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
.
4
,
A
u
g
u
s
t 2
0
2
6
,
p
p
.
3
8
5
2
~
3
8
6
4
I
SS
N:
2
2
5
2
-
8
9
3
8
,
DOI
: 1
0
.
1
1
5
9
1
/ijai.v
15
.i
4
.
p
p
3
8
5
2
-
3
8
6
4
3852
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
a
i
.
ia
esco
r
e.
co
m
T
he role o
f
big
da
ta in pre
cisio
n me
dicine and
hea
lth
ca
re
mo
nitoring
using
the
M
a
pRed
uce
f
ra
mewo
rk
M
ee
na
ks
hi Sa
n
ka
ra
s
ub
ra
m
a
nia
n
1
,
M
ee
na
Cha
v
a
n
2
,
G
o
v
ind
a
n M
a
no
ha
ra
n K
a
rt
hik
3,
J
ha
ns
i P
a
nd
iri
4
,
Aruma
lla
Na
g
a
ra
j
u
5
,
I
dim
a
da
ka
la
M
a
dh
a
v
ila
t
ha
6
1
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
,
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
H
u
ma
n
i
t
i
e
s,
S
R
M
I
n
st
i
t
u
t
e
o
f
S
c
i
e
n
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
K
a
t
t
a
n
k
u
l
a
t
h
u
r
,
I
n
d
i
a
2
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s
a
n
d
T
e
l
e
c
o
mm
u
n
i
c
a
t
i
o
n
,
B
h
a
r
a
t
i
V
i
d
y
a
p
e
e
t
h
(
D
e
e
me
d
t
o
b
e
U
n
i
v
e
r
s
i
t
y
)
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
,
P
u
n
e
,
I
n
d
i
a
3
S
c
h
o
o
l
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
V
e
l
l
o
r
e
I
n
st
i
t
u
t
e
o
f
Te
c
h
n
o
l
o
g
y
,
V
e
l
l
o
r
e
,
I
n
d
i
a
4
D
e
p
a
r
t
me
n
t
o
f
El
e
c
t
r
o
n
i
c
s a
n
d
C
o
m
mu
n
i
c
a
t
i
o
n
En
g
i
n
e
e
r
i
n
g
,
A
d
i
t
y
a
U
n
i
v
e
r
si
t
y
,
S
u
r
a
m
p
a
l
e
m
,
I
n
d
i
a
5
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
K
o
n
e
r
u
L
a
k
sh
m
a
i
a
h
E
d
u
c
a
t
i
o
n
F
o
u
n
d
a
t
i
o
n
,
V
i
j
a
y
a
w
a
d
a
,
I
n
d
i
a
6
D
e
p
a
r
t
me
n
t
o
f
C
o
m
p
u
t
e
r
S
c
i
e
n
c
e
a
n
d
E
n
g
i
n
e
e
r
i
n
g
,
S
r
i
V
e
n
k
a
t
e
sw
a
r
a
C
o
l
l
e
g
e
o
f
E
n
g
i
n
e
e
r
i
n
g
,
T
i
r
u
p
a
t
i
,
I
n
d
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 3
,
2
0
2
5
R
ev
is
ed
J
u
n
1
2
,
2
0
2
6
Acc
ep
ted
J
u
l 9
,
2
0
2
6
Da
ta
a
n
a
ly
ti
c
s
h
a
s
b
e
c
o
m
e
a
c
o
r
n
e
rsto
n
e
o
f
p
re
c
isio
n
m
e
d
icin
e
b
y
e
n
a
b
l
in
g
d
o
c
to
rs
a
n
d
sc
ien
ti
sts
t
o
e
x
trac
t
m
e
a
n
in
g
fu
l
in
si
g
h
ts
fro
m
v
a
st,
c
o
m
p
lex
d
a
t
a
se
ts.
M
o
st
h
e
a
lt
h
c
a
re
d
a
ta
a
re
h
i
g
h
-
d
ime
n
sio
n
a
l
d
a
ta
th
a
t
n
o
t
o
n
l
y
re
q
u
ir
e
lo
n
g
e
r
c
o
m
p
u
tatio
n
a
l
ti
m
e
b
u
t
a
l
so
a
ffe
c
t
th
e
a
c
c
u
ra
c
y
o
f
a
n
a
ly
si
s.
In
o
r
d
e
r
to
o
v
e
rc
o
m
e
th
e
se
issu
e
s,
th
e
m
a
p
re
d
u
c
e
b
a
se
d
b
ig
d
a
ta
h
e
a
lt
h
c
a
re
m
o
n
it
o
r
in
g
fra
m
e
wo
rk
is
p
ro
p
o
se
d
.
Th
e
p
ro
p
o
se
d
w
o
rk
c
o
m
p
rise
s
p
re
p
ro
c
e
ss
in
g
,
m
a
p
re
d
u
c
e
fra
m
e
wo
r
k
,
a
n
d
d
a
ta
c
las
sifica
t
io
n
.
Th
e
p
re
p
ro
c
e
ss
in
g
c
a
n
b
e
d
o
n
e
u
si
n
g
imp
ro
v
e
d
m
in
-
m
a
x
n
o
rm
a
li
z
a
ti
o
n
,
a
n
d
t
h
e
b
ig
d
a
ta
c
a
n
b
e
h
a
n
d
led
u
si
n
g
th
e
imp
ro
v
e
d
s
u
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
(S
VM)
-
re
c
u
rsiv
e
fe
a
tu
re
e
li
m
in
a
ti
o
n
(R
F
E)
m
e
th
o
d
.
