Ad
d
it
i
v
e
e
x
P
lan
a
ti
o
n
s
(S
HA
P
)
wa
s
e
m
p
lo
y
e
d
to
a
n
a
l
y
z
e
fe
a
tu
re
imp
o
rtan
c
e
a
n
d
n
o
n
l
in
e
a
r
in
tera
c
ti
o
n
s.
Re
su
lt
s
sh
o
w
th
a
t
S
TF
W
1
5
+
is
th
e
stro
n
g
e
st
p
re
d
icto
r
o
f
F
CL
P
(
m
e
a
n
|
S
H
AP|=0
.
5
1
4
).
It
e
x
e
rts
a
stro
n
g
p
o
siti
v
e
e
ffe
c
t
a
n
d
a
c
ts
a
s
a
p
o
we
rfu
l
m
o
d
e
ra
to
r
,
a
m
p
li
fy
in
g
b
e
n
e
fit
s
fro
m
in
c
o
m
e
a
n
d
e
c
o
n
o
m
ic
g
ro
wt
h
wh
il
e
m
it
ig
a
ti
n
g
t
h
e
n
e
g
a
ti
v
e
imp
a
c
t
o
f
in
fo
rm
a
l
e
m
p
lo
y
m
e
n
t.
T
h
e
fin
d
i
n
g
s
h
ig
h
li
g
h
t
t
h
e
c
rit
ica
l
ro
le
o
f
g
e
n
d
e
r
-
re
sp
o
n
si
v
e
tec
h
n
ica
l
a
n
d
v
o
c
a
ti
o
n
a
l
e
d
u
c
a
ti
o
n
a
n
d
trai
n
in
g
(T
VET)
.
Th
is
stu
d
y
re
c
o
m
m
e
n
d
s
targ
e
ted
v
o
c
a
t
io
n
a
l
trai
n
in
g
p
ro
g
ra
m
s,
g
e
n
d
e
r
i
n
teg
ra
ti
o
n
in
p
u
b
l
ic
in
v
e
stm
e
n
t,
a
n
d
p
u
b
li
c
-
p
riv
a
te
p
a
rtn
e
rs
h
ip
s
t
o
a
d
v
a
n
c
e
g
e
n
d
e
r
e
q
u
a
li
ty
a
n
d
su
p
p
o
r
t
su
sta
in
a
b
le d
e
v
e
lo
p
m
e
n
t
g
o
a
ls (
S
DG
s)
5
a
n
d
8
.
K
ey
w
o
r
d
s
:
C
o
n
s
tr
u
ctio
n
s
ec
to
r
Fem
ale
lab
o
r
L
ab
o
r
p
r
o
d
u
ctiv
ity
SHAP
Sy
n
th
etic
d
ata
au
g
m
en
tatio
n
Vo
ca
tio
n
al
tr
ain
in
g
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
:
L
am
Ng
u
y
e
n
So
n
Facu
lty
o
f
C
iv
il E
n
g
in
ee
r
in
g
,
I
n
d
u
s
tr
ial
Un
iv
er
s
ity
o
f
Ho
C
h
i M
in
h
C
ity
1
2
Ng
u
y
en
Van
B
ao
,
Han
h
T
h
o
n
g
W
ar
d
,
H
o
C
h
i M
in
h
,
Viet
n
am
E
m
ail: n
g
u
y
e
n
s
o
n
lam
@
iu
h
.
ed
u
.
v
n
1.
I
NT
RO
D
UCT
I
O
N
Vietn
am
’
s
co
n
s
tr
u
ctio
n
s
ec
to
r
h
as
g
r
o
wn
s
tr
o
n
g
ly
d
u
e
to
r
ap
id
u
r
b
a
n
izatio
n
an
d
lar
g
e
-
s
ca
le
in
f
r
astru
ctu
r
e
in
v
estme
n
t,
wi
th
g
r
o
wth
r
ates
o
f
9
–
9
.
6
%
in
2
0
2
5
an
d
c
o
n
tin
u
e
d
m
o
m
en
tu
m
e
x
p
ec
ted
in
2
0
2
6
[
1
]
.
Ho
wev
er
,
f
em
ale
wo
r
k
er
s
f
ac
e
m
ajo
r
b
ar
r
ier
s
:
in
f
o
r
m
al
em
p
lo
y
m
en
t
r
ates
o
f
6
0
–
6
5
%
(
ab
o
v
e
n
atio
n
al
av
er
a
g
e)
,
la
b
o
r
p
r
o
d
u
ctiv
ity
1
5
–
2
0
%
lo
wer
t
h
an
m
a
les,
ea
r
n
in
g
s
2
0
–
3
0
%
l
o
wer
in
m
an
u
al
r
o
les,
an
d
lim
ited
v
o
ca
tio
n
al
tr
ain
in
g
ac
ce
s
s
(
o
n
ly
1
9
–
2
5
%
h
o
ld
ce
r
tif
icate
s
,
v
er
s
u
s
th
e
n
atio
n
al
r
at
e
o
f
2
9
.
2
%
in
2
0
2
5
an
d
2
9
.
5
%
in
Q4
2
0
2
5
p
e
r
G
en
er
al
Statis
tics
Of
f
ice
(
GSO)
o
f
Vietn
am
.
T
ec
h
n
ical
an
d
v
o
ca
tio
n
al
ed
u
ca
tio
n
an
d
tr
ain
in
g
(
T
VE
T
)
p
lay
s
a
c
r
itical
r
o
le
in
ad
v
a
n
cin
g
g
en
d
er
eq
u
ality
an
d
r
e
d
u
cin
g
in
f
o
r
m
al
em
p
lo
y
m
e
n
t
in
d
ev
elo
p
in
g
co
u
n
tr
ies.
I
n
ter
n
a
tio
n
al
ev
id
en
ce
s
h
o
ws
th
at
g
en
d
er
-
r
esp
o
n
s
iv
e
T
VE
T
ca
n
d
r
iv
e
wid
er
g
en
d
e
r
tr
an
s
f
o
r
m
atio
n
an
d
s
u
p
p
o
r
t
th
e
d
ec
en
t w
o
r
k
ag
en
d
a
[
2
]
.
E
x
is
tin
g
s
tu
d
ies
o
n
f
em
ale
lab
o
r
in
Vietn
am
h
av
e
m
ain
ly
co
n
ce
n
tr
ated
o
n
tex
tiles
/g
ar
m
en
t
s
an
d
f
o
o
d
p
r
o
ce
s
s
in
g
s
ec
to
r
s
,
with
v
er
y
lim
ited
f
o
cu
s
o
n
co
n
s
tr
u
ctio
n
-
s
p
ec
if
ic
is
s
u
es
s
u
ch
as
p
h
y
s
ical
d
em
an
d
s
,
h
ig
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
2
2
I
n
t
J
E
v
al
&
R
es E
d
u
c
,
Vo
l
.
15
,
No
.
5
,
Octo
b
er
20
26
:
4
2
0
7
-
4219
4208
s
ea
s
o
n
ality
,
an
d
p
r
o
ject
-
b
ased
em
p
lo
y
m
en
t.
Mo
s
t
p
r
e
v
io
u
s
r
esear
ch
r
elies
o
n
tr
ad
itio
n
al
s
tatis
t
ical
m
eth
o
d
s
th
at
s
tr
u
g
g
le
with
s
m
all
tim
e
-
s
er
ies
s
am
p
les
an
d
s
ev
er
e
m
u
ltico
llin
ea
r
ity
.
Sev
e
r
al
s
tu
d
ies
h
av
e
a
p
p
lied
ar
tific
ial
in
tellig
en
ce
(
AI
)
tec
h
n
iq
u
es
to
p
r
ed
ict
lab
o
r
p
r
o
d
u
ctiv
ity
in
co
n
s
tr
u
ctio
n
p
r
o
ce
s
s
es,
in
clu
d
in
g
th
e
ap
p
licatio
n
o
f
ar
tific
ial
n
eu
r
a
l
n
etwo
r
k
(
ANN)
m
eth
o
d
o
lo
g
y
an
d
id
en
tific
atio
n
o
f
k
e
y
f
ac
to
r
s
in
f
lu
en
cin
g
f
o
r
m
wo
r
k
p
r
o
d
u
ctiv
ity
u
s
in
g
ANN.
Data
m
in
in
g
ap
p
r
o
ac
h
e
s
h
av
e
also
b
ee
n
u
s
ed
f
o
r
la
b
o
r
m
ar
k
et
f
o
r
ec
asti
n
g
[
3
]
.
T
h
is
s
tu
d
y
co
n
tr
ib
u
tes to
v
o
ca
tio
n
al
ed
u
ca
tio
n
an
d
co
n
s
tr
u
ctio
n
ed
u
ca
tio
n
b
y
p
r
o
v
id
in
g
an
ex
p
lain
a
b
le
AI
f
r
am
ewo
r
k
to
s
u
p
p
o
r
t
th
e
d
esig
n
o
f
g
en
d
er
-
r
esp
o
n
s
iv
e
T
VE
T
p
r
o
g
r
am
s
in
m
ale
-
d
o
m
in
ated
s
ec
to
r
s
.
I
t
ad
d
r
ess
es
cr
itical
g
ap
s
b
y
em
p
lo
y
in
g
in
ter
p
r
etab
le
e
n
s
em
b
le
lear
n
in
g
m
eth
o
d
s
[
4
]
,
m
et
a
-
en
s
em
b
le
m
o
d
els
f
o
r
c
o
n
s
tr
u
ctio
n
lab
o
r
p
r
o
d
u
c
tiv
ity
[
5
]
,
an
d
h
y
b
r
id
m
ac
h
i
n
e
lear
n
in
g
s
o
lu
tio
n
s
f
o
r
m
itig
atin
g
p
r
o
d
u
ctiv
ity
lo
s
s
es
[
6
]
.
Ad
v
an
ce
d
f
o
r
ec
asti
n
g
tech
n
iq
u
es
s
u
c
h
as
s
lip
f
o
r
m
lab
o
r
p
r
o
d
u
ctiv
ity
p
r
e
d
ictio
n
[
7
]
an
d
ANN
-
b
ased
p
r
ed
ictio
n
i
n
ter
v
als
[
8
]
f
u
r
th
er
d
em
o
n
s
tr
ate
th
e
p
o
ten
tial
o
f
th
ese
m
eth
o
d
s
,
wh
ile
c
o
m
p
ar
ativ
e
s
tu
d
ies
o
f
m
ac
h
in
e
lear
n
in
g
r
eg
r
ess
io
n
m
o
d
els h
ig
h
lig
h
t th
eir
ef
f
ec
t
iv
en
ess
in
co
n
s
tr
u
ctio
n
-
r
elate
d
p
r
ed
ictio
n
s
[
9
]
.
Gen
d
er
-
d
is
a
g
g
r
e
g
at
e
d
c
o
n
s
tr
u
cti
o
n
d
ata
in
Vie
tn
am
a
r
e
s
c
a
r
c
e,
l
im
ite
d
t
o
a
p
p
r
o
x
i
m
a
tel
y
1
5
a
n
n
u
al
o
b
s
e
r
v
ati
o
n
s
f
r
o
m
GSO
,
I
n
te
r
n
a
ti
o
n
al
L
a
b
o
r
O
r
g
a
n
i
z
ati
o
n
(
I
L
O)
,
a
n
d
W
o
r
l
d
B
an
k
,
wit
h
s
e
v
er
e
m
u
lti
co
lli
n
e
ar
it
y
(
r
≈
0
.
9
5
–
0
.
9
8
)
.
T
o
o
v
e
r
c
o
m
e
t
h
es
e
l
im
ita
ti
o
n
s
,
t
h
is
s
t
u
d
y
a
p
p
li
es t
h
e
Fo
r
es
tDif
f
u
s
i
o
n
s
y
n
th
eti
c
d
at
a
p
i
p
e
li
n
e
-
ti
m
e
-
s
e
r
i
es
t
o
g
e
n
e
r
a
te
7
5
h
i
g
h
-
q
u
al
it
y
s
y
n
t
h
eti
c
s
a
m
p
les
(
t
o
t
al
9
0
o
b
s
e
r
v
at
io
n
s
)
,
f
o
ll
o
w
ed
b
y
a
n
ANN
i
n
te
g
r
at
e
d
wit
h
SH
ap
l
ey
A
d
d
it
iv
e
e
x
Pl
an
ati
o
n
s
(
S
HAP)
f
o
r
i
n
t
er
p
r
e
ta
b
le
f
e
at
u
r
e
i
m
p
o
r
t
an
ce
an
d
n
o
n
li
n
ea
r
in
te
r
a
cti
o
n
a
n
al
y
s
is
[
1
0
]
.
T
h
is
w
o
r
k
al
ig
n
s
wit
h
Vie
tn
am
’
s
n
a
ti
o
n
al
s
tr
at
eg
y
o
n
g
e
n
d
e
r
e
q
u
a
lit
y
2
0
2
1
–
2
0
3
0
as
w
ell
as
g
l
o
b
a
l
em
p
h
asis
o
n
g
en
d
er
-
r
es
p
o
n
s
iv
e
T
VE
T
f
o
r
r
ed
u
ci
n
g
i
n
f
o
r
m
a
lit
y
a
n
d
ex
p
an
d
i
n
g
wo
m
e
n
’
s
o
p
p
o
r
tu
n
i
ties
i
n
t
h
e
co
n
s
t
r
u
ct
io
n
s
ec
t
o
r
.
T
h
e
n
o
v
e
l
ty
o
f
t
h
is
s
t
u
d
y
lies
i
n
b
e
in
g
t
h
e
f
i
r
s
t
t
o
c
o
m
b
i
n
e
Fo
r
estD
if
f
u
s
i
o
n
s
y
n
t
h
e
tic
d
a
ta
a
u
g
m
e
n
ta
ti
o
n
w
it
h
a
n
e
x
p
la
i
n
a
b
l
e
ANN
-
SHAP
f
r
a
m
ew
o
r
k
t
o
i
n
v
est
ig
ate
th
e
im
p
ac
t
o
f
v
o
ca
t
io
n
al
t
r
a
in
in
g
o
n
f
em
al
e
l
ab
o
r
p
r
o
d
u
ct
iv
it
y
in
Viet
n
am
’
s
co
n
s
t
r
u
ct
io
n
s
ec
to
r
.
T
h
is
a
p
p
r
o
a
ch
ef
f
e
cti
v
e
ly
a
d
d
r
ess
es
t
h
e
c
r
iti
c
al
c
h
a
lle
n
g
es
o
f
s
m
all
s
a
m
p
le
s
ize
an
d
s
ev
e
r
e
m
u
lti
co
lli
n
e
ar
i
ty
t
h
at
h
a
v
e
li
m
it
ed
p
r
e
v
i
o
u
s
r
es
ea
r
c
h
.
