I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
3
,
J
une
2025
, pp.
2146
~
2157
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
3
.pp
2146
-
2157
2146
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
L
u
n
g c
an
c
e
r
p
at
i
e
n
t
s s
u
r
vi
val
p
r
e
d
i
c
t
i
on
u
si
n
g o
u
t
l
i
e
r
d
e
t
e
c
t
i
on
an
d
op
t
i
m
i
z
e
d
X
G
B
oost
Wi
r
ot
Y
ot
s
a
w
at
, P
e
e
t
ip
h
ar
t
S
u
e
b
p
e
n
g
, S
ar
oc
h
P
u
r
i
s
an
gk
ah
a
, A
k
a
r
ap
o
n
P
o
on
s
a
w
ad
, K
an
y
al
a
g P
h
od
on
g
C
om
put
e
r
S
c
i
e
nc
e
P
r
ogr
a
m
, D
e
pa
r
t
m
e
nt
of
S
c
i
e
nc
e
, F
a
c
ul
t
y of
S
c
i
e
nc
e
a
nd T
e
c
hnol
ogy, P
hr
a
na
khon S
i
A
yut
t
ha
ya
R
a
j
a
bha
t
U
ni
ve
r
s
i
t
y,
P
hr
a
na
khon S
i
A
yut
t
ha
ya
, T
ha
i
l
a
nd
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
J
ul
2
,
2024
R
e
vi
s
e
d
F
e
b
14
,
2025
A
c
c
e
pt
e
d
M
a
r
15
,
2025
This
research
aims
to
improve
the
prediction’s
model
for
survival
time
of
lung
cancer
patients
by
using
outlier
detection,
hyper
-
par
ameter
optimization,
and
machine
learning
technique.
The
research
compar
es
the
performance
of
several
methods
including
multilayer
percep
tron
(
MLP),
decision
tree
(DT),
linear
regression
(LR),
Bagging,
XGBoost,
and
r
andom
forest
(RF).
The
dataset
used
for
the
experiment
is
obtained
fro
m
the
surveillance,
epidemiology,
and
end
result
(SEER)
cancer
database,
which
contains
diagno
ses
data
from
2004
to
2015.
The
total
number
of
r
ecords
used is 196,031 with 22 features.
10
-
fold cross
-
validation is used
for tr
aining
and
testing
sets.
The
evaluation
metrics
are
root
mean
square
error
(R
MSE),
mean
squared
error
(MSE),
R
-
s
quared
(R
2
),
and
mean
absolute
error
(
MAE).
The results show that the lung
cancer patient survival prediction
mode
l using
the
optimized
XGBoost
(O
-
XGBoost)
model
performs
the
best
w
ith
an
RMSE
of
13.74
and
outperforms
the
baseline
-
XGBoost
model
as
well
as
other
models.
T
his
research
will
be
useful
for
developing
a
clinical
d
ecision
support
system
for
the
care
of
lung
cancer
patients.
Physicians
can
use
the
developed
model
to
assess
the
patient’s
chance
of
survival
in
order
t
o
plan
more eff
ective tr
eatment.
K
e
y
w
o
r
d
s
:
H
ype
r
-
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on
L
ung c
a
nc
e
r
M
a
c
hi
ne
l
e
a
r
ni
ng
O
ut
li
e
r
de
te
c
ti
on
S
ur
vi
va
l
pr
e
di
c
ti
on
X
G
B
oos
t
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
K
a
nya
la
g P
hodong
C
om
put
e
r
S
c
ie
nc
e
P
r
ogr
a
m
, D
e
pa
r
tm
e
nt
of
S
c
ie
nc
e
, F
a
c
ul
ty
of
S
c
ie
nc
e
a
nd T
e
c
hnol
ogy
P
hr
a
na
khon S
i
A
yut
th
a
ya
R
a
ja
bha
t
U
ni
ve
r
s
it
y
P
hr
a
na
khon S
i
A
yut
th
a
ya
, T
ha
il
a
nd
E
m
a
il
:
ka
nya
la
gp@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
I
n
2022,
lu
ng
c
a
nc
e
r
w
a
s
th
e
s
e
c
ond
m
o
s
t
c
om
m
on
ty
pe
of
c
a
nc
e
r
a
nd
th
e
le
a
di
ng
c
a
us
e
of
c
a
nc
e
r
-
r
e
la
te
d
de
a
th
s
.
I
t
is
pa
r
ti
c
ul
a
r
ly
pr
e
va
le
nt
a
m
ong
m
e
n
[
1]
.
T
he
pr
im
a
r
y
c
a
us
e
of
lu
ng
c
a
nc
e
r
is
s
m
oki
ng, with s
m
oke
r
s
ha
vi
ng a
20
-
30 t
im
e
s
hi
ghe
r
r
is
k of
de
v
e
lo
pi
ng l
ung c
a
nc
e
r
c
om
pa
r
e
d t
o non
-
s
m
oke
r
s
[
2]
.
C
ig
a
r
e
tt
e
s
m
oke
c
ont
a
in
s
ove
r
4,000
c
he
m
ic
a
l
s
,
in
c
lu
di
ng
a
t
le
a
s
t
69
known
c
a
r
c
in
oge
n
s
a
nd
ot
he
r
to
xi
ns
.
O
th
e
r
c
a
us
e
s
of
lu
ng
c
a
nc
e
r
in
c
lu
de
a
hi
s
to
r
y
of
r
e
s
pi
r
a
to
r
y
di
s
e
a
s
e
s
(
s
uc
h
a
s
a
s
th
m
a
,
pne
um
oni
a
,
tu
be
r
c
ul
os
is
,
c
hr
oni
c
br
onc
hi
ti
s
,
a
nd
c
hr
oni
c
obs
tr
uc
ti
ve
p
ul
m
ona
r
y
di
s
e
a
s
e
)
,
e
xpos
ur
e
to
w
or
kpl
a
c
e
c
a
r
c
in
oge
ns
,
ge
ne
ti
c
f
a
c
to
r
s
[
3]
,
a
nd
e
nvi
r
onm
e
nt
a
l
a
ir
pol
l
ut
io
n,
e
s
pe
c
ia
ll
y
pr
ol
onge
d
e
xpos
ur
e
to
f
in
e
pa
r
ti
c
ul
a
te
m
a
tt
e
r
(
P
M
2.5)
, w
hi
c
h ha
s
be
e
n l
in
ke
d t
o a
n i
nc
r
e
a
s
e
d r
is
k of
l
ung c
a
nc
e
r
[
4]
.
P
r
e
di
c
ti
ng
th
e
s
ur
vi
va
l
ti
m
e
of
c
a
nc
e
r
p
a
ti
e
nt
s
f
r
om
th
e
in
it
ia
l
di
a
gnos
is
c
a
n
he
lp
pa
ti
e
nt
s
a
nd
c
a
r
e
gi
ve
r
s
pl
a
n t
he
ir
t
im
e
a
nd r
e
s
our
c
e
s
, a
s
w
e
ll
a
s
gui
de
t
he
m
e
di
c
a
l
te
a
m
’
s
c
a
r
e
a
nd t
r
e
a
tm
e
nt
a
ppr
oa
c
h
[
5]
.
H
ow
e
ve
r
,
pr
e
di
c
ti
ng
th
e
s
ur
vi
va
l
ti
m
e
of
c
a
nc
e
r
pa
ti
e
nt
s
r
e
qui
r
e
s
s
pe
c
ia
li
z
e
d
e
xpe
r
ti
s
e
,
pos
in
g
a
c
ha
ll
e
nge
in
a
r
e
a
s
la
c
ki
ng
s
uf
f
ic
ie
nt
s
pe
c
ia
li
s
ts
.
T
he
r
e
f
or
e
,
a
r
ti
f
ic
ia
l
in
te
ll
i
ge
nc
e
(
A
I
)
to
ol
s
bui
lt
w
it
h
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
c
a
n
a
ddr
e
s
s
th
is
is
s
u
e
.
M
a
c
hi
n
e
le
a
r
ni
ng
c
a
n
ut
il
iz
e
hi
s
to
r
ic
a
l
c
a
nc
e
r
pa
ti
e
nt
da
ta
,
s
to
r
e
d
in
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
L
ung c
anc
e
r
pat
ie
nt
s
s
ur
v
i
v
al
pr
e
di
c
ti
on us
in
g outl
ie
r
de
te
c
ti
on
and opti
m
iz
e
d X
G
B
oos
t
(
W
ir
ot
Y
ot
s
aw
at
)
2147
da
ta
ba
s
e
s
,
to
a
c
c
ur
a
te
ly
pr
e
di
c
t
pa
ti
e
nt
s
ur
vi
va
l
ti
m
e
s
[
6]
.
T
hi
s
not
onl
y
he
lp
s
a
voi
d
unne
c
e
s
s
a
r
y
tr
e
a
tm
e
nt
s
,
s
ur
ge
r
ie
s
, a
nd e
xpe
n
s
e
s
f
or
pa
ti
e
nt
s
but
a
ls
o r
e
duc
e
s
di
a
gnos
ti
c
ti
m
e
a
nd a
ll
e
vi
a
te
s
s
tr
e
s
s
f
or
doc
to
r
s
.
T
r
a
di
ti
ona
l
lu
ng
c
a
nc
e
r
s
ur
vi
va
l
pr
e
di
c
ti
on
r
e
s
e
a
r
c
h
ha
s
f
oc
us
e
d
pr
im
a
r
il
y
on
c
la
s
s
if
ic
a
ti
on
ta
s
ks
,
gr
oupi
ng
pa
ti
e
nt
s
in
to
b
r
oa
d
s
ur
vi
va
l
c
a
te
gor
ie
s
,
s
uc
h
a
s
s
ur
vi
ve
d/
not
s
ur
vi
ve
d
or
a
c
hi
e
vi
ng
f
iv
e
-
ye
a
r
s
ur
vi
va
l
[
7]
–
[
12]
.
H
ow
e
ve
r
,
c
la
s
s
if
ic
a
ti
on
of
te
n
la
c
k
s
s
uf
f
ic
ie
nt
de
ta
il
f
or
pr
e
c
is
e
s
ur
vi
va
l
e
s
ti
m
a
te
s
,
li
m
it
in
g
it
s
us
e
f
ul
ne
s
s
f
or
in
di
vi
dua
li
z
e
d
pa
ti
e
nt
c
a
r
e
.
A
s
a
r
e
s
ul
t,
in
te
r
e
s
t
ha
s
s
hi
f
te
d
to
w
a
r
d
r
e
gr
e
s
s
io
n
m
ode
ls
th
a
t
pr
e
di
c
t
c
ont
in
uous
s
ur
vi
va
l
ti
m
e
s
r
a
th
e
r
th
a
n
di
s
c
r
e
te
c
a
t
e
gor
ie
s
,
pr
ovi
di
ng
a
f
in
e
r
a
nd
m
or
e
a
c
c
ur
a
te
vi
e
w
of
pa
ti
e
nt
pr
ognos
is
.
T
e
c
hni
que
s
li
ke
li
ne
a
r
r
e
gr
e
s
s
io
n
(
L
R
)
,
de
c
is
io
n
tr
e
e
(
D
T
)
,
a
nd
r
a
ndom
f
or
e
s
t
(
R
F
)
a
r
e
f
r
e
que
nt
ly
us
e
d
to
e
nha
nc
e
m
ode
l
pe
r
f
or
m
a
nc
e
a
nd
r
e
duc
e
pr
e
di
c
ti
on
e
r
r
or
,
e
s
pe
c
ia
ll
y
in
te
r
m
s
of
r
oot
m
e
a
n
s
qua
r
e
e
r
r
or
(
R
M
S
E
)
[
6
]
.
E
ns
e
m
bl
e
m
e
th
ods
th
a
t
c
om
bi
ne
m
ul
ti
pl
e
a
lg
or
it
hm
s
f
ur
th
e
r
im
pr
ove
a
c
c
ur
a
c
y,
pr
oduc
in
g
m
ode
ls
m
or
e
r
e
s
is
ta
nt
to
ove
r
f
it
ti
ng
a
nd
m
or
e
c
lo
s
e
ly
a
li
gne
d
w
it
h
a
c
tu
a
l
s
ur
vi
va
l
out
c
om
e
s
[
5]
,
[
6
]
,
[
13
]
.
R
e
c
e
nt
ly
,
la
r
ge
da
ta
s
e
ts
li
ke
th
e
s
ur
ve
il
la
nc
e
,
e
pi
de
m
io
lo
gy,
a
nd
e
nd
r
e
s
ul
t
(
S
E
E
R
)
da
ta
ba
s
e
ha
ve
be
e
n
ut
il
iz
e
d
w
it
h
tr
a
di
ti
ona
l
m
a
c
hi
ne
le
a
r
ni
ng
m
e
th
od
s
,
in
c
lu
di
ng
L
R
,
gr
a
di
e
nt
boos
ti
ng
m
a
c
hi
ne
s
(
G
B
M
)
,
a
nd
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
s
(
S
V
M
)
,
to
pr
e
di
c
t
lu
ng
c
a
nc
e
r
s
ur
vi
va
l
ti
m
e
s
,
w
it
h
G
B
M
a
nd e
ns
e
m
bl
e
m
ode
ls
ge
ne
r
a
ll
y de
m
ons
tr
a
ti
ng t
he
hi
ghe
s
t
pr
e
di
c
ti
ve
a
c
c
ur
a
c
y
, pa
r
ti
c
ul
a
r
ly
f
or
pa
ti
e
nt
s
w
it
h s
hor
te
r
s
ur
vi
va
l
ti
m
e
s
[
6]
.