F
in
a
ll
y
,
t
h
e
c
las
sifica
ti
o
n
c
a
n
b
e
d
o
n
e
u
sin
g
a
d
e
e
p
Q
-
n
e
tw
o
rk
(DQ
N).
Th
e
p
e
rfo
rm
a
n
c
e
o
f
th
e
p
ro
p
o
se
d
m
e
th
o
d
is
a
n
a
l
y
z
e
d
i
n
term
s
o
f
a
c
c
u
ra
c
y
,
p
re
c
isio
n
,
f
-
m
e
a
su
re
,
a
n
d
M
a
tt
h
e
w'
s
c
o
rre
latio
n
c
o
e
fficie
n
t
(
M
CC
)
.
K
ey
w
o
r
d
s
:
B
ig
d
ata
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
Min
-
m
ax
n
o
r
m
aliza
tio
n
Pre
cisi
o
n
m
ed
icin
e
R
ec
u
r
s
iv
e
f
ea
tu
r
e
elim
in
atio
n
Su
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Me
en
ak
s
h
i San
k
ar
asu
b
r
am
an
i
an
Dep
ar
tm
en
t o
f
C
o
m
p
u
ter
Scie
n
ce
,
Facu
lty
o
f
Scien
ce
an
d
H
u
m
an
ities
S
R
M
I
n
s
titu
te
o
f
Scien
ce
an
d
T
ec
h
n
o
lo
g
y
SR
M
Nag
ar
,
Po
th
er
i,
Kattan
k
u
lath
u
r
–
6
0
3
2
0
3
,
T
am
il Na
d
u
,
I
n
d
ia
E
m
ail:
m
ee
n
ak
s
h
is
an
k
ar
2
0
1
3
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
aim
o
f
p
r
ec
is
io
n
m
ed
icin
e
is
to
allo
w
p
atien
t
-
f
o
cu
s
ed
c
h
o
ices
th
at
f
u
lf
il
th
e
tr
ip
le
o
b
j
ec
tiv
es
o
f
e
n
h
a
n
c
i
n
g
p
a
t
i
e
n
t
w
e
l
l
n
es
s
,
e
n
h
a
n
c
i
n
g
p
a
t
i
e
n
t
s
a
t
is
f
a
ct
i
o
n
,
a
n
d
l
o
w
e
r
i
n
g
e
x
p
e
n
s
e
s
[
1
]
.
To
r
e
a
c
h
t
h
i
s
o
b
jectiv
e,
p
r
ec
is
io
n
m
e
d
icin
e
d
ep
en
d
s
o
n
th
e
ex
a
m
in
atio
n
o
f
b
ig
d
ata
co
llectio
n
s
to
p
in
p
o
in
t
clin
ically
-
s
ig
n
if
ican
t
s
u
b
-
g
r
o
u
p
s
an
d
aid
in
u
n
c
o
v
er
in
g
d
is
ea
s
e
m
ec
h
an
is
m
s
.
Un
c
o
m
m
o
n
d
is
ea
s
es
p
r
esen
t
co
n
s
id
er
ab
le
d
if
f
icu
lties
in
id
en
tific
atio
n
an
d
m
an
ag
e
m
en
t
b
ec
au
s
e
o
f
th
eir
lim
ited
f
r
eq
u
en
cy
an
d
v
ar
ied
m
an
if
e
s
tatio
n
s
.
Desp
ite
th
e
in
co
r
p
o
r
atio
n
o
f
b
ig
d
ata
in
r
ar
e
d
is
ea
s
e
g
en
o
m
ics,
s
ig
n
if
ican
t
ch
allen
g
es
r
em
ain
,
s
u
ch
as
is
s
u
es
in
p
in
p
o
in
tin
g
ca
u
s
al
v
ar
ian
ts
an
d
ap
p
ly
in
g
d
is
co
v
e
r
ies
in
clin
i
ca
l
s
ettin
g
s
.
T
h
e
estab
lis
h
m
en
t
o
f
lar
g
e
co
n
s
o
r
tia
lead
s
to
ac
cu
m
u
latio
n
o
f
v
ast
am
o
u
n
ts
o
f
d
ata,
k
n
o
wn
as
b
i
g
d
ata,
h
ig
h
lig
h
tin
g
th
e
n
ee
d
f
o
r
b
ig
d
ata
-
d
r
i
v
en
an
aly
s
is
p
ip
elin
e.
Han
d
lin
g
lar
g
e
d
ata
s
ets p
o
s
es
ch
allen
g
es,
s
u
ch
as st
o
r
ag
e
co
n
s
tr
ain
ts
,
co
m
p
u
tatio
n
al
p
o
wer
n
ee
d
s
,
an
d
c
o
n
ce
r
n
s
o
v
er
d
ata
s
ec
u
r
ity
[
2
]
.
I
n
d
iv
id
u
als
with
r
ar
e
d
is
ea
s
es
f
r
eq
u
en
tly
en
co
u
n
ter
o
b
s
tacle
s
lik
e
d
elay
ed
d
iag
n
o
s
es
an
d
m
is
d
iag
n
o
s
es,
with
o
v
er
9
0
% o
f
r
ar
e
c
o
n
d
itio
n
s
h
av
in
g
n
o
e
f
f
ec
tiv
e
tr
ea
tm
en
ts
[
3
]
.