B
y
d
o
in
g
s
o
,
th
e
s
t
u
d
y
p
r
o
v
i
d
es
r
o
b
u
s
t
,
in
t
er
p
r
et
a
b
le
in
s
i
g
h
ts
i
n
t
o
t
h
e
m
o
d
e
r
at
o
r
r
o
le
o
f
v
o
ca
ti
o
n
al
t
r
ai
n
i
n
g
f
o
r
f
e
m
al
e
wo
r
k
er
s
i
n
a
m
ale
-
d
o
m
i
n
ate
d
i
n
d
u
s
t
r
y
.
T
h
e
m
ain
r
esear
c
h
q
u
esti
o
n
s
ar
e:
−
H
o
w
d
o
es
th
e
p
r
o
p
o
r
tio
n
o
f
f
em
ales
ag
ed
1
5
+
with
v
o
ca
tio
n
al
tr
ain
in
g
ce
r
tific
ates
(
STFW
1
5
+)
in
f
lu
e
n
ce
f
em
ale
co
n
s
tr
u
ctio
n
lab
o
r
p
r
o
d
u
ctiv
ity
(
FC
L
P)?
(
R
Q1
)
−
I
s
STFW
1
5
+
th
e
s
tr
o
n
g
est
p
r
ed
icto
r
o
f
FC
L
P
co
m
p
ar
ed
to
av
er
ag
e
m
o
n
th
ly
i
n
co
m
e
in
co
n
s
tr
u
ctio
n
(
AM
I
C
W
)
,
g
r
o
s
s
r
eg
io
n
al
d
o
m
esti
c
p
r
o
d
u
ct
p
er
ca
p
ita
(
GR
DPp
c)
,
an
d
th
e
s
h
ar
e
o
f
in
f
o
r
m
al
em
p
lo
y
m
e
n
t
in
co
n
s
tr
u
ctio
n
(
SIE
C
)
?
(
R
Q2
)
−
W
h
at
p
o
licy
m
ea
s
u
r
es
ca
n
im
p
r
o
v
e
f
em
ale
lab
o
r
p
r
o
d
u
ct
iv
ity
in
th
e
co
n
s
tr
u
ctio
n
s
ec
to
r
b
y
r
ed
u
cin
g
in
f
o
r
m
ality
a
n
d
s
tr
en
g
t
h
en
in
g
T
VE
T
?
(
R
Q3
)
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
Glo
b
a
l st
u
d
ies
h
i
g
h
li
g
h
t t
h
e
c
r
i
tica
l r
o
l
e
o
f
T
V
E
T
in
a
d
v
a
n
ci
n
g
g
e
n
d
e
r
e
q
u
al
it
y
a
n
d
r
e
d
u
ci
n
g
in
f
o
r
m
al
em
p
l
o
y
m
en
t in
d
ev
el
o
p
in
g
co
u
n
tr
ies
.
H
u
m
an
ca
p
it
al
t
h
e
o
r
y
p
o
s
its
t
h
a
t
i
n
v
est
m
e
n
ts
i
n
s
k
ill
s
e
n
h
a
n
c
e
i
n
d
i
v
i
d
u
al
p
r
o
d
u
ct
iv
it
y
a
n
d
ec
o
n
o
m
i
c
c
o
n
t
r
i
b
u
ti
o
n
s
[
1
1
]
.
Ge
n
d
e
r
e
d
la
b
o
r
m
ar
k
e
t
t
h
e
o
r
y
i
d
e
n
t
if
ies
s
tr
u
ct
u
r
al
b
a
r
r
ie
r
s
—
s
u
c
h
as
c
ar
e
g
i
v
i
n
g
r
es
p
o
n
s
ib
il
ities
a
n
d
li
m
it
e
d
ac
ce
s
s
t
o
s
p
e
cial
ize
d
s
k
il
ls
—
t
h
at
co
n
f
i
n
e
wo
m
e
n
t
o
in
f
o
r
m
a
l,
lo
w
-
p
r
o
d
u
ct
iv
it
y
r
o
l
es
[
1
2
]
.
Gen
d
er
-
r
es
p
o
n
s
i
v
e
T
VE
T
a
d
d
r
ess
es
s
k
i
ll
m
is
m
at
c
h
es
,
c
o
u
n
te
r
s
s
te
r
e
o
t
y
p
es
in
m
al
e
-
d
o
m
in
at
ed
s
ec
t
o
r
s
li
k
e
co
n
s
t
r
u
ct
io
n
,
r
ed
u
c
es
i
n
f
o
r
m
a
lit
y
,
a
n
d
e
n
h
a
n
c
es
i
n
c
o
m
e
a
n
d
s
o
cia
l
p
r
o
te
cti
o
n
.
Key
in
te
r
n
ati
o
n
a
l
r
ep
o
r
ts
s
u
p
p
o
r
t
th
is
v
i
ew
.
UN
E
SC
O
Gl
o
b
al
E
d
u
ca
t
io
n
M
o
n
it
o
r
i
n
g
R
e
p
o
r
t
(
G
E
M
r
e
p
o
r
t
)
em
p
h
asi
ze
g
en
d
er
-
r
es
p
o
n
s
iv
e
T
VE
T
f
o
r
c
lo
s
in
g
s
k
ill
a
n
d
e
m
p
l
o
y
m
en
t
g
a
p
s
in
n
o
n
-
t
r
a
d
i
ti
o
n
al
f
i
el
d
s
f
o
r
w
o
m
en
[
1
3
]
.
I
L
O
’
s
wo
r
l
d
em
p
l
o
y
m
e
n
t
a
n
d
s
o
ci
al
o
u
tl
o
o
k
:
tr
en
d
s
2
0
2
5
n
o
tes
th
at
v
o
ca
ti
o
n
a
l
tr
a
in
in
g
m
i
ti
g
at
es
in
f
o
r
m
al
e
m
p
lo
y
m
e
n
t
r
is
k
s
f
o
r
w
o
m
en
,
b
u
il
d
s
r
esil
i
e
n
ce
,
a
n
d
f
ac
i
lit
ates
t
r
a
n
s
iti
o
n
s
to
f
o
r
m
a
l
j
o
b
s
[
1
4
]
.
T
h
e
W
o
r
ld
B
a
n
k
(
2
0
2
4
–
2
0
2
5
)
u
n
d
e
r
s
c
o
r
es
T
V
E
T
’
s
v
a
lu
e
f
o
r
g
r
ee
n
a
n
d
d
i
g
it
al
tr
a
n
s
iti
o
n
s
i
n
V
iet
n
a
m
w
h
ile
h
i
g
h
li
g
h
ti
n
g
s
h
o
r
ta
g
es
o
f
g
e
n
d
er
-
d
is
a
g
g
r
e
g
at
ed
d
at
a
in
s
e
ct
o
r
s
li
k
e
c
o
n
s
t
r
u
cti
o
n
[
1
5
]
.
I
n
V
iet
n
a
m
’
s
co
n
s
t
r
u
c
ti
o
n
s
e
ct
o
r
,
i
n
f
o
r
m
al
e
m
p
lo
y
m
e
n
t
e
x
c
e
ed
s
t
h
e
n
at
io
n
a
l
a
v
e
r
a
g
e
,
wit
h
a
1
5
–
2
0
%
g
e
n
d
e
r
p
r
o
d
u
ct
i
v
it
y
an
d
in
c
o
m
e
g
a
p
an
d
lo
w
f
e
m
al
e
ce
r
ti
f
i
ca
tio
n
r
ates
(
1
9
–
2
5
%
v
er
s
u
s
ap
p
r
o
x
i
m
a
tel
y
2
9
%
in
2
0
2
5
;
GS
O)
[
1
6
]
.
Ur
b
a
n
i
za
ti
o
n
a
n
d
in
f
r
ast
r
u
ct
u
r
e
g
r
o
wt
h
o
f
f
e
r
o
p
p
o
r
tu
n
i
ties
,
y
et
w
o
m
e
n
r
em
a
in
co
n
ce
n
t
r
at
e
d
in
s
ea
s
o
n
a
l,
p
r
o
j
ec
t
-
b
as
ed
r
o
l
es
v
u
l
n
e
r
a
b
l
e
t
o
e
co
n
o
m
ic
c
y
cles
[
1
7
]
.
E
x
is
ti
n
g
d
o
m
es
tic
r
esea
r
c
h
o
n
f
e
m
al
e
lab
o
r
h
as
p
r
im
ar
il
y
f
o
c
u
s
e
d
o
n
te
x
ti
les/
g
a
r
m
e
n
ts
a
n
d
f
o
o
d
p
r
o
ce
s
s
i
n
g
s
ec
t
o
r
s
,
wi
th
v
e
r
y
lim
i
te
d
att
en
ti
o
n
t
o
co
n
s
t
r
u
ct
io
n
-
s
p
ec
if
i
c
c
h
all
en
g
es
s
u
ch
as
p
h
y
s
ica
l
d
e
m
a
n
d
s
,
h
i
g
h
s
e
aso
n
a
lit
y
,
a
n
d
p
r
o
j
ec
t
-
b
ase
d
em
p
l
o
y
m
en
t
[
1
8
]
.
I
n
t
h
e
d
o
m
ai
n
o
f
c
o
n
s
tr
u
ct
io
n
l
a
b
o
r
p
r
o
d
u
ct
iv
it
y
,
AI
a
n
d
m
ac
h
i
n
e
le
ar
n
i
n
g
t
ec
h
n
i
q
u
es
h
a
v
e
b
e
en
in
c
r
ea
s
i
n
g
l
y
a
p
p
lie
d
,
in
cl
u
d
in
g
ANN
f
o
r
p
r
o
d
u
cti
v
it
y
p
r
e
d
icti
o
n
[
1
9
]
.
E
n
s
e
m
b
le
a
n
d
i
n
t
er
p
r
et
ab
le
m
et
h
o
d
s
s
u
c
h
as
m
eta
-
e
n
s
e
m
b
l
e
m
o
d
e
ls
[
2
0
]
an
d
SHAP
-
b
as
ed
a
n
a
ly
s
is
[
2
1
]
h
a
v
e
g
a
in
e
d
t
r
ac
t
i
o
n
.
H
o
we
v
e
r
,
s
u
ch
ad
v
a
n
ce
d
ap
p
l
ica
ti
o
n
s
r
e
m
ai
n
lim
i
te
d
in
Vi
et
n
a
m
’
s
c
o
n
s
t
r
u
ct
io
n
s
ec
t
o
r
,
p
a
r
ti
cu
la
r
l
y
w
it
h
g
e
n
d
e
r
-
d
is
ag
g
r
e
g
a
te
d
d
at
a
a
n
d
s
m
all
tim
e
-
s
e
r
ies
d
a
t
asets
[
2
2
]
.
T
h
is
c
r
ea
t
es
a
s
i
g
n
i
f
ic
a
n
t
r
ese
ar
c
h
g
a
p
i
n
b
o
t
h
v
o
ca
ti
o
n
a
l
t
r
ai
n
i
n
g
a
n
d
co
n
s
t
r
u
ct
io
n
e
d
u
ca
ti
o
n
l
ite
r
a
tu
r
e
r
e
g
a
r
d
in
g
t
h
e
u
s
e
o
f
e
x
p
lai
n
ab
l
e
AI
f
o
r
f
em
ale
w
o
r
k
er
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t
J
E
v
al
&
R
es E
d
u
c
I
SS
N:
2252
-
8
8
2
2
Th
e
imp
a
ct
o
f v
o
c
a
tio
n
a
l tra
in
in
g
o
n
fema
le
la
b
o
r
p
r
o
d
u
ctivi
ty
a
n
ev
a
lu
a
tio
n
in
…
(
Tr
u
c
Th
a
i P
h
u
o
n
g
)
4209
3.
M
E
T
H
O
D
3
.
1
.
Da
t
a
s
o
urce
s
,
v
a
ri
a
bles
a
nd
des
cr
iptiv
e
a
na
ly
s
is
T
h
is
s
tu
d
y
u
tili
ze
s
an
an
n
u
al
t
im
e
-
s
er
ies
d
ataset
f
r
o
m
2
0
1
0
to
2
0
2
4
(
1
5
o
b
s
er
v
atio
n
s
)
s
o
u
r
ce
d
f
r
o
m
th
e
GSO
o
f
Vietn
am
,
I
L
O,
an
d
W
o
r
ld
B
an
k
.
T
h
ese
s
o
u
r
ce
s
o
f
f
er
co
n
s
is
ten
t,
n
atio
n
ally
r
ep
r
esen
tativ
e
in
d
icato
r
s
o
n
lab
o
r
,
ec
o
n
o
m
ic
d
ev
elo
p
m
en
t,
an
d
v
o
ca
tio
n
al
ed
u
ca
tio
n
.
Var
iab
les
ar
e
s
ele
cted
b
ased
o
n
th
eir
th
eo
r
etica
l
an
d
em
p
ir
ical
r
elev
an
ce
to
f
em
ale
lab
o
r
p
r
o
d
u
ctiv
ity
in
co
n
s
tr
u
ctio
n
,
en
co
m
p
ass
in
g
lab
o
r
-
s
p
ec
if
ic,
m
ac
r
o
ec
o
n
o
m
ic,
an
d
in
s
titu
tio
n
al
f
ac
to
r
s
.
R
ec
en
t
u
p
d
ates
f
r
o
m
GSO
2
0
2
5
–
2
0
2
6
r
ep
o
r
ts
in
d
icate
th
e
n
atio
n
a
l
tr
ain
ed
wo
r
k
f
o
r
ce
r
ate
r
ea
c
h
e
d
2
9
.
2
%
in
2
0
2
5
an
d
2
9
.
5
%
in
Q4
2
0
2
5
,
with
in
f
o
r
m
al
em
p
lo
y
m
en
t
at
~6
3
.
1
%.
T
h
e
d
e
p
en
d
e
n
t
v
a
r
iab
le
is
FC
L
P,
m
ea
s
u
r
ed
in
t
h
o
u
s
an
d
USD
p
er
f
em
ale
wo
r
k
er
p
er
y
ea
r
,
d
er
iv
e
d
f
r
o
m
s
ec
to
r
v
alu
e
ad
d
ed
ad
ju
s
ted
b
y
esti
m
ated
f
em
ale
la
b
o
r
s
h
ar
e.
T
h
e
i
n
d
ep
e
n
d
en
t
v
ar
iab
les
in
clu
d
e:
AM
I
C
W
:
(
USD/m
o
n
th
)
–
r
ef
lects wa
g
e
lev
els an
d
in
ce
n
tiv
es.
GR
DPp
c
: (
th
o
u
s
an
d
USD/p
er
s
o
n
/y
ea
r
)
–
p
r
o
x
ies’
ec
o
n
o
m
i
c
d
ev
elo
p
m
e
n
t
an
d
in
f
r
astru
ctu
r
e
in
ten
s
ity
.