B
y
in
c
or
por
a
ti
ng
ke
y
pa
ti
e
nt
a
tt
r
ib
ut
e
s
s
uc
h
a
s
a
ge
,
tu
m
or
s
iz
e
,
a
nd
s
ta
ge
th
e
s
e
m
ode
ls
of
te
n
pe
r
f
or
m
c
om
pa
r
a
bl
y
to
e
s
ta
bl
is
he
d
s
ta
ti
s
ti
c
a
l
a
ppr
oa
c
he
s
.
A
lt
h
ough
de
e
p
le
a
r
ni
ng
ha
s
a
l
s
o
be
e
n
e
xpl
or
e
d
f
or
s
ur
vi
va
l
pr
e
di
c
ti
on,
tr
a
di
ti
ona
l
m
a
c
hi
ne
le
a
r
ni
ng
m
e
th
ods
r
e
m
a
in
r
e
li
a
bl
e
a
nd
in
te
r
pr
e
ta
bl
e
,
m
a
ki
ng
th
e
m
e
f
f
e
c
ti
ve
a
lt
e
r
na
ti
ve
s
f
or
pr
e
di
c
ti
ng c
ont
in
uous
s
ur
vi
va
l
out
c
om
e
s
i
n l
ung c
a
nc
e
r
pa
ti
e
nt
s
[
5]
.
A
m
a
jo
r
c
ha
ll
e
nge
in
us
in
g
la
r
ge
da
ta
s
e
ts
f
or
pr
e
di
c
ti
on
is
th
a
t
th
e
c
ol
le
c
te
d
r
a
w
da
ta
is
of
te
n
not
r
e
a
dy
f
or
tr
a
in
in
g
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
s
.
I
t
r
e
qui
r
e
s
s
e
ve
r
a
l
pr
e
pr
oc
e
s
s
in
g
s
te
p
s
,
s
uc
h
a
s
ha
ndl
in
g
m
is
s
in
g
va
lu
e
s
,
s
ta
nd
a
r
di
z
in
g
th
e
da
ta
f
or
m
a
t,
tr
a
ns
f
or
m
in
g
da
ta
in
to
s
ui
ta
bl
e
f
or
m
s
f
or
th
e
a
lg
or
it
hm
s
,
f
e
a
tu
r
e
s
e
le
c
ti
on
a
nd
out
li
e
r
s
r
e
m
ova
l.
P
r
e
pr
oc
e
s
s
in
g
not
onl
y
pr
e
pa
r
e
s
th
e
da
ta
f
or
le
a
r
ni
ng
but
c
a
n
a
l
s
o
e
nha
nc
e
m
ode
l
pe
r
f
or
m
a
nc
e
.
F
or
in
s
ta
nc
e
,
out
li
e
r
s
r
e
m
ova
l
[
14]
,
[
15]
a
nd
f
e
a
tu
r
e
s
e
le
c
ti
on
[
16]
,
[
17]
c
a
n
s
ig
ni
f
ic
a
nt
ly
im
pr
ove
m
ode
l
pe
r
f
or
m
a
nc
e
.
N
e
gl
e
c
ti
ng
th
e
s
e
s
te
ps
c
a
n
r
e
s
ul
t
in
hi
ghl
y
in
a
c
c
ur
a
te
m
ode
ls
w
hi
c
h
uns
ui
ta
bl
e
f
or
pr
a
c
ti
c
a
l
us
e
.
A
ddi
ti
ona
ll
y,
du
r
in
g
th
e
le
a
r
ni
ng
pr
oc
e
s
s
,
m
ode
l
pe
r
f
or
m
a
nc
e
c
a
n
be
f
ur
th
e
r
opt
im
iz
e
d
by
hype
r
-
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on
[
18
]
–
[
21]
a
nd
us
in
g
s
ui
ta
bl
y
e
ns
e
m
bl
e
m
e
th
ods
.
E
ns
e
m
bl
e
m
e
th
ods
,
w
hi
c
h
de
ve
lo
p
by
m
ul
ti
pl
e
tr
a
in
in
g
it
e
r
a
ti
ons
,
ty
pi
c
a
ll
y
pr
ovi
de
be
tt
e
r
pe
r
f
or
m
a
nc
e
th
a
n
s
in
gl
e
t
e
c
hni
que
s
[
11]
, [
21]
–
[
24]
.
T
hi
s
r
e
s
e
a
r
c
h
pr
e
s
e
nt
s
a
n
im
pr
ove
m
e
nt
in
th
e
pe
r
f
or
m
a
nc
e
of
s
ur
vi
va
l
ti
m
e
p
r
e
di
c
ti
on
m
ode
ls
f
o
r
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
us
in
g
m
a
c
hi
ne
le
a
r
ni
ng
a
lg
or
it
hm
s
.
T
he
r
e
s
e
a
r
c
h
pr
oc
e
s
s
e
s
in
c
lu
de
out
li
e
r
s
r
e
m
ova
l
us
in
g
is
ol
a
ti
on
f
or
e
s
t
(
I
F
)
[
25]
a
nd
a
ut
o
e
nc
odi
ng
(
A
E
)
[
26]
,
hype
r
-
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on
us
in
g
B
a
ye
s
ia
n
m
e
th
ods
[
27]
,
a
s
w
e
ll
a
s
m
ode
l
c
ons
tr
uc
ti
on
u
s
in
g
va
r
io
us
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
. M
ode
ls
pe
r
f
or
m
a
nc
e
is
e
va
lu
a
te
d
us
in
g
10
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on,
a
nd
th
e
e
f
f
e
c
ti
ve
n
e
s
s
is
m
e
a
s
ur
e
d
by
R
M
S
E
,
m
e
a
n
s
qua
r
e
d
e
r
r
or
(
M
S
E
)
,
R
-
s
qua
r
e
d
(
R
2
)
,
a
nd
m
e
a
n a
bs
ol
ut
e
e
r
r
or
(
M
A
E
)
.
O
nc
e
t
he
m
ode
l
de
m
on
s
tr
a
te
s
th
e
be
s
t
pe
r
f
or
m
a
nc
e
,
it
w
il
l
be
us
e
d t
o de
ve
lo
p a
n a
ppl
ic
a
ti
on f
or
pr
e
di
c
ti
ng t
he
s
ur
vi
va
l
ti
m
e
of
l
ung c
a
nc
e
r
pa
ti
e
nt
s
.
T
he
obj
e
c
ti
ve
s
of
th
is
r
e
s
e
a
r
c
h
a
r
e
a
s
f
ol
lo
w
s
.
F
ir
s
t,
to
im
pr
ove
th
e
pe
r
f
or
m
a
nc
e
of
s
ur
vi
va
l
ti
m
e
pr
e
di
c
ti
on
m
ode
ls
f
or
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
c
ons
tr
uc
te
d
us
in
g
t
r
a
di
ti
ona
l
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
unde
r
r
e
gr
e
s
s
io
n
pr
obl
e
m
s
th
a
t
ge
ne
r
a
ti
ng
c
ont
in
uous
out
c
om
e
s
.
S
e
c
ond,
to
c
om
pa
r
e
th
e
e
f
f
ic
a
c
y
of
in
di
vi
dua
l
a
lg
or
it
hm
s
,
in
c
lu
di
ng
LR
,
DT
,
a
nd
m
ul
ti
la
ye
r
pe
r
c
e
pt
r
on
(
M
L
P
)
,
w
it
h
e
ns
e
m
bl
e
m
e
th
ods
s
uc
h
a
s
B
a
ggi
ng,
X
G
B
oos
t,
a
nd
RF
in
pr
e
di
c
ti
ng t
he
s
ur
vi
va
l
ti
m
e
of
l
ung c
a
nc
e
r
pa
ti
e
nt
s
.
2.
L
I
T
E
R
A
T
U
R
E
R
E
V
I
E
W
E
nha
nc
in
g
a
m
a
c
hi
ne
le
a
r
ni
ng
m
ode
l’
s
pe
r
f
or
m
a
nc
e
c
a
n
be
a
c
hi
e
ve
d
th
r
ough
e
f
f
e
c
ti
ve
da
ta
pr
e
pr
oc
e
s
s
in
g s
te
ps
.
O
ut
li
e
r
d
e
te
c
ti
on
is
th
e
pr
oc
e
s
s
of
id
e
nt
if
yi
ng
da
ta
th
a
t
di
f
f
e
r
s
s
ig
ni
f
ic
a
nt
ly
f
r
om
m
os
t
of
th
e
da
ta
in
a
da
ta
s
e
t.
T
he
s
e
out
li
e
r
s
m
a
y
r
e
s
ul
t
f
r
om
e
r
r
or
s
in
da
ta
c
ol
le
c
ti
on
or
f
r
om
da
ta
th
a
t
is
m
a
r
ke
dl
y
di
f
f
e
r
e
nt
f
r
om
nor
m
a
l
da
ta
[
25]
. T
he
I
F
a
lg
or
it
hm
i
s
one
m
e
th
o
d us
e
d f
or
out
li
e
r
de
te
c
ti
on, de
ve
lo
pe
d
in
[
14]
.
T
hi
s
m
e
th
od
r
e
li
e
s
on
th
e
pr
in
c
ip
le
s
of
DT
th
a
t
c
r
e
a
te
a
la
r
ge
num
be
r
of
DT
.
T
he
m
a
in
c
onc
e
pt
of
I
F
is
th
a
t
out
li
e
r
s
a
r
e
of
te
n
is
ol
a
t
e
d
poi
nt
s
th
a
t
di
f
f
e
r
f
r
om
m
os
t
da
ta
po
in
ts
,
w
hi
le
nor
m
a
l
da
ta
poi
nt
s
a
r
e
di
s
tr
ib
ut
e
d
th
r
oughout
th
e
m
a
in
da
ta
a
r
e
a
[
25]
.
A
not
he
r
popula
r
m
e
th
od
i
s
A
E
te
c
hni
que
s
w
hi
c
h
e
m
pl
oy
s
pe
c
ia
l
ty
pe
s
of
ne
ur
a
l
ne
twor
ks
c
om
pr
is
in
g
a
n
e
n
c
ode
r
pa
r
t
f
or
c
om
pr
e
s
s
in
g
d
a
ta
in
to
lo
w
e
r
di
m
e
n
s
io
ns
a
nd
a
d
e
c
ode
r
p
a
r
t
f
or
r
e
c
ons
tr
uc
ti
ng
th
e
da
ta
ba
c
k
to
it
s
or
ig
in
a
l
di
m
e
ns
io
ns
.
A
E
is
f
i
r
s
t
tr
a
in
e
d
w
it
h
no
r
m
a
l
da
ta
.
W
he
n
ne
w
da
ta
is
pa
s
s
e
d
th
r
ough
th
is
ne
twor
k,
th
e
r
e
c
ons
tr
uc
ti
on
e
r
r
or
is
c
a
lc
ul
a
te
d.
I
f
th
e
e
r
r
or
e
xc
e
e
ds
a
c
e
r
ta
in
th
r
e
s
hol
d,
th
e
da
ta
is
c
ons
id
e
r
e
d
a
n
out
li
e
r
be
c
a
u
s
e
it
c
a
nnot
be
c
om
pr
e
s
s
e
d
a
nd
r
e
c
on
s
tr
uc
te
d
a
c
c
ur
a
t
e
ly
.
T
hi
s
pr
in
c
ip
le
i
s
us
e
d t
o r
a
nk t
he
de
gr
e
e
of
da
ta
out
li
e
r
[
26]
.
O
pt
im
iz
in
g
hype
r
-
pa
r
a
m
e
te
r
s
is
a
not
he
r
m
e
th
od
to
e
nha
nc
e
t
he
pe
r
f
or
m
a
nc
e
of
m
a
c
hi
ne
le
a
r
ni
ng
m
ode
ls
[
18]
,
[
19]
.
T
he
B
a
ye
s
ia
n
m
e
th
od
e
m
pl
oys
B
a
y
e
s
ia
n
p
r
oba
bi
li
ty
th
e
or
y
to
it
e
r
a
ti
ve
ly
r
e
f
in
e
hype
r
-
pa
r
a
m
e
te
r
s
e
le
c
ti
on
f
or
AI
m
ode
l
c
ons
tr
uc
ti
on.
B
e
gi
nni
ng
w
it
h
a
pr
io
r
pr
oba
bi
li
ty
di
s
tr
ib
ut
io
n,
ty
pi
c
a
ll
y
ba
s
e
d
on
e
xpe
r
t
knowle
dge
or
a
s
s
um
pt
io
ns
a
bout
hype
r
-
pa
r
a
m
e
te
r
r
a
nge
s
,
th
e
m
e
th
od
s
a
m
pl
e
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
2146
-
2157
2148
va
lu
e
s
a
nd
e
va
lu
a
te
s
th
e
m
by
tr
a
in
in
g
th
e
m
ode
l
a
nd
a
s
s
e
s
s
in
g
pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
li
ke
e
r
r
or
r
a
te
s
or
a
c
c
ur
a
c
y.
U
s
in
g
B
a
ye
s
’
th
e
or
e
m
,
t
he
pr
io
r
di
s
tr
ib
ut
io
n
i
s
upda
t
e
d
to
a
po
s
t
e
r
io
r
di
s
tr
i
but
io
n
t
ha
t
b
e
tt
e
r
r
e
f
l
e
c
t
s
obs
e
r
ve
d
da
t
a
,
e
nh
a
n
c
in
g
th
e
pr
ob
a
bi
li
ty
e
s
ti
m
a
te
s
f
or
s
ubs
e
qu
e
nt
it
e
r
a
ti
ons
.
T
hi
s
it
e
r
a
ti
ve
pr
oc
e
s
s
c
ont
i
nue
s
,
w
it
h
e
a
c
h
c
yc
l
e
u
s
in
g
upd
a
te
d
pr
ob
a
bi
li
ti
e
s
to
gui
d
e
th
e
s
e
l
e
c
t
io
n
of
ne
w
hyp
e
r
-
pa
r
a
m
e
t
e
r
va
l
ue
s
,
a
im
in
g
to
opt
im
iz
e
m
ode
l
p
e
r
f
or
m
a
n
c
e
e
f
f
e
c
ti
ve
ly
.