Ar
tific
ial
in
tellig
en
ce
(
AI
)
an
d
m
ac
h
in
e
lear
n
in
g
(
ML
)
t
ec
h
n
o
lo
g
ies
aid
r
a
r
e
d
is
ea
s
e
r
esear
ch
b
y
f
ac
ilit
atin
g
th
e
e
x
am
in
atio
n
o
f
lar
g
e
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
Th
e
r
o
le
o
f b
ig
d
a
ta
in
p
r
ec
is
io
n
med
icin
e
a
n
d
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
… (
Meen
a
ksh
i
S
a
n
k
a
r
a
s
u
b
r
a
ma
n
ia
n
)
3853
v
o
lu
m
es
o
f
g
e
n
o
m
ic
an
d
cli
n
ical
d
ata
to
r
ec
o
g
n
ize
d
is
ea
s
e
p
atter
n
s
,
f
o
r
esee
tr
ea
tm
en
t
r
esu
lts
,
an
d
cr
ea
te
tailo
r
ed
th
er
ap
ies,
wh
ich
en
h
a
n
ce
s
d
iag
n
o
s
tic
p
r
ec
is
io
n
an
d
p
r
o
p
els
d
r
u
g
d
ev
el
o
p
m
en
t.
I
n
th
e
cu
r
r
en
t
d
i
g
ital
lan
d
s
ca
p
e,
a
s
ig
n
if
ican
t
r
is
e
i
n
d
ata
h
as
g
en
e
r
ated
n
u
m
er
o
u
s
o
p
p
o
r
tu
n
ities
an
d
ch
allen
g
es
f
o
r
g
o
v
er
n
m
e
n
ts
,
in
d
u
s
tr
ies,
an
d
o
r
g
a
n
izatio
n
s
.
R
ec
en
t
ad
v
an
ce
m
en
ts
in
tec
h
n
o
lo
g
y
a
n
d
n
etwo
r
k
in
g
h
a
v
e
led
to
a
g
r
o
win
g
d
ep
en
d
e
n
ce
o
n
d
ata,
n
ec
ess
it
atin
g
r
o
b
u
s
t
d
ata
s
ec
u
r
ity
a
n
d
class
if
icatio
n
s
y
s
tem
s
.
C
y
b
er
th
r
ea
ts
s
u
ch
as
attac
k
er
s
,
m
alwa
r
e,
esp
io
n
ag
e,
an
d
f
r
au
d
ar
e
in
cr
ea
s
in
g
l
y
co
m
m
o
n
,
n
ec
ess
itatin
g
ad
v
a
n
ce
d
ap
p
r
o
ac
h
es
to
p
r
o
tect
d
ata
in
te
g
r
ity
an
d
v
er
if
y
s
y
s
tem
s
[
4
]
.
Pre
cisi
o
n
m
ed
icin
e'
s
u
s
e
o
f
ML
alg
o
r
ith
m
s
h
as
th
e
p
o
ten
tial
to
s
ig
n
if
ican
tly
alter
m
ed
ical
p
r
o
ce
d
u
r
es.
Pre
cisi
o
n
m
ed
ici
n
e
b
en
ef
its
g
r
ea
tly
f
r
o
m
d
ee
p
lear
n
in
g
(
DL
)
alg
o
r
ith
m
s
'
ca
p
ac
ity
to
m
an
ag
e
co
m
p
lex
b
ig
d
at
a,
class
if
y
,
i
d
en
tify
p
atter
n
s
,
an
d
m
a
k
e
p
r
e
d
ictio
n
s
.
T
h
e
id
en
tific
atio
n
o
f
s
m
all
m
o
lecu
les
in
m
etab
o
lic
p
h
e
n
o
ty
p
in
g
,
p
r
o
g
n
o
s
is
,
p
atien
t
class
if
icatio
n
,
an
d
th
e
c
r
ea
tio
n
o
f
i
n
d
iv
id
u
alize
d
tr
ea
tm
en
t
p
lan
s
ar
e
o
n
ly
a
f
ew
o
f
th
e
n
u
m
er
o
u
s
co
n
tr
ib
u
tio
n
s
m
ad
e
b
y
t
h
ese
ap
p
licatio
n
s
.
Me
th
o
d
o
lo
g
ical
an
d
p
r
ac
tical
ch
allen
g
es,
s
u
ch
as
u
n
r
ea
lis
tic
d
ev
elo
p
m
e
n
t
en
v
ir
o
n
m
en
ts
,
th
e
r
eq
u
ir
em
en
t
f
o
r
ef
f
icien
t
d
iag
n
o
s
tic
m
o
d
els,
v
alid
atio
n
in
a
v
a
r
iety
o
f
co
n
t
ex
ts
,
an
d
s
m
o
o
th
in
te
g
r
atio
n
in
to
cu
r
r
en
t
h
ea
lth
ca
r
e
p
r
o
ce
s
s
es,
ar
e
th
e
ca
u
s
e
o
f
th
e
im
p
lem
en
tatio
n
g
ap
[
1
]
.
Hig
h
d
im
e
n
s
io
n
ality
in
b
ig
d
a
ta
with
p
r
ec
is
io
n
m
e
d
icin
e
is
o
n
e
o
f
t
h
e
cr
itical
ch
allen
g
es,
as
d
atasets
f
r
eq
u
e
n
tly
co
n
s
is
t
o
f
n
o
is
y
,
ir
r
elev
an
t
,
o
r
r
ed
u
n
d
an
t
f
ea
tu
r
es
th
at
ca
n
d
elay
th
e
ef
f
ec
tiv
en
ess
o
f
p
r
ed
ictiv
e
m
o
d
els
[
5
]
.