SIE
C
:
(
%)
–
in
d
icate
s
f
o
r
m
aliza
tio
n
d
eg
r
ee
a
n
d
r
is
k
s
(
esti
m
ated
60
–
6
5
%
in
2
0
2
5
,
ab
o
v
e
n
a
tio
n
al
av
er
a
g
e)
.
STFW
1
5
+:
s
h
ar
e
o
f
f
em
ales
a
g
ed
1
5
+
with
v
o
ca
tio
n
al
tr
ain
in
g
/cer
tific
ates
(
%)
–
m
ea
s
u
r
es
s
k
ill
d
ev
elo
p
m
en
t
ac
ce
s
s
.
Data
wer
e
cr
o
s
s
-
ch
ec
k
ed
f
o
r
co
n
s
is
ten
cy
;
an
y
m
is
s
in
g
v
alu
es
wer
e
lin
ea
r
ly
in
ter
p
o
lated
,
an
d
v
a
r
iab
les
s
tan
d
ar
d
ized
(
z
-
s
co
r
es)
f
o
r
co
m
p
ar
ab
ilit
y
.
Ou
tlier
s
wer
e
r
etain
ed
to
r
ef
lect
r
ea
l
e
co
n
o
m
ic
s
h
o
c
k
s
(
2
0
1
6
d
o
wn
t
u
r
n
an
d
2
0
2
2
p
o
s
t
-
C
OVI
D
r
ec
o
v
er
y
)
.
Descr
ip
tiv
e
s
tatis
t
ics
an
d
d
is
tr
ib
u
tio
n
tes
ts
ar
e
p
r
esen
ted
in
T
ab
le
1
.
All
v
ar
iab
les
ex
h
ib
it
n
ea
r
-
n
o
r
m
al
d
is
tr
ib
u
tio
n
s
(
Sh
ap
ir
o
-
W
ilk
p
>0
.
0
5
)
,
lo
w
s
k
ewn
ess
(
|
s
k
ew
|
<0
.
5
)
,
an
d
s
u
itab
ilit
y
f
o
r
n
o
n
lin
ea
r
m
o
d
elin
g
.
T
h
e
s
m
all
s
am
p
le
(
n
=1
5
)
an
d
h
ig
h
m
u
ltico
llin
ea
r
ity
ju
s
tify
s
y
n
th
etic
d
ata
a
u
g
m
en
tatio
n
later
.
Fig
u
r
e
1
illu
s
tr
ates
lo
n
g
-
ter
m
tr
en
d
s
i
n
th
e
k
ey
v
ar
ia
b
les
in
th
e
co
n
s
tr
u
ctio
n
s
ec
to
r
f
r
o
m
2
0
1
0
to
2
0
2
4
.
T
h
ese
v
is
u
aliza
tio
n
s
r
ev
ea
l
s
tr
o
n
g
p
ar
allel
u
p
war
d
tr
en
d
s
f
o
r
p
o
s
itiv
e
v
a
r
iab
le
s
(
FC
L
P,
AM
I
C
W
,
GR
DPp
c,
an
d
STFW
1
5
+)
,
wh
ile
th
e
SIE
C
s
h
o
ws
a
s
tead
y
d
ec
lin
e
.
T
h
e
p
atter
n
s
d
em
o
n
s
tr
ate
clea
r
co
-
m
o
v
e
m
en
t
b
etwe
en
v
o
ca
t
io
n
al
tr
ain
in
g
,
la
b
o
r
f
o
r
m
ali
za
tio
n
,
an
d
p
r
o
d
u
ctiv
ity
g
ain
s
,
h
ig
h
lig
h
tin
g
t
h
e
v
u
ln
er
ab
ilit
y
o
f
f
em
ale
wo
r
k
er
s
to
ec
o
n
o
m
ic
c
y
cles.
T
ab
le
1
.
Descr
ip
tiv
e
s
tatis
tics
an
d
d
is
tr
ib
u
tio
n
test
s
o
f
v
ar
iab
les
V
a
r
i
a
b
l
e
n
M
e
a
n
SD
M
e
d
i
a
n
R
a
n
g
e
(
mi
n
–
ma
x
)
I
Q
R
(Q1
–
Q
3
)
S
k
e
w
n
e
ss
p
-
v
a
l
u
e
(
S
h
a
p
i
r
o
-
W
i
l
k
)
K
S
s
t
a
t
i
st
i
c
K
S
p
-
v
a
l
u
e
W
a
ssers
t
e
i
n
d
i
s
t
a
n
c
e
F
C
LP
15
4
.
0
9
0
.
9
4
3
.
9
4
2
.
8
6
–
5
.
6
2
3
.
2
8
–
4
.
8
8
0
.
4
7
0
.
1
9
0
.
0
8
1
.
0
0
0
.
0
5
A
M
I
C
W
15
2
5
7
.
0
3
5
0
.
7
8
2
5
3
.
1
5
1
8
1
.
4
5
–
3
3
9
.
2
5
2
1
5
.
7
5
–
2
8
4
.
8
5
0
.
1
4
0
.
5
7
0
.
1
2
0
.
9
9
0
.
9
8
G
R
D
P
p
c
15
2
.
6
9
0
.
9
3
2
.
5
8
1
.
4
1
–
3
.
9
7
1
.
8
1
–
3
.
5
1
0
.
0
7
0
.
1
9
0
.
0
8
1
.
0
0
0
.
0
6
S
I
EC
15
5
8
.
4
6
7
.
8
0
5
8
.
3
7
4
6
.
9
0
–
7
0
.
2
3
5
3
.
7
5
–
6
4
.
4
5
0
.
0
5
0
.
3
9
0
.
0
8
1
.
0
0
0
.
5
5
S
TFW
1
5
+
15
1
8
.
9
7
3
.
8
4
1
9
.
0
0
1
2
.
8
0
–
2
4
.
6
7
1
5
.
7
0
–
2
3
.
3
3
-
0
.
0
3
0
.
5
4
0
.
1
1
1
.
0
0
0
.
2
1
N
o
t
e
:
M
u
l
t
i
-
a
x
i
s
c
h
a
r
t
c
o
mb
i
n
i
n
g
b
a
r
s
a
n
d
l
i
n
e
s
s
h
o
w
i
n
g
f
l
u
c
t
u
a
t
i
o
n
s
o
f
t
h
e
f
i
v
e
m
a
i
n
v
a
r
i
a
b
l
e
s
d
u
r
i
n
g
2
0
1
0
-
2
0
2
4
.
F
C
LP
(
l
i
g
h
t
b
l
u
e
h
a
t
c
h
e
d
b
a
r
s)
,
A
M
I
C
W
(
r
e
d
d
o
t
t
e
d
b
a
r
s)
,
G
R
D
P
p
c
(
so
l
i
d
g
r
e
e
n
l
i
n
e
)
,
S
I
EC
(
p
u
r
p
l
e
c
r
o
ss
-
h
a
t
c
h
e
d
b
a
r
s)
,
S
TFW
1
5
+
(
o
r
a
n
g
e
d
a
s
h
e
d
l
i
n
e
)
.
Fig
u
r
e
1
.
T
r
en
d
s
o
f
k
ey
v
ar
iab
les o
v
er
tim
e
(
2
0
1
0
–
2
0
2
4
)
Po
s
itiv
e
v
ar
iab
les
(
FC
L
P,
AM
I
C
W
,
G
R
DPp
c,
an
d
STFW
1
5
+)
ex
h
ib
it
s
tr
o
n
g
p
ar
allel
u
p
war
d
tr
en
d
s
,
wh
ile
SIE
C
d
ec
lin
es
s
tead
il
y
.
Fro
m
2
0
1
0
to
2
0
2
4
,
FC
L
P
in
cr
ea
s
ed
b
y
9
1
.
0
%,
GR
DPp
c
b
y
1
7
8
.
0
%
(
s
tr
o
n
g
est),
an
d
SIE
C
d
ec
r
ea
s
ed
b
y
3
3
.
2
%.
Key
p
e
r
io
d
s
in
c
lu
d
e
s
tead
y
g
r
o
wth
(
2
0
1
0
–
2
0
1
5
)
,
a
s
h
ar
p
d
r
o
p
in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
2
2
I
n
t
J
E
v
al
&
R
es E
d
u
c
,
Vo
l
.
15
,
No
.
5
,
Octo
b
er
20
26
:
4
2
0
7
-
4219
4210
2
0
1
6
(
FC
L
P
–
7
.
8
7
%),
r
a
p
i
d
r
e
c
o
v
e
r
y
(
2
0
1
7
–
2
0
2
1
)
,
a
p
e
a
k
i
n
2
0
2
2
(
FC
L
P
+1
5
.
1
6
%),
a
n
d
s
t
a
b
i
l
i
za
t
i
o
n
/
d
e
cl
i
n
e
in
2
0
2
3
–
2
0
2
4
.
T
h
ese
p
atter
n
s
d
em
o
n
s
tr
ate
c
o
-
m
o
v
em
en
t
b
etwe
en
v
o
ca
tio
n
al
tr
ai
n
in
g
,
lab
o
r
f
o
r
m
aliza
tio
n
,
an
d
p
r
o
d
u
ctiv
ity
g
ain
s
,
with
f
e
m
ale
wo
r
k
e
r
s
(
o
f
ten
s
ea
s
o
n
al/p
r
o
ject
-
b
ased
)
r
em
ai
n
in
g
v
u
l
n
er
ab
le
to
ec
o
n
o
m
ic
cy
cles.
Fig
u
r
e
2
illu
s
tr
ate
s
an
n
u
al
g
r
o
wth
r
ates o
f
th
e
k
ey
v
ar
iab
les an
d
p
er
io
d
s
o
f
s
tr
o
n
g
f
l
u
ctu
atio
n
.
Av
er
ag
e
an
n
u
al
g
r
o
wth
r
at
es:
GR
DPp
c_
G
R
h
ig
h
est
(
+7
.
6
8
%/y
ea
r
)
,
f
o
llo
wed
b
y
FC
L
P_
G
R
(
+4
.
8
9
%),
STFW
1
5
+_
GR
(
+4
.
8
4
%),
AM
I
C
W
_
GR
(
+4
.
7
5
%),
an
d
SIE
C
_
GR
d
ec
lin
in
g
(
–
2
.
8
0
%/y
ea
r
)
.
Ma
jo
r
o
u
tlier
s
in
clu
d
e
2
0
1
6
(
FC
L
P_
GR
–
7
.
8
7
%,
s
ec
to
r
d
o
wn
tu
r
n
)
an
d
2
0
2
2
(
FC
L
P_
GR
+1
5
.
1
6
%,
p
o
s
t
-
C
OVI
D
r
ec
o
v
er
y
)
.
Gr
o
wth
s
lo
wed
in
2
0
2
3
–
2
0
2
4
,
b
u
t
v
o
ca
tio
n
al
t
r
ain
in
g
s
u
s
tain
ed
p
o
s
itiv
e
m
o
m
en
tu
m
.
T
h
is
ch
ar
t
co
m
p
lem
en
ts
Fig
u
r
e
1
b
y
h
ig
h
lig
h
tin
g
r
elativ
e
f
lu
ctu
atio
n
s
an
d
th
e
ca
tal
y
tic
r
o
le
o
f
tr
ai
n
in
g
/f
o
r
m
aliza
tio
n
d
u
r
in
g
r
ec
o
v
er
y
.
Fig
u
r
e
3
illu
s
tr
ates
th
e
Pear
s
o
n
co
r
r
elatio
n
m
atr
ix
an
d
s
ev
e
r
e
m
u
ltico
llin
e
ar
ity
.
N
o
t
e
:
Li
n
e
c
h
a
r
t
s
h
o
w
i
n
g
y
e
a
r
-
o
v
e
r
-
y
e
a
r
g
r
o
w
t
h
r
a
t
e
s
(
%)
f
r
o
m
2
0
1
1
t
o
2
0
2
4
.
F
C
LP
_
G
R
(
b
l
u
e
d
o
t
t
e
d
l
i
n
e
)
,
A
M
I
C
W
_
G
R
(
r
e
d
d
a
s
h
e
d
-
d
o
t
t
e
d
l
i
n
e
)
,
G
R
D
P
p
c
_
G
R
(
so
l
i
d
g
r
e
e
n
l
i
n
e
)
,
S
I
EC
_
G
R
(
p
u
r
p
l
e
l
i
n
e
)
,
S
TF
W
1
5
+
_
G
R
(
o
r
a
n
g
e
d
o
t
t
e
d
l
i
n
e
)
.
G
r
a
y
d
a
s
h
e
d
h
o
r
i
z
o
n
t
a
l
l
i
n
e
a
t
0
%
ser
v
e
s
a
s ref
e
r
e
n
c
e
.
Fig
u
r
e
2
.
An
n
u
al
g
r
o
wth
r
ates
o
f
k
ey
v
ar
iab
les (
%)
N
o
t
e
:
H
e
a
t
ma
p
o
f
P
e
a
r
s
o
n
c
o
r
r
e
l
a
t
i
o
n
s
c
o
m
b
i
n
e
d
w
i
t
h
sca
t
t
e
r
p
l
o
t
s a
n
d
h
i
st
o
g
r
a
ms o
n
t
h
e
d
i
a
g
o
n
a
l
.
D
a
r
k
r
e
d
i
n
d
i
c
a
t
e
s
st
r
o
n
g
p
o
si
t
i
v
e
c
o
r
r
e
l
a
t
i
o
n
s,
d
a
r
k
b
l
u
e
st
r
o
n
g
n
e
g
a
t
i
v
e
;
c
i
r
c
l
e
si
z
e
p
r
o
p
o
r
t
i
o
n
a
l
t
o
|
r
|
.
Fig
u
r
e
3
.
Pear
s
o
n
co
r
r
elatio
n
h
ea
tm
ap
o
f
v
ar
iab
les
An
aly
s
is
s
h
o
ws
v
er
y
s
tr
o
n
g
co
r
r
elatio
n
s
:
FC
L
P
with
ST
FW
1
5
+
(
+0
.
9
6
)
,
AM
I
C
W
(
+0
.
9
5
)
,
GR
DPp
c
(
+0
.
9
6
)
,
an
d
SIE
C
(
–
0
.
9
6
)
.
Sev
er
e
m
u
ltico
llin
ea
r
ity
am
o
n
g
in
d
e
p
en
d
e
n
t
v
ar
ia
b
les
(
GR
DPp
c
–
STF
W
1
5
+
,
r
=+
0
.
9
8
)
r
en
d
e
r
s
tr
ad
itio
n
al
lin
ea
r
r
eg
r
ess
io
n
u
n
s
tab
le.
Scatter
p
lo
ts
co
n
f
ir
m
lin
ea
r
tr
e
n
d
s
b
etwe
en
FC
L
P
an
d
STFW
1
5
+,
an
d
in
v
er
s
e
p
atte
r
n
s
with
SIE
C
.