T
hi
s
a
ppr
oa
c
h
s
ig
ni
f
ic
a
nt
ly
r
e
duc
e
s
th
e
c
om
p
ut
a
ti
ona
l
bur
d
e
n
c
om
pa
r
e
d t
o
e
xh
a
u
s
ti
v
e
s
e
a
r
c
h m
e
th
od
s
,
e
n
s
ur
in
g m
or
e
e
f
f
ic
ie
nt
a
nd
e
f
f
e
c
t
iv
e
hype
r
-
p
a
r
a
m
e
te
r
t
uni
ng
[
27]
.
I
n
r
e
c
e
nt
s
tu
di
e
s
,
va
r
io
us
m
a
c
hi
n
e
l
e
a
r
ni
n
g
m
o
de
l
s
h
a
ve
b
e
e
n
a
ppl
i
e
d
t
o
pr
e
di
c
t
c
a
nc
e
r
p
a
ti
e
nt
s
ur
vi
v
a
l
pe
r
io
d
s
us
in
g
th
e
S
E
E
R
da
t
a
s
e
t.
T
hom
gk
a
m
e
t
al
.
[
2
8]
c
o
nduc
t
e
d
a
c
o
m
pa
r
a
ti
ve
s
tu
dy
on
m
od
e
l
s
pr
e
di
c
ti
ng
br
e
a
s
t
c
a
n
c
e
r
s
ur
vi
va
l,
a
n
a
ly
z
i
ng
115,
184
r
e
c
or
ds
f
r
om
2004
to
2014
.
T
h
e
y
c
a
te
g
or
iz
e
d
s
ur
v
iv
a
l
out
c
om
e
s
i
nt
o
two
c
la
s
s
e
s
:
le
s
s
th
a
n
5
ye
a
r
s
a
nd
m
or
e
th
a
n
5
ye
a
r
s
.
T
h
e
m
a
c
hi
ne
le
a
r
ni
ng
t
e
c
hni
que
s
c
om
pa
r
e
d
i
nc
lu
d
e
d
n
a
ïv
e
B
a
y
e
s
,
p
a
r
ti
a
l
de
c
is
io
n
tr
e
e
(
P
A
R
T
)
,
M
L
P
,
S
V
M
,
a
nd
B
a
ggi
ng
of
t
he
f
o
ur
m
e
th
od
s
.
T
he
ir
f
in
di
ngs
in
di
c
a
te
d
th
a
t
B
a
g
gi
ng
of
P
A
R
T
ha
d
th
e
b
e
s
t
pe
r
f
or
m
a
nc
e
,
w
it
h
a
s
e
n
s
it
iv
it
y
of
99.
39%
,
s
pe
c
if
ic
it
y
of
96.85%
,
a
nd
a
c
c
ur
a
c
y
of
98.
89%
.
T
he
r
e
s
e
a
r
c
h
e
r
s
s
ug
ge
s
te
d
th
a
t
th
e
m
ode
l
a
nd
de
c
is
io
n
r
ul
e
s
c
oul
d
b
e
de
v
e
lo
p
e
d
in
t
o
a
di
s
e
a
s
e
s
ur
ve
il
l
a
nc
e
s
ys
t
e
m
f
or
e
a
r
ly
r
is
k
s
c
r
e
e
ni
ng
by
ph
ys
i
c
ia
ns
.
S
im
i
la
r
ly
,
D
oppa
l
a
pud
i
e
t
al
.
[
5]
us
e
d
S
E
E
R
da
ta
to
unde
r
s
ta
nd
f
a
c
to
r
s
r
e
la
te
d
to
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
ur
vi
va
l
a
nd
de
ve
lo
pe
d
pr
e
di
c
ti
ve
m
ode
ls
us
in
g
de
e
p
le
a
r
ni
ng
te
c
hni
q
ue
s
li
ke
a
r
ti
f
ic
ia
l
ne
ur
a
l
ne
twor
ks
(
A
N
N
)
,
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
)
,
a
nd
r
e
c
ur
r
e
nt
ne
ur
a
l
n
e
twor
ks
(
R
N
N
)
.
T
h
e
y
de
f
in
e
d
th
r
e
e
s
ur
vi
va
l
c
a
te
gor
ie
s
f
or
c
la
s
s
if
ic
a
ti
on:
le
s
s
th
a
n
or
e
qu
a
l
to
6
m
ont
hs
,
m
o
r
e
th
a
n
6
m
ont
hs
up
to
2
ye
a
r
s
, a
nd
m
or
e
th
a
n
2
ye
a
r
s
,
w
it
h
A
N
N
a
c
hi
e
vi
ng
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
a
t
71.1
8%
.
F
or
r
e
gr
e
s
s
io
n,
C
N
N
s
how
e
d
th
e
be
s
t
pe
r
f
or
m
a
nc
e
w
it
h
a
n
R
M
S
E
of
13.50
a
nd
a
n
R
²
of
0.5066.
L
ync
h
e
t
al
.
[
13]
pr
opos
e
d
a
li
ne
a
r
m
ode
l
f
or
pr
e
di
c
ti
ng
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
ur
vi
va
l,
c
om
pa
r
in
g
it
w
it
h
S
V
M
,
DT
,
RF
,
G
B
M
,
a
nd
a
c
us
to
m
e
n
s
e
m
bl
e
.
T
he
ir
s
tu
dy,
w
hi
c
h
a
na
ly
z
e
d
da
ta
f
r
om
2004
to
2009
w
it
h
18
f
e
a
tu
r
e
s
a
nd
10,442
r
e
c
or
ds
,
f
ound
th
e
c
us
to
m
e
ns
e
m
bl
e
t
o be
t
he
m
o
s
t
e
f
f
e
c
ti
ve
, w
it
h a
n R
M
S
E
of
15.30.
B
a
r
th
ol
om
a
i
a
nd
F
r
ie
boe
s
[
6]
s
tu
di
e
d
th
e
pr
e
di
c
ti
on
o
f
lu
n
g
c
a
nc
e
r
pa
ti
e
nt
s
’
s
ur
vi
va
l
ti
m
e
s
in
m
ont
hs
us
in
g
r
e
gr
e
s
s
io
n
a
nd
c
la
s
s
if
ic
a
ti
on
m
ode
ls
ba
s
e
d
on
S
E
E
R
da
ta
ba
s
e
da
ta
.
T
he
y
obs
e
r
ve
d
th
a
t
w
hi
le
m
ode
ls
w
e
r
e
hi
ghl
y
a
c
c
ur
a
te
f
or
pr
e
di
c
ti
ng
s
hor
t
s
ur
vi
va
l
ti
m
e
s
(
le
s
s
th
a
n
6
m
ont
hs
)
,
th
e
ir
a
c
c
ur
a
c
y
di
m
in
is
he
d
f
or
lo
nge
r
s
ur
vi
va
l
pe
r
io
ds
.
T
he
ir
a
ppr
oa
c
h
ut
il
iz
e
d
RF
f
or
c
la
s
s
if
ic
a
ti
on
a
nd
a
c
om
bi
na
ti
on
of
LR
,
G
B
M
,
a
nd
RF
f
or
r
e
gr
e
s
s
io
n.
M
od
e
l
a
c
c
ur
a
c
y
w
a
s
e
va
lu
a
te
d
us
in
g
a
c
onf
u
s
io
n
m
a
tr
ix
a
nd
R
M
S
E
.
T
he
s
tu
dy
f
ound
th
a
t
RF
pe
r
f
or
m
e
d
be
s
t
f
or
pr
e
di
c
ti
ng
s
ur
vi
va
l
t
im
e
s
of
≤6
m
ont
hs
(
R
M
S
E
10.52)
a
nd
>
24
m
ont
hs
(
R
M
S
E
20.51)
,
w
hi
le
G
B
M
w
a
s
m
os
t
e
f
f
e
c
ti
ve
f
or
7
–
24
m
ont
hs
(
R
M
S
E
15.65)
.
T
he
ir
f
in
di
ngs
s
ugge
s
t
th
a
t
r
e
gr
e
s
s
io
n
m
ode
ls
a
r
e
m
or
e
r
e
li
a
bl
e
f
or
s
hor
te
r
s
ur
vi
va
l
pr
e
di
c
ti
ons
th
a
n
R
M
S
E
va
lu
e
s
m
ig
ht
in
di
c
a
te
.
O
ve
r
a
ll
,
th
e
s
e
s
tu
di
e
s
hi
ghl
ig
ht
th
e
e
f
f
e
c
ti
ve
ne
s
s
o
f
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
in
pr
e
di
c
ti
ng
c
a
nc
e
r
pa
ti
e
nt
s
ur
vi
va
l
a
nd s
ugge
s
t
f
ur
th
e
r
de
ve
lo
pm
e
nt
a
nd
in
t
e
gr
a
ti
on
of
th
e
s
e
m
ode
ls
in
to
c
li
ni
c
a
l
pr
a
c
ti
c
e
f
or
e
a
r
ly
r
is
k s
c
r
e
e
ni
ng a
nd i
m
pr
ove
d pa
ti
e
nt
m
a
na
ge
m
e
nt
.
3.
M
E
T
H
O
D
3.1.
R
e
s
e
ar
c
h
f
r
a
m
e
w
or
k
T
hi
s
r
e
s
e
a
r
c
h
a
ppl
ie
s
th
e
c
r
os
s
-
in
dus
tr
y
s
t
a
nda
r
d
pr
oc
e
s
s
f
or
da
ta
m
in
in
g
(
C
R
I
S
P
-
D
M
)
.
T
he
C
R
I
S
P
-
D
M
pr
oc
e
s
s
c
ons
is
t
s
of
th
e
f
ol
lo
w
in
g
s
te
p
s
:
bus
in
e
s
s
und
e
r
s
ta
ndi
ng,
da
ta
und
e
r
s
ta
ndi
ng,
da
ta
pr
e
pa
r
a
ti
on,
m
ode
li
ng,
m
ode
l
e
va
lu
a
ti
on,
a
nd
de
pl
oym
e
nt
[
29]
.
I
n
th
is
r
e
s
e
a
r
c
h,
th
e
w
or
kf
lo
w
is
di
vi
de
d
in
t
o
s
e
ve
n s
ta
g
e
s
, a
s
s
how
n i
n F
ig
ur
e
1, w
it
h t
he
f
ol
lo
w
in
g de
ta
il
s
.
F
ig
ur
e
1. R
e
s
e
a
r
c
h f
r
a
m
e
w
or
k
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
L
ung c
anc
e
r
pat
ie
nt
s
s
ur
v
i
v
al
pr
e
di
c
ti
on us
in
g outl
ie
r
de
te
c
ti
on
and opti
m
iz
e
d X
G
B
oos
t
(
W
ir
ot
Y
ot
s
aw
at
)
2149
3.
1
.1.
D
at
a gat
h
e
r
in
g
T
he
c
ol
le
c
ti
on
of
lu
ng
c
a
nc
e
r
pa
ti
e
nt
da
ta
f
r
om
th
e
S
E
E
R
da
ta
ba
s
e
,
w
hi
c
h
i
s
a
c
om
pr
e
he
n
s
iv
e
c
a
nc
e
r
r
e
gi
s
tr
y
a
nd
e
pi
de
m
io
lo
gi
c
s
ur
ve
il
la
nc
e
pr
ogr
a
m
of
th
e
U
ni
te
d
S
ta
te
s
.
D
a
ta
s
pa
nni
ng
f
r
om
1973
to
2015
is
ga
th
e
r
e
d
us
in
g
th
e
S
E
E
R
*S
ta
t
a
ppl
ic
a
ti
on
[
30]
.
T
he
S
E
E
R
da
ta
ba
s
e
c
ont
a
in
s
a
s
uf
f
ic
ie
nt
a
m
ount
of
da
ta
f
or
de
ve
lo
pi
ng
m
od
e
ls
a
nd
a
ppl
ic
a
ti
ons
to
pr
e
di
c
t
s
ur
vi
va
l
ti
m
e
.
F
r
om
th
e
da
t
a
c
ol
le
c
ti
on,
it
w
a
s
f
ound
th
a
t
th
e
r
e
a
r
e
a
t
ot
a
l
of
721,697 r
e
c
or
ds
of
l
ung c
a
nc
e
r
pa
ti
e
nt
s
,
e
a
c
h c
om
pr
is
in
g 35 a
tt
r
ib
ut
e
s
.
3.1.2. Dat
a c
le
an
in
g
T
hi
s
pr
oc
e
s
s
c
on
s
is
t
s
of
two
s
ub
-
s
te
p
s
:
i
)
m
is
s
in
g
va
lu
e
ha
n
dl
in
g:
m
a
na
ge
m
e
nt
of
m
is
s
in
g
da
ta
,
w
he
r
e
i
t
w
a
s
f
ound tha
t
70%
o
f
t
he
c
ol
le
c
te
d l
ung c
a
nc
e
r
pa
ti
e
nt
da
ta
c
ont
a
in
e
d m
is
s
in
g va
lu
e
s
t
ha
t
c
a
nnot
be
us
e
d
f
or
m
ode
l
bui
ld
in
g.
T
he
r
e
f
or
e
,
da
ta
w
it
h
m
is
s
in
g
va
lu
e
s
w
a
s
r
e
m
ove
d
f
r
om
th
e
da
ta
s
e
t.
A
ddi
ti
ona
ll
y,
f
e
a
tu
r
e
s
w
it
h
m
is
s
in
g
va
lu
e
s
e
xc
e
e
di
ng
70%
w
e
r
e
a
ls
o
e
xc
lu
de
d.
ii
)
o
ut
l
ie
r
s
r
e
m
ova
l:
r
e
m
ova
l
of
a
bno
r
m
a
l
da
ta
f
r
om
th
e
da
ta
s
e
t.
