I
AE
S In
t
er
na
t
io
na
l J
o
urna
l o
f
Art
if
icia
l In
t
ellig
ence
(
I
J
-
AI
)
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
,
p
p
.
2247
~
2
2
5
9
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
3
.
p
p
2
2
4
7
-
2
2
5
9
2247
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
Stacking
ensembl
e t
ech
niques
for a
utom
a
ted
pe
riphe
ra
l bloo
d
cell clas
sifica
tion
using
Inception
v
3
f
ea
tures
M
a
rwa
M
a
wf
a
q M
o
ha
m
eds
heet
Al
-
Hatab
1
,
M
a
y
s
a
lo
o
n
Abed Q
a
s
im
2
,
Na
wa
r
A.
Su
lt
a
n
2
1
Te
c
h
n
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
C
o
l
l
e
g
e
,
N
o
r
t
h
e
r
n
T
e
c
h
n
i
c
a
l
U
n
i
v
e
r
s
i
t
y
,
M
o
s
u
l
,
I
r
a
q
2
Te
c
h
n
i
c
a
l
E
n
g
i
n
e
e
r
i
n
g
C
o
l
l
e
g
e
f
o
r
C
o
mp
u
t
e
r
a
n
d
A
r
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
N
o
r
t
h
e
r
n
Te
c
h
n
i
c
a
l
U
n
i
v
e
r
s
i
t
y
,
M
o
s
u
l
,
I
r
a
q
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
2
2
,
2
0
2
5
R
ev
is
ed
Feb
2
1
,
2
0
2
6
Acc
ep
ted
Ap
r
2
2
,
2
0
2
6
Ro
b
u
st
d
isti
n
c
ti
o
n
o
f
b
lo
o
d
c
e
ll
s
is
c
ru
c
ial
in
c
li
n
ica
l
e
v
a
l
u
a
ti
o
n
.
M
a
n
u
a
l
e
x
a
m
in
a
ti
o
n
is
sl
o
w
a
n
d
e
x
p
o
se
d
t
o
e
rr
o
rs.
T
h
is
w
o
rk
i
n
v
e
stig
a
tes
u
sin
g
m
a
c
h
in
e
lea
rn
in
g
(
ML
)
tec
h
n
i
q
u
e
s
fo
r
a
u
t
o
m
a
ted
c
las
sifica
ti
o
n
o
f
e
ig
h
t
c
a
teg
o
ries
o
f
p
e
rip
h
e
ra
l
b
l
o
o
d
c
e
ll
ty
p
e
s
fro
m
m
u
lt
i
-
c
o
lo
r
im
a
g
e
s.
Th
e
I
n
c
e
p
ti
o
n
v
3
n
e
two
rk
wa
s
u
se
d
t
o
e
x
trac
t
fe
a
tu
re
s,
a
sp
li
t
o
f
6
6
%
/
3
4
%
we
re
u
se
d
to
e
v
a
l
u
a
te
th
e
m
o
d
e
l
a
lo
n
g
with
2
0
-
f
o
ld
c
ro
ss
-
v
a
li
d
a
ti
o
n
.
To
re
d
u
c
e
c
o
m
p
u
tati
o
n
a
l
c
o
m
p
lex
it
y
,
p
rin
c
i
p
a
l
c
o
m
p
o
n
e
n
t
a
n
a
ly
sis
(
P
CA)
w
a
s
u
se
d
to
re
d
u
c
e
th
e
2
0
4
8
-
d
ime
n
si
o
n
a
l
fe
a
t
u
re
v
e
c
to
rs
to
1
0
0
c
o
m
p
o
n
e
n
ts.
Am
o
n
g
a
ll
c
las
sifiers
u
se
d
,
th
e
h
ig
h
e
st
p
e
rf
o
rm
a
n
c
e
with
o
u
t
u
sin
g
P
CA
wa
s
a
c
h
iev
e
d
u
sin
g
t
h
e
su
p
p
o
r
t
v
e
c
to
r
m
a
c
h
in
e
(S
VM)
with
a
n
a
c
c
u
ra
c
y
e
q
u
a
l
to
9
3
.
4
%
a
n
d
a
n
a
re
a
u
n
d
e
r
th
e
c
u
r
v
e
(AU
C)
o
f
0
.
9
9
6
.
Us
i
n
g
P
C
A,
a
ffe
c
ted
m
o
n
o
c
y
tes
a
n
d
imm
a
tu
re
g
ra
n
u
lo
c
y
tes
m
o
st d
u
e
to
th
e
slig
h
t
re
d
u
c
ti
o
n
i
n
t
h
e
a
c
c
u
ra
c
y
a
n
d
AU
C
wh
ich
b
e
c
a
m
e
9
0
.
1
%
a
n
d
to
0
.
9
9
1
re
sp
e
c
ti
v
e
l
y
.
Re
su
lt
s
we
re
fu
rth
e
r
e
n
h
a
n
c
e
d
wh
e
n
a
sta
c
k
e
d
e
n
se
m
b
le
o
f
n
e
u
ra
l
n
e
two
rk
(NN
)
,
lo
g
isti
c
re
g
re
ss
io
n
(LR)
,
a
n
d
S
VM
we
re
u
se
d
,
a
c
h
iev
in
g
a
n
a
c
c
u
ra
c
y
o
f
9
5
.
2
%
a
n
d
a
n
AU
C
o
f
0
.
9
9
8
.
Th
e
o
b
tain
e
d
fi
n
d
i
n
g
s
c
o
n
fi
rm
e
d
th
e
e
ffe
c
ti
v
e
n
e
ss
o
f
u
sin
g
sta
c
k
e
d
e
n
se
m
b
les
in
p
ro
v
id
i
n
g
a
ro
b
u
st,
h
ig
h
-
a
c
c
u
ra
c
y
fra
m
e
wo
rk
fo
r
a
u
t
o
m
a
ted
b
lo
o
d
c
e
ll
c
las
sifica
ti
o
n
,
w
h
il
e
P
CA
e
fficie
n
tl
y
re
d
u
c
e
d
d
ime
n
sio
n
s w
i
th
m
in
ima
l
p
e
rfo
rm
a
n
c
e
lo
ss
.
K
ey
w
o
r
d
s
:
C
r
o
s
s
v
alid
atio
n
E
n
s
em
b
le
lear
n
in
g
I
n
ce
p
tio
n
v
3
Ma
ch
in
e
lear
n
in
g
Prin
cip
al
co
m
p
o
n
en
t a
n
al
y
s
is
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
:
Ma
r
wa
Ma
wf
aq
Mo
h
am
e
d
s
h
e
et
Al
-
Hata
b
T
ec
h
n
ical
E
n
g
i
n
ee
r
in
g
C
o
lleg
e,
No
r
th
er
n
T
ec
h
n
ical
Un
i
v
er
s
ity
Mo
s
u
l,
I
r
aq
E
m
ail:
m
ar
wa.
alh
atab
@
n
tu
.
ed
u
.
iq
1.
I
NT
RO
D
UCT
I
O
N
B
lo
o
d
ce
ll
an
aly
s
is
is
an
es
s
en
tial
is
s
u
e
in
clin
ical
d
iag
n
o
s
tics
b
ec
au
s
e
it
p
r
o
v
id
es
s
ig
n
if
ican
t
in
f
o
r
m
atio
n
ab
o
u
t
im
m
u
n
e
s
t
atu
s
,
h
em
ato
lo
g
ical
d
is
o
r
d
er
s
,
an
d
d
is
ea
s
e
p
r
o
g
r
ess
io
n
[
1
]
.
Ho
wev
er
,
alth
o
u
g
h
th
at
th
e
tr
ad
itio
n
al
m
an
u
al
t
est
wh
ich
d
ep
en
d
o
n
m
icr
o
s
co
p
e
wid
esp
r
ea
d
,
it
s
u
f
f
er
s
f
r
o
m
in
ter
-
o
b
s
er
v
e
r
v
ar
iab
ilit
y
,
s
u
b
jectiv
ity
,
an
d
lo
w
th
r
o
u
g
h
p
u
t,
esp
ec
ially
in
r
eso
u
r
ce
-
co
n
s
tr
ain
ed
e
n
v
ir
o
n
m
e
n
ts
.
T
h
ese
d
is
ad
v
an
tag
es
led
to
th
e
d
e
v
elo
p
m
en
t
o
f
au
to
m
ated
a
n
d
s
em
i
-
au
to
m
ated
im
ag
e
-
b
ase
d
s
y
s
tem
s
th
at
u
s
e
co
m
p
u
ter
v
is
io
n
an
d
m
ac
h
in
e
lear
n
in
g
(
ML
)
to
im
p
r
o
v
e
r
e
p
r
o
d
u
cib
ilit
y
,
s
p
ee
d
,
an
d
d
ia
g
n
o
s
tic
ac
cu
r
ac
y
[
2
]
.
E
ar
ly
co
m
p
u
tatio
n
al
m
eth
o
d
s
u
s
ed
th
e
co
m
b
in
atio
n
o
f
h
a
n
d
cr
af
ted
f
ea
tu
r
es
s
u
ch
as
s
h
ap
e
,
tex
tu
r
e,
a
n
d
co
l
o
r
an
d
class
ical
m
ac
h
in
e
-
lear
n
in
g
alg
o
r
ith
m
s
s
u
ch
as
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVMs)
o
r
r
a
n
d
o
m
f
o
r
ests
(
R
Fs
)
.
T
h
ese
tech
n
iq
u
es
we
r
e
v
e
r
y
s
u
cc
ess
f
u
l
in
s
im
p
le
class
if
icatio
n
task
s
,
b
u
t
th
ey
f
ailed
in
c
o
m
p
lex
m
o
r
p
h
o
lo
g
ical
v
a
r
iatio
n
s
,
o
v
e
r
lap
p
in
g
ce
lls
,
an
d
s
tain
in
g
in
c
o
n
s
is
ten
cies
[
3
]
,
[
4
]
.
I
n
co
n
tr
ast,
m
o
d
er
n
ad
v
a
n
ce
s
in
d
ee
p
lear
n
in
g
,
h
elp
e
d
th
e
class
if
icatio
n
s
y
s
tem
s
an
d
h
av
e
b
ec
o
m
e
th
e
m
ain
p
ar
ad
ig
m
f
o
r
id
e
n
tify
in
g
r
ed
b
lo
o
d
ce
lls
(
R
B
C
s
)
,
wh
ite
b
l
o
o
d
ce
lls
(
W
B
C
s
)
,
an
d
p
latelets
[
5
]
.
Mo
r
eo
v
er
,
u
s
in
g
th
e
c
o
m
b
i
n
atio
n
o
f
en
s
em
b
le
lear
n
in
g
s
tr
ateg
ies,
f
o
r
ex
am
p
le
th
o
s
e
u
s
in
g
m
u
ltip
le
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
2
4
7
-
2
2
5
9
2248
c
o
n
v
o
l
u
tio
n
al
n
e
u
r
al
n
etwo
r
k
(
C
NN
)
ar
ch
itectu
r
es
in
class
if
icatio
n
s
y
s
tem
s
im
p
r
o
v
e
d
r
o
b
u
s
tn
ess
,
ac
cu
r
ac
y
,
an
d
g
e
n
er
aliza
tio
n
p
e
r
f
o
r
m
an
ce
co
m
p
ar
e
d
to
s
in
g
le
-
m
o
d
el
ap
p
r
o
ac
h
es
[
6
]
.