T
h
is
m
u
ltico
l
lin
ea
r
ity
is
th
e
m
ain
r
ea
s
o
n
f
o
r
s
elec
tin
g
ANN
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t
J
E
v
al
&
R
es E
d
u
c
I
SS
N:
2252
-
8
8
2
2
Th
e
imp
a
ct
o
f v
o
c
a
tio
n
a
l tra
in
in
g
o
n
fema
le
la
b
o
r
p
r
o
d
u
ctivi
ty
a
n
ev
a
lu
a
tio
n
in
…
(
Tr
u
c
Th
a
i P
h
u
o
n
g
)
4211
co
m
b
in
ed
with
SHAP,
wh
ich
h
an
d
les
n
o
n
lin
ea
r
r
elatio
n
s
h
ip
s
an
d
p
r
o
v
id
es
r
o
b
u
s
t
f
ea
tu
r
e
im
p
o
r
tan
ce
an
aly
s
is
[
2
3
]
.
Fig
u
r
e
3
p
r
o
v
id
e
s
a
v
is
u
al
f
o
u
n
d
atio
n
f
o
r
tr
e
n
d
s
,
f
lu
ctu
atio
n
s
,
an
d
in
te
r
-
v
ar
i
ab
le
r
elatio
n
s
h
ip
s
,
wh
ile
h
i
g
h
li
g
h
ti
n
g
li
m
i
tati
o
n
s
o
f
s
m
all
s
a
m
p
le
s
ize
a
n
d
m
u
lti
co
lli
n
e
ar
it
y
.
T
h
is
m
o
ti
v
ates
s
y
n
t
h
e
tic
d
ata
au
g
m
e
n
ta
ti
o
n
a
n
d
a
d
v
a
n
ce
d
m
o
d
eli
n
g
i
n
s
u
b
s
eq
u
en
t
s
ec
ti
o
n
s
.
Se
v
e
r
a
l
s
t
u
d
i
es
h
a
v
e
e
m
p
lo
y
ed
h
y
b
r
i
d
AI
ap
p
r
o
ac
h
es
t
o
p
r
e
d
i
ct
a
n
d
o
p
t
im
i
ze
c
o
n
s
t
r
u
cti
o
n
la
b
o
r
p
r
o
d
u
c
ti
v
it
y
u
n
d
e
r
s
im
ila
r
co
n
s
t
r
ai
n
ts
.
D
ata
au
g
m
e
n
ta
ti
o
n
te
c
h
n
iq
u
es,
s
u
ch
as
g
e
n
e
r
ati
v
e
m
o
d
els
c
o
m
b
in
ed
wit
h
e
x
p
lai
n
a
b
l
e
AI
,
h
a
v
e
a
ls
o
b
ee
n
ap
p
l
ie
d
i
n
o
t
h
e
r
d
o
m
ai
n
s
t
o
im
p
r
o
v
e
m
o
d
el
p
e
r
f
o
r
m
a
n
c
e
o
n
li
m
it
ed
d
a
tasets
[
2
4
]
.
R
e
ce
n
t
GSO
2
0
2
5
-
2
0
2
6
d
ata
c
o
n
f
ir
m
p
o
s
iti
v
e
n
ati
o
n
al
tr
en
d
s
(
r
is
i
n
g
t
r
ai
n
i
n
g
r
at
es,
d
e
cli
n
i
n
g
i
n
f
o
r
m
a
lit
y
)
,
y
e
t
p
e
r
s
is
te
n
t
g
ap
s
i
n
c
o
n
s
t
r
u
ct
io
n
—
p
a
r
ti
c
u
la
r
l
y
f
o
r
f
e
m
a
le
w
o
r
k
e
r
s
—
u
n
d
er
s
c
o
r
e
t
h
e
n
e
e
d
f
o
r
d
e
ep
er
,
s
ec
to
r
-
s
p
ec
i
f
i
c
a
n
al
y
s
is
.
3
.
2
.
Sy
nthet
ic
da
t
a
a
ug
m
ent
a
t
io
n a
nd
m
o
del sp
ec
if
ica
t
io
n
Giv
en
th
e
lim
ited
o
r
ig
in
al
o
b
s
er
v
atio
n
s
(
n
=1
5
tim
e
p
o
in
ts
)
,
ap
p
ly
in
g
co
m
p
lex
n
o
n
lin
ea
r
m
o
d
els
lik
e
ANN
r
is
k
s
s
ev
er
e
o
v
e
r
f
it
ti
n
g
an
d
u
n
s
tab
le
p
r
ed
ictio
n
s
.
T
h
e
Fo
r
estDiff
u
s
io
n
s
y
n
th
etic
d
ata
p
ip
elin
e
-
tim
e
-
s
er
ies,
ad
ap
ted
f
r
o
m
d
if
f
u
s
i
o
n
-
b
ased
g
en
er
ati
v
e
m
o
d
els
f
o
r
tab
u
lar
tim
e
-
s
er
ies
d
ata
,
was
em
p
lo
y
e
d
to
p
r
eser
v
e
tem
p
o
r
al
s
tr
u
ctu
r
e
a
n
d
k
ey
ec
o
n
o
m
ic
s
h
o
c
k
s
wh
ile
ex
p
an
d
in
g
th
e
d
ataset.
T
o
p
r
eser
v
e
tem
p
o
r
al
s
tr
u
ctu
r
e
an
d
m
ac
r
o
ec
o
n
o
m
ic
f
ea
tu
r
es
wh
ile
e
x
p
an
d
in
g
th
e
d
ataset,
th
is
s
tu
d
y
em
p
lo
y
s
th
e
Fo
r
estDiff
u
s
io
n
s
y
n
th
etic
d
ata
p
ip
elin
e
-
tim
e
-
s
er
ies,
a
d
if
f
u
s
io
n
-
b
ased
tec
h
n
iq
u
e
f
o
r
tim
e
-
s
er
ies
au
g
m
en
t
atio
n
.
T
h
e
m
et
h
o
d
ad
d
s
p
r
o
g
r
ess
iv
e
n
o
is
e
an
d
r
ev
er
s
es
d
en
o
is
in
g
to
g
en
er
ate
s
y
n
th
etic
s
am
p
les
m
at
ch
in
g
th
e
o
r
ig
i
n
al
d
is
tr
ib
u
tio
n
,
m
ain
tain
i
n
g
tem
p
o
r
al
co
n
tin
u
ity
a
n
d
r
e
p
r
o
d
u
cin
g
k
ey
b
r
ea
k
p
o
in
ts
(
2
0
1
6
d
ec
lin
e
an
d
2
0
2
2
s
u
r
g
e)
.
Fro
m
1
5
r
ea
l
o
b
s
er
v
atio
n
s
,
7
5
h
ig
h
-
q
u
ality
s
y
n
th
eti
c
s
am
p
les
wer
e
cr
ea
ted
,
y
ield
in
g
a
m
ix
e
d
d
ataset
o
f
9
0
s
am
p
les
f
o
r
t
r
ain
in
g
a
n
d
ev
alu
atio
n
.
Sy
n
th
etic
d
ata
q
u
ality
was
v
alid
ated
r
ig
o
r
o
u
s
ly
:
W
ass
er
s
tein
d
is
tan
ce
s
:
0
.
0
5
–
0
.
9
8
(
m
in
im
a
l
s
h
if
t)
.
AC
F/P
A
C
F
m
atch
ed
o
r
ig
in
al
at
r
ele
v
an
t
lag
s
.
A
DF
test
co
n
f
ir
m
ed
p
r
eser
v
ed
s
tatio
n
ar
ity
.
Ma
jo
r
s
h
o
ck
s
(
2
0
1
6
d
o
wn
tu
r
n
,
2
0
2
2
p
o
s
t
-
C
OVI
D
s
u
r
g
e)
f
aith
f
u
lly
r
ep
licated
.
Ab
latio
n
s
tu
d
y
:
7
5
s
y
n
th
etic
s
am
p
les
y
ield
ed
b
est
g
en
er
ali
za
tio
n
(
h
i
g
h
est
test
R
s
q
u
ar
e,
s
m
allest
tr
ain
-
test
g
ap
)
.
T
h
e
p
r
im
ar
y
m
o
d
el
is
a
f
ee
d
f
o
r
war
d
ANN
with
4
-
3
-
1
ar
ch
itectu
r
e
(
4
i
n
p
u
ts
:
AM
I
C
W
,
GR
DPp
c,
SIE
C
,
STFW
1
5
+;
o
n
e
h
id
d
en
lay
er
with
3
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
n
eu
r
o
n
s
;
lin
ea
r
o
u
tp
u
t
f
o
r
r
eg
r
ess
io
n
)
.
T
h
e
p
r
ed
ictio
n
f
o
r
m
u
la
is
:
th
e
m
ain
m
o
d
el
is
a
f
ee
d
f
o
r
war
d
ANN
with
ar
ch
itectu
r
e
4
-
3
-
1
(
4
in
p
u
ts
,
3
h
id
d
en
R
eL
U
n
eu
r
o
n
s
,
an
d
1
lin
ea
r
o
u
tp
u
t)
.
Pre
d
ictio
n
f
o
r
m
u
la
(
e
x
p
licitly
wr
itten
)
:
̂
=
2
+
∑
2
,
3
=
1
(
1
,
+
∑
1
,
4
=
1
)
(
1
)
W
h
er
e
,
(
j
=
1
…4
)
:
in
p
u
t
v
a
r
iab
les
(
AM
I
C
W
,
G
R
DPp
c,
SI
E
C
,
STFW
1
5
⁺)
.
(
⋅
)
:
R
eL
U
ac
tiv
atio
n
.
1
,
,
,
1
,
:
h
id
d
en
lay
er
weig
h
ts
an
d
b
iases
.
2
,
,
2
:
o
u
tp
u
t
lay
er
weig
h
t
s
an
d
b
ias
.
T
h
e
lo
s
s
f
u
n
ctio
n
is
m
ea
n
s
q
u
ar
ed
er
r
o
r
(
MSE
)
:
M
SE
=
1
∑
(
−
̂
)
2
=
1
(
2
)
Per
f
o
r
m
an
ce
m
etr
ics in
clu
d
e:
2
=
1
−
∑
(
−
̂
)
2
=
1
∑
(
−
̄
)
2
=
1
;
=
√
1
∑
(
−
̂
)
2
=
1
(
3
)
T
r
ain
in
g
:
Ad
am
o
p
tim
izer
(
in
itial
lr
=0
.
0
0
1
,
ad
ap
tiv
e
)
,
ea
r
ly
s
to
p
p
in
g
(
p
atien
ce
=1
0
)
,
b
atc
h
s
ize=
1
6
,
d
r
o
p
o
u
t
(
0
.
2
–
0
.
3
)
,
L
2
r
eg
u
lar
izatio
n
.
S
HAP
v
alu
e
f
o
r
m
u
la:
=
∑
∣
∣
!
(
∣
∣
−
∣
∣
−
1
)
!
∣
∣
!
[
(
∪
{
}
)
−
(
)
]
⊆
∖
{
}
(
4)
W
h
er
e
,
:
s
et
o
f
all
f
ea
tu
r
es.
:
s
u
b
s
et
ex
clu
d
in
g
f
ea
tu
r
e
i.
(
⋅
)
:
m
o
d
el
p
r
e
d
ictio
n
f
u
n
ctio
n
.
SHAP
en
ab
les
g
lo
b
al
f
ea
tu
r
e
im
p
o
r
ta
n
ce
,
lo
ca
l
ex
p
lan
atio
n
s
,
an
d
v
is
u
aliza
tio
n
o
f
n
o
n
lin
ea
r
i
n
ter
ac
tio
n
s
(
v
ia
3
D
s
u
r
f
ac
e
p
lo
ts
)
.
Mo
d
el
p
er
f
o
r
m
an
ce
a
n
d
tr
ain
e
d
p
ar
am
eter
s
ar
e
s
h
o
wn
in
T
a
b
le
s
2
an
d
3
.
T
h
e
ANN
m
o
d
el
ac
h
iev
ed
s
tr
o
n
g
p
e
r
f
o
r
m
an
ce
o
n
th
e
m
i
x
ed
d
ataset
o
f
9
0
o
b
s
er
v
atio
n
s
.
On
th
e
tr
ain
in
g
s
et,
th
e
m
o
d
e
l
r
ec
o
r
d
e
d
a
n
R
²
o
f
0
.
8
7
3
,
r
o
o
t
m
ea
n
s
q
u
a
r
ed
er
r
o
r
(
R
MSE
)
o
f
0
.
2
9
0
,
an
d
m
e
an
ab
s
o
lu
te
er
r
o
r
(
MA
E
)
o
f
0
.
2
0
6
.
T
h
e
v
alid
atio
n
an
d
test
in
g
r
esu
lts
r
em
ain
e
d
s
tab
le
with
R
²
v
alu
es
o
f
0
.
7
2
3
an
d
0
.
6
1
1
r
esp
ec
tiv
el
y
,
in
d
icatin
g
g
o
o
d
g
en
er
aliza
tio
n
ab
ilit
y
.
T
h
e
lo
w
o
v
er
f
itti
n
g
s
co
r
e
o
f
0
.
0
5
7
co
n
f
ir
m
s
th
at
th
e
co
m
b
i
n
atio
n
o
f
s
y
n
th
etic
d
ata
au
g
m
en
tatio
n
,
L
2
r
eg
u
lar
izati
o
n
,
d
r
o
p
o
u
t,
a
n
d
ea
r
l
y
s
to
p
p
i
n
g
was
h
ig
h
l
y
ef
f
ec
tiv
e.
T
est
er
r
o
r
s
r
an
g
in
g
f
r
o
m
0
.
5
t
o
0
.
6
th
o
u
s
an
d
USD
p
e
r
wo
r
k
e
r
p
e
r
y
ea
r
a
r
e
r
ea
s
o
n
a
b
le
wh
en
co
m
p
ar
ed
to
th
e
m
ea
n
FC
L
P
v
alu
e
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
2
2
I
n
t
J
E
v
al
&
R
es E
d
u
c
,
Vo
l
.
15
,
No
.
5
,
Octo
b
er
20
26
:
4
2
0
7
-
4219
4212
ap
p
r
o
x
im
ately
4
.
0
9
.
I
n
co
n
tr
a
s
t,
m
o
d
els
tr
ain
ed
o
n
ly
o
n
th
e
o
r
ig
in
al
1
5
o
b
s
er
v
atio
n
s
ex
h
i
b
ited
s
ig
n
if
ican
tly
lo
wer
R
²
v
alu
es
an
d
s
ev
er
e
o
v
er
f
itti
n
g
.
T
h
ese
p
er
f
o
r
m
a
n
ce
m
etr
ics
d
em
o
n
s
tr
ate
th
e
r
o
b
u
s
tn
ess
o
f
th
e
p
r
o
p
o
s
ed
ANN
m
o
d
el
in
h
an
d
lin
g
s
m
all
-
s
am
p
le,
h
ig
h
-
m
u
lti
co
llin
ea
r
ity
tim
e
-
s
er
ies
d
ata.