I
n
th
is
r
e
s
e
a
r
c
h,
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
two
out
li
e
r
de
te
c
ti
on
te
c
hni
que
s
,
IF
a
nd
AE
,
w
a
s
c
om
pa
r
e
d. A
m
a
xi
m
um
t
hr
e
s
hol
d of
20%
of
t
he
t
ot
a
l
da
ta
w
a
s
s
e
t
f
or
out
li
e
r
s
’
r
e
m
ova
l.
3.1.3. Dat
a p
r
e
p
r
oc
e
s
s
in
g
P
r
e
pa
r
a
ti
on of
da
ta
pr
io
r
t
o
pr
oc
e
s
s
in
g t
o e
ns
ur
e
r
e
a
di
ne
s
s
f
or
m
a
c
hi
ne
l
e
a
r
ni
ng t
e
c
hni
que
s
i
nc
lu
de
s
:
i)
tr
a
ns
f
or
m
a
ti
on
us
in
g
one
-
hot
e
nc
odi
ng,
a
ppl
ie
d
to
nom
in
a
l
c
a
te
gor
ic
a
l
f
e
a
tu
r
e
s
th
a
t
c
a
nnot
be
pr
oc
e
s
s
e
d
m
a
th
e
m
a
ti
c
a
ll
y,
a
nd
ii
)
s
c
a
li
ng
of
num
e
r
ic
f
e
a
tu
r
e
s
(
num
e
r
ic
)
th
r
ough
f
e
a
tu
r
e
s
c
a
li
ng,
s
pe
c
if
ic
a
ll
y
us
in
g
nor
m
a
li
z
a
ti
on
to
s
ta
nda
r
di
z
e
va
lu
e
s
w
it
hi
n
th
e
r
a
ng
e
of
0
to
1.
T
hi
s
pr
e
ve
nt
s
bi
a
s
e
s
to
w
a
r
ds
f
e
a
tu
r
e
s
w
it
h
la
r
ge
r
or
s
m
a
ll
e
r
va
lu
e
s
th
a
t
c
oul
d
a
f
f
e
c
t
th
e
f
in
a
l
m
ode
l
ou
tc
om
e
s
.
T
hi
s
s
tu
dy
f
ol
lo
w
s
th
e
da
ta
s
c
a
li
ng
pr
oc
e
dur
e
s
s
im
il
a
r
in
[
13]
to
e
xpe
di
te
m
ode
l
a
dj
us
tm
e
nt
s
a
nd a
c
hi
e
ve
opt
im
a
l
r
e
s
e
a
r
c
h outc
om
e
s
.
3.1.4. Dat
a s
p
li
t
t
in
g
D
a
ta
s
pl
it
ti
ng
is
a
c
r
it
ic
a
l
s
te
p
in
m
a
c
hi
ne
le
a
r
ni
ng
w
he
r
e
th
e
pr
e
pr
oc
e
s
s
e
d
da
ta
s
e
t
is
di
vi
de
d
in
to
two
m
a
in
s
ubs
e
ts
:
a
tr
a
in
in
g
da
ta
s
e
t
c
om
pr
is
in
g
80%
of
th
e
da
t
a
us
e
d
to
tr
a
in
th
e
m
ode
l
by
l
e
a
r
ni
ng
pa
tt
e
r
ns
a
nd
r
e
la
ti
ons
hi
ps
,
a
nd
a
te
s
ti
ng
da
ta
s
e
t
c
om
pr
is
in
g
20%
us
e
d
to
e
va
lu
a
te
th
e
m
ode
l’
s
pe
r
f
or
m
a
nc
e
on
uns
e
e
n
da
ta
.
T
hi
s
a
ppr
oa
c
h
e
n
s
ur
e
s
th
e
m
ode
l’
s
a
bi
li
ty
to
ge
n
e
r
a
li
z
e
w
e
ll
a
nd
m
a
ke
a
c
c
ur
a
te
pr
e
di
c
ti
on
s
be
yond
th
e
tr
a
in
in
g
da
ta
.
B
y
s
e
pa
r
a
ti
ng
tr
a
in
in
g
a
nd
te
s
ti
ng
da
ta
,
m
a
c
hi
ne
le
a
r
ni
ng
pr
a
c
ti
ti
one
r
s
c
a
n
a
s
s
e
s
s
th
e
m
ode
l’
s
e
f
f
e
c
ti
ve
ne
s
s
in
r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
ons
,
gu
a
r
di
ng
a
ga
in
s
t
is
s
u
e
s
li
ke
ove
r
f
it
ti
ng
or
unde
r
f
it
ti
ng,
a
nd
e
ns
ur
in
g
r
obus
t
pe
r
f
or
m
a
nc
e
a
c
r
os
s
di
f
f
e
r
e
nt
da
ta
s
e
t
s
.
3.1.5. M
od
e
li
n
g
I
n t
hi
s
pr
oc
e
s
s
, t
h
e
t
r
a
in
in
g da
t
a
s
e
t
is
u
s
e
d t
o a
ll
ow
m
a
c
hi
ne
l
e
a
r
ni
ng a
lg
or
it
hm
s
t
o l
e
a
r
n a
nd di
s
c
ov
e
r
pa
tt
e
r
n
s
c
ol
le
c
ti
v
e
ly
w
it
hi
n
th
e
da
t
a
.
T
hi
s
in
vol
ve
s
c
r
o
s
s
-
va
li
da
ti
on
w
it
h
1
0
f
ol
d
s
a
nd
a
dj
u
s
ti
n
g
hyp
e
r
-
pa
r
a
m
e
te
r
s
of
th
e
a
lg
or
it
hm
s
to
f
in
d
th
e
m
o
s
t
s
ui
ta
bl
e
va
l
ue
s
.
T
he
le
a
r
ni
ng
m
e
t
hod
s
e
m
pl
oy
e
d
i
nc
lu
de
s
in
gl
e
m
e
th
od
s
s
uc
h
a
s
DT
,
LR
,
a
n
d
M
L
P
,
a
s
w
e
ll
a
s
e
ns
e
m
bl
e
m
e
t
ho
ds
s
uc
h
a
s
B
a
ggi
ng,
X
G
B
oo
s
t,
a
nd
RF
.
H
yp
e
r
-
pa
r
a
m
e
te
r
tu
ni
ng
f
or
th
e
s
e
m
e
t
hod
s
ut
il
i
z
e
s
B
a
y
e
s
ia
n
m
e
th
o
ds
to
de
t
e
r
m
in
e
th
e
opt
im
a
l
h
ype
r
-
p
a
r
a
m
e
te
r
va
lu
e
s
,
a
s
de
m
ons
tr
a
te
d i
n
T
a
bl
e
1.
3.1.6. M
od
e
l
e
val
u
at
io
n
M
ode
l
e
va
lu
a
ti
on
in
vol
ve
s
a
s
s
e
s
s
in
g
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
th
e
tr
a
in
e
d
m
ode
l
us
in
g
th
e
te
s
ti
ng
da
ta
s
e
t.
T
hi
s
e
v
a
lu
a
ti
on
is
c
r
uc
ia
l
f
or
de
te
r
m
in
in
g
how
w
e
ll
th
e
m
ode
l
ge
ne
r
a
li
z
e
s
to
ne
w
,
un
s
e
e
n
da
ta
.
M
e
tr
ic
s
s
uc
h
a
s
R
M
S
E
, M
S
E
, M
A
E
, a
nd
R
2
a
r
e
ut
il
iz
e
d. R
M
S
E
, M
S
E
,
a
nd
M
A
E
m
e
a
s
ur
e
t
he
a
ve
r
a
ge
m
a
gni
tu
de
of
e
r
r
or
s
be
twe
e
n
pr
e
di
c
te
d
a
nd
a
c
tu
a
l
va
lu
e
s
,
w
it
h
lo
w
e
r
va
lu
e
s
in
di
c
a
ti
ng
be
tt
e
r
a
c
c
ur
a
c
y.
R
2
,
on
th
e
ot
he
r
ha
nd,
qua
nt
if
ie
s
th
e
pr
opor
ti
on
of
th
e
va
r
ia
nc
e
in
th
e
de
pe
nde
nt
va
r
ia
bl
e
th
a
t
is
pr
e
di
c
ta
bl
e
f
r
om
th
e
in
d
e
pe
nde
nt
va
r
ia
bl
e
s
;
a
hi
ghe
r
R
2
va
lu
e
(
c
lo
s
e
r
to
1)
s
ugge
s
t
s
th
a
t
th
e
m
od
e
l
e
xpl
a
in
s
a
la
r
ge
r
por
ti
on
of
th
e
va
r
ia
nc
e
in
th
e
da
ta
. T
he
s
e
m
e
tr
ic
s
c
ol
le
c
ti
ve
ly
pr
ovi
de
i
ns
ig
ht
s
i
nt
o t
he
m
ode
l’
s
pr
e
di
c
ti
ve
pe
r
f
or
m
a
nc
e
a
nd i
ts
a
bi
li
ty
to
m
a
ke
r
e
li
a
bl
e
pr
e
di
c
ti
ons
i
n pr
a
c
ti
c
a
l
a
ppl
ic
a
ti
on
s
.
3.1.7. Ap
p
li
c
at
io
n
d
e
ve
lo
p
m
e
n
t
T
he
la
s
t
s
ta
ge
in
vol
ve
s
tr
a
ns
f
or
m
in
g
th
e
hi
ghe
s
t
-
pe
r
f
or
m
in
g
m
ode
l
in
to
a
s
ophi
s
ti
c
a
te
d
w
e
b
a
ppl
ic
a
ti
on
ta
il
o
r
e
d
s
pe
c
if
ic
a
ll
y
f
or
pr
e
di
c
ti
ng
th
e
s
ur
vi
va
l
dur
a
ti
on
of
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
.
T
hi
s
a
ppl
ic
a
ti
on
a
im
s
to
pr
ovi
de
he
a
lt
hc
a
r
e
pr
of
e
s
s
io
na
ls
w
it
h
c
om
pr
e
he
ns
iv
e
pr
e
li
m
in
a
r
y
da
ta
e
s
s
e
nt
ia
l
f
or
m
a
ki
ng
in
f
or
m
e
d
m
e
di
c
a
l
a
s
s
e
s
s
m
e
nt
s
a
nd
tr
e
a
tm
e
nt
de
c
is
io
n
s
.
B
y
le
v
e
r
a
gi
ng
a
dva
nc
e
d
m
a
c
hi
ne
le
a
r
ni
ng
te
c
hni
que
s
a
nd
in
tu
it
iv
e
us
e
r
in
te
r
f
a
c
e
s
,
th
e
a
ppl
ic
a
ti
on
s
e
e
ks
to
s
tr
e
a
m
li
ne
t
he
pr
e
di
c
ti
on
pr
oc
e
s
s
,
e
ns
ur
in
g
a
c
c
ur
a
c
y
a
nd
r
e
li
a
bi
li
ty
in
pr
ognos
is
.
E
m
pha
s
is
w
il
l
be
pl
a
c
e
d
on
in
te
gr
a
ti
ng
s
c
a
la
bl
e
a
r
c
hi
te
c
tu
r
e
a
nd
r
obus
t
ba
c
ke
nd
s
ys
te
m
s
to
s
uppor
t
s
e
a
m
le
s
s
d
e
pl
oym
e
nt
a
nd
ut
il
iz
a
ti
on
in
c
li
n
ic
a
l
s
e
tt
in
gs
.
U
lt
im
a
te
ly
,
th
e
a
ppl
ic
a
ti
on
a
im
s
to
e
nha
nc
e
th
e
e
f
f
ic
ie
nc
y
of
m
e
di
c
a
l
pr
a
c
ti
ti
one
r
s
in
de
li
ve
r
in
g
pe
r
s
ona
li
z
e
d
c
a
r
e
a
nd
im
pr
ovi
ng
pa
ti
e
nt
out
c
om
e
s
t
hr
ough ti
m
e
ly
a
nd a
c
c
ur
a
te
pr
ognos
ti
c
i
n
s
ig
ht
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
2146
-
2157
2150
T
a
bl
e
1
.
H
ype
r
-
pa
r
a
m
e
te
r
s
c
a
pe
f
or
e
a
c
h m
ode
li
ng a
ppr
oa
c
h u
s
e
d t
o f
in
d t
he
opt
im
a
l
hype
r
-
pa
r
a
m
e
te
r
va
lu
e
s
A
l
gor
i
t
hm
s
H
ype
r
-
pa
r
a
m
e
t
e
r
s
S
pa
c
e
of
hype
r
-
pa
r
a
m
e
t
e
r
s
DT
c
c
p_a
l
pha
[
0, 10]
c
r
i
t
e
r
i
on
[
‘
s
qua
r
e
d_e
r
r
or
’
,
‘
f
r
i
e
dm
a
n_m
s
e
’
, ‘
poi
s
s
on’
]
m
a
x_de
pt
h
[
1, 20]
m
i
n_s
a
m
pl
e
s
_l
e
a
f
[
1, 100]
m
i
n_s
a
m
pl
e
s
_s
pl
i
t
[
0, 0.9]
s
pl
i
t
t
e
r
[
‘
be
s
t
’
, ‘
r
a
ndom
’
]
M
L
P
a
l
pha
[
0, 1]
ba
t
c
h_s
i
z
e
[
1, 800]
hi
dde
n_l
a
ye
r
_s
i
z
e
s
[
(
100)
, (
200
)
, (
300)
,
(
400)
, (
123
)
, (
246)
,
(
61)
]
a
c
t
i
va
t
i
on
[
‘
t
a
nh’
, ‘
r
e
l
u’
, ‘
i
de
nt
i
t
y
’
,
‘
l
ogi
s
t
i
c
’
]
l
e
a
r
ni
ng_r
a
t
e
_i
ni
t
[
0, 1]
m
a
x_i
t
e
r
[
10, 200]
l
e
a
r
ni
ng_r
a
t
e
[
‘
c
ons
t
a
nt
’
, ‘
a
da
pt
i
ve
’
, ‘
i
nvs
c
a
l
i
ng’
]
m
om
e
nt
um
[
0, 1]
XGB
m
a
x_de
pt
h
[
1, 30]
ga
m
m
a
[
0, 1]
l
e
a
r
ni
ng_r
a
t
e
[
0, 1.0]
m
i
n_c
hi
l
d_w
e
i
ght
[
1, 20]
a
l
pha
[
0, 1]
s
ubs
a
m
pl
e
[
0, 1]
c
ol
s
a
m
pl
e
_byt
r
e
e
[
0, 1]
m
a
x_de
l
t
a
_s
t
e
p
[
0, 10]
n_e
s
t
i
m
a
t
or
s
[
100, 800]
B
a
g
n_e
s
t
i
m
a
t
or
s
[
1, 300]
m
a
x_s
a
m
pl
e
s
[
0.2, 1.0]
m
a
x_f
e
a
t
ur
e
s
[
0, 1]
RF
n_e
s
t
i
m
a
t
or
s
[
1, 350]
m
a
x_de
pt
h
[
1, 20]
m
i
n_s
a
m
pl
e
s
_l
e
a
f
[
1, 100]
m
i
n_s
a
m
pl
e
s
_s
pl
i
t
[
0.001, 0.9]
m
a
x_f
e
a
t
ur
e
s
[
1, 21]
3.2.