Dim
en
s
io
n
a
lity
r
ed
u
ctio
n
a
n
d
s
elec
tio
n
ap
p
r
o
ac
h
es
s
u
ch
as
p
r
in
cip
al
co
m
p
o
n
e
n
t
an
aly
s
is
(
PC
A)
,
p
ar
ticle
s
war
m
o
p
tim
izatio
n
(
PS
O)
,
en
tr
o
p
y
m
in
im
izatio
n
,
a
n
d
m
em
etic
alg
o
r
ith
m
s
ar
e
o
th
er
t
ec
h
n
iq
u
es
a
p
p
lied
to
d
ee
p
f
ea
tu
r
e
em
b
ed
d
in
g
s
in
o
r
d
er
to
d
ec
r
ea
s
e
r
e
d
u
n
d
an
cy
an
d
k
ee
p
d
is
cr
im
in
ativ
e
in
f
o
r
m
atio
n
[
7
]
.
I
n
ad
d
itio
n
to
th
es
e
m
eth
o
d
s
,
d
o
m
ain
ad
ap
tatio
n
an
d
d
ata
au
g
m
en
ta
tio
n
tech
n
iq
u
es
r
a
n
g
in
g
f
r
o
m
s
tain
n
o
r
m
aliza
tio
n
to
s
em
i
-
s
u
p
er
v
is
ed
lear
n
i
n
g
p
lay
ed
a
b
ig
r
o
le
in
p
er
f
o
r
m
an
ce
im
p
r
o
v
em
en
t,
esp
ec
ially
f
o
r
r
ar
e
ce
ll
ty
p
es
an
d
d
atasets
f
r
o
m
d
if
f
er
en
t
s
o
u
r
ce
s
[
8
]
.
M
u
lti
-
class
b
lo
o
d
-
ce
ll
class
if
icat
io
n
s
y
s
tem
s
w
h
ich
u
s
e
h
ig
h
-
ca
p
ac
ity
b
ac
k
b
o
n
es
s
u
ch
as
R
e
s
Ne
t
an
d
E
f
f
icien
tNet
to
g
eth
e
r
with
en
s
em
b
le
m
eth
o
d
s
an
d
a
d
v
an
ce
d
f
ea
tu
r
e
an
aly
s
is
ac
h
iev
ed
class
if
icatio
n
ac
cu
r
ac
ies
ex
ce
ed
in
g
9
9
%
[
9
]
.
Desp
ite
all
o
f
th
e
ab
o
v
e
b
lo
o
d
ce
ll
class
if
icat
io
n
s
y
s
tem
s
h
av
e
m
an
ch
allen
g
es
n
o
wad
ay
s
,
s
u
c
h
as
h
an
d
lin
g
im
ag
e
n
o
is
e,
en
s
u
r
in
g
r
ea
l
-
tim
e
d
ep
lo
y
m
en
t,
e
n
h
an
ci
n
g
in
ter
p
r
etab
ilit
y
in
clin
ical
co
n
tex
ts
,
an
d
ac
h
iev
i
n
g
m
o
d
el
g
en
er
aliza
tio
n
ac
r
o
s
s
lab
o
r
ato
r
ies
an
d
im
ag
in
g
p
r
o
to
co
ls
[
1
0
]
.
I
n
th
is
wo
r
k
a
n
o
v
el
f
r
am
ewo
r
k
co
m
b
in
in
g
ad
v
an
ce
d
f
ea
tu
r
e
d
im
en
s
io
n
ality
an
aly
s
is
with
e
n
s
em
b
le
lear
n
in
g
to
o
b
tain
h
ig
h
ac
c
u
r
ac
y
,
c
o
m
p
ac
t
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
,
an
d
p
r
ac
tical
v
alu
e
in
r
ea
l
-
wo
r
l
d
clin
ical
s
ettin
g
s
h
as
b
ee
n
p
r
o
p
o
s
ed
.
R
ec
en
tly
d
ee
p
lear
n
in
g
,
en
s
em
b
le
m
eth
o
d
s
,
a
n
d
f
ea
tu
r
e
o
p
tim
izatio
n
tech
n
iq
u
es
h
av
e
b
ee
n
u
s
ed
wid
ely
in
m
o
d
e
r
n
b
lo
o
d
ce
ll
class
if
icatio
n
to
ac
h
iev
e
r
o
b
u
s
tn
ess
,
an
d
h
ig
h
ac
c
u
r
ac
y
.
Par
ay
il
et
a
l
.
[
1
1
]
p
r
o
p
o
s
ed
a
h
ier
ar
c
h
ical
ar
tifi
cial
in
tellig
en
ce
(
AI
)
p
ip
elin
e
f
o
r
m
u
lti
-
class
R
B
C
m
o
r
p
h
o
lo
g
y
a
n
aly
s
is
,
th
eir
m
eth
o
d
illu
s
tr
ates
th
e
p
o
ten
t
ial
o
f
s
tag
ed
class
if
icatio
n
m
o
d
els
.
B
an
er
jee
an
d
C
h
au
d
h
u
r
i
in
t
r
o
d
u
ce
d
a
f
r
am
ewo
r
k
b
ased
o
n
ap
p
ly
in
g
two
-
s
tag
e
f
ea
tu
r
e
s
elec
tio
n
b
a
s
ed
o
n
to
tal
co
n
tr
ib
u
tio
n
s
co
r
e
an
d
f
u
zz
y
e
n
tr
o
p
y
,
th
eir
wo
r
k
ac
h
iev
ed
h
ig
h
ac
c
u
r
ac
y
an
d
r
ed
u
ce
d
f
ea
tu
r
e
d
im
e
n
s
io
n
ality
.
Usi
n
g
en
s
em
b
le
lear
n
in
g
in
class
if
icat
io
n
s
y
s
tem
f
u
r
th
er
en
h
an
ce
d
class
if
icatio
n
p
e
r
f
o
r
m
a
n
ce
.
Gh
o
s
h
et
a
l.
[
1
3
]
a
ch
ie
v
e
d
9
6
.
6
7
%
ac
c
u
r
a
cy
f
o
r
l
e
u
k
o
c
y
t
e
c
lass
i
f
i
ca
t
io
n
b
in
te
g
r
ati
n
g
De
n
s
eNe
t
1
2
1
,
R
esNet
5
0
,
a
n
d
a
c
u
s
t
o
m
C
N
N
v
i
a
a
leas
t
en
tr
o
p
y
c
o
m
b
i
n
e
r
.
I
c
h
i
m
et
a
l
.
[
1
4
]
p
r
o
p
o
s
e
d
a
m
eth
o
d
o
e
n
h
a
n
c
e
h
e
r
o
b
u
s
t
n
ess
ac
r
o
s
s
e
ig
h
t
b
l
o
o
d
ce
ll
cl
ass
es
u
s
i
n
g
d
ec
is
i
o
n
f
u
s
io
n
a
cr
o
s
s
m
u
lti
p
l
e
n
e
two
r
k
s
(
VGG1
6
,
X
ce
p
ti
o
n
,
R
esNet
5
0
,
a
n
d
NasN
etL
a
r
g
e)
.
T
en
g
u
a
m
et
a
l
.
[
1
5
]
in
tr
o
d
u
c
ed
a
m
o
d
e
l
wh
ic
h
in
te
g
r
at
e
d
e
ep
a
n
d
h
an
d
c
r
af
t
e
d
f
e
at
u
r
es
wit
h
e
n
s
e
m
b
le
le
a
r
n
i
n
g
f
o
r
H
&
E
h
is
t
o
p
at
h
o
l
o
g
y
i
m
a
g
es
.
R
eRNet
an
d
R
o
u
t
t
et
a
l
.
[
1
6
]
u
s
e
d
C
N
N
en
s
em
b
l
es
an
d
s
e
g
m
e
n
ta
ti
o
n
-
b
ase
d
t
r
ac
k
i
n
g
f
o
r
R
B
C
m
o
r
p
h
o
lo
g
y
an
al
y
s
is
,
t
h
ey
o
b
tai
n
e
d
9
9
.
9
7
%
a
cc
u
r
ac
y
.
Featu
r
e
o
p
tim
izatio
n
u
s
in
g
m
eta
-
h
eu
r
is
tics
,
en
tr
o
p
y
co
n
tr
o
l,
PC
A,
an
d
PS
O
ar
e
o
th
er
t
ec
h
n
iq
u
es
wh
ich
p
r
o
v
ed
ef
f
ec
tiv
en
ess
.
Ah
m
ad
et
a
l.
[
1
7
]
p
r
o
p
o
s
ed
a
W
B
C
class
if
icatio
n
u
s
in
g
s
ig
n
if
ican
t
f
ea
tu
r
e
r
ed
u
ctio
n
,
ac
h
iev
in
g
9
9
.
9
%
a
cc
u
r
ac
y
.
Awa
is
et
a
l
.
[
1
8
]
in
t
r
o
d
u
ce
d
an
ac
u
te
ly
m
p
h
o
b
last
ic
leu
k
em
ia
(
AL
L
)
s
u
b
ty
p
e
u
s
in
g
m
em
etic
d
ee
p
f
ea
tu
r
e
o
p
tim
izatio
n
,
h
eir
m
eth
o
d
o
b
tain
e
d
ap
p
r
o
x
im
at
ely
9
9
%
ac
cu
r
ac
y
.
C
ai
et
a
l.
[
1
9
]
p
r
o
p
o
s
ed
a
f
r
am
ewo
r
k
co
m
b
in
in
g
lo
w
-
r
a
n
k
ad
ap
tio
n
(
L
o
R
A)
-
ad
a
p
ted
,
s
eg
m
en
t
an
y
th
in
g
m
o
d
els
(
SAM)
with
u
n
s
u
p
e
r
v
i
s
ed
au
to
en
co
d
er
s
f
o
r
cr
o
s
s
-
d
o
m
ain
s
in
g
le
-
ce
ll im
ag
e
class
if
icatio
n
.
Sh
ar
m
a
[
2
0
]
d
e
v
elo
p
e
d
a
C
N
N
m
o
d
el
f
o
r
m
u
lti
-
class
b
lo
o
d
ce
ll
id
en
tific
atio
n
ac
r
o
s
s
b
en
ig
n
an
d
m
alig
n
an
t
ca
teg
o
r
ies.
His
m
o
d
el
r
ec
o
r
d
ed
m
o
r
e
th
an
9
8
%
ac
cu
r
ac
y
.
T
ep
ak
h
an
et
a
l.
[
2
1
]
p
r
o
p
o
s
ed
a
s
y
s
tem
to
d
if
f
er
en
tiate
b
etwe
en
ir
o
n
d
ef
icien
cy
an
em
ia
an
d
th
alass
em
ia,
th
ey
ap
p
lied
e
n
s
em
b
le
s
tack
in
g
with
R
F
an
d
g
r
ad
ien
t
b
o
o
s
tin
g
alg
o
r
ith
m
s
.
T
h
eir
s
y
s
tem
d
em
o
n
s
tr
ated
h
ig
h
class
if
icatio
n
p
er
f
o
r
m
an
ce
an
d
th
e
ef
f
ec
tiv
en
ess
o
f
en
s
em
b
le
-
b
a
s
ed
m
eth
o
d
s
f
o
r
co
m
p
lex
cli
n
ical
d
atasets
.
Uzm
a
an
d
Af
s
ar
[
2
2
]
p
r
o
p
o
s
ed
a
s
em
i
-
s
u
p
er
v
is
ed
f
r
am
ewo
r
k
to
im
p
r
o
v
e
W
B
C
s
u
b
ty
p
e
class
if
icatio
n
with
m
in
im
al
lab
eled
d
ata
.