T
h
e
r
esu
lts
v
alid
ate
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
Fo
r
es
tDif
f
u
s
io
n
s
y
n
th
etic
d
ata
p
ip
elin
e
-
tim
e
-
s
er
ies
au
g
m
en
tatio
n
tech
n
iq
u
e
ad
o
p
te
d
in
th
is
s
tu
d
y
.
Ov
er
all,
th
e
m
o
d
el
p
r
o
v
i
d
es
r
eliab
le
p
r
ed
ictio
n
s
an
d
s
er
v
es
as
a
s
o
lid
f
o
u
n
d
at
io
n
f
o
r
s
u
b
s
eq
u
en
t
SHAP in
ter
p
r
etab
ilit
y
an
aly
s
is
.
Sm
all
m
o
d
el
(
1
6
p
a
r
am
eter
s
)
wo
r
k
s
well
to
ca
p
tu
r
e
n
o
n
li
n
ea
r
in
te
r
ac
tio
n
s
.
STFW
1
5
+
d
o
m
in
ates
th
r
o
u
g
h
in
ter
ac
tio
n
s
(
later
e
x
p
lain
s
SHAP
r
esu
lt
s
)
.
T
h
is
s
etu
p
o
v
e
r
co
m
es
s
m
all
-
d
ata
a
n
d
m
u
ltico
llin
ea
r
ity
lim
itatio
n
s
allo
win
g
r
eliab
le
S
HAP
an
aly
s
is
an
d
v
is
u
aliza
tio
n
o
f
n
o
n
lin
ea
r
in
ter
ac
tio
n
s
a
n
d
p
av
es
th
e
way
f
o
r
th
e
r
esu
lts
.
C
o
n
s
tr
u
ctio
n
s
p
ec
if
ic
ap
p
licatio
n
s
o
f
h
y
b
r
id
AI
o
p
tim
izatio
n
m
eth
o
d
s
[
2
5
]
an
d
h
y
b
r
id
m
o
d
els
co
m
b
in
in
g
v
ar
io
u
s
tech
n
iq
u
es
f
o
r
lab
o
r
p
r
o
d
u
ctiv
ity
p
r
ed
ict
io
n
an
d
o
p
tim
izatio
n
[
2
6
]
h
av
e
b
ee
n
s
u
cc
ess
f
u
lly
ap
p
lied
to
s
im
ilar
p
r
o
b
lem
s
.
On
e
s
p
ec
if
ic
ac
tiv
ity
,
f
o
r
w
h
ich
ca
s
e
s
tu
d
ies
s
h
o
w
th
e
p
r
a
ctica
l
u
s
ef
u
ln
ess
o
f
f
o
r
ec
asti
n
g
f
o
r
lab
o
r
p
r
o
d
u
ctiv
ity
,
is
f
o
r
m
wo
r
k
in
h
ig
h
-
r
is
e
b
u
ild
in
g
s
[
2
7
]
.
Gen
e
r
ativ
e
m
o
d
els
an
d
ex
p
lain
ab
le
AI
b
ased
d
ata
a
u
g
m
en
tatio
n
s
tr
ateg
ies
h
av
e
also
b
ee
n
s
h
o
wn
ef
f
ec
tiv
e
at
im
p
r
o
v
in
g
m
o
d
el
r
o
b
u
s
tn
ess
o
n
lim
ited
o
r
d
o
m
a
in
s
p
ec
if
ic
d
atasets
.
T
ab
le
2
.
Per
f
o
r
m
an
ce
m
etr
ics o
f
th
e
ANN
m
o
d
el
M
o
d
e
l
Tr
a
i
n
i
n
g
V
a
l
i
d
a
t
i
o
n
Te
st
i
n
g
O
v
e
r
f
i
t
t
i
n
g
s
c
o
r
e
R
sq
u
a
r
e
R
M
S
E
M
A
E
R
sq
u
a
r
e
R
M
S
E
M
A
E
R
sq
u
a
r
e
R
M
S
E
M
A
E
ANN
0
.
8
7
3
0
.
2
9
0
0
.
2
0
6
0
.
7
2
3
0
.
3
4
7
0
.
2
8
2
0
.
6
1
1
0
.
6
1
7
0
.
4
8
5
0
.
0
5
7
T
ab
le
3
.
W
eig
h
ts
an
d
b
iases
o
f
th
e
tr
ain
ed
ANN
m
o
d
el
H
i
d
d
e
n
n
e
u
r
o
n
s (
i
)
H
i
d
d
e
n
b
i
a
s
(
b
1
i
)
H
i
d
d
e
n
w
e
i
g
h
t
s
(
w
1
,
j)
O
u
t
p
u
t
w
e
i
g
h
t
s
(
w
2
,
i)
A
M
I
C
W
(
j
=
1
)
G
R
D
P
p
c
(
j
=
2
)
S
I
EC
(
j
=
3
)
S
TFW
1
5
⁺
(
j
=
4
)
1
-
1
2
.
9
1
0
.
0
2
0
.
2
2
0
.
2
0
-
0
.
2
0
-
1
.
3
4
2
4
.
6
5
-
0
.
0
1
-
1
.
3
2
0
.
0
8
-
0
.
1
9
1
.
0
4
3
-
4
.
4
0
0
.
0
1
0
.
5
4
0
.
0
4
-
0
.
0
3
2
.
6
4
O
u
t
p
u
t
n
ode
(
k
=
1
)
O
u
t
p
u
t
b
i
a
s
(
b
2
)
O
u
t
p
u
t
w
e
i
g
h
t
s (w
2
,
i)
i
=
1
i
=
2
i
=
3
1
4
.
2
0
-
1
.
3
4
1
.
0
4
2
.
6
4
3
.
3
.
T
im
e
s
er
ies dia
g
no
s
t
ics,
ba
s
eline
co
m
pa
riso
n,
a
nd
o
utlier
a
na
ly
s
is
T
o
ass
ess
th
e
ANN
m
o
d
el
’
s
v
alid
ity
in
a
tim
e
-
s
er
ies
co
n
te
x
t,
d
iag
n
o
s
tics
,
b
aselin
e
co
m
p
ar
is
o
n
s
,
an
d
o
u
tlier
an
aly
s
is
wer
e
p
er
f
o
r
m
ed
.
Diag
n
o
s
tics
o
n
o
r
i
g
in
al
(
n
=1
5
)
an
d
m
ix
ed
(
n
=
9
0
)
d
at
asets
:
Stat
io
n
ar
ity
:
Mo
s
t
v
ar
iab
les
(
FC
L
P,
AM
I
C
W
,
G
R
DPp
c,
an
d
STFW
1
5
+)
s
tatio
n
ar
y
o
r
s
tatio
n
ar
y
a
f
ter
f
ir
s
t
d
if
f
er
e
n
cin
g
(
ADF
p
<0
.
0
5
;
KPSS
p
>0
.
0
5
)
[
2
8
]
.
R
esid
u
als:
L
ju
n
g
-
B
o
x
te
s
t
p
>0
.
0
5
(
u
p
t
o
lag
1
0
)
,
c
o
n
f
ir
m
in
g
wh
ite
-
n
o
is
e
an
d
n
o
au
to
c
o
r
r
elatio
n
.
T
h
ese
alig
n
with
r
ec
en
t
GSO
2
0
2
5
–
2
0
2
6
d
ata
(
tr
ai
n
ed
wo
r
k
f
o
r
ce
r
ate
2
9
.
2
%
in
2
0
2
5
,
2
9
.
5
%
in
Q4
2
0
2
5
;
in
f
o
r
m
al
em
p
lo
y
m
e
n
t
~6
3
.
1
%;
co
n
s
tr
u
ctio
n
g
r
o
wth
9
.
6
2
%).
B
aselin
e
co
m
p
ar
is
o
n
o
n
h
o
ld
-
o
u
t
s
et
(
2
0
2
0
–
2
0
2
4
)
:
B
a
s
elin
es:
OL
S/
R
id
g
e/L
ass
o
(
u
n
s
tab
le
d
u
e
to
m
u
ltico
llin
ea
r
ity
,
VI
F>1
0
;
h
o
ld
-
o
u
t
R
s
q
u
ar
e<
0
.
4
5
)
;
AR
I
MA
/AR
I
MA
X/VA
R
(
R
s
q
u
ar
e≈
0
.
5
0
–
0
.
5
5
,
m
is
s
n
o
n
lin
ea
r
ities
)
.
A
NN+
au
g
m
en
tatio
n
:
h
o
ld
-
o
u
t
R
s
q
u
ar
e≈
0
.
6
1
1
,
R
MSE
≈
0
.
6
1
7
,
MA
E
≈
0
.
4
8
5
—
o
u
tp
er
f
o
r
m
s
b
aselin
es
b
y
1
0
–
2
0
p
p
i
n
R
s
q
u
ar
e,
th
an
k
s
to
n
o
n
lin
ea
r
ca
p
ab
ilit
y
an
d
r
ed
u
ce
d
o
v
er
f
itti
n
g
.
Ou
tlier
an
aly
s
is
(
k
ey
an
o
m
a
lo
u
s
y
ea
r
s
)
:
2
0
1
6
:
FC
L
P
–
7
.
8
7
%
(
s
ec
to
r
d
o
wn
tu
r
n
f
r
o
m
s
talled
in
v
estme
n
t
/b
an
k
in
g
is
s
u
es).
2
0
2
2
:
FC
L
P
+1
5
.
1
6
%
(
p
o
s
t
-
C
OVI
D
r
ec
o
v
er
y
,
h
i
g
h
in
v
e
s
tm
en
t/F
DI
)
.
2
0
2
4
:
Stalled
g
r
o
wth
in
s
o
m
e
v
ar
iab
les
(
i
n
co
m
e
-
p
r
o
d
u
ctiv
ity
m
is
m
atch
)
.
I
n
2
0
2
5
,
co
n
s
tr
u
ctio
n
r
ec
o
v
er
e
d
s
tr
o
n
g
ly
(
9
.
6
2
%
g
r
o
wth
;
in
d
u
s
tr
ial
-
co
n
s
tr
u
ctio
n
9
.
7
3
%;
n
et
em
p
lo
y
m
e
n
t g
ain
s
~1
8
3
–
2
6
4
t
h
o
u
s
an
d
v
s
.
2
0
2
4
)
,
u
n
d
er
s
co
r
i
n
g
tr
ain
in
g
’
s
b
u
f
f
er
in
g
r
o
le.
T
ab
le
4
,
wh
ich
ap
p
lies
th
e
n
ea
r
est
n
eig
h
b
o
r
s
m
eth
o
d
o
n
y
ea
r
-
o
v
er
-
y
ea
r
g
r
o
wth
r
ates,
id
en
tifie
s
2
0
2
4
as
th
e
m
o
s
t
p
r
o
m
in
en
t
o
u
tlier
with
a
d
is
tan
ce
o
f
2
.
8
2
4
8
2
4
.
T
h
is
o
u
tlier
is
p
r
im
ar
ily
d
r
iv
en
b
y
s
ig
n
i
f
ican
t
d
ev
iatio
n
s
in
AM
I
C
W
,
SIE
C
,
an
d
FC
L
P.
T
h
e
an
aly
s
is
h
ig
h
lig
h
ts
th
e
ex
ce
p
tio
n
al
ec
o
n
o
m
ic
co
n
d
itio
n
s
in
2
0
2
4
co
m
p
a
r
ed
to
p
r
ev
i
o
u
s
y
ea
r
s
.
T
h
e
Au
g
m
e
n
ted
ANN
m
o
d
el
h
a
n
d
les
th
ese
o
u
tlier
s
r
o
b
u
s
tly
.
Sy
n
th
etic
d
ata
ef
f
ec
tiv
ely
r
e
p
r
o
d
u
ce
s
th
e
o
b
s
er
v
e
d
f
lu
ctu
atio
n
s
,
wh
ile
th
e
n
o
n
li
n
ea
r
s
tr
u
ct
u
r
e
co
m
b
in
ed
with
r
eg
u
lar
izatio
n
tech
n
iq
u
es
m
in
im
izes
th
e
n
eg
ativ
e
im
p
ac
t
o
f
ec
o
n
o
m
ic
s
h
o
ck
s
.
T
h
is
ap
p
r
o
ac
h
en
s
u
r
es
m
o
d
el
s
tab
ilit
y
ev
en
d
u
r
i
n
g
ab
n
o
r
m
al
p
er
io
d
s
[
2
9
]
.
T
ab
le
4
.
An
aly
s
is
o
f
o
u
tlier
s
in
an
n
u
al
g
r
o
wth
r
ates (
lar
g
est
o
u
tlier
s
f
r
o
m
8
n
ea
r
est n
ei
g
h
b
o
r
s
)
R
o
w
D
i
st
a
n
c
e
N
e
a
r
n
e
i
g
h
b
o
r
s
C
o
l
1
R
M
S
1
C
o
l
2
R
M
S
2
C
o
l
3
R
M
S
3
C
o
l
4
R
M
S
4
C
o
l
5
R
M
S
5
15
2
.
8
2
4
8
2
4
1
3
1
4
1
2
1
1
1
0
9
8
7
A
M
I
C
W
1
.
1
4
8
5
S
I
EC
0
.
9
2
4
5
F
C
LP
0
.
8
7
8
4
S
TFW
1
5
+
0
.
6
4
2
7
G
R
D
P
p
c
0
.
6
3
3
2
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t
J
E
v
al
&
R
es E
d
u
c
I
SS
N:
2252
-
8
8
2
2
Th
e
imp
a
ct
o
f v
o
c
a
tio
n
a
l tra
in
in
g
o
n
fema
le
la
b
o
r
p
r
o
d
u
ctivi
ty
a
n
ev
a
lu
a
tio
n
in
…
(
Tr
u
c
Th
a
i P
h
u
o
n
g
)
4213
4.
RE
SU
L
T
S
4
.
1
.
M
o
del pre
dict
io
n f
it
a
n
d a
ug
m
ent
a
t
io
n i
m
pa
ct
T
h
e
ANN
m
o
d
el
o
n
th
e
m
ix
ed
d
ataset
(
9
0
s
am
p
les)
ac
h
ie
v
ed
test
R
s
q
u
ar
e=
0
.
6
1
1
,
R
M
SE=
0
.
6
1
7
,
MA
E
=0
.
4
8
5
,
as
s
h
o
wn
in
T
ab
le
2
.
Au
g
m
e
n
tatio
n
m
ar
k
ed
ly
im
p
r
o
v
ed
s
m
o
o
th
n
ess
an
d
ac
cu
r
ac
y
o
v
e
r
r
ea
l
-
o
n
ly
tr
ai
n
in
g
(
1
5
s
am
p
les
)
,
b
etter
ca
p
tu
r
in
g
f
l
u
ctu
atio
n
s
an
d
tr
en
d
s
.