D
at
as
e
t
d
e
t
ai
l
T
he
da
ta
ut
il
iz
e
d
f
or
th
is
r
e
s
e
a
r
c
h
or
ig
in
a
te
s
f
r
om
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
s
our
c
e
d
f
r
om
th
e
S
E
E
R
da
ta
ba
s
e
of
th
e
U
ni
te
d
S
ta
te
s
of
A
m
e
r
ic
a
.
A
f
te
r
unde
r
goi
ng
c
om
pr
e
he
ns
iv
e
m
is
s
in
g
da
t
a
ha
ndl
in
g
pr
oc
e
dur
e
s
, t
he
da
ta
s
e
t
e
nc
om
pa
s
s
e
s
a
t
ot
a
l
of
245,034 pa
ti
e
nt
r
e
c
or
ds
, e
a
c
h c
om
pr
is
in
g 22 dis
ti
nc
t
a
tt
r
ib
ut
e
s
.
T
he
s
e
r
e
c
or
ds
s
p
a
n
th
e
pe
r
io
d
f
r
om
2004
to
2015,
w
it
h
s
ur
vi
va
l
dur
a
ti
ons
r
a
ngi
ng
f
r
om
0
to
191
m
ont
hs
.
T
he
22
a
tt
r
ib
ut
e
s
in
c
lu
de
de
m
ogr
a
phi
c
in
f
or
m
a
ti
on,
c
li
ni
c
a
l
c
ha
r
a
c
te
r
is
ti
c
s
,
a
nd
s
ur
vi
va
l
out
c
om
e
s
,
a
ll
m
e
ti
c
ul
ous
ly
doc
um
e
nt
e
d
a
nd
s
to
r
e
d
w
it
hi
n
th
e
S
E
E
R
da
ta
ba
s
e
.
T
h
e
s
e
d
e
ta
il
s
s
e
r
ve
a
s
d
a
ta
f
or
de
ve
lo
pi
ng
pr
e
di
c
ti
ve
m
ode
ls
a
im
e
d
a
t
f
or
e
c
a
s
ti
ng
s
ur
vi
va
l
ti
m
e
s
f
or
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
,
th
e
r
e
by
s
uppor
ti
ng
c
li
ni
c
a
l
de
c
is
io
n
-
m
a
ki
ng pr
oc
e
s
s
e
s
i
n he
a
lt
h
c
a
r
e
s
e
tt
in
gs
.
T
he
d
e
ta
il
s
of
a
ll
22 f
e
a
tu
r
e
s
a
r
e
pr
e
s
e
nt
e
d i
n
T
a
bl
e
2.
T
a
bl
e
2. C
h
a
r
a
c
te
r
is
ti
c
s
of
l
ung c
a
n
c
e
r
pa
ti
e
nt
da
ta
f
r
om
t
he
S
E
E
R
da
ta
ba
s
e
#
V
a
r
i
a
bl
e
na
m
e
O
r
i
gi
na
l
na
m
e
D
a
t
a
t
ype
1
AGE
A
ge
N
om
i
na
l
2
S
E
X
G
e
nde
r
B
i
na
r
y
3
R
A
C
E
R
a
c
e
s
r
e
c
od
e
(
W
, B
, A
I
,
A
P
I
)
N
om
i
na
l
4
P
R
I
M
A
R
Y
_S
I
T
E
P
r
i
m
a
r
y S
i
t
e
N
om
i
na
l
5
G
R
A
D
E
G
R
A
D
E
N
om
i
na
l
6
H
I
S
T
_B
R
O
A
D
_G
R
O
U
P
H
i
s
t
ol
ogy R
e
c
or
d
–
B
r
oa
d G
r
oupi
ngs
N
om
i
na
l
7
S
E
Q
U
E
N
C
E
_N
U
M
B
E
R
S
e
que
nc
e
N
um
be
r
B
i
na
r
y
8
R
E
G
I
O
N
A
L
_N
O
D
E
S
_P
O
S
I
T
I
V
E
R
e
gi
ona
l
node
s
po
s
i
t
i
ve
(
1988+)
N
um
e
r
i
c
9
R
E
G
I
O
N
A
L
_N
O
D
E
S
_E
X
A
M
I
N
E
D
R
e
gi
ona
l
node
s
e
x
a
m
i
ne
d (
1988+)
N
um
e
r
i
c
10
R
E
A
S
O
N
_S
U
R
G
R
e
a
s
on no c
a
nc
e
r
-
di
r
e
c
t
e
d s
ur
ge
r
y
N
om
i
na
l
11
R
X
S
U
M
M
_S
U
R
G
_P
R
I
M
_S
I
T
E
R
X
S
um
m
–
S
ur
g P
r
i
m
S
i
t
e
(
1998+)
N
om
i
na
l
12
R
X
S
U
M
M
_S
C
O
P
E
_R
E
G
_L
N
_S
U
R
R
X
S
um
m
–
S
c
ope
R
e
g L
N
S
ur
(
2003+)
N
om
i
na
l
13
D
I
A
N
O
S
T
I
C
_C
O
N
F
I
R
M
D
i
a
gnos
t
i
c
C
onf
i
r
m
a
t
i
on
N
om
i
na
l
14
C
S
E
X
T
E
N
S
I
O
N
C
S
e
xt
e
ns
i
on (
2004
-
2015)
N
om
i
na
l
15
C
S
T
U
M
O
R
_S
I
Z
E
S
C
S
t
um
or
s
i
z
e
(
2004
-
2015)
N
um
e
r
i
c
16
C
S
L
Y
M
P
H
_N
O
D
E
S
C
S
l
ym
ph node
s
(
2004
-
2015)
N
om
i
na
l
17
S
U
M
M
A
R
Y
_S
T
A
G
E
S
um
m
a
r
y
s
t
a
ge
2000 (
1998
-
2017)
N
om
i
na
l
18
T
D
e
r
i
ve
d A
J
C
C
T
, 6
th
e
d (
2004
-
2015)
N
om
i
na
l
19
N
D
e
r
i
ve
d A
J
C
C
N
, 6
th
e
d (
2004
-
2015)
N
om
i
na
l
20
M
D
e
r
i
ve
d A
J
C
C
M
, 6
th
e
d (
2004
-
2015)
N
om
i
na
l
21
S
T
A
G
E
D
e
r
i
ve
d A
J
C
C
S
t
a
ge
G
r
oup, 6
th
e
d (
2004
-
2015)
N
om
i
na
l
22
S
U
R
V
I
V
A
L
_M
O
N
T
H
s
ur
vi
va
l
m
ont
hs
N
um
e
r
i
c
N
ot
e
:
F
ur
t
he
r
s
t
udy de
t
a
i
l
s
c
a
n be
f
ound on t
he
w
e
bs
i
t
e
ht
t
p
s
:
/
/
s
e
e
r
.c
a
nc
e
r
.gov
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
L
ung c
anc
e
r
pat
ie
nt
s
s
ur
v
i
v
al
pr
e
di
c
ti
on us
in
g outl
ie
r
de
te
c
ti
on
and opti
m
iz
e
d X
G
B
oos
t
(
W
ir
ot
Y
ot
s
aw
at
)
2151
4.
R
E
S
U
L
T
S
A
N
D
D
I
S
C
U
S
S
I
O
N
4
.1.
R
e
s
u
lt
s
of
ou
t
li
e
r
s
r
e
m
oval
T
he
r
e
s
ul
ts
a
f
te
r
r
e
m
ovi
ng
out
li
e
r
s
da
ta
f
r
om
th
e
da
ta
s
e
t
us
in
g
th
e
IF
a
nd
AE
m
e
th
ods
,
e
a
c
h
w
it
h
a
r
e
m
ova
l
r
a
te
of
20
%
,
s
how
th
a
t
th
e
in
it
ia
l
da
ta
s
e
t,
w
hi
c
h
c
ont
a
in
e
d
245,034
r
e
c
or
ds
,
w
a
s
r
e
duc
e
d
to
196,031
r
e
c
or
ds
us
in
g
bot
h
m
e
th
ods
.
T
hi
s
r
e
duc
ti
on
in
da
ta
w
il
l
be
us
e
d
f
or
s
ubs
e
que
nt
e
xpe
r
im
e
nt
a
l
s
te
ps
.
D
e
s
pi
te
th
e
two
da
ta
s
e
ts
be
in
g
e
qua
l
in
s
iz
e
,
th
e
y
di
f
f
e
r
s
ig
ni
f
ic
a
nt
ly
in
c
ont
e
nt
be
c
a
us
e
th
e
IF
a
nd
AE
m
e
th
ods
e
m
pl
oy
di
s
ti
nc
t
pr
oc
e
s
s
e
s
to
id
e
nt
if
y
a
nd
r
e
m
ove
out
li
e
r
s
.
T
he
IF
m
e
th
od
de
te
c
ts
a
nom
a
li
e
s
ba
s
e
d
on
how
is
ol
a
te
d
th
e
da
ta
poi
nt
s
a
r
e
in
a
f
e
a
tu
r
e
s
pa
c
e
,
w
hi
le
AE
us
e
s
ne
ur
a
l
ne
twor
ks
to
e
n
c
ode
d
a
ta
a
nd
id
e
nt
if
y
de
vi
a
ti
ons
f
r
om
th
e
nor
m
.
T
he
r
e
f
or
e
,
th
e
da
ta
s
e
ts
r
e
s
ul
ti
ng
f
r
om
th
e
s
e
m
e
th
ods
,
a
lt
hough
num
e
r
ic
a
ll
y
e
qui
va
le
nt
,
va
r
y
in
th
e
s
pe
c
if
ic
da
ta
th
e
y
c
ont
a
in
due
to
th
e
s
e
m
e
th
odol
ogi
c
a
l
di
f
f
e
r
e
nc
e
s
.
T
he
de
ta
il
s
of
th
e
c
om
pa
r
is
on be
twe
e
n t
he
i
ni
ti
a
l
da
ta
s
e
t
a
nd t
he
da
t
a
s
e
t
s
a
f
te
r
out
li
e
r
s
r
e
m
ova
l
is
pr
e
s
e
nt
e
d i
n
T
a
bl
e
3.
T
a
bl
e
3.
C
om
pa
r
is
on
th
e
r
e
s
ul
ts
of
t
he
da
ta
b
e
f
or
e
a
nd a
f
te
r
r
e
m
ovi
ng outl
ie
r
s
A
l
gor
i
t
hm
s
O
r
i
gi
na
l
i
ns
t
a
nt
s
G
ood i
ns
t
a
nt
s
A
nom
a
l
y i
ns
t
a
nt
s
%
R
e
m
ova
l
I
s
ol
a
t
i
on
f
or
e
s
t
245
,
034
196
,
031
49
,
003
20.0
A
ut
o
e
nc
odi
ng
245
,
034
196
,
031
49
,
003
20.0
W
he
n t
r
a
in
in
g t
he
a
lg
or
it
hm
s
w
it
h both da
ta
s
e
ts
, i
t
w
a
s
obs
e
r
ve
d t
ha
t
th
e
m
ode
ls
e
xhi
bi
te
d i
m
pr
ove
d
r
e
s
ul
ts
a
nd
pr
e
di
c
ti
ve
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d
to
th
os
e
tr
a
in
e
d
w
it
h
da
ta
th
a
t
r
e
ta
in
e
d
out
li
e
r
s
.
T
he
r
e
du
c
e
d
da
ta
s
e
t
s
iz
e
a
ls
o
c
ont
r
ib
ut
e
d
to
f
a
s
te
r
m
ode
l
c
r
e
a
ti
on.
A
de
ta
il
e
d
c
om
pa
r
is
on
of
M
S
E
,
R
M
S
E
,
M
A
E
,
a
nd
R
2
va
lu
e
s
r
e
ve
a
le
d
th
a
t
th
e
da
ta
s
e
t
pr
oc
e
s
s
e
d
w
it
h
th
e
I
F
m
e
th
o
d
yi
e
ld
e
d
s
upe
r
io
r
r
e
s
ul
ts
a
c
r
os
s
a
ll
m
e
tr
ic
s
,
e
xc
e
pt
f
or
th
e
R
2
va
lu
e
in
th
e
DT
a
lg
or
it
hm
.