Desp
ite
th
ese
ad
v
an
ce
s
,
b
alan
cin
g
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
f
ea
tu
r
e
d
im
en
s
io
n
ality
,
an
d
class
-
lev
el
d
is
cr
im
in
ativ
e
p
o
wer
s
till
ch
allen
g
es.
T
h
is
s
tu
d
y
a
d
d
r
ess
ed
th
ese
d
if
f
icu
lties
b
y
c
o
m
b
in
in
g
I
n
ce
p
tio
n
V3
f
ea
tu
r
e
ex
tr
ac
tio
n
with
a
s
tack
ed
en
s
em
b
le
o
f
n
eu
r
al
n
etwo
r
k
(
NN
)
,
lo
g
is
tic
r
eg
r
ess
io
n
(
LR
)
,
an
d
SVM.
T
h
e
ef
f
ec
tiv
en
ess
o
f
ap
p
l
y
in
g
PC
A
o
n
class
if
icatio
n
p
er
f
o
r
m
an
ce
was
s
y
s
tem
atica
l
ly
ev
alu
ated
.
C
o
m
p
ar
ed
t
o
th
e
ea
r
ly
s
tu
d
ies,
th
is
ap
p
r
o
ac
h
o
b
tain
s
h
ig
h
er
ac
cu
r
ac
y
a
n
d
r
o
b
u
s
tn
ess
,
h
ig
h
lig
h
tin
g
th
e
ad
v
an
tag
es
o
f
in
teg
r
atin
g
d
ee
p
-
f
ea
t
u
r
e
an
d
e
n
s
em
b
le
lear
n
in
g
in
b
lo
o
d
ce
ll
class
if
icatio
n
.
2.
M
E
T
H
O
D
AND
M
AT
E
R
I
A
L
S
T
h
is
s
tu
d
y
p
r
o
p
o
s
ed
b
lo
o
d
ce
ll
class
if
ica
tio
n
s
y
s
tem
b
ased
o
n
ML
,
u
s
in
g
I
n
ce
p
tio
n
v
3
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
.
T
h
e
s
y
s
tem
ev
alu
a
tes
th
e
ef
f
ec
tiv
en
ess
o
f
PC
A
f
o
r
d
im
en
s
io
n
ality
r
e
d
u
ctio
n
.
Fo
u
r
class
if
ier
s
:
L
R
,
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
,
NN,
an
d
SVM
wer
e
u
s
ed
w
ith
a
6
6
%/3
4
%
tr
ain
in
g
-
test
in
g
d
iv
is
io
n
an
d
2
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
.
T
wo
d
if
f
er
en
t
test
s
:
p
air
ed
t
-
test
s
an
d
W
ilc
o
x
o
n
s
ig
n
e
d
-
r
a
n
k
wer
e
u
s
ed
t
o
ass
ess
th
e
im
p
ac
t
o
f
PC
A
o
n
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
m
o
d
el.
Fin
ally
,
a
s
tack
ed
en
s
em
b
le
o
f
th
e
b
est
p
er
f
o
r
m
a
n
ce
class
if
ier
s
wa
s
im
p
lem
en
ted
to
im
p
r
o
v
e
th
e
a
cc
u
r
ac
y
.
Fig
u
r
e
1
illu
s
tr
ates th
e
co
m
p
lete
wo
r
k
f
lo
w.
2
.
1
.
I
ma
g
e
a
cquis
it
io
n
T
o
p
e
r
f
o
r
m
th
e
d
ata
s
et,
a
to
t
al
o
f
1
1
,
0
9
2
p
e
r
ip
h
er
al
b
lo
o
d
ce
ll
im
ag
es
(
3
6
0
×3
6
3
p
ix
els,
J
PG)
wer
e
co
llected
u
s
in
g
th
e
C
ellaVisio
n
DM
9
6
a
n
aly
ze
r
at
B
ar
ce
lo
n
a
Ho
s
p
ital
C
lin
ic.
T
h
e
im
ag
es
ca
teg
o
r
ized
eig
h
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ta
ck
in
g
e
n
s
emb
le
tech
n
iq
u
es
fo
r
a
u
to
ma
ted
p
erip
h
era
l b
lo
o
d
…
(
Ma
r
w
a
Ma
w
fa
q
Mo
h
a
m
ed
s
h
ee
t A
l
-
Ha
ta
b
)
2249
class
es
:
n
eu
tr
o
p
h
ils
,
eo
s
in
o
p
h
ils
,
b
aso
p
h
ils
,
ly
m
p
h
o
cy
tes,
m
o
n
o
c
y
tes,
im
m
atu
r
e
g
r
an
u
lo
cy
tes,
er
y
th
r
o
b
last
s
,
an
d
p
latelets
[
2
3
]
.
T
h
e
ce
ll
s
ize
d
is
tr
ib
u
tio
n
s
s
h
o
w
th
at
p
latelets
ex
h
ib
it
th
e
m
o
s
t
u
n
if
o
r
m
m
o
r
p
h
o
lo
g
y
,
wh
er
ea
s
m
o
n
o
c
y
tes an
d
im
m
a
tu
r
e
g
r
a
n
u
lo
cy
tes d
is
p
lay
t
h
e
g
r
ea
test
v
ar
iab
ilit
y
.
2
.
2
.
I
ncept
io
n v
3
cla
s
s
if
ica
t
i
o
n
I
n
ce
p
t
io
n
v
3
is
d
e
e
p
C
N
N
wit
h
4
2
-
la
y
er
d
e
v
e
lo
p
ed
f
o
r
i
m
a
g
e
r
e
c
o
g
n
i
ti
o
n
b
y
Go
o
g
l
e
R
ese
ar
ch
[
2
4
]
,
an
d
is
p
r
et
r
ai
n
e
d
o
n
I
m
a
g
eN
e
t
f
o
r
1
,
0
0
0
c
lass
es
.
I
t
im
p
r
o
v
e
s
class
if
ica
ti
o
n
,
d
ete
cti
o
n
[
2
5
]
,
a
n
d
s
eg
m
e
n
t
ati
o
n
p
e
r
f
o
r
m
a
n
c
e
[
2
6
]
.
I
ts
ar
ch
ite
c
tu
r
e
c
o
n
s
is
ts
o
f
t
h
r
e
e
m
ai
n
I
n
ce
p
ti
o
n
m
o
d
u
l
es:
A
,
B
,
an
d
C
wit
h
p
ar
all
el
1
×
1
,
3
×
3
,
a
n
d
5
×
5
c
o
n
v
o
lu
ti
o
n
s
,
s
u
p
p
o
r
te
d
b
y
c
o
n
v
o
lu
ti
o
n
,
p
o
o
l
in
g
,
a
n
d
f
u
l
ly
c
o
n
n
ec
t
e
d
la
y
er
s
wit
h
a
n
au
x
i
lia
r
y
class
i
f
i
er
u
s
e
d
f
o
r
r
eg
u
l
ar
iza
ti
o
n
.
F
i
g
u
r
e
2
s
h
o
ws
t
h
e
I
n
c
e
p
tio
n
v
3
ar
ch
ite
ct
u
r
e
.
Fa
ct
o
r
iz
e
d
an
d
as
y
m
m
et
r
i
c
co
n
v
o
lu
ti
o
n
s
a
r
e
u
s
e
d
t
o
d
ec
r
ea
s
e
p
a
r
a
m
et
er
s
a
n
d
en
h
an
ce
f
e
at
u
r
e
e
x
tr
ac
t
io
n
[
2
7
]
,
wh
ile
r
e
cti
f
ie
d
li
n
e
ar
u
n
it
(
R
eL
U
)
a
cti
v
a
ti
o
n
,
a
n
d
s
o
f
tm
a
x
la
y
e
r
a
r
e
u
s
ed
b
y
i
n
p
u
t
im
ag
es t
o
p
r
o
d
u
ce
p
r
o
b
a
b
il
ities
f
o
r
eig
h
t
t
ar
g
et
cl
ass
es.
Fig
u
r
e
1
.
Pro
p
o
s
ed
s
y
s
tem
f
o
r
b
lo
o
d
ce
ll c
lass
if
icatio
n
Fig
u
r
e
2
.
I
n
ce
p
tio
n
V3
ar
c
h
itectu
r
e
2
.
3
.
Cla
s
s
if
ica
t
io
n a
lg
o
rit
hm
in m
a
chine le
a
rning
Fo
u
r
ML
class
if
ier
s
wer
e
ev
alu
ated
:
L
R
,
NN,
SVM,
an
d
K
NN.
L
R
u
s
es
th
e
lo
g
is
tic
f
u
n
ctio
n
(
1
)
to
m
o
d
els th
e
p
r
o
b
ab
ilit
y
o
f
a
ca
t
eg
o
r
ical
r
esu
lt
[
2
8
]
,
[
2
9
]
.
(
)
=
(
0
+
1
∗
)
1
+
(
0
+
1
∗
)
(
1
)
W
h
er
e
(
)
is
p
r
ed
icted
o
u
tp
u
t,
0
is
in
ter
ce
p
t te
r
m
,
1
∗
co
ef
f
icien
t
f
o
r
th
e
s
in
g
le
in
p
u
t
v
alu
e
(
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
2
4
7
-
2
2
5
9
2250
T
h
e
NN
class
if
ier
is
a
m
u
ltil
a
y
er
p
e
r
ce
p
tr
o
n
co
n
tain
in
g
a
f
u
lly
co
n
n
ec
ted
h
id
d
e
n
lay
er
s
i
n
ad
d
itio
n
to
tu
n
a
b
le
p
a
r
am
eter
s
s
u
ch
as
ac
tiv
atio
n
f
u
n
ctio
n
s
,
s
o
lv
er
t
y
p
e,
a
n
d
tr
ain
in
g
ep
o
ch
s
[
3
0
]
,
[
3
1
]
.
SVM
m
o
d
el
,
class
if
ies
s
am
p
le
s
b
y
cr
ea
tin
g
a
m
ax
im
u
m
-
m
ar
g
i
n
h
y
p
er
p
lan
e
in
a
h
ig
h
-
d
im
en
s
io
n
al
f
ea
tu
r
e
s
p
ac
e
[
3
2
]
,
[
3
3
]
,
u
s
in
g
(
2
).
(
)
=
.
+
(
2
)
W
h
er
e
is
th
e
in
p
u
t
f
ea
tu
r
e,
is
weig
h
t
v
ec
to
r
,
an
d
is
th
e
b
ias
ter
m
.
KNN
d
ef
in
es
class
la
b
els
d
ep
en
d
in
g
on
KNNs
u
s
in
g
E
u
clid
ea
n
d
is
tan
ce
(
3
)
[
3
4
]
,
[
3
5
]
.
(
,
)
=
√
∑
(
−
)
2
=
1
(
3
)
W
h
er
e
,
is
t
h
e
f
ea
tu
r
e
v
ec
to
r
s
o
f
two
d
ata
p
o
in
ts
an
d
n
is
t
h
e
n
u
m
b
er
o
f
f
ea
tu
r
es.
2
.
4
.