Fig
u
r
e
4
illu
s
tr
ate
s
th
e
ef
f
ec
tiv
en
ess
o
f
s
y
n
th
etic
d
ata
a
u
g
m
e
n
tatio
n
in
en
h
an
cin
g
f
o
r
ec
ast s
m
o
o
t
h
n
ess
an
d
ac
cu
r
ac
y
.
All
s
er
ies
s
h
o
w
lo
n
g
-
ter
m
r
i
s
e
(
~2
.
9
in
2
0
1
0
to
~5
.
5
in
2
0
2
2
–
2
0
2
4
)
,
ac
ce
ler
atio
n
2
0
1
9
–
2
0
2
2
.
Sy
n
th
etic
s
am
p
les
clo
s
ely
s
u
r
r
o
u
n
d
r
ea
l
d
ata
(
W
ass
er
s
tei
n
0
.
0
5
f
o
r
FC
L
P).
Pre
d
icted
R
ea
l
is
s
tep
-
lik
e,
f
latten
s
2
0
1
6
d
r
o
p
,
u
n
d
er
esti
m
ates
2
0
2
2
s
u
r
g
e
[
3
0
]
.
Pre
d
ic
ted
m
ix
is
s
m
o
o
th
er
,
tr
ac
k
s
r
ea
l
clo
s
ely
,
ca
p
tu
r
es
k
ey
tu
r
n
s
(
m
o
d
er
ated
2
0
1
6
d
e
clin
e,
s
tr
o
n
g
2
0
2
2
p
ea
k
,
2
0
2
3
–
2
0
2
4
s
tab
ilizatio
n
)
[
3
1
]
.
Key
s
tag
es:
2
0
1
0
–
2
0
1
5
:
n
ea
r
o
v
e
r
lap
.
2
0
1
6
:
m
ix
ca
p
t
u
r
es
m
o
d
er
ated
d
ec
lin
e;
r
ea
l
n
ea
r
ly
f
lat.
2
0
1
9
–
2
0
2
2
:
m
ix
f
o
llo
ws
ac
ce
ler
atio
n
(
p
ea
k
5
.
6
2
)
.
2
0
2
3
–
2
0
2
4
:
m
ix
s
m
o
o
th
er
.
Fig
u
r
e
4
h
ig
h
lig
h
ts
au
g
m
en
tatio
n
’
s
v
alu
e:
f
ills
g
ap
s
,
lear
n
s
co
m
p
lex
p
atter
n
s
,
r
ed
u
ce
s
o
v
er
f
itti
n
g
(
s
co
r
e
0
.
0
5
7
)
,
s
u
p
p
o
r
ts
r
eliab
le
SHAP
an
aly
s
is
,
a
s
s
h
o
wn
i
n
Fig
u
r
es
5
–
7
.
Mix
p
r
ed
ictio
n
alig
n
s
with
2
0
2
4
–
2
0
2
5
tr
en
d
s
(
GSO
2
0
2
5
–
2
0
2
6
)
:
co
n
s
tr
u
ctio
n
g
r
o
wth
9
.
6
2
%,
in
d
u
s
tr
ial
-
co
n
s
tr
u
ctio
n
9
.
7
3
%,
tr
ain
e
d
wo
r
k
f
o
r
ce
2
9
.
2
%
(
2
0
2
5
)
→2
9
.
5
%
(
Q4
2
0
2
5
)
,
r
ein
f
o
r
cin
g
S
T
FW
1
5
+
r
o
le
am
id
cy
cles.
Piv
o
tal
ev
en
ts
:
2
0
1
6
:
FC
L
P
–
7
.
8
7
%
(
d
o
wn
tu
r
n
,
s
ea
s
o
n
al
f
em
ale
wo
r
k
er
s
af
f
e
cted
)
.
2
0
2
2
:
FC
L
P
+1
5
.
1
6
%
(
r
ec
o
v
er
y
,
f
o
r
m
al
s
h
if
t
b
o
o
s
ted
p
r
o
d
u
ctiv
ity
)
.
I
m
p
licatio
n
s
:
r
etr
ain
i
n
g
f
u
n
d
s
,
f
lex
ib
le
T
VE
T
,
g
en
d
er
-
in
teg
r
ated
i
n
v
estme
n
t
p
lan
n
in
g
,
cy
cle
m
o
n
ito
r
in
g
.
A
u
g
m
en
tatio
n
e
n
s
u
r
es
m
o
d
el
r
e
liab
ilit
y
,
lead
in
g
to
SHAP f
ea
tu
r
e
im
p
o
r
tan
ce
a
n
d
STFW
1
5
+
d
o
m
in
an
ce
ex
p
lan
a
tio
n
[
3
2
]
.
N
o
t
e
:
Li
n
e
c
h
a
r
t
o
f
F
C
LP
(
t
h
o
u
sa
n
d
U
S
D
/
f
e
mal
e
w
o
r
k
e
r
/
y
e
a
r
,
2
0
1
0
–
2
0
2
4
)
:
R
e
a
l
(
n
=
1
5
)
:
so
l
i
d
g
r
e
e
n
w
i
t
h
b
l
a
c
k
d
o
t
s
(
g
r
o
u
n
d
t
r
u
t
h
)
.
S
y
n
t
h
e
t
i
c
(
n
=
7
5
)
:
r
e
d
c
i
r
c
l
e
s.
P
r
e
d
i
c
t
e
d
r
e
a
l
:
b
l
u
e
d
a
s
h
e
d
-
-
-
(
t
r
a
i
n
e
d
o
n
1
5
r
e
a
l
s
o
n
l
y
)
.
P
r
e
d
i
c
t
e
d
m
i
x
:
r
e
d
d
a
s
h
e
d
-
-
-
(
t
r
a
i
n
e
d
o
n
9
0
mi
x
s
a
m
p
l
e
s)
.
Fig
u
r
e
4
.
C
o
m
p
a
r
is
o
n
o
f
ac
tu
a
l,
s
y
n
th
etic,
m
ix
ed
p
r
ed
icted
,
an
d
r
ea
l
-
o
n
ly
p
r
ed
icted
tim
e
s
er
ies
4
.
2
.
SH
AP
ba
s
ed
f
ea
t
ure
i
mp
o
rt
a
nce
SHAP
was
ap
p
lied
to
in
ter
p
r
et
th
e
ANN
m
o
d
el,
p
r
o
v
id
in
g
cl
ea
r
ass
ess
m
en
t
o
f
v
ar
iab
le
co
n
tr
ib
u
tio
n
s
to
FC
L
P
p
r
ed
ictio
n
.
STFW
1
5
+
em
er
g
es
as
th
e
d
o
m
in
an
t
f
a
cto
r
with
th
e
s
tr
o
n
g
est
p
o
s
itiv
e
n
o
n
lin
ea
r
im
p
ac
t.
Fig
u
r
e
5
illu
s
tr
ate
s
th
e
r
an
k
in
g
o
f
f
ea
tu
r
e
im
p
o
r
tan
ce
b
a
s
e
d
o
n
m
e
a
n
a
b
s
o
l
u
t
e
S
HA
P
v
al
u
es
.
STF
W
1
5
+
h
as th
e
h
ig
h
est
m
ea
n
|
SHAP
|
(
0
.
5
1
4
)
—
n
ea
r
ly
d
o
u
b
le
AM
I
C
W
an
d
4
–
6
×
h
ig
h
e
r
th
an
GR
DPp
c
/SIE
C
—
co
n
f
ir
m
in
g
f
em
ale
v
o
ca
tio
n
al
tr
ain
in
g
as
th
e
m
o
s
t
in
f
lu
en
tial
d
r
iv
er
o
f
FC
L
P.
Desp
ite
s
im
ilar
Pe
ar
s
o
n
co
r
r
elatio
n
s
(
~0
.
9
6
)
,
SHAP
h
ig
h
lig
h
ts
STFW
1
5
+’
s
in
d
ep
en
d
en
t
n
o
n
lin
ea
r
s
tr
en
g
th
,
o
v
er
co
m
in
g
m
u
ltico
llin
ea
r
ity
.
AM
I
C
W
is
s
ec
o
n
d
ar
y
;
GR
DPp
c
an
d
SIE
C
s
h
o
w
m
u
ch
wea
k
er
av
e
r
ag
e
e
f
f
ec
ts
(
SIE
C
’
s
n
eg
ativ
e
im
p
ac
t
m
itig
ated
b
y
STFW
1
5
+
in
la
ter
p
lo
ts
)
.
T
h
is
d
ir
ec
tly
an
s
wer
s
R
Q1
an
d
R
Q2
:
STFW
1
5
+
d
o
m
in
ates
o
v
er
AM
I
C
W
,
G
R
DPp
c
,
SIE
C
.
S
HAP
o
u
tp
er
f
o
r
m
s
lin
ea
r
m
eth
o
d
s
,
lin
k
in
g
d
escr
ip
tiv
e
tr
e
n
d
s
(
Fig
u
r
es
1
–
4
)
to
d
ee
p
er
in
ter
p
r
etab
ilit
y
in
Fig
u
r
es
6
–
7
,
em
p
h
asizin
g
STFW
1
5
+’
s
d
u
al
r
o
le
as
to
p
p
r
e
d
icto
r
an
d
m
o
d
er
ato
r
.
Natio
n
al
alig
n
m
e
n
t:
tr
ain
in
g
r
ate
r
ea
ch
e
d
2
9
.
5
%
in
Q4
2
0
2
5
(
+0
.
9
–
1
.
0
p
p
;
GSO
2
0
2
6
)
,
b
u
t
co
n
s
tr
u
ctio
n
f
em
ale
ce
r
tific
atio
n
r
e
m
ain
s
lo
w
(
1
9
–
2
5
%),
u
r
g
in
g
s
p
ec
i
alize
d
g
en
d
e
r
-
r
esp
o
n
s
iv
e
T
V
E
T
[
3
3
]
.
Fig
u
r
e
6
illu
s
tr
ate
s
th
e
d
is
tr
ib
u
tio
n
o
f
i
m
p
ac
ts
(
SHAP v
alu
es)
o
f
ea
ch
in
d
ep
en
d
en
t v
a
r
iab
le
o
n
FC
L
P p
r
ed
ictio
n
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
2
2
I
n
t
J
E
v
al
&
R
es E
d
u
c
,
Vo
l
.
15
,
No
.
5
,
Octo
b
er
20
26
:
4
2
0
7
-
4219
4214
Fig
u
r
e
5
.
Featu
r
e
im
p
o
r
ta
n
ce
r
an
k
in
g
b
ased
o
n
m
ea
n
a
b
s
o
lu
t
e
SHAP
v
alu
es
N
o
t
e
:
b
e
e
sw
a
r
m
s
h
o
w
s
S
H
A
P
d
i
st
r
i
b
u
t
i
o
n
a
c
r
o
ss
9
0
s
a
mp
l
e
s
:
h
o
r
i
z
o
n
t
a
l
a
x
i
s:
S
H
A
P
v
a
l
u
e
(
p
o
s
i
t
i
v
e
→i
n
c
r
e
a
s
e
F
C
LP;
n
e
g
a
t
i
v
e
→d
e
c
r
e
a
s
e
)
.
C
o
l
o
r
:
v
a
r
i
a
b
l
e
v
a
l
u
e
(
h
i
g
h
:
r
e
d
/
y
e
l
l
o
w
;
l
o
w
:
p
u
r
p
l
e
/
b
l
u
e
)
.
D
a
s
h
e
d
l
i
n
e
a
t
0
.
Fig
u
r
e
6
.
SHAP su
m
m
ar
y
p
lo
t
(
b
ee
s
war
m
)
s
h
o
win
g
f
ea
tu
r
e
im
p
ac
ts
Key
o
b
s
er
v
atio
n
s
:
STFW
1
5
+:
wid
est
p
o
s
itiv
e
s
k
ew,
d
en
s
e
+0
.
2
to
+0
.
6
,
ex
tr
em
es
+
0
.
8
at
h
ig
h
v
alu
es
(
~2
0
–
2
4
%)→
s
tr
o
n
g
est
s
tab
le
d
r
iv
er
(
m
ea
n
|
SHAP
|
=0
.
5
1
4
)
.
AM
I
C
W
:
m
ain
ly
p
o
s
itiv
e
(
0
to
+0
.
2
)
,
s
m
aller
im
p
ac
t.
GR
D
Pp
c:
n
ar
r
o
w,
m
o
d
est
p
o
s
itiv
e
(
~0
to
+0
.
1
)
,
wea
k
lev
er
.
SIE
C
:
m
o
s
tly
n
eg
ativ
e
(
–
0
.
2
5
to
0
)
,
d
e
n
s
e
–
0
.
1
to
–
0
.
2
;
tr
ain
in
g
o
f
f
s
ets
m
u
c
h
o
f
n
eg
ativ
e
ef
f
ec
t.
Fig
u
r
e
6
u
n
d
er
s
co
r
es
STFW
1
5
+
p
r
o
m
i
n
en
c
e
an
d
s
ig
n
als
s
tr
o
n
g
in
ter
ac
tio
n
s
(
lead
in
g
to
3
D
p
lo
ts
,
Fig
u
r
es
7
(
a
)
–
(
c
)
)
.
R
ec
en
t
GSO
2
0
2
5
–
2
0
2
6
d
ata
(
t
r
ain
in
g
2
9
.
5
%,
co
n
s
tr
u
ctio
n
g
r
o
wth
9
.
6
2
%,
in
f
o
r
m
al
~6
3
.
1
%)
r
ei
n
f
o
r
ce
STFW
1
5
+
as
k
e
y
le
v
er
t
o
clo
s
e
g
en
d
e
r
g
ap
s
am
id
s
ea
s
o
n
ality
an
d
s
k
ill b
ar
r
ier
s
.
4
.
3
.
No
nli
nea
r
inte
ra
ct
io
n e
f
f
ec
t
s
T
h
e
th
r
ee
3
D
s
u
r
f
ac
e
p
lo
ts
(
a
r
r
an
g
e
d
lef
t
to
r
ig
h
t:
Fig
u
r
e
7
(
a
)
–
AM
I
C
W
×ST
FW
1
5
+,
Fi
g
u
r
e
7
(
b
)
–
GR
DPp
c×
ST
FW
1
5
+,
Fig
u
r
e
7
(
c
)
–
SIE
C
×ST
FW
1
5
+)
illu
s
tr
ate
n
o
n
lin
ea
r
in
ter
ac
tio
n
s
b
et
wee
n
STFW
1
5
+
an
d
o
th
er
v
ar
iab
les
o
n
p
r
e
d
icted
FC
L
P
(
th
o
u
s
an
d
USD
p
er
f
e
m
ale
wo
r
k
er
p
er
y
ea
r
)
.
T
h
ey
co
m
p
lem
en
t
SHAP
r
esu
lts
(
Fig
u
r
es
5
–
6
)
,
s
h
o
win
g
STFW
1
5
+
as
b
o
th
a
s
tr
o
n
g
m
ain
d
r
iv
er
a
n
d
p
o
wer
f
u
l
m
o
d
er
ato
r
.