A
ddi
ti
ona
ll
y,
w
he
n
th
e
X
G
B
oos
t
a
lg
or
it
hm
w
a
s
tr
a
in
e
d
w
it
h
th
e
IF
-
pr
oc
e
s
s
e
d da
ta
s
e
t,
i
t
de
m
ons
tr
a
te
d t
he
hi
ghe
s
t
pr
e
di
c
ti
ve
pe
r
f
or
m
a
nc
e
a
m
ong a
ll
m
ode
ls
a
nd da
ta
s
e
ts
. T
hi
s
in
di
c
a
te
s
th
a
t
th
e
I
F
m
e
th
od
not
onl
y e
nha
nc
e
s
pr
e
di
c
ti
on a
c
c
ur
a
c
y
but
a
l
s
o
s
tr
e
a
m
li
ne
s
th
e
m
od
e
li
ng
pr
oc
e
s
s
by
e
f
f
e
c
ti
ve
ly
id
e
nt
if
yi
ng
a
nd
r
e
m
ovi
ng
out
li
e
r
s
. T
he
c
om
pr
e
he
ns
iv
e
c
om
pa
r
is
on
a
nd
pe
r
f
or
m
a
nc
e
de
ta
il
s
a
r
e
il
lu
s
tr
a
te
d i
n F
ig
ur
e
2.
F
r
om
F
ig
ur
e
2,
w
he
n
c
ons
id
e
r
in
g
th
e
ove
r
a
ll
pi
c
tu
r
e
,
th
e
e
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
c
a
n
be
e
xpl
a
in
e
d
a
s
f
ol
lo
w
s
:
i)
r
e
m
ovi
ng
out
li
e
r
s
us
in
g
th
e
I
F
m
e
th
od
s
uf
f
ic
ie
nt
ly
r
e
duc
e
s
M
S
E
a
s
s
how
n
in
F
ig
ur
e
2(
a
)
,
R
M
S
E
in
F
ig
ur
e
2(
b)
,
a
nd
M
A
E
in
F
ig
ur
e
2(
c
)
a
c
r
os
s
a
ll
m
e
th
odol
ogi
c
a
l
s
te
p
s
w
h
e
n
c
om
pa
r
e
d
to
bot
h
d
a
ta
s
e
t
s
:
th
e
da
ta
s
e
t
w
it
hout
out
li
e
r
s
r
e
m
ova
l
a
nd
th
e
da
ta
s
e
t
w
it
h
out
li
e
r
s
r
e
m
ova
l
us
in
g
th
e
A
E
m
e
th
od.
ii
)
out
li
e
r
s
r
e
m
ova
l
w
it
h
th
e
A
E
m
e
th
od,
how
e
ve
r
,
s
li
ght
ly
de
c
r
e
a
s
e
s
th
e
ove
r
a
ll
pe
r
f
or
m
a
nc
e
of
th
e
m
ode
ls
,
e
vi
de
nc
e
d
by
hi
ghe
r
M
S
E
a
s
s
how
n
in
F
ig
ur
e
2(
a
)
,
R
M
S
E
in
F
ig
ur
e
2(
b
)
,
a
nd
M
A
E
in
F
ig
ur
e
2(
c
)
va
lu
e
s
,
a
s
w
e
ll
a
s
de
c
r
e
a
s
e
d
R
2
in
F
ig
ur
e
2(
d)
va
lu
e
s
w
he
n
c
om
pa
r
e
d
to
th
e
d
a
ta
s
e
t
w
it
hout
out
li
e
r
s
r
e
m
ova
l.
ii
i)
a
lt
hough
out
li
e
r
s
r
e
m
ova
l
w
it
h
th
e
I
F
m
e
th
od
le
a
d
s
to
de
c
r
e
a
s
e
d
M
S
E
a
s
s
how
n
in
F
ig
ur
e
2(
a
)
,
R
M
S
E
in
F
ig
ur
e
2(
b)
,
a
nd
M
A
E
in
F
ig
ur
e
2(
c
)
va
lu
e
s
in
D
T
,
th
e
R
2
va
lu
e
de
c
r
e
a
s
e
s
a
s
w
e
ll
a
s
s
how
n
in
F
ig
ur
e
2(
d)
.
T
hi
s
is
due
to
th
e
D
T
m
e
th
od’
s
in
it
ia
l
hype
r
-
pa
r
a
m
e
te
r
s
,
w
hi
c
h
do
not
s
p
e
c
if
y
a
m
a
xi
m
um
de
pt
h
a
nd
le
a
f
node
s
,
r
e
s
ul
ti
ng
in
a
n
unne
c
e
s
s
a
r
y
pr
ol
if
e
r
a
ti
on
of
le
a
f
node
s
dur
in
g
tr
a
in
in
g.
C
ons
e
que
nt
ly
,
th
e
f
in
a
l
le
a
r
ne
d
r
e
s
ul
ts
s
how
a
n
in
a
c
c
ur
a
te
a
ve
r
a
ge
of
e
a
c
h
r
e
c
or
d
m
ix
e
d
w
it
hi
n
th
e
e
xc
e
s
s
iv
e
le
a
f
node
s
.
iv
)
de
s
pi
te
th
e
de
c
r
e
a
s
e
d
R
2
a
s
s
how
n
in
F
ig
ur
e
2
(
d)
va
lu
e
in
D
T
a
f
te
r
out
li
e
r
s
r
e
m
ova
l
w
it
h
th
e
I
F
m
e
th
od,
ot
he
r
m
e
th
ods
s
how
a
n
im
pr
ove
m
e
nt
i
n R
2 va
lu
e
s
i
n F
ig
ur
e
2(
d)
.
B
a
s
e
d
on
th
e
e
xpe
r
im
e
nt
a
l
f
in
di
ngs
,
it
w
a
s
e
vi
de
nt
th
a
t
th
e
da
ta
s
e
t
tr
e
a
te
d
w
it
h
th
e
I
F
out
li
e
r
s
r
e
m
ova
l
m
e
th
od c
ons
is
te
nt
ly
e
xhi
bi
te
d s
upe
r
io
r
pe
r
f
or
m
a
nc
e
c
om
pa
r
e
d t
o both
t
he
da
ta
s
e
t
pr
oc
e
s
s
e
d w
it
h t
he
A
E
m
e
th
od
a
nd
th
e
or
ig
in
a
l
da
ta
s
e
t
w
it
hout
ou
tl
ie
r
s
r
e
m
ova
l.
C
ons
e
que
nt
ly
,
th
e
I
F
-
pr
oc
e
s
s
e
d
da
ta
s
e
t
w
a
s
s
e
le
c
te
d
f
or
f
ur
th
e
r
m
ode
l
de
ve
lo
pm
e
nt
s
ta
g
e
s
.
S
pe
c
if
ic
a
ll
y,
a
c
r
os
s
th
e
s
ix
m
ode
ls
e
va
lu
a
te
d
(
M
L
P
,
L
R
,
D
T
,
X
G
B
oos
t,
a
nd
B
a
ggi
ng)
it
w
a
s
not
e
d
th
a
t
f
iv
e
of
th
e
s
e
m
ode
ls
de
m
ons
tr
a
te
d
e
nha
nc
e
d
pr
e
di
c
ti
ve
c
a
pa
bi
li
ti
e
s
w
he
n
tr
a
in
e
d
on
th
e
I
F
-
pr
oc
e
s
s
e
d
d
a
ta
s
e
t.
T
hi
s
e
nha
nc
e
m
e
nt
w
a
s
r
e
f
le
c
te
d
in
im
pr
ove
d
m
e
tr
ic
s
s
uc
h
a
s
r
e
duc
e
d
R
M
S
E
,
M
A
E
,
a
nd
M
S
E
,
a
s
w
e
ll
a
s
hi
ghe
r
R
2
v
a
lu
e
s
,
in
di
c
a
ti
ng
be
tt
e
r
m
ode
l
a
c
c
ur
a
c
y
a
nd
r
obus
tn
e
s
s
.
T
he
s
e
r
e
s
ul
ts
unde
r
s
c
or
e
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
th
e
I
F
m
e
th
od
in
e
nha
nc
in
g
da
ta
qua
li
ty
by
e
f
f
e
c
ti
ve
ly
id
e
nt
if
yi
ng
a
nd
r
e
m
ovi
ng
out
li
e
r
s
,
th
e
r
e
by
c
ont
r
ib
ut
in
g
to
im
pr
ove
d
m
ode
l
pe
r
f
or
m
a
nc
e
a
c
r
os
s
va
r
io
us
m
a
c
hi
ne
l
e
a
r
ni
ng a
lg
or
it
hm
s
.
4.2. T
h
e
r
e
s
u
lt
s
o
f
h
yp
e
r
-
p
ar
am
e
t
e
r
s
’
op
t
im
iz
at
io
n
I
n
T
a
bl
e
4,
th
e
hype
r
-
pa
r
a
m
e
te
r
va
lu
e
s
obt
a
in
e
d
th
r
ough
th
e
B
a
ye
s
ia
n
s
e
a
r
c
h
m
e
th
od,
u
s
in
g
th
e
da
ta
s
e
t
pr
oc
e
s
s
e
d
w
it
h
I
F
out
li
e
r
s
r
e
m
ova
l,
w
e
r
e
us
e
d
to
bui
l
d
m
ode
ls
.
T
he
m
ode
l
s
w
e
r
e
e
va
lu
a
te
d
u
s
in
g
K
-
f
ol
ds
c
r
os
s
-
va
li
da
ti
on w
it
h K
s
e
t
to
10
.
T
hi
s
r
e
s
ul
te
d
in
a
ve
r
a
ge
pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
a
c
r
os
s
a
ll
10 f
ol
ds
f
o
r
e
a
c
h of
t
he
6 m
e
th
ods
, a
s
s
how
n i
n
T
a
bl
e
5.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
2146
-
2157
2152
(
a
)
(
b)
(
c
)
(
d)
F
ig
ur
e
2. C
om
pa
r
is
on
th
e
r
e
s
ul
ts
of
t
he
m
ode
ls
t
r
a
in
e
d on da
ta
be
f
or
e
a
nd a
f
te
r
out
li
e
r
r
e
m
ova
l
on (
a
)
M
S
E
va
lu
e
, (
b)
R
M
S
E
va
lu
e
, (
c
)
M
A
E
va
lu
e
, a
nd (
d)
R
2
va
lu
e
T
a
bl
e
4. R
e
s
ul
ts
of
hype
r
-
pa
r
a
m
e
te
r
s
opt
im
iz
a
ti
on us
in
g B
a
ye
s
i
a
n s
e
a
r
c
h
A
l
gor
i
t
hm
s
H
ype
r
-
pa
r
a
m
e
t
e
r
B
e
s
t
pa
r
a
m
e
t
e
r
s
DT
c
c
p_a
l
pha
0.00339225
c
r
i
t
e
r
i
on
s
qua
r
e
d_e
r
r
or
m
a
x_de
pt
h
8
m
i
n_s
a
m
pl
e
s
_l
e
a
f
61
m
i
n_s
a
m
pl
e
s
_s
pl
i
t
0.860951952
s
pl
i
t
t
e
r
r
a
ndom
XGB
m
a
x_de
pt
h
24
ga
m
m
a
0.173941856
l
e
a
r
ni
ng_r
a
t
e
0.05028551
m
i
n_c
hi
l
d_w
e
i
ght
9.006166167
a
l
pha
0.26153907
s
ubs
a
m
pl
e
0.700633533
c
ol
s
a
m
pl
e
_byt
r
e
e
0.403669676
m
a
x_de
l
t
a
_s
t
e
p
6
n_e
s
t
i
m
a
t
or
s
480
M
L
P
a
l
pha
0.203240288
ba
t
c
h_s
i
z
e
798
hi
dde
n_l
a
ye
r
_s
i
z
e
s
200
a
c
t
i
va
t
i
on
r
e
l
u
l
e
a
r
ni
ng_r
a
t
e
_i
ni
t
0.000542794
m
a
x_i
t
e
r
82
l
e
a
r
ni
ng_r
a
t
e
c
ons
t
a
nt
m
om
e
nt
um
0.637558948
B
a
ggi
ng
n_e
s
t
i
m
a
t
or
s
296
m
a
x_s
a
m
pl
e
s
0.695150026
m
a
x_f
e
a
t
ur
e
s
0.207177187
RF
n_e
s
t
i
m
a
t
or
s
114
m
a
x_de
pt
h
12
m
i
n_s
a
m
pl
e
s
_l
e
a
f
82
m
i
n_s
a
m
pl
e
s
_s
pl
i
t
0.003641991
m
a
x_f
e
a
t
ur
e
s
16
I
n
T
a
bl
e
5,
it
is
e
vi
de
nt
th
a
t
a
f
te
r
a
dj
us
ti
ng
th
e
hype
r
-
pa
r
a
m
e
t
e
r
s
to
opt
im
a
l
va
lu
e
s
,
th
e
D
T
,
M
L
P
,
L
R
,
R
F
,
B
a
ggi
ng,
a
nd
opt
im
iz
e
d
X
G
B
oos
t
(
O
-
X
G
B
oos
t)
m
ode
ls
a
ll
s
how
e
d
va
r
yi
ng
de
gr
e
e
s
of
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
L
ung c
anc
e
r
pat
ie
nt
s
s
ur
v
i
v
al
pr
e
di
c
ti
on us
in
g outl
ie
r
de
te
c
ti
on
and opti
m
iz
e
d X
G
B
oos
t
(
W
ir
ot
Y
ot
s
aw
at
)
2153
im
pr
ove
m
e
nt
.
N
ot
a
bl
y,
th
e
O
-
X
G
B
oos
t
m
ode
l
e
m
e
r
ge
d
a
s
th
e
to
p
pe
r
f
or
m
e
r
a
c
r
os
s
a
ll
e
va
lu
a
ti
on
m
e
tr
ic
s
(
M
S
E
,
R
M
S
E
,
M
A
E
,
a
nd
R
2
)
in
di
c
a
ti
ng
it
s
r
obus
tn
e
s
s
a
n
d
s
upe
r
io
r
pr
e
di
c
ti
ve
c
a
pa
bi
li
ti
e
s
.