Selectin
g
th
e
m
o
s
t
r
elev
an
t
f
ea
tu
r
es
ca
n
ad
d
r
ess
th
ese
ch
allen
g
es
b
y
im
p
r
o
v
in
g
th
e
class
if
icatio
n
ac
cu
r
ac
y
as
well
as
r
ed
u
cin
g
co
m
p
u
tatio
n
al
n
ec
ess
ities
.
Du
e
to
th
eir
h
ig
h
co
m
p
u
tatio
n
al
r
eq
u
ir
em
en
ts
,
co
n
v
en
tio
n
al
ML
m
o
d
els
lik
e
d
ec
is
io
n
tr
ees
(
DT
)
,
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
es
(
SVM)
,
a
n
d
ar
tific
ial
n
e
u
r
al
n
etwo
r
k
s
(
ANNs)
f
ac
e
s
ca
l
ab
ilit
y
p
r
o
b
lem
s
with
b
ig
d
at
a.
Du
e
t
o
th
eir
ef
f
icien
c
y
an
d
co
m
p
ar
ab
le
p
er
f
o
r
m
an
ce
,
alg
o
r
ith
m
s
lik
e
k
-
n
ea
r
est
n
eig
h
b
o
r
s
(
KNN)
h
av
e
ex
p
o
s
ed
ca
p
ac
ity
f
o
r
ce
r
tain
class
if
icatio
n
task
s
[
6
]
.
B
y
au
to
m
atin
g
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
class
if
icatio
n
b
y
n
eu
r
al
n
etwo
r
k
s
(
NN)
,
r
ec
en
t
DL
d
ev
elo
p
m
e
n
ts
h
av
e
s
ig
n
if
ican
tly
en
h
an
ce
d
ab
ilit
ies
in
p
h
y
s
ical
lay
er
m
o
d
u
latio
n
class
if
icatio
n
ar
ea
s
[
7
]
,
[
8
]
.
B
y
ad
d
r
ess
in
g
th
e
ch
allen
g
es f
ac
ed
in
h
a
n
d
lin
g
b
ig
d
ata
:
−
A
n
ew
class
if
icatio
n
m
o
d
el
is
p
r
o
p
o
s
ed
in
th
is
wo
r
k
,
an
d
its
p
r
im
a
r
y
co
n
tr
i
b
u
tio
n
is
an
im
p
r
o
v
ed
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
,
wh
ic
h
is
u
s
ed
to
h
an
d
le
b
ig
d
ata.
−
T
h
e
p
r
e
p
r
o
ce
s
s
in
g
is
d
o
n
e
u
s
in
g
im
p
r
o
v
ed
m
in
-
m
a
x
n
o
r
m
al
izatio
n
to
r
em
o
v
e
o
u
tlier
s
.
−
Her
e,
an
i
m
p
r
o
v
ed
f
ea
tu
r
e
f
u
s
io
n
m
eth
o
d
u
s
in
g
an
SV
M
-
r
ec
u
r
s
iv
e
f
ea
t
u
r
e
elim
in
at
io
n
(
R
FE
)
is
em
p
lo
y
ed
in
t
h
e
m
ap
p
e
r
p
h
as
e.
Par
ticu
lar
ly
,
f
ea
tu
r
es
ar
e
e
x
tr
ac
ted
u
s
in
g
s
tatis
tical
m
et
h
o
d
s
,
an
d
r
aw
d
ata
ex
tr
ac
ts
th
e
p
r
ec
is
io
n
m
e
d
icin
e
-
o
r
ien
te
d
attr
ib
u
tes to
en
h
an
ce
th
e
class
if
icatio
n
ac
cu
r
a
cy
.
T
h
e
r
em
ain
d
er
o
f
th
is
p
ap
er
is
s
tr
u
ctu
r
ed
as
f
o
llo
ws.
A
li
ter
atu
r
e
s
u
r
v
ey
p
r
esen
ted
in
s
ec
tio
n
2
.
A
p
r
o
p
o
s
ed
b
i
g
d
ata
class
if
icatio
n
m
o
d
el
p
r
esen
te
d
in
s
ec
tio
n
3
.
E
x
p
e
r
im
en
tal
v
alid
atio
n
d
escr
ib
ed
in
s
ec
tio
n
4
.
Fin
ally
,
a
co
n
clu
s
io
n
o
f
th
e
r
esear
ch
in
s
ec
tio
n
5
.
2.
L
I
T
E
R
AT
U
RE
SU
RVE
Y
I
n
2
0
2
2
,
J
ay
asri
an
d
Ar
u
n
a
[
9
]
d
e
v
elo
p
ed
a
m
o
d
el
f
o
r
e
v
alu
atin
g
th
e
d
iab
etic
p
atien
t
h
ea
lth
ca
r
e
d
ataset
b
y
th
e
co
m
b
in
atio
n
o
f
th
e
m
u
lticlas
s
o
u
tlier
ca
teg
o
r
izatio
n
,
h
ier
ar
ch
ical
d
ec
is
io
n
ass
o
ciatio
n
r
u
le
(
AR
)
as
well
as
a
clas
s
if
icat
io
n
-
b
ased
Ma
p
R
ed
u
ce
m
o
d
el.