C
o
n
s
is
ten
t
s
ty
le:
Z
-
ax
is
=FC
L
P;
co
lo
r
f
r
o
m
b
lu
e
(
lo
w)
to
r
ed
(
h
ig
h
)
.
T
o
illu
s
tr
ate
th
e
n
o
n
lin
ea
r
in
ter
ac
tio
n
s
th
r
o
u
g
h
3
D
s
u
r
f
ac
e
p
lo
ts
,
as
s
ee
n
in
Fi
g
u
r
e
7
(
a
)
–
(
c
)
.
Fig
u
r
e
7
(
a
)
(
AM
I
C
W
×ST
F
W
1
5
+)
:
Su
r
f
ac
e
l
o
w
(
~4
.
0
,
b
lu
e
)
at
lo
w
AM
I
C
W
/lo
w
ST
FW
1
5
+;
r
is
es
s
h
ar
p
ly
with
STFW
1
5
+.
Hig
h
est
p
ea
k
(
~
6
.
0
)
at
h
i
g
h
A
MI
C
W
(
>4
5
0
–
5
0
0
)
+h
i
g
h
ST
FW
1
5
+(
>3
9
–
4
2
%).
Po
s
itiv
e
s
lo
p
e
with
s
lig
h
t
wav
es→s
tr
o
n
g
s
y
n
er
g
y
:
tr
ai
n
in
g
a
m
p
lifie
s
in
co
m
e’
s
p
r
o
d
u
ctiv
ity
ef
f
ec
t
.
I
m
p
licatio
n
:
T
r
ain
ed
wo
m
e
n
lev
er
ag
e
h
ig
h
e
r
wag
es
m
o
r
e
ef
f
ec
tiv
ely
.
Fig
u
r
e
7
(
b
)
(
GR
DPp
c×
STFW
1
5
+)
:
Hig
h
est
p
ea
k
(
~7
.
0
–
7
.
5
)
at
h
ig
h
GR
DPp
c
(
>4
.
0
–
5
.
0
)
+h
i
g
h
STFW
1
5
+
(
>4
5
–
5
5
%).
L
o
w
r
eg
io
n
(
GR
DPp
c
<2
.
5
,
STFW
1
5
+<
3
0
%)
f
lat
(
~4
.
0
–
4
.
5
,
b
lu
e)
.
Steep
u
p
war
d
s
lo
p
e
in
h
ig
h
r
e
g
io
n
→s
h
ar
p
r
e
d
p
ea
k
.
I
m
p
licatio
n
:
Stro
n
g
ec
o
n
o
m
ic
g
r
o
wth
m
u
ltip
lies
tr
ain
in
g
’
s
p
r
o
d
u
ctiv
ity
im
p
ac
t
—
clea
r
est
ev
id
en
ce
to
p
r
io
r
itize
f
em
ale
tr
ain
i
n
g
in
i
n
f
r
astru
ctu
r
e
b
o
o
m
s
.
Fig
u
r
e
7
(
c
)
(
STFW
1
5
+×
SIE
C
)
:
lo
west
(
~4
.
0
–
4
.
5
,
b
lu
e)
a
t
h
ig
h
SIE
C
(
>6
0
–
7
0
%)+
lo
w
STF
W
1
5
+(
<3
5
%).
Su
r
f
ac
e
r
is
es
s
ig
n
if
ican
tly
with
S
T
FW
1
5
+(
>4
5
–
5
0
%),
r
ea
ch
in
g
~6
.
0
–
6
.
5
ev
en
at
h
ig
h
SIE
C
.
L
o
ca
l
p
ea
k
at
m
ed
iu
m
SIE
C
(
~4
5
–
5
5
%);
s
lig
h
t
d
ec
lin
e
at
v
er
y
h
ig
h
SIE
C
b
u
t
r
em
ain
s
h
i
g
h
.
I
m
p
l
icatio
n
:
t
r
ain
in
g
s
tr
o
n
g
ly
m
itig
ates
in
f
o
r
m
ality
’
s
n
eg
ativ
e
ef
f
ec
t;
p
r
o
d
u
ctiv
it
y
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t
J
E
v
al
&
R
es E
d
u
c
I
SS
N:
2252
-
8
8
2
2
Th
e
imp
a
ct
o
f v
o
c
a
tio
n
a
l tra
in
in
g
o
n
fema
le
la
b
o
r
p
r
o
d
u
ctivi
ty
a
n
ev
a
lu
a
tio
n
in
…
(
Tr
u
c
Th
a
i P
h
u
o
n
g
)
4215
r
is
es
with
tr
ain
in
g
d
esp
ite
p
e
r
s
is
ten
t
in
f
o
r
m
ality
—
p
r
o
tectiv
e/co
m
p
en
s
ato
r
y
r
o
le.
STFW
1
5
+’
s
m
o
d
er
ato
r
r
o
le
is
k
ey
in
2
0
2
5
r
ec
o
v
er
y
a
m
id
s
ea
s
o
n
al
r
is
k
s
f
o
r
f
em
ale
wo
r
k
er
s
.
GSO
2
0
2
5
–
2
0
2
6
d
ata
(
in
d
u
s
tr
ial
-
co
n
s
tr
u
ctio
n
+9
.
7
3
%,
c
o
n
s
tr
u
ctio
n
+9
.
6
2
%;
tr
ain
in
g
r
ate
2
9
.
2
%
in
2
0
2
5
,
2
9
.
5
%
Q4
;
in
f
o
r
m
al
~
6
3
.
1
%)
co
n
f
ir
m
r
aisi
n
g
STFW
1
5
+
lifts
p
r
o
d
u
ctiv
ity
d
ir
ec
tly
,
am
p
l
if
ies
G
R
DPp
c
/AMI
C
W
g
ain
s
,
an
d
co
m
p
en
s
ates
SIE
C
b
ar
r
ier
s
—
v
ital
f
o
r
d
ig
ita
l/s
u
s
tain
ab
le
co
n
s
tr
u
ctio
n
n
ee
d
in
g
h
ig
h
er
s
k
ills
.
T
h
e
p
lo
ts
c
o
n
f
ir
m
STFW
1
5
+’
s
ce
n
tr
al
n
o
n
lin
ea
r
r
o
le:
am
p
lif
y
in
g
p
o
s
itiv
e
d
r
i
v
er
s
(
AM
I
C
W
,
GR
DPp
c)
an
d
o
f
f
s
ettin
g
i
n
f
o
r
m
ality
b
a
r
r
ier
s
(
SIE
C
)
.
T
h
is
tr
an
s
itio
n
s
t
o
t
h
e
d
is
cu
s
s
io
n
s
ec
tio
n
f
o
r
s
y
n
th
esis
an
d
p
o
licy
r
ec
o
m
m
e
n
d
atio
n
s
o
n
g
en
d
e
r
eq
u
ality
an
d
f
em
ale
p
r
o
d
u
ctiv
ity
in
Vietn
am
’
s
co
n
s
tr
u
ctio
n
s
ec
to
r
[
3
4
]
.
(
a)
(
b
)
(
c)
N
o
t
e
:
H
o
r
i
z
o
n
t
a
l
l
a
y
o
u
t
f
o
r
i
n
t
e
r
a
c
t
i
o
n
s
o
n
p
r
e
d
i
c
t
e
d
F
C
LP;
c
o
mm
o
n
Z
-
a
x
i
s=FC
LP
;
b
l
u
e
-
to
-
r
e
d
g
r
a
d
i
e
n
t
f
o
r
p
r
e
d
i
c
t
e
d
l
e
v
e
l
.
Fig
u
r
e
7
.
3
D
s
u
r
f
ac
e
p
l
o
t: (
a)
i
n
ter
ac
tio
n
b
etwe
en
AM
I
C
W
an
d
STFW
1
5
+
o
n
FC
L
P,
(
b
)
i
n
ter
ac
tio
n
b
etwe
en
GR
DPp
c
an
d
STFW
1
5
+
o
n
FC
L
P,
an
d
(
c)
in
ter
ac
tio
n
b
etw
ee
n
STFW
1
5
+
an
d
SIE
C
o
n
F
C
L
P
5.
DIS
CU
SS
I
O
N
T
h
e
f
in
d
in
g
s
estab
lis
h
STF
W
1
5
+
as
th
e
s
tr
o
n
g
est
p
r
ed
icto
r
o
f
f
em
ale
lab
o
r
p
r
o
d
u
ctiv
ity
in
Vietn
am
’
s
co
n
s
tr
u
ctio
n
s
ec
to
r
(
FC
L
P)
o
v
er
2
0
1
0
–
2
0
2
4
.
T
h
e
m
ea
n
a
b
s
o
lu
te
SHAP
v
alu
e
o
f
0
.
5
1
4
f
o
r
STFW
1
5
+,
wh
ich
is
s
u
b
s
tan
t
ially
h
ig
h
er
th
a
n
o
th
e
r
v
ar
ia
b
les,
in
d
icate
s
th
at
v
o
ca
tio
n
a
l
tr
ain
in
g
ex
er
ts
a
n
in
d
ep
en
d
en
t
an
d
d
o
m
in
an
t
in
f
lu
en
ce
ev
en
in
th
e
p
r
esen
ce
o
f
s
ev
er
e
m
u
ltico
llin
ea
r
ity
[
3
5
]
.
Un
lik
e
tr
ad
itio
n
al
lin
ea
r
r
eg
r
ess
io
n
o
r
s
tan
d
ar
d
ANN
m
o
d
els
th
at
o
f
ten
f
ail
to
d
is
en
tan
g
le
h
ig
h
ly
c
o
r
r
elate
d
p
r
ed
icto
r
s
,
th
e
in
teg
r
atio
n
o
f
SHAP
with
s
y
n
th
etic
d
ata
au
g
m
en
tatio
n
a
llo
wed
th
is
s
tu
d
y
to
r
e
v
ea
l
th
e
n
o
n
lin
ea
r
s
tr
en
g
th
an
d
m
o
d
er
ato
r
r
o
le
o
f
STFW
1
5
+
[
3
6
]
.
Sp
ec
if
ically
,
th
e
3
D
s
u
r
f
ac
e
p
lo
ts
d
em
o
n
s
tr
ate
th
at
STFW
1
5
+
n
o
t
o
n
l
y
d
r
iv
es
d
ir
ec
t
p
r
o
d
u
ctiv
ity
g
ai
n
s
b
u
t
also
am
p
lifie
s
th
e
p
o
s
itiv
e
ef
f
ec
ts
o
f
in
co
m
e
(
AM
I
C
W
)
an
d
ec
o
n
o
m
ic
g
r
o
wth
(
GR
DPp
c
)
,
wh
ile
ef
f
ec
tiv
ely
b
u
f
f
e
r
in
g
th
e
n
eg
ativ
e
im
p
ac
t
o
f
in
f
o
r
m
al
em
p
l
o
y
m
en
t
(
SIE
C
)
.
T
h
is
m
o
d
er
ato
r
ef
f
ec
t
h
ig
h
lig
h
ts
a
cr
itical
m
ec
h
an
is
m
:
v
o
ca
tio
n
a
l
tr
ain
in
g
eq
u
i
p
s
f
em
ale
w
o
r
k
er
s
with
s
k
ills
th
at
en
ab
le
th
em
to
b
etter
lev
er
a
g
e
ec
o
n
o
m
ic
o
p
p
o
r
tu
n
ities
an
d
m
itig
ate
s
tr
u
ctu
r
al
d
is
ad
v
an
tag
es
[
3
7
]
.
T
h
ese
r
esu
lts
ex
ten
d
h
u
m
an
ca
p
ital
th
eo
r
y
an
d
g
en
d
e
r
ed
lab
o
r
m
ar
k
et
th
eo
r
y
b
y
p
r
o
v
id
in
g
em
p
ir
ical,
ex
p
lain
ab
le
ev
id
en
ce
in
a
m
ale
-
d
o
m
in
ated
,
d
ata
-
s
ca
r
ce
s
ec
to
r
[
3
8
]
.
W
h
i
le
p
r
ev
io
u
s
s
tu
d
ies
h
av
e
ap
p
li
ed
ANN
to
p
r
ed
ict
co
n
s
tr
u
ctio
n
p
r
o
d
u
ctiv
ity
,
th
e
y
r
ar
ely
ad
d
r
ess
ed
g
en
d
er
-
s
p
e
cif
ic
d
im
en
s
io
n
s
o
r
o
f
f
er
e
d
in
ter
p
r
etab
le
in
s
ig
h
ts
in
to
f
ea
tu
r
e
in
ter
ac
tio
n
s
[
3
9
]
.
T
h
e
cu
r
r
en
t
ap
p
r
o
ac
h
th
er
e
f
o
r
e
f
ills
a
m
eth
o
d
o
lo
g
ical
g
a
p
a
n
d
s
tr
en
g
th
e
n
s
th
e
s
cien
tific
f
o
u
n
d
atio
n
f
o
r
g
en
d
er
-
r
esp
o
n
s
iv
e
T
VE
T
r
esear
c
h
[
4
0
]
.
T
h
e
s
ig
n
if
ican
ce
o
f
t
h
ese
f
in
d
in
g
s
g
o
es
b
ey
o
n
d
th
e
co
n
s
tr
u
ctio
n
s
ec
to
r
.
T
h
e
y
o
f
f
er
im
p
o
r
tan
t
r
am
i
f
icatio
n
s
f
o
r
T
VE
T
p
o
licy
in
d
e
v
elo
p
in
g
co
u
n
tr
ies,
d
em
o
n
s
tr
atin
g
th
at
ta
r
g
eted
s
k
ill
d
ev
elo
p
m
e
n
t
f
o
r
wo
m
e
n
ca
n
s
er
v
e
as
a
p
o
wer
f
u
l
lev
er
f
o
r
r
ed
u
ci
n
g
in
f
o
r
m
ality
,
n
ar
r
o
win
g
g
e
n
d
er
p
r
o
d
u
ctiv
ity
g
a
p
s
,
an
d
s
u
p
p
o
r
tin
g
s
u
s
tain
ab
le
ec
o
n
o
m
ic
g
r
o
wth
.
B
y
alig
n
in
g
with
Vietn
am
’
s
n
atio
n
al
s
tr
ateg
y
o
n
g
e
n
d
er
eq
u
ality
2
0
2
1
–
2
0
3
0
an
d
th
e
v
o
ca
tio
n
al
ed
u
ca
tio
n
d
ev
elo
p
m
en
t
s
tr
ateg
y
,
as
well
as
co
n
tr
ib
u
tin
g
to
S
DG
5
(
g
en
d
er
e
q
u
ality
)
a
n
d
S
DG
8
(
d
ec
en
t
w
o
r
k
an
d
ec
o
n
o
m
ic
g
r
o
wth
)
,
th
is
s
tu
d
y
p
r
o
v
id
es
ac
tio
n
ab
le
ev
id
en
ce
f
o
r
ad
v
a
n
cin
g
in
clu
s
iv
e
wo
r
k
f
o
r
ce
d
ev
elo
p
m
e
n
t.