F
ol
lo
w
in
g
X
G
B
oos
t,
th
e
M
L
P
,
B
a
ggi
ng,
R
F
,
L
R
,
a
nd
D
T
m
ode
ls
s
e
que
nt
ia
ll
y
e
xhi
bi
te
d
c
om
m
e
nda
bl
e
pe
r
f
or
m
a
nc
e
im
pr
ove
m
e
nt
s
.
S
pe
c
if
ic
a
ll
y,
th
e
O
-
X
G
B
oos
t
m
ode
l
a
c
hi
e
ve
d
im
pr
e
s
s
iv
e
m
e
tr
ic
s
w
it
h
R
M
S
E
a
t
13.74,
R
2
of
0.4747, M
A
E
o
f
0.0876, a
nd M
S
E
of
0.0190.
C
om
pa
r
a
ti
v
e
ly
, w
he
n c
om
pa
r
e
d t
o t
he
B
a
s
e
li
ne
-
X
G
B
oos
t
m
ode
l
tr
a
in
e
d
on
th
e
in
it
ia
l
da
ta
s
e
t,
th
e
X
G
B
oos
t
m
ode
l
r
e
f
in
e
d
th
r
ough
th
e
da
ta
m
in
in
g
p
r
oc
e
s
s
s
how
c
a
s
e
d
s
uf
f
ic
ie
nt
ly
e
nha
nc
e
d
pe
r
f
or
m
a
nc
e
.
T
he
s
e
f
in
di
ngs
unde
r
s
c
or
e
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
hype
r
-
pa
r
a
m
e
te
r
tu
ni
ng
a
nd
da
ta
pr
e
pr
oc
e
s
s
in
g
te
c
hni
que
s
in
opt
im
iz
in
g
m
ode
l
a
c
c
ur
a
c
y
a
nd
r
e
li
a
bi
li
ty
a
c
r
os
s
di
ve
r
s
e
m
a
c
hi
ne
le
a
r
ni
ng a
lg
or
it
hm
s
.
T
a
bl
e
5. R
e
s
ul
ts
of
t
he
m
ode
ls
a
f
te
r
a
dj
us
ti
ng hype
r
-
pa
r
a
m
e
te
r
s
t
o a
ppr
opr
ia
te
va
lu
e
s
M
ode
l
s
M
S
E
R
M
S
E
M
A
E
R
2
DT
0.0233
15.2200
0.1007
0.3554
M
L
P
0.0193
13.8600
0.0880
0.4656
LR
0.0197
13.9900
0.0906
0.4555
RF
0.0196
13.9600
0.0899
0.4579
B
a
g
gi
ng
0.0195
13.9100
0.0908
0.4610
O
-
XGB
oos
t
0.0190
13.7400
0.0876
0.4747
4
.
3
.
R
e
s
u
lt
s
d
is
c
u
s
s
io
n
I
n
T
a
bl
e
6,
th
e
c
om
pa
r
a
ti
ve
a
na
ly
s
is
be
twe
e
n
th
e
O
-
X
G
B
oos
t
a
nd
B
a
s
e
li
ne
-
X
G
B
oos
t
m
ode
ls
s
ho
w
s
th
a
t
th
e
O
-
X
G
B
oos
t
m
ode
l
out
pe
r
f
or
m
s
th
e
ba
s
e
li
ne
a
c
r
o
s
s
a
l
l
e
va
lu
a
ti
on
m
e
tr
ic
s
.
T
he
O
-
X
G
B
oos
t
m
ode
l
a
c
hi
e
ve
d
a
lo
w
e
r
M
S
E
a
t
0.0190
c
om
pa
r
e
d
to
0.0231,
a
nd
a
lo
w
e
r
R
M
S
E
a
t
13.74
c
om
pa
r
e
d
to
15.13
,
in
di
c
a
ti
ng
m
or
e
pr
e
c
is
e
a
nd
a
c
c
ur
a
te
pr
e
di
c
ti
ons
.
A
ddi
ti
ona
ll
y
,
th
e
M
A
E
is
r
e
duc
e
d
f
r
om
0.0999
to
0.0876,
f
ur
th
e
r
c
onf
ir
m
in
g
th
e
m
ode
l’
s
im
pr
ove
d
a
c
c
ur
a
c
y.
T
he
R
²
va
lu
e
in
c
r
e
a
s
e
d
f
r
om
0.4537
to
0.4747,
de
m
ons
tr
a
ti
ng
th
a
t
th
e
opt
im
iz
e
d
m
ode
l
e
xpl
a
in
s
a
gr
e
a
te
r
pr
opor
ti
on
of
va
r
ia
nc
e
in
th
e
da
ta
.
O
ve
r
a
ll
,
th
e
O
-
X
G
B
oos
t
m
ode
l
pr
ovi
de
s
m
or
e
r
e
li
a
bl
e
pr
e
di
c
ti
ons
f
or
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
ur
vi
va
l
ti
m
e
s
.
T
hi
s
a
li
gns
w
it
h
th
e
r
e
s
e
a
r
c
h
in
[
18]
, [
19]
, w
hi
c
h us
e
d hype
r
-
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on t
o e
nha
nc
e
t
he
pe
r
f
or
m
a
nc
e
of
X
G
B
oos
t.
T
a
bl
e
6. C
om
pa
r
is
on of
X
G
B
oos
t
m
ode
l,
B
a
s
e
li
ne
-
X
G
B
oos
t
a
n
d ot
he
r
m
ode
ls
M
ode
l
s
I
ns
t
a
nc
e
s
A
t
t
r
i
but
e
s
S
ur
vi
va
l
t
i
m
e
(
m
ont
hs
)
M
S
E
R
M
S
E
M
A
E
R
2
C
us
t
om
e
ns
e
m
bl
e
[
13
]
10,442
19
0
-
72
-
15.30
-
-
R
F
[
6]
10,442
25
0
-
6
-
10.52
-
-
G
B
M
[
6]
10,442
25
7
-
24
-
15.65
-
-
R
F
[
6]
10,442
25
>24
-
20.51
-
-
G
B
M
[
5]
702,411
15
0
-
60
0.023
15.30
-
0.3664
R
F
[
5]
702,411
15
0
-
60
0.022
14.87
-
0.4015
C
N
N
[
5]
702,411
15
0
-
60
0.018
13.50
-
0.5066
B
a
s
e
l
i
ne
-
X
G
B
oo
s
t
196
,
031
22
0
-
191
0.023
15.13
0.099
0.4537
O
-
X
G
B
oos
t
196
,
031
22
0
-
191
0.019
13.74
0.087
0.4747
T
he
f
in
di
ngs
of
th
is
r
e
s
e
a
r
c
h
a
li
gn
w
it
h
pr
e
vi
ous
s
tu
di
e
s
th
a
t
ha
ve
de
m
on
s
tr
a
te
d
th
e
e
f
f
ic
a
c
y
of
m
a
c
hi
ne
le
a
r
ni
ng
m
ode
ls
in
pr
e
di
c
ti
ng
c
a
nc
e
r
pa
ti
e
nt
s
ur
vi
va
l
ti
m
e
s
us
in
g
S
E
E
R
da
ta
.
S
im
il
a
r
to
th
e
s
tu
dy
[
28]
on
br
e
a
s
t
c
a
nc
e
r
,
a
nd
s
im
il
a
r
ly
to
th
e
s
tu
dy
[
5]
,
[
13]
r
e
s
e
a
r
c
h
on
lu
ng
c
a
nc
e
r
,
our
s
tu
dy
c
onf
ir
m
s
th
e
s
upe
r
io
r
pe
r
f
or
m
a
nc
e
of
e
ns
e
m
bl
e
m
e
th
ods
a
nd
a
dv
a
n
c
e
d
a
lg
or
it
hm
s
.
N
ot
a
bl
y,
our
u
s
e
of
out
li
e
r
de
te
c
ti
on
a
nd
hype
r
-
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on
ha
s
f
ur
th
e
r
e
nha
nc
e
d
m
ode
l
a
c
c
ur
a
c
y,
w
it
h
X
G
B
oos
t
out
pe
r
f
or
m
in
g
ot
he
r
m
ode
ls
,
a
c
hi
e
vi
ng
R
M
S
E
of
13.74.
T
hi
s
r
e
s
ul
t
s
ur
pa
s
s
e
s
th
e
pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
r
e
por
te
d
by
B
a
r
th
ol
om
a
i
a
nd
F
r
ie
boe
s
[
6]
,
w
ho
hi
ghl
ig
ht
e
d
th
e
va
r
ia
bi
li
ty
in
a
c
c
ur
a
c
y
a
c
r
os
s
di
f
f
e
r
e
nt
s
ur
vi
va
l
pe
r
io
ds
.
T
he
s
uc
c
e
s
s
of
X
G
B
oos
t
in
our
s
tu
dy
und
e
r
s
c
or
e
s
it
s
pot
e
nt
ia
l
f
or
c
li
ni
c
a
l
a
ppl
ic
a
ti
on,
pr
ovi
di
ng
phys
ic
ia
ns
w
it
h
a
r
obus
t
to
ol
f
or
a
s
s
e
s
s
in
g
pa
ti
e
nt
pr
ognos
is
.
T
hi
s
a
li
gns
w
it
h
th
e
s
ugge
s
ti
on
by
T
hom
gka
m
e
t
al
.
[
28]
a
nd
ot
he
r
r
e
s
e
a
r
c
he
r
s
th
a
t
pr
e
di
c
ti
ve
m
o
de
ls
c
a
n
s
uf
f
ic
ie
nt
ly
a
id
in
e
a
r
ly
r
is
k s
c
r
e
e
ni
ng
a
nd
th
e
f
or
m
ul
a
ti
on
of
m
or
e
e
f
f
e
c
ti
ve
tr
e
a
tm
e
nt
pl
a
ns
.
O
ur
r
e
s
e
a
r
c
h
c
ont
r
ib
ut
e
s
to
th
e
ongoing
de
ve
lo
pm
e
nt
of
c
li
ni
c
a
l
de
c
is
io
n
s
uppor
t
s
ys
te
m
s
,
e
nha
nc
in
g
th
e
pr
e
c
is
io
n
a
nd
r
e
li
a
bi
li
ty
of
s
ur
vi
va
l
pr
e
di
c
ti
ons
f
or
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
.
O
-
X
G
B
oos
t
is
a
pow
e
r
f
ul
to
ol
f
or
s
ur
vi
va
l
a
na
ly
s
is
,
w
hi
le
C
N
N
c
a
n
a
ls
o
be
e
f
f
e
c
ti
ve
,
pa
r
ti
c
ul
a
r
ly
w
it
h
im
a
ge
or
s
pa
ti
a
l
da
ta
.
C
N
N
e
xc
e
l
a
t
e
xt
r
a
c
ti
ng
e
s
s
e
nt
ia
l
f
e
a
tu
r
e
s
f
r
om
c
om
pl
e
x
da
ta
s
e
ts
,
m
a
ki
ng
th
e
m
s
ui
ta
bl
e
f
or
a
ppl
ic
a
ti
ons
l
ik
e
m
e
di
c
a
l
im
a
gi
ng a
nd ge
ogr
a
phi
c
d
a
ta
a
na
ly
s
is
. H
o
w
e
ve
r
, C
N
N
ge
ne
r
a
ll
y r
e
qui
r
e
la
r
ge
r
da
ta
s
e
ts
a
nd
gr
e
a
te
r
c
om
put
a
ti
ona
l
r
e
s
our
c
e
s
f
or
e
f
f
e
c
ti
ve
tr
a
in
in
g
[
31]
.
I
n
c
ont
r
a
s
t,
O
-
X
G
B
oos
t
is
ty
pi
c
a
ll
y m
or
e
i
nt
e
r
p
r
e
ta
bl
e
, ha
ndl
in
g t
a
bul
a
r
da
ta
e
f
f
ic
ie
nt
ly
a
nd of
te
n r
e
qui
r
in
g
f
e
w
e
r
r
e
s
our
c
e
s
. T
he
c
hoi
c
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
2146
-
2157
2154
be
twe
e
n
O
-
X
G
B
oos
t
a
nd
C
N
N
ul
ti
m
a
te
ly
de
pe
nds
on
th
e
pr
obl
e
m
,
da
ta
c
ha
r
a
c
te
r
is
ti
c
s
,
a
nd
r
e
qui
r
e
d
in
te
r
pr
e
ta
bi
li
ty
. I
n T
a
bl
e
6, i
t’
s
a
ls
o note
d t
ha
t
th
is
e
xpe
r
im
e
nt
a
l
c
om
pa
r
is
on w
it
h pr
e
vi
ous
s
tu
di
e
s
i
s
ba
s
e
d on
di
f
f
e
r
in
g
da
ta
s
e
ts
,
s
o
no
de
f
in
it
iv
e
c
onc
lu
s
io
n
c
a
n
be
m
a
de
a
b
out
w
hi
c
h
m
ode
l
is
s
upe
r
io
r
.
I
n
th
is
s
tu
dy,
w
e
a
ls
o
c
r
e
a
te
d
a
w
e
b
a
ppl
ic
a
ti
on
th
a
t
pr
e
di
c
t
s
th
e
s
ur
vi
va
l
ti
m
e
of
lu
ng
c
a
nc
e
r
p
a
ti
e
nt
s
u
s
in
g
th
e
O
-
X
G
B
oos
t
m
ode
ls
, a
s
i
ll
us
tr
a
te
d i
n F
ig
ur
e
3.
F
ig
ur
e
3. A
w
e
b
a
ppl
ic
a
ti
on of
l
ung c
a
nc
e
r
s
ur
vi
va
l
pr
e
di
c
ti
on
us
in
g O
-
X
G
B
oos
t
m
ode
l
5.