I
n
a
Ma
p
R
ed
u
ce
ar
ch
itectu
r
e,
th
e
AR
-
ap
r
io
r
i
alg
o
r
ith
m
co
n
s
id
er
e
d
m
ed
ica
l
d
ata
wh
en
g
en
er
atin
g
r
u
les.
T
h
is
was
u
s
ed
to
d
eter
m
in
e
th
e
r
elatio
n
s
h
ip
b
etwe
en
th
e
d
is
ea
s
es
an
d
th
eir
s
y
m
p
to
m
s
.
UC
I
-
ML
d
iab
etic
d
atasets
wi
th
5
0
attr
ib
u
tes
we
r
e
u
s
ed
to
co
n
d
u
c
t
th
is
an
aly
s
is
.
I
n
2
0
2
1
,
Mo
h
am
m
e
d
et
a
l.
[
1
0
]
d
ev
el
o
p
ed
a
n
o
v
el
o
p
tim
iz
atio
n
tech
n
iq
u
e
n
am
ed
E
-
B
at
in
teg
r
ated
th
e
b
at
alg
o
r
ith
m
(
BA
)
as
well
as
ex
p
o
n
en
tially
weig
h
te
d
m
o
v
in
g
a
v
er
ag
e
(
E
W
MA
)
.
I
n
th
is
p
ap
er
,
a
b
ig
d
ata
class
if
icatio
n
was
p
er
f
o
r
m
ed
b
y
em
p
l
o
y
in
g
th
e
Ma
p
R
ed
u
ce
m
o
d
el
,
in
w
h
ich
th
e
m
a
p
p
er
as
well
as
th
e
r
ed
u
ce
r
f
u
n
ctio
n
d
ep
en
d
ed
o
n
th
e
d
ev
elo
p
ed
m
o
d
el.
C
h
o
o
s
in
g
th
e
f
ea
tu
r
es
was
th
e
task
o
f
th
e
m
ap
p
er
f
u
n
ctio
n
,
wh
ich
was
d
ev
el
o
p
ed
f
o
r
th
e
ca
teg
o
r
izatio
n
in
t
h
e
r
ed
u
ce
r
p
h
ase
,
u
tili
zin
g
N
N.
T
h
u
s
,
th
e
E
-
B
at
m
eth
o
d
,
wh
ic
h
ass
is
ted
in
u
s
in
g
Ma
p
R
ed
u
ce
,
was u
s
ed
to
cla
s
s
if
y
b
ig
d
ata.
I
n
2
0
2
4
,
Hu
an
g
et
a
l.
[
1
1
]
s
u
g
g
ested
a
d
ee
p
n
eu
r
al
n
etwo
r
k
(
DNN
)
-
b
ased
s
to
ck
p
r
ed
icti
o
n
m
o
d
el
en
h
an
ce
d
f
o
r
th
e
b
ig
d
ata
s
y
s
tem
s
u
s
ed
b
y
f
in
an
cial
o
r
g
a
n
izatio
n
s
an
d
in
ter
n
et
-
b
ased
f
in
an
cial
f
ac
ilit
ies.
B
ased
o
n
b
ig
d
ata
tec
h
n
o
lo
g
ies,
th
is
wo
r
k
o
f
f
er
e
d
ex
is
tin
g
ten
d
en
cies
in
p
r
o
ce
s
s
in
g
a
s
well
as
an
aly
s
i
s
p
latf
o
r
m
s
th
r
o
u
g
h
t
h
e
wo
r
ld
,
p
r
esen
tin
g
in
s
ig
h
ts
in
to
a
m
o
d
u
lar
ly
ca
lcu
lated
b
ig
d
ata
s
tag
e
f
o
r
c
o
m
m
er
cia
l
b
an
k
s
.
Als
o
,
it
p
r
o
p
o
s
ed
a
c
o
n
v
o
lu
ti
o
n
al
n
eu
r
al
n
etwo
r
k
(
C
NN
)
m
o
d
el,
f
o
cu
s
in
g
o
n
th
eir
ex
p
er
im
e
n
tal
p
ar
am
eter
o
p
tim
izatio
n
,
p
r
in
ci
p
les
,
an
d
s
tr
u
ctu
r
es.
Fin
ally
,
t
h
e
p
r
o
p
o
s
ed
s
to
ck
p
r
e
d
ictio
n
m
o
d
el
d
em
o
n
s
tr
ates
b
etter
p
er
f
o
r
m
a
n
ce
t
h
an
co
n
v
e
n
tio
n
al
m
eth
o
d
s
with
r
esp
ec
t
t
o
ac
cu
r
ac
y
,
v
er
if
y
in
g
its
ef
f
ici
en
cy
a
n
d
f
o
cu
s
in
g
on
its
d
ep
en
d
ab
le
ar
r
a
n
g
em
e
n
t w
ith
ex
p
er
im
en
tal
r
esu
lts
.
I
n
2
0
2
4
,
W
an
g
et
a
l.
[
1
2
]
f
o
c
u
s
ed
o
n
th
e
d
y
n
a
m
ics
am
o
n
g
b
lo
ck
ch
ain
,
n
etwo
r
k
tr
u
s
t,
u
s
e
r
id
en
tity
,
in
f
o
r
m
atio
n
s
ec
u
r
ity
,
a
n
d
b
ig
d
ata
in
th
e
d
ig
ital
ec
o
n
o
m
y
.