Fro
m
a
p
r
ac
tical
p
er
s
p
ec
tiv
e,
th
e
f
in
d
in
g
s
s
u
g
g
est
s
ev
er
al
p
ath
way
s
f
o
r
s
tak
eh
o
ld
er
s
.
Fo
r
ed
u
ca
to
r
s
an
d
T
VE
T
in
s
titu
tio
n
s
,
th
er
e
is
a
n
ee
d
to
d
esig
n
f
lex
ib
le,
s
h
o
r
t
-
ter
m
g
e
n
d
er
-
r
esp
o
n
s
iv
e
p
r
o
g
r
am
s
th
at
em
p
h
asize
g
r
ee
n
co
n
s
tr
u
ctio
n
s
k
ills
,
b
u
ild
in
g
in
f
o
r
m
a
tio
n
m
o
d
elin
g
(
B
I
M)
,
o
cc
u
p
atio
n
al
s
af
ety
,
an
d
p
r
o
ject
m
a
n
ag
em
en
t
tailo
r
ed
to
f
e
m
ale
wo
r
k
er
s
’
r
ea
liti
e
s
[
4
1
]
.
Po
licy
m
ak
e
r
s
s
h
o
u
ld
in
teg
r
ate
g
en
d
e
r
co
n
s
id
er
atio
n
s
in
to
p
u
b
lic
in
v
estme
n
t
p
lan
n
in
g
b
y
i
n
tr
o
d
u
cin
g
m
an
d
ato
r
y
tr
ain
in
g
clau
s
es
an
d
s
o
cial
in
s
u
r
an
ce
r
e
q
u
ir
em
e
n
ts
in
in
f
r
astru
ctu
r
e
co
n
tr
ac
ts
,
p
ar
t
icu
lar
ly
in
p
r
o
jects
with
h
ig
h
f
em
ale
lab
o
r
p
ar
ticip
atio
n
[
4
2
]
.
T
r
ain
in
g
i
n
s
titu
tio
n
s
ca
n
ad
o
p
t
s
im
ilar
ANN
-
SHAP
f
r
am
ewo
r
k
s
f
o
r
r
eg
u
lar
m
o
n
ito
r
in
g
an
d
ev
alu
atio
n
o
f
T
VE
T
p
r
o
g
r
am
ef
f
ec
ti
v
en
ess
,
en
ab
lin
g
d
ata
-
d
r
iv
e
n
im
p
r
o
v
em
e
n
ts
in
cu
r
r
icu
lu
m
d
esig
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
8
2
2
I
n
t
J
E
v
al
&
R
es E
d
u
c
,
Vo
l
.
15
,
No
.
5
,
Octo
b
er
20
26
:
4
2
0
7
-
4219
4216
an
d
o
u
tco
m
es
[
4
3
]
.
E
v
alu
atio
n
s
o
f
lea
n
to
o
ls
an
d
d
ig
ital
te
ch
n
o
lo
g
ies
h
av
e
also
s
h
o
wn
t
h
eir
ef
f
ec
tiv
e
n
ess
in
b
o
o
s
tin
g
o
v
er
all
lab
o
r
p
r
o
d
u
c
tiv
ity
in
c
o
n
s
tr
u
ctio
n
,
f
u
r
th
er
s
u
p
p
o
r
tin
g
th
e
in
teg
r
atio
n
o
f
ad
v
an
ce
d
s
k
ills
in
T
VE
T
p
r
o
g
r
am
s
[
4
4
]
.
Desp
ite
th
e
m
eth
o
d
o
lo
g
ical
a
d
v
an
ce
m
en
ts
,
s
ev
er
al
lim
itatio
n
s
s
h
o
u
ld
b
e
ac
k
n
o
wled
g
ed
.
T
h
e
o
r
ig
in
al
d
ataset
was
s
m
al
l
(
n
=1
5
)
,
an
d
alth
o
u
g
h
s
y
n
th
etic
d
ata
au
g
m
en
tatio
n
im
p
r
o
v
e
d
m
o
d
el
g
en
er
aliza
tio
n
,
th
e
r
esu
lts
d
em
o
n
s
tr
ate
s
tr
o
n
g
ass
o
ciatio
n
s
r
ath
er
th
an
s
tr
ict
ca
u
s
al
r
elatio
n
s
h
ip
s
.
Natio
n
al
-
le
v
el
d
ata
m
a
y
m
ask
im
p
o
r
tan
t
r
eg
io
n
al
v
ar
iatio
n
s
in
lab
o
r
m
a
r
k
ets
an
d
tr
ain
i
n
g
ac
ce
s
s
.
Ad
d
itio
n
ally
,
s
y
n
th
etic
d
ata,
wh
ile
r
ig
o
r
o
u
s
ly
v
alid
ated
,
ca
r
r
ies
in
h
er
en
t
r
is
k
s
o
f
in
tr
o
d
u
cin
g
a
r
tific
ial
p
atter
n
s
.
T
h
e
s
tu
d
y
al
s
o
co
v
er
s
d
ata
o
n
ly
u
p
to
2
0
2
4
,
p
o
ten
tially
m
is
s
in
g
m
o
r
e
r
ec
en
t
p
o
licy
im
p
ac
t
s
in
2
0
2
5
-
2
0
2
6
.
Fu
t
u
r
e
r
esear
ch
s
h
o
u
ld
a
d
d
r
ess
th
ese
lim
itatio
n
s
b
y
u
tili
zin
g
p
r
o
v
in
cial
o
r
f
ir
m
-
le
v
el
d
at
asets
to
ca
p
tu
r
e
h
eter
o
g
en
eit
y
,
ap
p
ly
in
g
ca
u
s
al
in
f
er
en
ce
m
et
h
o
d
s
s
u
ch
as
d
if
f
er
en
ce
-
in
-
d
if
f
e
r
en
ce
s
(
DiD)
o
r
in
s
tr
u
m
en
tal
v
ar
ia
b
les
(
I
V)
f
o
llo
win
g
n
ew
T
VE
T
p
o
licy
im
p
lem
en
tatio
n
s
,
an
d
in
co
r
p
o
r
atin
g
th
e
ef
f
e
cts
o
f
clim
ate
ch
an
g
e,
d
ig
ital
au
to
m
atio
n
,
an
d
I
n
d
u
s
tr
y
4
.
0
tech
n
o
l
o
g
ies
o
n
f
em
ale
c
o
n
s
tr
u
ctio
n
w
o
r
k
er
s
[
4
5
]
.
Ov
er
all,
th
is
s
tu
d
y
p
r
o
v
id
es
r
o
b
u
s
t,
in
ter
p
r
etab
le
ev
i
d
en
ce
th
at
f
e
m
ale
v
o
ca
tio
n
al
tr
ain
i
n
g
is
th
e
p
iv
o
tal
lev
er
f
o
r
r
aisi
n
g
p
r
o
d
u
ctiv
ity
,
n
a
r
r
o
win
g
g
en
d
er
g
ap
s
,
a
n
d
s
u
p
p
o
r
tin
g
s
u
s
tain
ab
le
g
r
o
wth
in
Vietn
am
’
s
co
n
s
tr
u
ctio
n
s
ec
to
r
am
i
d
d
ig
ital
a
n
d
p
o
s
t
-
C
OVI
D
tr
an
s
itio
n
s
.
6.
CO
NCLU
SI
O
N
T
h
is
s
tu
d
y
d
e
m
o
n
s
tr
ates
th
e
p
iv
o
tal
r
o
le
o
f
v
o
ca
tio
n
al
tr
ai
n
in
g
in
en
h
an
cin
g
FC
L
P
in
Vietn
am
’
s
co
n
s
tr
u
ctio
n
s
ec
to
r
f
r
o
m
2
0
1
0
to
2
0
2
4
.
B
y
co
m
b
in
in
g
Fo
r
estDiff
u
s
io
n
s
y
n
th
etic
d
ata
p
ip
elin
e
-
tim
e
-
s
er
ies
au
g
m
en
tatio
n
with
a
n
ANN
a
n
d
SHAP
in
ter
p
r
etatio
n
,
th
e
r
esear
ch
s
u
cc
ess
f
u
lly
o
v
er
ca
m
e
th
e
c
h
allen
g
es
o
f
a
s
m
all
s
am
p
le
s
ize
(
n
=1
5
)
an
d
s
ev
er
e
m
u
ltico
llin
ea
r
ity
,
d
eli
v
er
in
g
r
o
b
u
s
t
an
d
i
n
ter
p
r
eta
b
le
r
esu
lts
.
T
h
e
k
ey
co
n
tr
ib
u
tio
n
lies
in
id
e
n
tify
in
g
STFW
1
5
+
as
th
e
s
tr
o
n
g
est
p
r
ed
icto
r
o
f
FC
L
P
(
m
ea
n
|
SHAP
|
=0
.
5
1
4
)
a
n
d
r
ev
ea
lin
g
its
d
u
al
r
o
le
as
b
o
th
a
d
i
r
ec
t
d
r
iv
er
an
d
a
p
o
wer
f
u
l
m
o
d
er
ato
r
.
I
t
am
p
lifie
s
th
e
b
en
ef
its
o
f
i
n
co
m
e
an
d
ec
o
n
o
m
ic
g
r
o
wth
wh
ile
m
itig
atin
g
th
e
n
eg
ativ
e
ef
f
ec
t
s
o
f
in
f
o
r
m
al
em
p
lo
y
m
en
t.
T
h
ese
f
in
d
in
g
s
p
r
o
v
i
d
e
n
o
v
el,
ex
p
lain
a
b
le
ev
id
en
ce
o
n
th
e
ef
f
ec
tiv
en
ess
o
f
g
en
d
e
r
-
r
esp
o
n
s
iv
e
T
VE
T
in
a
m
ale
-
d
o
m
in
ated
s
ec
to
r
.
Pra
ctica
lly
,
th
e
r
esu
lts
s
tr
o
n
g
ly
s
u
p
p
o
r
t
th
e
f
o
llo
win
g
p
o
licy
r
ec
o
m
m
en
d
atio
n
s
:
p
r
io
r
itize
s
p
ec
ialized
,
f
lex
ib
le
g
en
d
er
-
r
esp
o
n
s
iv
e
T
VE
T
p
r
o
g
r
am
s
f
o
c
u
s
in
g
o
n
g
r
ee
n
s
k
ills
,
s
af
ety
,
B
I
M,
an
d
p
r
o
ject
m
a
n
ag
em
en
t;
in
te
g
r
ate
g
en
d
er
co
n
s
id
er
atio
n
s
in
to
p
u
b
lic
in
v
estme
n
t
p
lan
n
i
n
g
w
ith
m
an
d
at
o
r
y
tr
ain
in
g
an
d
f
o
r
m
al
co
n
tr
ac
ts
in
in
f
r
astru
ctu
r
e
p
r
o
jects;
an
d
f
o
s
ter
p
u
b
lic
-
p
r
iv
ate
p
ar
tn
e
r
s
h
ip
s
to
ex
p
an
d
tr
ain
in
g
o
p
p
o
r
tu
n
ities
f
o
r
f
em
al
e
co
n
s
tr
u
ctio
n
wo
r
k
er
s
.
Su
ch
m
ea
s
u
r
es
ca
n
h
elp
ac
h
iev
e
t
h
e
tar
g
ets
o
f
Vietn
am
’
s
n
atio
n
al
g
en
d
er
eq
u
ality
s
tr
ateg
y
2021
–
2
0
3
0
an
d
co
n
tr
ib
u
te
m
ea
n
in
g
f
u
lly
to
SDG
5
an
d
SDG
8
.
Ov
er
all,
th
is
s
tu
d
y
o
f
f
e
r
s
a
s
o
lid
f
o
u
n
d
atio
n
f
o
r
a
d
v
an
cin
g
g
e
n
d
er
eq
u
ality
,
im
p
r
o
v
in
g
wo
r
k
f
o
r
ce
q
u
ality
,
an
d
p
r
o
m
o
tin
g
SDG
in
Vietn
am
’
s
r
ap
id
ly
g
r
o
win
g
co
n
s
tr
u
ctio
n
s
ec
to
r
.
ACK
NO
WL
E
DG
M
E
N
T
S
T
h
e
au
th
o
r
s
th
an
k
I
n
d
u
s
tr
ial
Un
iv
er
s
ity
o
f
Ho
C
h
i
Min
h
C
ity
f
o
r
r
esear
ch
f
ac
ilit
ies,
Gen
er
al
Statis
t
ics Of
f
ice
,
I
n
ter
n
atio
n
al
L
ab
o
r
Or
g
an
izatio
n
,
a
n
d
W
o
r
l
d
B
an
k
f
o
r
d
ata
s
u
p
p
o
r
t.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
Au
th
o
r
s
s
tate
th
er
e
is
n
o
f
u
n
d
i
n
g
in
v
o
lv
ed
.
AUTHO
R
CO
NT
RI
B
UT
I
O
NS ST
A
T
E
M
E
N
T
T
h
is
jo
u
r
n
al
u
s
es
th
e
C
o
n
t
r
ib
u
to
r
R
o
les
T
a
x
o
n
o
m
y
(
C
R
ed
iT)
to
r
ec
o
g
n
ize
in
d
iv
i
d
u
al
au
th
o
r
co
n
tr
ib
u
tio
n
s
,
r
ed
u
ce
au
th
o
r
s
h
ip
d
is
p
u
tes,
an
d
f
ac
ilit
ate
co
llab
o
r
atio
n
.
Na
m
e
o
f
Aut
ho
r
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
T
r
u
c
T
h
ai
Ph
u
o
n
g
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
L
am
Ng
u
y
e
n
So
n
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
o
n
c
e
p
t
u
a
l
i
z
a
t
i
o
n
M
:
M
e
t
h
o
d
o
l
o
g
y
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
d
a
t
i
o
n
Fo
:
Fo
r
mal
a
n
a
l
y
s
i
s
I
:
I
n
v
e
s
t
i
g
a
t
i
o
n
R
:
R
e
so
u
r
c
e
s
D
:
D
a
t
a
C
u
r
a
t
i
o
n
O
:
W
r
i
t
i
n
g
-
O
r
i
g
i
n
a
l
D
r
a
f
t
E
:
W
r
i
t
i
n
g
-
R
e
v
i
e
w
&
E
d
i
t
i
n
g
Vi
:
Vi
su
a
l
i
z
a
t
i
o
n
Su
:
Su
p
e
r
v
i
s
i
o
n
P
:
P
r
o
j
e
c
t
a
d
mi
n
i
st
r
a
t
i
o
n
Fu
:
Fu
n
d
i
n
g
a
c
q
u
i
si
t
i
o
n
Evaluation Warning : The document was created with Spire.PDF for Python.