C
O
N
C
L
U
S
I
O
N
T
hi
s
r
e
s
e
a
r
c
h,
a
n
e
ns
e
m
bl
e
m
ode
l
is
de
ve
lo
pe
d
to
e
nha
nc
e
th
e
pr
e
di
c
ti
on
of
lu
ng
c
a
nc
e
r
pa
ti
e
nt
s
ur
vi
va
l
ti
m
e
s
by
in
c
or
por
a
ti
ng
bot
h
out
li
e
r
de
te
c
ti
on
a
nd
h
ype
r
-
pa
r
a
m
e
te
r
opt
im
iz
a
ti
on
te
c
hni
que
s
.
T
h
e
e
ns
e
m
bl
e
a
ppr
oa
c
h
c
om
bi
ne
s
m
ul
ti
pl
e
tr
a
in
in
g
it
e
r
a
ti
ons
o
f
ba
s
e
a
lg
or
it
hm
w
hi
c
h
e
a
c
h
tr
a
in
e
d
th
r
ough
m
ul
ti
pl
e
it
e
r
a
ti
ons
to
opt
im
iz
e
th
e
ir
p
e
r
f
or
m
a
nc
e
.
F
or
out
li
e
r
de
te
c
ti
on,
th
e
IF
a
lg
or
it
hm
is
e
m
pl
oye
d
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
L
ung c
anc
e
r
pat
ie
nt
s
s
ur
v
i
v
al
pr
e
di
c
ti
on us
in
g outl
ie
r
de
te
c
ti
on
and opti
m
iz
e
d X
G
B
oos
t
(
W
ir
ot
Y
ot
s
aw
at
)
2155
id
e
nt
if
y
a
nd
m
a
na
ge
da
ta
poi
nt
s
th
a
t
de
vi
a
te
f
r
om
th
e
nor
m
,
e
ns
ur
in
g
th
a
t
th
e
m
ode
l
is
tr
a
in
e
d
on
c
le
a
n
a
nd
r
e
le
va
nt
da
ta
.
A
ddi
ti
ona
ll
y,
B
a
ye
s
ia
n
opt
im
iz
a
ti
on
is
us
e
d
to
f
in
e
-
tu
ne
hype
r
-
pa
r
a
m
e
te
r
s
,
e
f
f
ic
ie
nt
ly
r
e
f
in
in
g
th
e
m
ode
l
pa
r
a
m
e
te
r
s
th
r
ough
it
e
r
a
ti
ve
upd
a
te
s
ba
s
e
d
on
p
e
r
f
or
m
a
nc
e
m
e
tr
ic
s
.
T
he
da
ta
s
e
t
f
or
th
is
s
tu
dy
is
s
our
c
e
d f
r
om
t
he
S
E
E
R
c
a
nc
e
r
da
ta
ba
s
e
, i
nc
lu
di
ng 196,031 r
e
c
or
ds
w
it
h 22 a
tt
r
ib
ut
e
s
c
ol
le
c
te
d be
twe
e
n 2004
a
nd
2015.
T
h
e
r
e
s
e
a
r
c
h
ut
il
iz
e
s
10
-
f
ol
d
c
r
os
s
-
va
li
da
ti
on
f
or
c
om
pr
e
he
ns
iv
e
m
ode
l
tr
a
in
in
g
a
nd
te
s
ti
ng
to
e
ns
ur
e
r
obus
t
pe
r
f
or
m
a
nc
e
e
va
lu
a
ti
on. T
he
e
ns
e
m
bl
e
m
ode
l’
s
e
f
f
e
c
ti
ve
ne
s
s
i
s
m
e
a
s
ur
e
d us
in
g m
e
tr
ic
s
s
uc
h a
s
R
M
S
E
,
M
S
E
,
R
²,
a
nd
M
A
E
.
R
e
s
ul
ts
in
di
c
a
te
th
a
t
th
e
X
G
B
oos
t
a
lg
or
it
hm
a
c
hi
e
ve
s
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
w
it
h
a
n
R
M
S
E
of
13.74,
s
ur
pa
s
s
in
g
ot
he
r
m
od
e
ls
a
nd
th
e
b
a
s
e
li
ne
-
X
G
B
oos
t.
T
hi
s
r
e
s
e
a
r
c
h
unde
r
s
c
or
e
s
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
ut
il
iz
in
g
th
e
e
ns
e
m
bl
e
m
e
th
ods
w
it
h
out
li
e
r
de
te
c
ti
on
a
nd
B
a
y
e
s
ia
n
opt
im
iz
a
ti
on
to
im
pr
ove
pr
e
di
c
ti
on
pe
r
f
or
m
a
nc
e
,
pr
ovi
di
ng
s
uf
f
ic
ie
nt
c
ont
r
ib
ut
io
ns
to
th
e
de
ve
lo
pm
e
nt
of
c
li
ni
c
a
l
de
c
i
s
io
n
s
uppor
t
s
ys
te
m
s
th
a
t
c
a
n
e
nha
nc
e
tr
e
a
tm
e
nt
pl
a
nni
ng
a
nd
pa
ti
e
nt
c
a
r
e
.
F
ut
ur
e
w
or
k
c
oul
d
e
xpl
or
e
m
or
e
c
om
pl
e
x
a
lg
or
it
hm
s
, t
hor
ough f
e
a
tu
r
e
s
e
le
c
ti
on, or
f
e
a
tu
r
e
w
e
ig
ht
in
g t
e
c
hni
que
s
t
o e
nha
nc
e
t
h
e
m
ode
l’
s
pe
r
f
or
m
a
nc
e
.
F
U
N
D
I
N
G
I
N
F
O
R
M
A
T
I
O
N
T
he
r
e
s
e
a
r
c
he
r
s
w
oul
d
li
ke
to
th
a
nk
th
e
R
e
s
e
a
r
c
h
a
nd
D
e
ve
lo
pm
e
nt
I
ns
ti
tu
te
,
P
hr
a
na
khon
S
i
A
yut
th
a
ya
R
a
ja
bha
t
U
ni
ve
r
s
it
y,
f
or
th
e
ir
r
e
s
e
a
r
c
h
f
undi
ng
s
uppor
t
(
f
in
a
nc
ia
l
c
ode
:
Z
03301010642
)
,
w
hi
c
h
gr
e
a
tl
y
c
ont
r
ib
ut
e
d
to
th
e
s
uc
c
e
s
s
f
ul
c
om
pl
e
ti
on
of
th
is
s
tu
dy.
S
pe
c
ia
l
th
a
nks
a
r
e
a
ls
o e
xt
e
nde
d
to
th
e
e
xpe
r
ts
f
or
t
he
ir
e
va
lu
a
ti
on of
t
he
r
e
s
e
a
r
c
h pr
opos
a
l
a
nd t
he
ir
va
lu
a
bl
e
r
e
c
om
m
e
nda
ti
ons
f
or
i
m
pr
ovi
ng t
he
r
e
s
e
a
r
c
h.
A
U
T
H
O
R
C
O
N
T
R
I
B
U
T
I
O
N
S
S
T
A
T
E
M
E
N
T
T
hi
s
jo
ur
na
l
us
e
s
th
e
C
ont
r
ib
ut
or
R
ol
e
s
T
a
xonomy
(
C
R
e
di
T
)
to
r
e
c
ogni
z
e
in
di
vi
dua
l
a
ut
hor
c
ont
r
ib
ut
io
ns
, r
e
duc
e
a
ut
hor
s
hi
p di
s
put
e
s
,
a
nd f
a
c
il
it
a
te
c
ol
la
bo
r
a
ti
on.
N
am
e
o
f
A
u
t
h
or
C
M
So
Va
Fo
I
R
D
O
E
Vi
Su
P
Fu
W
ir
ot
Y
ot
s
a
w
a
t
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
P
e
e
ti
pha
r
t
S
ue
bpe
ng
✓
✓
✓
✓
✓
✓
✓
✓
✓
✓
S
a
r
oc
h
P
ur
is
a
ngka
ha
✓
✓
✓
A
ka
r
a
pon P
oons
a
w
a
d
✓
✓
✓
K
a
nya
la
g P
hodong
✓
✓
✓
✓
✓
✓
✓
✓
✓
C
:
C
onc
e
pt
ua
l
i
z
a
t
i
on
M
:
M
e
t
hodol
ogy
So
:
So
f
t
w
a
r
e
Va
:
Va
l
i
da
t
i
on
Fo
:
Fo
r
m
a
l
a
na
l
ys
i
s
I
:
I
nve
s
t
i
ga
t
i
on
R
:
R
e
s
our
c
e
s
D
:
D
a
t
a
C
ur
a
t
i
on
O
:
W
r
i
t
i
ng
-
O
r
i
gi
na
l
D
r
a
f
t
E
:
W
r
i
t
i
ng
-
R
e
vi
e
w
&
E
di
t
i
ng
Vi
:
Vi
s
ua
l
i
z
a
t
i
on
Su
:
Su
pe
r
vi
s
i
on
P
:
P
r
oj
e
c
t
a
dm
i
ni
s
t
r
a
t
i
on
Fu
:
Fu
ndi
ng a
c
qui
s
i
t
i
on
C
O
N
F
L
I
C
T
O
F
I
N
T
E
R
E
S
T
S
T
A
T
E
M
E
N
T
A
ut
hor
s
s
ta
te
no c
onf
li
c
t
of
i
nt
e
r
e
s
t.
D
A
T
A
A
V
A
I
L
A
B
I
L
I
T
Y
T
he
da
ta
s
uppor
ti
ng
th
e
f
in
di
ngs
of
th
is
s
tu
dy
a
r
e
publ
ic
ly
a
c
c
e
s
s
ib
le
th
r
ough
th
e
S
E
E
R
*S
ta
t
a
ppl
ic
a
ti
on
[
30]
, a
va
il
a
bl
e
on t
he
of
f
ic
ia
l
w
e
bs
it
e
:
ht
tp
s
:/
/s
e
e
r
.c
a
nc
e
r
.gov.
R
E
F
E
R
E
N
C
E
S
[
1]
B
.
S
.
C
hhi
ka
r
a
a
nd
K
.
P
a
r
a
ng,
“
G
l
oba
l
c
a
nc
e
r
s
t
a
t
i
s
t
i
c
s
2022:
t
he
t
r
e
nds
pr
oj
e
c
t
i
on
a
na
l
ys
i
s
,”
C
h
e
m
i
c
al
B
i
ol
ogy
L
e
t
t
e
r
s
,
vol
.
10,
no. 1, pp. 1
–
16, 2022.
[
2]
J
.
D
.
M
i
nna
,
J
.
A
.
R
ot
h,
a
nd
A
.
F
.
G
a
z
da
r
,
“
F
oc
u
s
on
l
ung
c
a
nc
e
r
,”
C
anc
e
r
C
e
l
l
,
vol
.
1,
no.
1,
pp.
49
–
52,
2002,
doi
:
10.1016/
S
1535
-
6108(
02
)
00027
-
2.
[
3]
M
.
B
.
S
c
ha
ba
t
h
a
nd
M
.
L
.
C
ot
e
,
“
C
a
nc
e
r
pr
ogr
e
s
s
a
nd
pr
i
or
i
t
i
e
s
:
l
ung
c
a
nc
e
r
,”
C
anc
e
r
E
pi
de
m
i
ol
ogy
B
i
om
ar
k
e
r
s
and
P
r
e
v
e
nt
i
on
,
vol
. 28, no. 10, pp. 1563
–
1579, 2019, doi
:
10.1158/
1055
-
9965.E
P
I
-
19
-
0221.
[
4]
B
.
K
os
a
npi
pa
t
e
t
al
.
,
“
I
m
pa
c
t
of
P
M
2.5
e
xpos
ur
e
on
m
or
t
a
l
i
t
y
a
nd
t
um
or
r
e
c
ur
r
e
nc
e
i
n
r
e
s
e
c
t
a
bl
e
non
-
s
m
a
l
l
c
e
l
l
l
ung
c
a
r
c
i
nom
a
,”
Sc
i
e
nt
i
f
i
c
r
e
por
t
s
, vol
. 14, no. 1, 2024, doi
:
10.1038/
s
41598
-
024
-
76696
-
0.
[
5]
S
.
D
oppa
l
a
pudi
,
R
.
G
.
Q
i
u,
a
nd
Y
.
B
a
dr
,
“
L
ung
c
a
nc
e
r
s
ur
vi
va
l
pe
r
i
od
pr
e
di
c
t
i
on
a
nd
unde
r
s
t
a
ndi
ng:
de
e
p
l
e
a
r
ni
ng
a
ppr
oa
c
he
s
,”
I
nt
e
r
nat
i
onal
J
our
nal
of
M
e
di
c
al
I
nf
or
m
at
i
c
s
, vol
. 148, 2021, doi
:
10.1016/
j
.i
j
m
e
di
nf
.2020.104371.
[
6]
J
.
A
.
B
a
r
t
hol
om
a
i
a
nd
H
.
B
.
F
r
i
e
boe
s
,
“
L
ung
c
a
nc
e
r
s
ur
vi
va
l
pr
e
di
c
t
i
on
vi
a
m
a
c
hi
ne
l
e
a
r
ni
ng
r
e
gr
e
s
s
i
on,
c
l
a
s
s
i
f
i
c
a
t
i
on,
a
nd
s
t
a
t
i
s
t
i
c
a
l
t
e
c
hni
que
s
,”
i
n
2018
I
E
E
E
I
nt
e
r
nat
i
onal
Sy
m
pos
i
um
on
Si
gnal
P
r
oc
e
s
s
i
ng
and
I
nf
or
m
at
i
on
T
e
c
hnol
ogy
(
I
SSP
I
T
)
,
pp. 632
–
637
,
2018,
doi
:
10.1109/
I
S
S
P
I
T
.2018.8642753.
Evaluation Warning : The document was created with Spire.PDF for Python.