P
rincipa
l
co
m
po
nent
a
na
ly
s
is
PC
A
is
a
lin
ea
r
d
im
en
s
io
n
ality
-
r
ed
u
ctio
n
tech
n
iq
u
e,
co
v
e
r
ts
h
ig
h
-
d
im
e
n
s
io
n
al
d
ata
t
o
a
s
m
aller
s
et
o
f
u
n
co
r
r
elate
d
co
m
p
o
n
en
ts
m
ain
tain
in
g
t
h
e
m
o
s
t
in
f
o
r
m
ativ
e
f
ea
tu
r
es.
I
n
th
is
s
t
u
d
y
,
PC
A
r
ed
u
ce
d
2
,
048
-
d
im
en
s
io
n
al
f
ea
tu
r
e
v
ec
to
r
s
ex
tr
ac
ted
b
y
I
n
ce
p
tio
n
v
3
to
1
0
0
p
r
in
cip
al
c
o
m
p
o
n
en
ts
c
alled
f
ea
tu
r
e
b
lo
o
d
ce
lls
(
FB
C
s
)
[
3
6
]
.
Featu
r
es we
r
e
f
ir
s
t stan
d
ar
d
ized
u
s
in
g
(
4
).
=
−
(
4
)
T
h
en
co
v
ar
ian
ce
b
etwe
en
f
ea
t
u
r
e
p
air
s
co
m
p
u
ted
u
s
in
g
(
5
).
(
1
,
2
)
=
∑
(
1
−
1
̀
)
=
1
(
2
−
2
̀
)
(
5
)
W
h
er
e
n
is
th
e
t
o
tal
n
u
m
b
er
o
f
s
am
p
les
an
d
1
̀
an
d
2
̀
is
th
e
m
e
an
v
al
u
e.
Fin
ally
,
th
e
p
r
i
n
cip
a
l
co
m
p
o
n
en
ts
o
b
tain
ed
u
s
in
g
th
e
eig
e
n
v
alu
e
(
6
)
.
=
(
6
)
W
h
er
e
is
elem
en
ts
o
f
f
ea
tu
r
es
as sq
u
ar
e
m
atr
ix
.
2
.
5
.
K
-
f
o
ld cr
o
s
s
v
a
lid
a
t
io
n
m
et
ho
d
K
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
d
i
v
id
e
s
th
e
d
ataset
in
to
n
f
o
ld
s
; e
ac
h
f
o
ld
is
u
s
ed
o
n
ce
as v
alid
atio
n
s
et
wh
ile
th
e
r
em
ain
in
g
ar
e
u
s
ed
as
tr
ain
in
g
d
ata.
T
h
is
p
r
o
ce
s
s
is
r
ep
ea
ted
till
all
f
o
ld
s
h
av
e
b
ee
n
u
s
ed
f
o
r
v
alid
atio
n
.
B
y
tr
ain
in
g
th
e
m
o
d
el
to
d
i
f
f
er
en
t
d
ata
d
is
tr
ib
u
tio
n
s
,
K
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
r
e
d
u
ce
s
th
e
p
r
o
b
a
b
ilit
y
o
f
o
v
er
f
itti
n
g
,
en
h
a
n
ce
s
g
en
er
ali
za
tio
n
,
an
d
p
r
o
v
id
es
a
m
o
r
e
s
tab
le
esti
m
atio
n
o
f
m
o
d
el
p
er
f
o
r
m
an
ce
b
ased
o
n
th
e
av
er
ag
e
ac
c
u
r
ac
y
ac
r
o
s
s
all
f
o
ld
s
[
3
7
]
.
2
.
6
.
P
er
f
o
r
m
a
nce
ev
a
lua
t
io
n m
et
rics
Stan
d
ar
d
m
etr
ics
wer
e
u
s
ed
to
ev
alu
ate
th
e
p
er
f
o
r
m
an
ce
o
f
th
e
p
r
o
p
o
s
ed
b
lo
o
d
ce
ll
cla
s
s
if
icatio
n
.
Acc
u
r
ac
y
as in
(
7
)
d
eter
m
in
es
th
e
o
v
er
all
co
r
r
ec
tn
ess
o
f
p
r
ed
ictio
n
s
.
=
+
+
+
+
(
7
)
R
ec
all
o
r
s
en
s
itiv
ity
as in
(
8
)
r
ef
er
s
to
th
e
p
r
o
p
o
r
tio
n
o
f
ac
tu
al
p
o
s
itiv
es c
o
r
r
ec
tly
id
en
tifie
d
.
=
+
(
8
)
Pre
cisi
o
n
as in
(
9
)
q
u
an
tifie
s
th
e
ac
cu
r
ac
y
o
f
p
o
s
itiv
e
p
r
e
d
ictio
n
s
.
=
+
(
9
)
F1
-
s
co
r
e
as in
(
1
0
)
is
th
e
h
ar
m
o
n
ic
m
ea
n
o
f
p
r
ec
is
io
n
an
d
r
e
ca
ll.
I
t is u
s
ef
u
l f
o
r
im
b
alan
ce
d
d
atasets
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ta
ck
in
g
e
n
s
emb
le
tech
n
iq
u
es
fo
r
a
u
to
ma
ted
p
erip
h
era
l b
lo
o
d
…
(
Ma
r
w
a
Ma
w
fa
q
Mo
h
a
m
ed
s
h
ee
t A
l
-
Ha
ta
b
)
2251
1
−
=
2
×
×
+
(
1
0
)
A
r
ea
u
n
d
e
r
th
e
c
u
r
v
e
(
AUC)
-
r
ec
eiv
er
o
p
er
atin
g
ch
a
r
ac
ter
is
tic
(
R
O
C
)
ev
alu
ates
th
e
a
b
ilit
y
o
f
m
o
d
el
to
d
is
tin
g
u
is
h
b
etwe
en
p
o
s
itiv
e
an
d
n
eg
ativ
e
class
es.
T
h
e
v
alu
es
o
f
AUC
r
an
g
in
g
f
r
o
m
0
.
5
wh
ich
r
ep
r
esen
ts
r
an
d
o
m
class
if
icatio
n
to
1
r
e
f
er
r
in
g
p
er
f
ec
t
class
if
icatio
n
,
wh
ile
th
e
R
OC
cu
r
v
e
p
lo
t
o
f
tr
u
e
p
o
s
itiv
e
r
ate
ag
ain
s
t th
e
f
alse p
o
s
itiv
e
r
ate
ac
r
o
s
s
d
if
f
er
en
t th
r
esh
o
ld
s
[
3
4
]
.
2
.
7
.
St
a
t
is
t
ica
l
s
ig
nifica
nce
a
na
ly
s
is
T
o
ass
ess
wh
eth
er
th
e
o
b
s
er
v
ed
im
p
r
o
v
em
en
ts
in
class
if
icatio
n
p
er
f
o
r
m
an
ce
af
te
r
ap
p
ly
in
g
PC
A
wer
e
s
tati
s
tically
s
ig
n
if
ican
t,
b
o
th
p
air
ed
t
-
test
an
d
W
ilco
x
o
n
s
ig
n
ed
-
r
a
n
k
test
wer
e
co
n
d
u
cted
.
T
h
ese
test
s
co
m
p
ar
e
th
e
p
er
f
o
r
m
an
ce
o
f
m
o
d
els
b
ef
o
r
e
an
d
af
te
r
PC
A
ac
r
o
s
s
m
u
ltip
le
cr
o
s
s
-
v
alid
atio
n
f
o
ld
s
,
d
eter
m
in
in
g
wh
eth
er
p
e
r
f
o
r
m
an
ce
g
ain
s
w
er
e
d
u
e
to
g
en
u
i
n
e
im
p
r
o
v
em
e
n
ts
r
ath
er
th
a
n
r
an
d
o
m
v
a
r
iatio
n
.
Fo
r
ea
ch
m
o
d
el,
class
if
icatio
n
ac
cu
r
ac
y
f
r
o
m
s
ev
er
al
in
d
ep
e
n
d
en
t
f
o
ld
s
was
u
s
ed
as p
air
ed
s
am
p
les
:
i)
Pair
ed
t
-
test
:
th
e
p
air
ed
t
-
test
ev
alu
ates
th
e
m
ea
n
d
if
f
e
r
en
ce
b
etwe
en
two
p
air
ed
s
am
p
les
u
n
d
er
th
e
ass
u
m
p
tio
n
o
f
n
o
r
m
al
d
is
tr
ib
u
tio
n
.
T
h
e
test
s
tatis
tic
is
co
m
p
u
ted
as
(
1
1
)
.
=
̅
∕
√
(
1
1
)
W
h
er
e
̅
is
th
e
m
ea
n
d
if
f
er
en
ce
b
etwe
en
p
air
ed
o
b
s
er
v
atio
n
s
,
is
co
m
p
u
ted
as
(
1
2
)
.
̅
=
1
∑
(
−
)
1
=
1
⋅
(
1
2
)
T
h
e
s
tan
d
er
d
e
v
iatio
n
o
f
d
if
f
er
en
ce
s
r
ep
r
esen
ts
b
y
is
co
m
p
u
t
ed
as
(
1
3
)
.
=
√
1
−
1
∑
(
(
=
1
−
)
−
̅
)
2
(
1
3
)
W
h
er
e
n
is
th
e
n
u
m
b
e
r
o
f
p
ai
r
ed
o
b
s
er
v
atio
n
s
(
f
o
ld
s
)
,
,
ar
e
th
e
ac
cu
r
a
cies
b
ef
o
r
e
an
d
a
f
ter
PC
A,
r
esp
ec
tiv
ely
.
A
s
m
all
p
-
v
alu
e
(
p
<0
.
0
5
)
in
d
icate
s
th
at
t
h
e
im
p
r
o
v
em
e
n
t
af
ter
PC
A
is
s
ta
tis
t
ically
s
ig
n
if
ican
t
[
3
8
]
.
ii)
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
:
th
e
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
i
s
a
n
o
n
-
p
a
r
am
etr
ic
alter
n
ativ
e
to
t
h
e
p
air
ed
t
-
test
,
wh
ich
d
o
es
n
o
t
ass
u
m
e
n
o
r
m
ality
.
I
t
e
v
alu
ates
wh
eth
e
r
th
e
m
ed
ian
o
f
d
if
f
er
en
ce
s
b
e
twee
n
p
air
ed
s
am
p
les is
ze
r
o
.
T
h
e
test
s
tatis
tic
is
ca
lcu
lated
as
(
1
4
)
.
=
∑
+
=
1
(
1
4
)
W
h
er
e
=
−
is
th
e
d
if
f
er
en
ce
b
etwe
en
p
air
ed
s
am
p
les,
+
is
th
e
r
an
k
o
f
th
e
ab
s
o
lu
te
d
if
f
er
en
ce
s
f
o
r
p
o
s
itiv
e
d
if
f
er
e
n
ce
s
(
>0
)
,
t
ies
ar
e
h
an
d
led
b
y
ass
ig
n
in
g
av
er
ag
e
r
an
k
s
.
A
s
m
all
p
-
v
alu
e
(
p
<0
.
0
5
)
in
d
icate
s
th
at
th
e
d
if
f
er
e
n
ce
s
a
r
e
s
ig
n
if
ican
t,
s
u
p
p
o
r
tin
g
th
at
PC
A
im
p
r
o
v
ed
p
er
f
o
r
m
a
n
ce
[
3
9
]
.
2
.
8
.
St
a
ck
ing
ens
em
ble m
e
t
ho
d
Stack
in
g
is
an
en
s
em
b
le
lear
n
in
g
ap
p
r
o
ac
h
th
at
im
p
r
o
v
es
th
e
class
if
icatio
n
p
er
f
o
r
m
an
ce
b
y
in
teg
r
atin
g
th
e
o
u
tp
u
ts
o
f
m
u
ltip
le
p
r
ed
ictiv
e
m
o
d
els.