A
co
m
b
in
ed
m
o
d
el
o
f
d
ig
ital
id
en
tity
attr
ib
u
tes
is
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
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
8
5
2
-
3
8
6
4
3854
d
ev
elo
p
e
d
,
with
th
r
ee
co
r
e
an
d
n
in
e
ad
d
itio
n
al
d
im
en
s
io
n
s
,
an
d
a
d
eleg
ated
p
r
o
x
y
s
y
s
tem
en
ab
lin
g
co
o
p
er
atio
n
b
etwe
en
g
o
v
er
n
m
en
t
an
d
p
r
i
v
ate
b
lo
ck
c
h
ain
all
ian
ce
s
.
T
h
is
p
ap
er
also
cr
ea
te
d
a
d
ig
ital
id
en
tity
m
o
d
el
with
th
r
ee
m
ain
d
im
en
s
io
n
s
an
d
n
in
e
s
u
p
p
o
r
tin
g
d
im
en
s
io
n
s
,
wh
ich
in
tr
o
d
u
ce
d
a
p
r
o
x
y
-
b
ased
m
o
d
e
l
am
o
n
g
g
o
v
e
r
n
m
en
tal
a
n
d
n
o
n
-
g
o
v
er
n
m
e
n
tal
b
lo
ck
c
h
ain
o
r
g
an
izatio
n
s
.
I
t
r
e
d
esig
n
s
an
ex
ten
s
iv
e
b
ig
d
ata
s
u
p
p
ly
ch
ain
m
o
d
el
an
d
cr
ea
t
es
a
s
ec
u
r
e
b
lo
ck
ch
ain
d
i
g
ital
id
en
titi
es
p
r
o
ce
d
u
r
e
u
n
d
e
r
th
e
ze
r
o
-
tr
u
s
t
m
o
d
el,
wh
ich
o
f
f
e
r
s
n
ew
s
ec
u
r
ity
g
o
v
er
n
an
ce
s
o
lu
tio
n
s
f
o
r
th
e
d
ig
ital
ec
o
n
o
m
y
,
m
etav
er
s
e
,
a
n
d
d
ig
ital
g
o
v
e
r
n
m
en
t
on
an
i
n
ter
n
atio
n
al
s
ca
le.
I
n
2
0
2
3
,
Al
B
an
n
a
et
a
l.
[
1
3
]
p
r
o
p
o
s
ed
to
ex
am
in
e
th
e
m
en
t
al
h
ea
lth
co
n
s
eq
u
e
n
ce
s
o
f
th
e
C
OVI
D
-
19
p
an
d
em
ic
.
T
h
e
s
tu
d
y
was
f
o
c
u
s
ed
o
n
d
e
p
r
ess
io
n
,
o
v
er
th
e
X
d
ata
an
al
y
s
is
.
T
h
is
wo
r
k
o
b
tain
ed
an
im
p
r
ess
iv
e
h
ig
h
ac
cu
r
ac
y
in
id
en
tif
y
in
g
d
ep
r
ess
iv
e
twee
ts
b
y
em
p
lo
y
i
n
g
an
in
n
o
v
ativ
e
p
ip
elin
e,
wh
ich
co
m
b
in
es
C
NN
an
d
lo
n
g
s
h
o
r
t
-
te
r
m
m
em
o
r
y
(
L
STM
)
.
I
n
2
0
2
3
,
J
aisw
al
et
a
l
.
[
1
4
]
p
r
o
p
o
s
e
d
b
i
g
d
at
a
f
u
s
i
o
n
m
et
h
o
d
s
an
d
e
n
s
e
m
b
le
le
ar
n
i
n
g
f
o
r
b
r
e
ast
c
a
n
ce
r
r
is
k
p
r
e
d
i
cti
o
n
as
we
ll
as
c
lass
if
i
ca
t
io
n
.
A
n
o
v
el
I
-
e
x
t
r
e
m
e
g
r
ad
i
e
n
t
b
o
o
s
ti
n
g
(
XGBo
o
s
t
)
m
e
th
o
d
was
p
r
o
p
o
s
e
d
in
t
h
is
wo
r
k
,
wh
ic
h
c
o
n
ta
in
s
a
d
at
a
p
r
ep
r
o
c
ess
i
n
g
p
h
as
e,
f
ea
t
u
r
e
e
x
t
r
a
cti
o
n
t
ec
h
n
iq
u
es
,
an
d
an
i
d
en
ti
f
ie
d
t
ar
g
et
r
o
le
t
o
i
m
p
r
o
v
e
t
h
e
a
cc
u
r
a
cy
o
f
d
i
ag
n
o
s
i
n
g
b
r
ea
s
t
ca
n
c
er
.
T
h
e
s
u
g
g
este
d
m
o
d
el
s
h
o
we
d
s
u
p
e
r
i
o
r
p
e
r
f
o
r
m
a
n
c
e
in
te
r
m
s
o
f
h
i
g
h
e
r
ac
cu
r
a
c
y
t
h
a
n
o
t
h
er
cl
ass
i
f
ic
ati
o
n
m
o
d
els
li
k
e
DT
,
r
a
n
d
o
m
f
o
r
est
(
RF
)
,
a
n
d
S
VM
.
I
n
2
0
2
5
,
T
h
a
t
h
a
et
a
l
.
[
1
5
]
p
r
o
p
o
s
e
d
d
e
c
i
s
i
v
e
r
e
d
f
o
x
-
b
a
s
e
d
f
e
a
t
u
r
e
s
e
l
e
ct
i
o
n
(
DR
FB
PS
)
w
h
i
c
h
w
a
s
d
e
v
e
l
o
p
e
d
i
n
P
y
t
h
o
n
,
u
s
i
n
g
d
a
t
a
p
r
e
p
r
o
c
e
s
s
i
n
g
a
n
d
r
e
d
f
o
x
o
p
t
i
m
i
z
a
ti
o
n
f
o
r
f
e
a
t
u
r
e
s
e
l
ec
t
i
o
n
.
I
t
a
n
a
l
y
z
e
d
r
i
s
k
f
a
c
t
o
r
s
a
n
d
m
a
d
e
p
r
e
d
i
c
t
i
o
n
s
,
v
a
l
i
d
a
t
e
d
t
h
r
o
u
g
h
m
e
t
r
i
c
s
l
i
k
e
p
r
e
c
i
s
i
o
n
,
a
r
ea
u
n
d
e
r
t
h
e
c
u
r
v
e
(
A
UC
)
,
F
1
-
s
c
o
r
e
,
r
e
c
a
l
l
,
a
cc
u
r
a
c
y
,
a
n
d
e
r
r
o
r
r
a
t
e
.
T
h
e
s
y
s
t
e
m
d
e
m
o
n
s
t
r
a
t
e
d
it
s
p
o
t
e
n
t
i
a
l
i
n
h
e
a
l
t
h
c
a
r
e
a
n
a
l
y
ti
cs
,
o
f
f
e
r
i
n
g
a
c
c
u
r
a
t
e
,
r
e
l
i
a
b
l
e
p
r
e
d
i
ct
i
o
n
s
.
O
v
e
r
a
l
l
,
DR
FB
P
S
p
r
o
v
e
d
t
o
b
e
a
p
r
o
m
i
s
in
g
t
o
o
l
f
o
r
h
e
a
r
t
d
i
s
e
as
e
r
is
k
p
r
e
d
i
c
t
i
o
n
.
I
n
2
0
2
5
,
Palan
is
am
y
et
al
.
[
1
6
]
d
ev
elo
p
ed
h
y
b
r
id
f
r
am
ewo
r
k
co
m
b
in
i
n
g
lo
c
al
h
eter
o
g
en
eit
y
-
o
r
ien
te
d
(
L
HO
)
f
o
r
f
ea
tu
r
e
ex
tr
ac
tio
n
an
d
h
eter
o
g
e
n
eo
u
s
g
r
a
p
h
em
b
ed
d
in
g
-
s
im
ilar
ity
n
etwo
r
k
f
u
s
io
n
(
HGE
-
SNF
)
f
o
r
r
o
b
u
s
t
d
is
ea
s
e
p
r
ed
ictio
n
.
D
ev
elo
p
ed
i
n
MA
T
L
AB
R
2
0
2
3
b
,
it
was
test
ed
o
n
d
atas
ets
f
o
r
C
OVI
D
-
1
9
.
I
t
attain
ed
h
ig
h
p
er
f
o
r
m
a
n
c
e
with
an
AUC
u
p
t
o
0
.
9
9
5
a
n
d
a
s
en
s
itiv
ity
/
s
p
ec
if
icity
o
f
ab
o
v
e
9
8
%.
A
co
m
p
ar
ativ
e
a
n
aly
s
is
s
h
o
wed
its
s
u
p
er
io
r
ity
o
v
er
co
n
v
en
ti
o
n
al
m
o
d
els.
I
n
2
0
2
5
,
Qi
[
1
7
]
in
te
g
r
ated
ML
m
eth
o
d
s
with
cl
o
u
d
-
b
ased
in
ter
n
et
o
f
t
h
in
g
s
(
C
-
I
o
T
)
s
y
s
tem
s
f
o
r
h
ea
lth
d
ata
s
ec
u
r
ity
.
I
t
u
s
ed
a
lig
h
tweig
h
t
e
n
cr
y
p
tio
n
b
lo
ck
an
d
h
elp
ed
d
iag
n
o
s
e
h
ea
lth
i
s
s
u
es
b
ased
o
n
t
h
e
p
atien
t'
s
h
i
s
to
r
ical
d
atab
ase.
I
t
attain
ed
ap
p
r
o
x
im
ately
9
1
%
ac
cu
r
ac
y
v
ia
ANN
alg
o
r
i
th
m
s
.
T
h
e
s
y
s
tem
p
r
o
v
id
e
d
a
s
ec
u
r
e
an
d
ef
f
icien
t
way
to
m
a
n
ag
e
h
ea
lth
d
ata.
T
h
is
ex
p
an
s
io
n
r
ep
r
esen
ted
a
s
tep
f
o
r
war
d
in
th
e
m
o
d
er
n
izatio
n
o
f
s
o
ci
ety
5
.
0
.
I
n
2
0
2
5
,
Ma
n
i
k
an
d
a
n
et
a
l.
[
1
8
]
d
ev
elo
p
e
d
d
ee
p
co
n
v
o
lu
tio
n
al
n
eu
r
al
n
etwo
r
k
(
DC
NN)
-
L
STM
,
wh
ich
au
to
m
ated
d
iv
er
s
e
h
e
alth
ca
r
e
f
u
n
ctio
n
s
lik
e
in
f
r
a
s
tr
u
ctu
r
e
m
an
ag
e
m
en
t,
p
atie
n
t
m
o
n
ito
r
i
n
g
,
an
d
d
is
ea
s
e
p
r
ed
ictio
n
.