I
ts
ar
ch
itectu
r
e
co
m
p
r
is
es o
f
two
lay
er
s
.
I
n
th
e
f
ir
s
t o
n
e
,
d
if
f
er
en
t
b
ase
m
o
d
els
ar
e
tr
ain
ed
s
ep
ar
ately
u
s
in
g
th
e
s
am
e
d
ataset,
ea
ch
m
o
d
el
ca
p
tu
r
e
d
if
f
er
en
t
ch
ar
ac
ter
is
tics
s
u
ch
as
lin
ea
r
r
elatio
n
s
h
ip
s
,
n
o
n
-
lin
ea
r
d
ec
is
io
n
b
o
u
n
d
ar
ies,
o
r
co
m
p
le
x
f
ea
tu
r
e
in
ter
ac
tio
n
s
.
T
h
en
th
e
class
if
icatio
n
s
wh
ich
g
en
er
ated
b
y
th
ese
m
o
d
els ar
e
f
o
r
war
d
e
d
as in
p
u
t t
o
th
e
s
ec
o
n
d
lay
e
r
[
4
0
]
.
T
h
e
s
ec
o
n
d
lay
er
co
n
s
is
ts
o
f
a
m
eta
-
m
o
d
el
wh
ich
lear
n
s
h
o
w
to
o
p
tim
ally
in
teg
r
ate
th
e
b
ase
-
m
o
d
el
o
u
tp
u
ts
.
T
h
is
m
o
d
el
co
llects
th
e
ad
v
an
ta
g
es
o
f
th
e
b
ase
class
if
ier
s
wh
ile
tr
ea
tin
g
th
eir
in
d
i
v
id
u
al
wea
k
n
ess
es.
T
o
en
s
u
r
e
g
e
n
er
aliza
tio
n
an
d
p
r
ev
e
n
t
o
v
er
f
itti
n
g
,
th
e
cr
o
s
s
-
v
alid
ated
p
r
e
d
ictio
n
s
o
b
tain
ed
f
r
o
m
t
h
e
b
ase
m
o
d
els ar
e
u
s
ed
to
tr
ain
th
e
m
eta
-
m
o
d
el
o
n
u
n
s
ee
n
f
o
ld
s
[
4
1
]
.
Ma
th
em
atica
lly
,
th
e
s
tack
in
g
p
r
o
ce
s
s
ca
n
b
e
e
x
p
r
ess
ed
as
(
1
5
)
.
1
=
1
(
)
,
2
=
2
(
)
,
…
,
=
(
)
(
1
5
)
W
h
er
e
b
e
t
h
e
in
p
u
t
f
ea
tu
r
es,
an
d
1
,
2
,
…,
r
ep
r
esen
ts
th
e
n
b
a
s
e
m
o
d
els.
E
ac
h
b
ase
m
o
d
el
g
en
er
ates
p
r
ed
ictio
n
s
.
T
h
ese
p
r
ed
ictio
n
s
ar
e
co
n
ca
ten
ate
d
to
f
o
r
m
a
n
e
w
f
ea
tu
r
e
s
et
(
1
6
)
.
=
[
1
,
2
,
…
,
]
(
1
6
)
A
m
eta
-
m
o
d
el
is
th
en
tr
ain
ed
o
n
to
p
r
o
d
u
ce
th
e
f
in
al
p
r
e
d
ic
tio
n
(
1
7
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
2
4
7
-
2
2
5
9
2252
̂
=
(
)
(
1
7
)
Stack
in
g
tech
n
i
q
u
es
en
h
an
ce
s
p
r
ed
ictiv
e
ac
cu
r
ac
y
b
y
tak
in
g
ad
v
an
tag
e
o
f
v
ar
io
u
s
lear
n
i
n
g
m
ec
h
an
is
m
s
an
d
in
teg
r
ated
th
em
in
to
a
u
n
if
ie
d
m
o
d
el,
p
r
o
v
id
in
g
r
o
b
u
s
tn
ess
,
r
ed
u
ce
o
v
er
f
itti
n
g
,
an
d
im
p
r
o
v
e
g
en
er
aliza
tio
n
th
r
o
u
g
h
cr
o
s
s
-
v
alid
atio
n
,
an
d
t
h
ey
u
s
u
ally
o
u
tp
er
f
o
r
m
in
d
iv
i
d
u
al
class
if
ier
s
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
h
e
r
esu
lts
illu
s
tr
ate
p
er
f
o
r
m
an
ce
d
if
f
er
e
n
ce
s
b
etwe
en
m
o
d
els
b
ef
o
r
e
an
d
af
ter
ap
p
ly
in
g
PC
A.
T
h
e
h
ig
h
est
ac
cu
r
ac
y
was
ac
h
iev
e
d
b
y
SVM
ac
r
o
s
s
m
o
s
t
ce
ll
ty
p
es,
an
d
th
e
s
tack
ed
en
s
em
b
le
f
u
r
th
er
im
p
r
o
v
ed
o
v
er
all
p
er
f
o
r
m
a
n
ce
.
Deta
iled
f
in
d
in
g
s
ar
e
s
u
m
m
ar
ized
.
3
.
1
.
I
m
pa
c
t
o
f
f
ea
t
ure
dim
e
ns
io
na
lity
o
n m
o
del per
f
o
rma
nce
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
ML
class
if
ier
s
was
an
aly
ze
d
th
r
o
u
g
h
two
f
ea
tu
r
e
co
n
d
itio
n
s
:
f
u
ll
d
ee
p
f
ea
tu
r
e
with
2
,
048
-
D
an
d
,
PC
A
r
ed
u
ce
d
f
ea
tu
r
es
with
1
0
0
-
D.
R
esu
lts
wer
e
ass
e
s
s
ed
u
s
in
g
two
v
alid
atio
n
s
tr
ateg
ies:
a
6
6
%/3
4
% sp
lit.
I
n
ad
d
itio
n
,
2
0
-
f
o
ld
cr
o
s
s
-
v
alid
a
tio
n
was a
p
p
lied
f
o
r
ea
ch
f
ea
t
u
r
e
co
n
d
itio
n
.
3
.
1
.
1
.
P
er
f
o
r
m
a
nce
wit
ho
ut
P
CA
T
a
b
le
1
i
ll
u
s
tr
at
es
t
h
e
r
es
u
lts
a
ch
i
e
v
e
d
u
s
i
n
g
b
o
t
h
ev
al
u
at
io
n
s
tr
a
te
g
ies
wi
th
o
u
t
PC
A.
F
o
r
6
6
/
3
4
s
p
lit
,
th
e
h
i
g
h
est
p
er
f
o
r
m
an
ce
was
c
o
n
s
is
te
n
t
ly
a
ch
ie
v
e
d
b
y
SV
M
wi
th
9
2
.
6
%
ac
c
u
r
a
cy
a
n
d
an
AUC
o
f
0
.
9
9
5
,
f
o
ll
o
we
d
b
y
NN
an
d
L
R
wit
h
9
2
.
2
%
an
d
9
2
.
4
%
a
cc
u
r
ac
y
r
es
p
e
cti
v
e
ly
a
n
d
an
AUC
o
f
0
.
9
9
4
f
o
r
e
ac
h
class
i
f
i
er
.
I
n
co
n
t
r
as
t,
K
NN
o
b
t
ai
n
e
d
t
h
e
l
o
wes
t
p
er
f
o
r
m
a
n
ce
wi
th
7
3
.
4
%
a
cc
u
r
ac
y
,
0
.
9
3
5
AUC
d
u
e
t
o
h
ig
h
-
d
im
en
s
io
n
al
n
o
is
e
.
Si
m
il
ar
tr
en
d
s
w
er
e
o
b
s
er
v
e
d
u
s
i
n
g
2
0
-
f
o
l
d
c
r
o
s
s
-
v
ali
d
at
io
n
,
w
h
e
r
e
S
VM
a
ch
ie
v
e
d
9
3
.
4
%
a
cc
u
r
ac
y
w
it
h
an
A
UC
o
f
0
.
9
9
6
.
R
es
u
lts
c
o
n
f
i
r
m
ed
th
at
SV
M
ca
n
ef
f
e
cti
v
e
ly
h
a
n
d
les
h
i
g
h
-
d
i
m
en
s
i
o
n
a
l,
m
u
lti
-
cl
ass
f
ea
tu
r
es
,
wh
i
le
KNN
s
t
r
u
g
g
l
es i
n
t
h
is
is
s
u
e
.
3
.
1
.
2
.
P
er
f
o
r
m
a
nce
wit
h P
C
A
PC
A
r
ed
u
ce
d
th
e
f
ea
tu
r
e
s
p
a
ce
f
r
o
m
2
,
0
4
8
to
1
0
0
co
m
p
o
n
en
ts
,
r
esu
ltin
g
in
s
lig
h
tly
d
e
cr
ea
s
ed
in
ac
cu
r
ac
y
o
f
th
e
to
p
p
er
f
o
r
m
an
ce
class
if
ier
s
wh
ile
en
h
an
cin
g
co
m
p
u
tatio
n
al
e
f
f
icien
cy
.
Fo
r
6
6
/3
4
s
p
lit,
SVM
ac
h
iev
ed
8
9
.
3
%
ac
cu
r
ac
y
an
d
0
.
9
9
0
AUC,
NN
8
8
.
0
%
an
d
0
.
9
8
7
AUC,
L
R
8
7
.
8
%
a
n
d
0
.
9
8
4
AUC,
an
d
KNN
7
2
.
9
%
an
d
0
.
9
2
0
AUC.
I
n
2
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
,
SVM
a
ch
iev
ed
9
0
.
1
%
ac
cu
r
ac
y
an
d
0
.
9
9
1
AUC,
wh
ile
KNN
s
lig
h
tly
im
p
r
o
v
ed
to
7
4
.
2
%
ac
cu
r
ac
y
d
u
e
to
n
o
is
e
r
ed
u
ctio
n
i
n
th
e
d
is
tan
ce
-
b
a
s
ed
f
ea
tu
r
e
s
p
ac
e.
Desp
ite
th
e
s
lig
h
t
d
ec
r
ea
s
e,
SVM
r
em
ain
ed
t
h
e
to
p
-
p
er
f
o
r
m
in
g
class
if
ier
ac
r
o
s
s
all
PC
A
-
r
ed
u
ce
d
s
ettin
g
s
,
co
n
f
ir
m
in
g
its
r
eliab
ilit
y
f
o
r
b
lo
o
d
ce
ll
class
if
icatio
n
ev
en
with
r
ed
u
ce
d
d
im
en
s
io
n
ality
.
T
ab
le
2
s
u
m
m
ar
izes
th
e
co
m
p
lete
ev
al
u
atio
n
u
s
in
g
two
ev
alu
ated
s
tr
ateg
ies.
T
ab
le
1
.
ML
m
o
d
el
p
e
r
f
o
r
m
an
ce
with
o
u
t PC
A
(
6
6
%/3
4
% sp
lit v
s
.
2
0
-
f
o
l
d
C
V)
M
o
d
e
l
A
c
c
u
r
a
c
y
6
/
3
4
(
%)
A
c
c
u
r
a
c
y
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
(
%)
A
U
C
6
6
/
3
4
A
U
C
c
r
o
s
s
-
v
a
l
i
d
a
t
i
o
n
%
F1
-
s
c
o
r
e
6
6
/
3
4
F1
-
s
c
o
r
e
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
%
P
r
e
c
i
s
i
o
n
6
6
/
3
4
(
%)
P
r
e
c
i
s
i
o
n
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
%
R
e
c
a
l
l
6
6
/
3
4
(
%)
R
e
c
a
l
l
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
%
NN
9
2
.
2
9
3
.
1
0
.
9
9
4
0
.
9
9
5
9
2
.
2
9
3
.
1
9
2
.
2
9
3
.
1
9
2
.
2
9
3
.
1
LR
9
2
.
4
9
2
.
9
0
.
9
9
4
0
.
9
9
5
9
2
.
4
9
2
.
9
9
2
.
4
9
3
.
0
9
2
.
4
9
2
.
9
S
V
M
9
2
.
6
9
3
.
4
0
.
9
9
5
0
.
9
9
6
9
2
.
6
9
3
.
4
9
2
.
7
9
3
.
4
9
2
.
6
9
3
.
4
K
N
N
7
3
.
4
7
3
.
4
0
.
9
3
5
0
.
9
2
5
7
2
.
9
7
3
.
1
7
4
.
2
7
4
.
0
7
3
.
4
7
3
.
4
T
ab
le
2
.
ML
m
o
d
el
p
e
r
f
o
r
m
an
ce
with
PC
A
(
6
6
%/3
4
% sp
lit v
s
.
2
0
-
f
o
ld
C
V)
M
o
d
e
l
A
c
c
u
r
a
c
y
6
6
/
3
4
(
%)
A
c
c
u
r
a
c
y
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
(
%)
A
U
C
6
6
/
3
4
A
U
C
c
r
o
s
s
-
v
a
l
i
d
a
t
i
o
n
%
F1
-
s
c
o
r
e
6
6
/
3
4
F1
-
s
c
o
r
e
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
%
P
r
e
c
i
s
i
o
n
6
6
/
3
4
(
%)
P
r
e
c
i
s
i
o
n
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
%
R
e
c
a
l
l
6
6
/
3
4
(
%)
R
e
c
a
l
l
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
%
NN
8
8
.
0
8
8
.
1
0
.
9
8
7
0
.
9
8
9
8
7
.
9
8
8
.
0
8
8
.
0
8
8
.
1
8
8
.
0
8
8
.
1
LR
8
7
.
8
8
8
.
1
0
.
9
8
7
0
.
9
8
5
8
7
.
7
8
8
.
1
8
7
.
7
8
8
.
1
8
7
.
8
8
8
.
1
S
V
M
8
9
.
3
9
0
.
1
0
.
9
9
0
0
.
9
9
1
8
9
.
2
9
0
.
0
8
9
.
3
9
0
.
1
8
9
.
3
9
0
.
1
K
N
N
7
2
.
9
7
4
.
2
0
.
9
2
0
0
.
9
2
8
7
2
.
5
7
3
.
9
7
3
.
3
7
4
.
5
7
2
.
9
7
4
.
2
T
h
e
s
cr
ee
n
p
lo
t
a
n
d
c
u
m
u
lati
v
e
v
ar
ia
n
ce
cu
r
v
e
o
f
t
h
e
PC
A
co
m
p
o
n
en
ts
ar
e
s
h
o
wn
in
f
ig
u
r
e
(
3
)
,
t
h
e
m
o
s
t o
f
th
e
v
ar
ian
ce
ar
e
co
n
ce
n
tr
ated
in
th
e
f
ir
s
t f
ew
p
r
in
cip
al
co
m
p
o
n
en
ts
as sh
o
wn
in
f
ig
u
r
e
3
(
a)
,
f
o
r
m
in
g
a
clea
r
elb
o
w
p
o
in
t.
Fig
u
r
e
3
(
b
)
illu
s
tr
ated
th
at
th
e
f
ir
s
t
1
0
0
co
m
p
o
n
en
ts
p
r
eser
v
e
d
ap
p
r
o
x
im
ately
9
5
-
9
8
%
o
f
th
e
to
tal
v
ar
ian
ce
wh
ich
r
ef
lec
ts
th
e
ef
f
ec
tiv
en
ess
o
f
PC
A
in
r
etain
in
g
th
e
m
o
s
t
m
ea
n
in
g
f
u
l
in
f
o
r
m
atio
n
wh
ile
r
em
o
v
in
g
r
ed
u
n
d
a
n
t o
r
n
o
is
y
f
ea
tu
r
es
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ta
ck
in
g
e
n
s
emb
le
tech
n
iq
u
es
fo
r
a
u
to
ma
ted
p
erip
h
era
l b
lo
o
d
…
(
Ma
r
w
a
Ma
w
fa
q
Mo
h
a
m
ed
s
h
ee
t A
l
-
Ha
ta
b
)
2253
(
a)
(
b
)
Fig
u
r
e
3
.
PC
A
a
n
aly
s
is
f
o
r
(
a)
s
cr
ee
p
lo
t o
f
p
r
in
cip
al
c
o
m
p
o
n
en
ts
an
d
(
b
)
c
u
m
u
lativ
e
v
a
r
ian
ce
p
lo
t sh
o
win
g
~9
5
–
9
8
% o
f
v
a
r
ian
ce
r
etain
e
d
3
.
2
.
Cla
s
s
-
wis
e
a
na
ly
s
is
a
nd
pro
ba
bil
it
y
t
hresh
o
ld
s
Fig
u
r
e
4
an
d
T
ab
le
3
s
h
o
w
t
h
e
class
-
wis
e
p
r
o
b
ab
ilit
y
th
r
e
s
h
o
ld
s
o
f
SVM
f
o
r
all
b
lo
o
d
ce
ll
ty
p
es
b
ef
o
r
e
an
d
af
ter
PC
A.
T
h
e
r
e
s
u
lts
r
ev
ea
led
th
at
m
o
r
p
h
o
l
o
g
ically
s
im
ilar
ce
lls
wer
e
af
f
ec
ted
b
y
PC
A
m
o
r
e
th
an
d
is
tin
ct
ty
p
es.
Fo
r
e
x
am
p
le,
p
r
o
b
ab
ilit
y
t
h
r
esh
o
ld
s
f
o
r
n
eu
tr
o
p
h
ils
,
eo
s
in
o
p
h
ils
,
an
d
er
y
th
r
o
b
last
s
wer
e
s
lig
h
tly
r
ed
u
ce
d
with
PC
A
(
0
.
9
7
5
v
s
.
0
.
9
8
7
,
0
.
9
7
6
v
s
.
0
.
9
9
4
,
0
.
9
6
9
v
s
.
0
.
9
8
3
,
r
esp
ec
tiv
el
y
)
.
M
o
n
o
c
y
tes
an
d
im
m
atu
r
e
g
r
an
u
lo
c
y
tes
r
ec
o
r
d
ed
th
e
m
o
s
t
n
o
ta
b
le
d
r
o
p
s
with
(
0
.
8
6
2
v
s
.
0
.
9
5
0
)
an
d
(
0
.
9
1
3
v
s
.
0
.
9
5
5
)
r
esp
ec
tiv
ely
,
d
u
e
to
th
eir
m
o
r
p
h
o
lo
g
ical
s
im
ilar
ity
to
o
th
er
W
B
C
s
.
T
h
ese
r
e
s
u
lts
in
d
icate
PC
A
ca
p
ab
ilit
y
to
p
r
eser
v
e
th
e
m
aj
o
r
ity
o
f
d
is
cr
i
m
in
ativ
e
f
ea
tu
r
es b
u
t it
m
ay
d
is
ca
r
d
s
u
b
tle
d
etails f
o
r
s
im
ilar
ce
ll ty
p
es.
Fig
u
r
e
4
.
T
h
e
class
-
wis
e
p
r
ed
icted
p
r
o
b
ab
ilit
ies o
f
th
e
SVM
m
o
d
el
b
ef
o
r
e
a
n
d
af
ter
PC
A
T
ab
le
3
.
Pro
b
ab
ilit
y
o
f
SVM
m
o
d
el
with
all
class
es
Cl
a
s
ses
N
e
u
t
r
o
p
h
i
l
Eo
si
n
o
p
h
i
l
Er
y
t
h
r
o
b
l
a
st
I
mm
a
t
u
r
e
g
r
a
n
u
l
o
c
y
t
e
s
Ly
m
p
h
o
c
y
t
e
B
a
s
o
p
h
i
l
P
l
a
t
e
l
e
t
M
o
n
o
c
y
t
e
S
V
M
+
P
C
A
0
.
9
7
5
0
.
9
7
6
0
.
9
6
9
0
.
9
1
3
0
.
9
4
9
0
.
9
3
7
0
.
9
9
6
0
.
8
6
2
S
V
M
0
.
9
8
7
0
.
9
9
4
0
.
9
8
3
0
.
9
5
5
0
.
9
7
2
0
.
9
7
0
.
9
9
8
0
.
9
5
Fig
u
r
e
5
illu
s
tr
ates
class
-
wis
e
R
OC
cu
r
v
es
an
d
AUC
v
a
lu
es
f
o
r
SVM
b
ef
o
r
e
a
n
d
a
f
ter
PC
A
u
s
in
g
2
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
.
Fig
u
r
e
5
(
a)
c
o
r
r
esp
o
n
d
s
to
e
r
y
th
r
o
b
last
,
Fig
u
r
e
5
(
b
)
to
n
eu
t
r
o
p
h
il,
Fig
u
r
e
5
(
c)
to
eo
s
in
o
p
h
il,
Fig
u
r
e
5
(
d
)
to
im
m
atu
r
e
g
r
an
u
l
o
cy
tes,
Fig
u
r
e
5
(
e
)
to
l
y
m
p
h
o
cy
te,
Fig
u
r
e
5
(
f
)
to
b
aso
p
h
il,
Fig
u
r
e
5
(
g
)
to
p
latelet,
an
d
Fig
u
r
e
5
(
h
)
to
m
o
n
o
c
y
te.
T
h
is
f
ig
u
r
e
h
ig
h
lig
h
ts
th
e
ab
ilit
y
o
f
th
e
m
o
d
el
to
d
is
cr
im
in
ate
b
etwe
en
b
lo
o
d
ce
ll
ty
p
e
an
d
co
n
f
ir
m
s
th
at
PC
A
s
lig
h
tly
r
ed
u
ce
s
p
er
f
o
r
m
an
ce
f
o
r
m
o
r
p
h
o
lo
g
ically
s
im
ilar
ce
lls
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
2
4
7
-
2
2
5
9
2254
(
a)
(
b
)
(
c)
(
d
)
(
e)
(f)
(
g
)
(
h
)
Fig
u
r
e
5
.
C
lass
-
wis
e
R
O
C
cu
r
v
es a
n
d
AUC v
alu
es f
o
r
SVM
b
ef
o
r
e
a
n
d
af
te
r
PC
A
u
s
in
g
2
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
f
o
r
(
a)
er
y
th
r
o
b
last
,
(
b
)
n
eu
tr
o
p
h
il,
(
c
)
eo
s
in
o
p
h
il
,
(
d
)
I
G,
(
e)
l
y
m
p
h
o
cy
te
,
(
f
)
b
aso
p
h
il,
(
g
)
p
latelet,
an
d
(
h
)
m
o
n
o
cy
te
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J Ar
tif
I
n
tell
I
SS
N:
2252
-
8
9
3
8
S
ta
ck
in
g
e
n
s
emb
le
tech
n
iq
u
es
fo
r
a
u
to
ma
ted
p
erip
h
era
l b
lo
o
d
…
(
Ma
r
w
a
Ma
w
fa
q
Mo
h
a
m
ed
s
h
ee
t A
l
-
Ha
ta
b
)
2255
3
.
3
.
St
a
t
is
t
ica
l e
v
a
lua
t
i
o
n
3
.
3
.
1
.
Sim
ula
t
ed
pa
ired
s
a
mp
le
a
s
s
es
s
m
ent
T
o
ac
cu
r
atel
y
ev
al
u
ate
th
e
e
f
f
ec
t
o
f
PC
A
o
n
th
e
p
er
f
o
r
m
a
n
ce
o
f
th
e
m
o
d
els,
a
s
im
u
lated
p
air
e
d
s
am
p
le
ev
alu
atio
n
:
m
ea
n
ac
cu
r
ac
y
an
d
s
tan
d
ar
d
d
ev
iatio
n
(
m
ea
n
±
SD)
was
co
n
d
u
c
ted
.
T
h
is
m
eth
o
d
ap
p
r
o
x
im
ates
th
e
ex
p
ec
ted
v
a
r
iab
ilit
y
o
f
ea
c
h
class
if
ier
b
e
f
o
r
e
a
n
d
af
te
r
PC
A.
T
h
is
s
im
u
latio
n
p
r
o
v
id
es
a
s
tatis
t
ical
s
ig
n
if
ican
ce
test
in
g
,
ev
en
w
h
en
o
n
ly
s
u
m
m
a
r
y
s
tatis
tic
s
ar
e
av
ailab
le,
p
r
o
v
id
in
g
i
n
s
ig
h
t
in
to
wh
eth
er
th
e
d
if
f
er
e
n
ce
s
in
o
b
s
er
v
ed
ac
c
u
r
ac
y
ar
e
c
o
n
s
is
ten
t
o
r
d
u
e
to
r
an
d
o
m
v
ar
iatio
n
.
T
ab
le
4
s
u
m
m
ar
izes
th
e
p
air
ed
e
v
alu
atio
n
s
.
T
h
e
r
esu
lts
d
em
o
n
s
tr
ated
th
at
PC
A
af
f
ec
ted
o
n
th
e
p
er
f
o
r
m
an
ce
o
n
m
o
s
t
m
o
d
els.
Af
ter
ap
p
ly
in
g
PC
A
NN,
L
R
,
an
d
SVM
s
h
o
wed
an
ap
p
r
o
x
im
ately
3
-
5
%
co
n
s
is
ten
t
d
ec
r
ea
s
e
in
ac
cu
r
a
cy
.
Pair
ed
t
-
test
an
d
W
ilco
x
o
n
an
aly
s
es
em
p
h
asize
d
th
e
s
tatis
tical
im
p
o
r
tan
ce
o
f
th
is
r
ed
u
ctio
n
with
p
<0
.
0
5
.
T
h
is
illu
s
tr
ates
th
at
alth
o
u
g
h
PC
A
is
s
u
cc
es
s
f
u
l
in
r
ed
u
ci
n
g
f
ea
t
u
r
e
r
e
d
u
n
d
an
cy
an
d
co
m
p
u
tatio
n
al
co
m
p
l
ex
ity
,
it
m
ay
also
r
em
o
v
e
d
is
cr
im
in
ativ
e
f
ea
tu
r
e
s
wh
ich
ar
e
n
ec
ess
ar
y
f
o
r
o
p
tim
al
class
if
icat
io
n
.
Acc
o
r
d
in
g
to
KNN,
it
wa
s
n
o
s
ig
n
if
ican
t
ch
an
g
e
in
its
p
e
r
f
o
r
m
an
ce
,
c
o
n
f
ir
m
i
n
g
its
r
o
b
u
s
t
n
ess
to
d
im
en
s
io
n
ality
r
ed
u
cti
o
n
.
T
h
ese
f
in
d
in
g
s
u
n
d
er
s
co
r
e
th
at
b
ala
n
cin
g
th
e
d
im
en
s
io
n
ality
r
ed
u
ctio
n
with
class
if
icatio
n
p
er
f
o
r
m
a
n
ce
is
es
s
en
tial
in
b
io
m
ed
ical
im
ag
e
a
n
aly
s
is
.
T
ab
le
4
.
Simu
lated
p
air
ed
s
am
p
le
ev
alu
atio
n
b
ef
o
r
e
an
d
af
ter
PC
A
M
o
d
e
l
W
i
t
h
o
u
t
P
C
A
6
6
/
3
4
s
p
l
i
t
W
i
t
h
o
u
t
P
C
A
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
W
i
t
h
P
C
A
6
6
/
3
4
s
p
l
i
t
W
i
t
h
P
C
A
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
NN
9
2
.
2
±
0
.
8
9
3
.
1
±
0
.
6
8
8
.
0
±
1
.
0
8
8
.
1
±
0
.
9
LR
9
2
.
4
±
0
.
7
9
2
.
9
±
0
.
5
8
7
.
8
±
1
.
1
8
8
.
1
±
0
.
8
S
V
M
9
2
.
6
±
0
.
6
9
3
.
4
±
0
.
4
8
9
.
3
±
0
.
9
9
0
.
1
±
0
.
7
K
N
N
7
3
.
4
±
1
.
2
7
3
.
4
±
1
.
0
7
2
.
9
±
1
.
5
7
4
.
2
±
1
.
3
3
.
3
.
2
.
Sta
t
is
t
ica
l sig
nifica
nce
t
ests
I
n
o
r
d
er
to
ac
cu
r
ately
ev
al
u
ate
th
e
im
p
ac
t
o
f
PC
A,
p
air
ed
t
-
test
s
an
d
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
s
wer
e
ap
p
lied
to
th
e
r
esu
lts
o
b
tain
ed
f
r
o
m
2
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
.
T
ab
le
5
s
u
m
m
ar
izes
th
e
s
tatis
tical
s
ig
n
if
ican
ce
r
esu
lts
.
I
t
illu
s
tr
ates
th
at
th
e
ac
cu
r
ac
y
o
f
NN,
L
R
,
an
d
SVM
was
s
ig
n
if
ican
tly
d
ec
r
ea
s
ed
,
wh
ile
th
e
p
er
f
o
r
m
an
ce
o
f
KNN
r
em
a
in
s
u
n
af
f
ec
ted
.
T
ab
le
5
.
Statis
tical
s
ig
n
if
ican
ce
an
aly
s
is
o
f
PC
A
im
p
ac
t o
n
ML
m
o
d
el
p
e
r
f
o
r
m
an
ce
M
o
d
e
l
M
e
a
n
a
c
c
u
r
a
c
y
(
w
i
t
h
o
u
t
P
C
A
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
)
M
e
a
n
a
c
c
u
r
a
c
y
(
w
i
t
h
P
C
A
c
r
o
ss
-
v
a
l
i
d
a
t
i
o
n
)
P
a
i
r
e
d
t
-
t
e
s
t
p
W
i
l
c
o
x
o
n
p
S
i
g
n
i
f
i
c
a
n
t
NN
9
3
.
1
8
8
.
1
0
.
0
1
3
0
.
0
1
2
Y
e
s
LR
9
2
.
9
8
8
.
1
0
.
0
1
6
0
.
0
1
4
Y
e
s
S
V
M
9
3
.
4
9
0
.
1
0
.
0
4
5
0
.
0
4
1
Y
e
s
K
N
N
7
3
.
4
7
4
.
2
0
.
4
6
2
0
.
4
5
6
No
3
.
4
.
St
a
ck
ing
ens
em
ble per
f
o
rm
a
nce
A
s
tack
in
g
en
s
em
b
le
was
im
p
lem
en
ted
to
im
p
r
o
v
e
th
e
r
o
b
u
s
tn
ess
an
d
class
if
icatio
n
ac
cu
r
ac
y
o
f
th
e
m
o
d
el
.
T
h
e
s
tack
ed
e
n
s
em
b
le
u
s
ed
in
th
is
s
tu
d
y
co
m
b
in
es
N
N,
L
R
,
an
d
SVM.
U
s
in
g
th
is
t
ec
h
n
iq
u
e
m
i
n
im
ize
b
ias
an
d
v
ar
ian
ce
b
y
lev
er
a
g
i
n
g
th
e
co
m
p
lem
en
tar
y
s
tr
en
g
th
s
o
f
th
e
b
ase
m
o
d
els.
A
co
m
p
ar
is
o
n
b
etwe
en
in
d
iv
id
u
al
class
if
ier
s
an
d
th
e
s
tack
ed
en
s
em
b
le
is
p
r
esen
ted
in
T
ab
le
6
.
T
ab
le
6
.
Per
f
o
r
m
an
ce
co
m
p
a
r
is
o
n
o
f
in
d
iv
id
u
al
a
n
d
s
tack
ed
ML
m
o
d
els
M
o
d
e
l
AUC
A
c
c
u
r
a
c
y
(
%)
F1
-
sc
o
r
e
(
%)
P
r
e
c
i
s
i
o
n
(
%)
R
e
c
a
l
l
(
%)
NN
0
.
9
9
5
9
3
.
1
9
3
.
1
9
3
.
1
9
3
.
1
LR
0
.
9
9
5
9
2
.
9
9
2
.
9
9
3
.
0
9
2
.
9
S
V
M
0
.
9
9
6
9
3
.
4
9
3
.
4
9
3
.
4
9
3
.
4
S
t
a
c
k
e
d
(
N
N
+
L
R
+
S
V
M
)
0
.
9
9
8
9
5
.
2
9
5
.
1
9
5
.
4
9
5
.
0
T
h
e
s
ta
ck
ed
en
s
em
b
l
e
s
u
r
p
a
s
s
ed
a
ll
i
n
d
iv
id
u
a
l
c
la
s
s
if
ie
r
s
,
ac
h
i
ev
in
g
an
AU
C
o
f
0
.
9
9
8
an
d
an
ac
cu
r
ac
y
o
f
9
5
.
2
%,
wi
th
co
n
s
is
ten
t
im
p
r
o
v
em
en
ts
a
cr
o
s
s
F1
-
s
co
r
e
,
p
r
ec
i
s
io
n
,
an
d
r
ec
a
ll
.
T
h
e
s
e
r
e
s
u
l
ts
in
d
i
ca
te
th
at
t
h
e
p
er
f
o
r
m
an
c
e
o
f
th
e
en
s
em
b
le
is
b
a
la
n
c
ed
in
th
e
c
la
s
s
if
i
ca
tio
n
o
f
b
lo
o
d
ce
ll
c
la
s
s
e
s
,
in
c
lu
d
in
g
r
ar
e
ty
p
e
s
.
Fig
u
r
e
6
il
lu
s
tr
at
e
s
th
e
ev
alu
at
io
n
o
f
m
o
d
e
l
p
er
f
o
r
m
an
ce
f
o
r
th
e
s
t
ac
k
ed
en
s
e
m
b
l
e
co
m
p
ar
ed
to
in
d
iv
id
u
a
l
m
o
d
el
s
,
u
s
in
g
b
o
th
t
h
e
p
r
ec
i
s
i
o
n
–
r
ec
a
l
l
c
u
r
v
e
s
h
o
wn
in
Fig
u
r
e
6
(
a
)
an
d
th
e
l
if
t
p
r
o
b
ab
i
l
ity
th
r
e
s
h
o
ld
p
lo
t
a
s
in
F
ig
u
r
e
6
(
b
)
,
s
h
o
wi
n
g
th
at
t
h
e
s
t
ac
k
ed
r
e
ta
in
ed
th
e
s
t
ac
k
ed
en
s
em
b
l
e
s
u
r
p
as
s
ed
a
l
l
in
d
iv
id
u
a
l
c
la
s
s
if
i
er
s
,
ac
h
iev
in
g
a
n
A
U
C
o
f
0
.
9
9
8
an
d
an
ac
cu
r
ac
y
o
f
9
5
.
2
%,
w
ith
co
n
s
i
s
ten
t
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
9
3
8
I
n
t J Ar
tif
I
n
tell
,
Vo
l.
1
5
,
No
.
3
,
J
u
n
e
2
0
2
6
:
2
2
4
7
-
2
2
5
9
2256
im
p
r
o
v
em
en
t
s
ac
r
o
s
s
F
1
-
s
co
r
e,
p
r
ec
i
s
io
n
,
an
d
r
ec
al
l.
T
h
e
s
e
r
e
s
u
lt
s
in
d
i
ca
te
th
at
th
e
p
er
f
o
r
m
an
c
e
o
f
t
h
e
en
s
em
b
le
s
u
p
er
i
o
r
p
r
ec
is
io
n
a
cr
o
s
s
v
ar
y
in
g
r
ec
al
l
th
r
e
s
h
o
ld
s
a
n
d
a
ch
i
ev
ed
a
g
r
ea
ter
l
if
t
,
wh
i
ch
co
n
f
ir
m
s
th
at
s
t
ac
k
in
g
ef
f
e
ct
iv
e
ly
in
t
eg
r
at
e
s
th
e
p
r
ed
ic
tiv
e
ad
v
an
t
ag
e
s
o
f
N
N,
L
R
,
an
d
S
VM
in
to
a
m
o
r
e
s
ta
b
le
an
d
ac
cu
r
at
e
cl
as
s
if
ica
ti
o
n
f
r
am
e
wo
r
k
.
(
a)
(
b
)
Fig
u
r
e
6
.
Per
f
o
r
m
an
c
e
ev
alu
at
io
n
o
f
t
h
e
s
tack
ed
en
s
em
b
le
c
o
m
p
ar
ed
to
in
d
i
v
id
u
al
m
o
d
els
f
o
r
(
a)
p
r
ec
is
io
n
-
r
ec
all
cu
r
v
e
an
d
(
b
)
lift
p
r
o
b
a
b
ilit
y
th
r
esh
o
ld
3
.
5
.
Co
m
pa
riso
n wit
h pre
v
io
us
s
t
ud
ie
s
T
ab
le
7
s
u
m
m
ar
izes
r
elate
d
b
lo
o
d
ce
ll
class
if
icatio
n
s
tu
d
ies.
T
h
e
p
r
o
p
o
s
ed
s
y
s
tem
o
u
tp
er
f
o
r
m
s
p
r
ev
io
u
s
s
tu
d
ies,
ac
h
iev
in
g
9
5
.
2
%
ac
c
u
r
ac
y
an
d
an
AUC
o
f
0
.
9
9
8
u
s
in
g
I
n
ce
p
tio
n
V3
f
ea
tu
r
es
with
a
s
tack
ed
NN,
L
R
,
an
d
SVM
e
n
s
em
b
le.
I
t
d
e
m
o
n
s
tr
ates
s
tr
o
n
g
r
o
b
u
s
tn
ess
u
n
d
e
r
2
0
-
f
o
ld
c
r
o
s
s
-
v
alid
atio
n
,
an
d
PC
A
-
b
ased
f
ea
tu
r
e
r
ed
u
ctio
n
ca
u
s
ed
m
in
im
al
p
e
r
f
o
r
m
an
c
e
lo
s
s
,
h
ig
h
lig
h
tin
g
its
ac
cu
r
ac
y
,
ef
f
icien
c
y
,
an
d
p
r
ac
tical
r
elev
an
ce
.
T
ab
le
7
.
C
o
m
p
a
r
ativ
e
s
u
m
m
ar
y
o
f
b
lo
o
d
ce
ll c
lass
if
icatio
n
s
t
u
d
ies
R
e
f
e
r
e
n
c
e
M
e
t
h
o
d
/
M
o
d
e
l
D
i
scu
ssi
o
n
F
e
r
r
e
i
r
a
e
t
a
l
.
[
4
2
]
,
2
0
2
5
C
N
N
w
i
t
h
c
o
n
t
r
a
s
t
l
i
m
i
t
e
d
a
d
a
p
t
i
v
e
h
i
s
t
o
g
r
a
m
e
q
u
a
l
i
z
a
t
i
o
n
(
C
LA
H
E
)
A
c
h
i
e
v
e
d
u
p
t
o
9
8
.
7
2
%
t
e
st
a
c
c
u
r
a
c
y
f
o
r
l
y
mp
h
o
c
y
t
e
s
a
n
d
mo
n
o
c
y
t
e
s;
C
LA
H
E
i
m
p
r
o
v
e
d
i
mag
e
c
o
n
t
r
a
st
a
n
d
e
n
h
a
n
c
e
d
C
N
N
c
l
a
ssi
f
i
c
a
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
.
S
a
z
a
k
a
n
d
K
o
t
a
n
[
4
3
]
,
2
0
2
4
Y
O
LO
v
1
1
w
i
t
h
o
p
t
i
m
i
z
e
d
w
e
i
g
h
t
s
A
c
h
i
e
v
e
d
mea
n
a
v
e
r
a
g
e
p
r
e
c
i
s
i
o
n
(
mA
P
)
o
f
9
3
.
8
%;
d
a
t
a
a
u
g
me
n
t
a
t
i
o
n
a
n
d
o
p
t
i
mi
z
e
d
w
e
i
g
h
t
s
e
n
h
a
n
c
e
d
mo
d
e
l
r
o
b
u
st
n
e
ss
a
n
d
p
r
e
c
i
s
i
o
n
.
M
a
h
a
l
e
[
4
4
]
,
2
0
2
5
D
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
(
I
n
c
e
p
t
i
o
n
,
V
G
G
1
6
,
c
a
p
s
u
l
e
n
e
t
w
o
r
k
s
)
I
n
c
e
p
t
i
o
n
a
c
h
i
e
v
e
d
t
h
e
b
e
s
t
a
c
c
u
r
a
c
y
a
n
d
p
r
e
c
i
si
o
n
a
mo
n
g
t
e
s
t
e
d
m
o
d
e
l
s,
h
i
g
h
l
i
g
h
t
i
n
g
t
h
e
st
r
e
n
g
t
h
o
f
d
e
e
p
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
f
o
r
b
l
o
o
d
c
e
l
l
c
l
a
ss
i
f
i
c
a
t
i
o
n
.
Tse
n
g
a
n
d
H
u
a
n
g
[
4
5
]
,
2
0
2
2
En
se
mb
l
e
o
f
C
N
N
s
A
c
h
i
e
v
e
d
a
v
e
r
a
g
e
t
e
st
i
n
g
a
c
c
u
r
a
c
y
o
f
9
0
.
1
%
f
o
r
n
e
u
t
r
o
p
h
i
l
c
l
a
ssi
f
i
c
a
t
i
o
n
;
e
n
sem
b
l
e
l
e
a
r
n
i
n
g
i
mp
r
o
v
e
d
g
e
n
e
r
a
l
i
z
a
t
i
o
n
a
c
r
o
ss
mu
l
t
i
p
l
e
d
a
t
a
set
s.
Th
i
s
w
o
r
k
I
n
c
e
p
t
i
o
n
V
3
f
e
a
t
u
r
e
e
x
t
r
a
c
t
i
o
n
+
S
t
a
c
k
e
d
(
N
N
+
LR
+
S
V
M
)
,
P
C
A
,
2
0
-
f
o
l
d
C
V
Th
e
st
a
c
k
e
d
e
n
s
e
mb
l
e
a
c
h
i
e
v
e
d
t
h
e
b
e
s
t
p
e
r
f
o
r
m
a
n
c
e
w
i
t
h
9
5
.
2
%
a
c
c
u
r
a
c
y
a
n
d
a
n
A
U
C
o
f
0
.
9
9
8
.
4.
CO
NCLU
SI
O
N
T
h
e
s
t
u
d
y
i
ll
u
s
t
r
at
es
t
h
e
ef
f
e
cti
v
e
n
ess
o
f
u
s
in
g
S
VM
c
o
m
b
in
e
d
wit
h
2
0
-
f
o
l
d
c
r
o
s
s
-
v
ali
d
at
io
n
i
n
ac
h
ie
v
i
n
g
h
i
g
h
ac
cu
r
a
c
y
f
o
r
p
er
i
p
h
e
r
a
l
b
lo
o
d
c
ell
ty
p
es
cl
as
s
if
i
ca
t
io
n
.
W
it
h
o
u
t
u
s
i
n
g
PC
A
,
SVM
ac
h
i
ev
ed
th
e
h
i
g
h
est
p
r
e
d
ic
ti
o
n
p
r
o
b
a
b
i
lit
i
es
ac
r
o
s
s
m
o
s
t
cl
ass
es
,
i
n
cl
u
d
i
n
g
p
l
ate
lets
(
0
.
9
9
8
)
,
e
o
s
i
n
o
p
h
ils
(
0
.
9
9
4
)
,
an
d
n
e
u
t
r
o
p
h
i
ls
(
0
.
9
8
7
)
.
Ap
p
l
y
i
n
g
PC
A
r
esu
lte
d
i
n
d
i
f
f
e
r
e
n
t
i
m
p
a
cts
ac
r
o
s
s
d
i
f
f
e
r
e
n
t
c
ell
t
y
p
es.
M
o
n
o
cy
tes
a
n
d
im
m
at
u
r
e
g
r
a
n
u
l
o
c
y
t
es
r
ec
o
r
d
ed
t
h
e
m
o
s
t
n
o
t
ab
le
d
ec
l
in
es
(
0
.
8
6
2
v
s
.
0
.
9
5
0
)
a
n
d
(
0
.
9
1
3
v
s
.
0
.
9
5
5
)
r
esp
ec
ti
v
el
y
,
wh
ich
in
d
icate
th
eir
h
ig
h
er
s
en
s
itiv
ity
to
d
im
en
s
io
n
ality
r
ed
u
ctio
n
d
u
e
to
th
eir
m
o
r
p
h
o
lo
g
ical
s
im
ilar
ity
to
o
th
er
le
u
k
o
c
y
tes.
Mo
d
e
r
ate
r
e
d
u
ctio
n
s
wer
e
also
o
b
s
er
v
e
d
f
o
r
b
aso
p
h
ils
,
ly
m
p
h
o
cy
tes
an
d
er
y
th
r
o
b
last
s
with
(
0
.
9
3
7
v
s
.
0
.
9
7
0
)
,
(
0
.
9
4
9
v
s
.
0
.
9
7
2
)
,
a
n
d
(
0
.
9
6
9
v
s
.
0
.
9
8
3
)
r
esp
ec
tiv
ely
.
T
h
ese
r
ed
u
ctio
n
s
wer
e
co
n
f
ir
m
ed
to
b
e
s
tatis
tica
lly
s
ig
n
if
ican
t
f
o
r
SVM,
NN,
an
d
L
R
(
p
<0
.
0
5
)
,
in
d
icatin
g
th
at
PC
A
m
ay
r
em
o
v
e
s
u
b
tle
Evaluation Warning : The document was created with Spire.PDF for Python.