Mo
r
eo
v
er
,
d
iv
er
s
e
d
ata
ty
p
es
s
u
ch
as
alp
h
an
u
m
er
ic,
n
u
m
er
ic,
s
ig
n
als,
v
id
eo
,
an
d
im
a
g
es
ar
e
an
aly
ze
d
b
y
th
is
m
o
d
el
t
o
ac
cu
r
ately
f
in
d
o
u
t
th
e
ab
n
o
r
m
al
co
n
d
itio
n
s
an
d
p
r
ed
ic
t
th
e
d
is
ea
s
es.
Fo
r
m
ed
ical
p
er
s
o
n
al,
th
e
r
ea
l
-
tim
e
aler
ts
wer
e
also
g
en
e
r
ated
b
y
th
is
s
y
s
tem
,
wh
ich
en
s
u
r
es
tim
ely
in
ter
v
en
tio
n
.
I
n
ad
d
itio
n
,
9
6
% c
ateg
o
r
izatio
n
ac
cu
r
ac
y
was a
ttain
ed
b
y
t
h
is
DC
NN
-
L
STM
m
o
d
el.
I
n
2
0
2
5
,
J
am
eil
an
d
Al
-
R
awe
s
h
id
y
[
1
9
]
d
e
v
elo
p
e
d
a
r
ea
l
-
tim
e
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
s
y
s
tem
,
wh
ich
u
tili
ze
s
th
e
d
ig
ital
twin
.
T
h
is
m
o
d
el
u
tili
ze
d
th
e
in
ter
n
et
o
f
th
in
g
s
(
I
o
T
)
f
o
r
d
ata
co
llectio
n
,
s
ca
lab
le
d
ata
p
r
o
ce
s
s
in
g
u
s
in
g
clo
u
d
c
o
m
p
u
tin
g
,
a
n
d
ac
cu
r
ate
h
ea
lth
p
r
e
d
ictio
n
u
s
in
g
ad
v
an
ce
d
alg
o
r
ith
m
s
.
T
h
is
s
ea
m
less
in
teg
r
atio
n
was
r
ea
lized
v
ia
b
o
th
d
ig
ital
an
d
p
h
y
s
ical
ar
ch
itectu
r
es
,
in
clu
d
in
g
s
en
s
in
g
eq
u
ip
m
e
n
t,
m
u
lti
-
o
b
jectiv
e
alg
o
r
ith
m
s
,
cl
o
u
d
-
b
ased
s
to
r
ag
e,
r
ea
l
-
tim
e
d
ata
s
y
n
th
esis
,
an
d
in
tu
itiv
e
d
ash
b
o
ar
d
a
n
aly
tics
.
A
X
G
B
o
o
s
t
h
y
b
r
id
m
o
d
el
wa
s
in
tr
o
d
u
ce
d
,
wh
ich
was
th
e
co
m
b
in
atio
n
o
f
b
o
th
m
er
g
es
b
o
th
XGBo
o
s
t
an
d
ML
P m
o
d
els,
th
at
ad
d
r
ess
es t
h
e
ch
allen
g
es r
elate
d
t
o
h
is
to
r
i
ca
l a
n
d
r
ea
l
-
tim
e
d
ata
an
aly
s
is
.
3.
P
RO
P
O
SE
D
B
I
G
DATA C
L
ASSI
F
I
CA
T
I
O
N
F
RA
M
E
W
O
RK
I
n
th
is
p
r
o
p
o
s
ed
wo
r
k
,
a
n
o
v
el
b
ig
d
ata
h
ea
lth
ca
r
e
m
o
d
e
l
is
p
r
o
p
o
s
ed
b
ased
o
n
th
e
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
m
o
d
el
u
n
d
e
r
th
r
ee
p
h
ases
:
i)
I
n
itially
,
th
e
in
p
u
t h
ea
lth
ca
r
e
d
ata
is
p
r
ep
r
o
ce
s
s
ed
u
s
in
g
im
p
r
o
v
ed
m
in
-
m
a
x
n
o
r
m
aliza
tio
n
.
ii)
Seco
n
d
ly
,
th
e
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
is
em
p
lo
y
ed
to
h
a
n
d
le
b
i
g
d
ata
.
T
h
e
m
a
p
p
er
p
h
ase
co
n
tain
s
an
im
p
r
o
v
e
d
f
ea
tu
r
e
f
u
s
io
n
t
ec
h
n
iq
u
e
u
s
in
g
im
p
r
o
v
e
d
SVM
-
R
FE
an
d
th
e
r
ed
u
ce
r
p
h
ase
m
er
g
in
g
th
e
f
u
s
ed
f
ea
tu
r
es.
iii)
L
astl
y
,
th
e
f
u
s
ed
f
ea
tu
r
es
ar
e
in
p
u
t
in
to
th
e
d
e
e
p
Q
-
n
e
t
w
o
r
k
(
DQN
)
cla
s
s
if
icat
io
n
m
o
d
el
to
o
b
tain
th
e
f
in
al
class
if
icatio
n
o
u
tp
u
t.
T
h
e
en
tire
b
ig
d
ata
class
if
icatio
n
p
r
o
ce
s
s
is
d
ep
icted
in
Fig
u
r
e
1.
3
.
1
.
P
re
pro
ce
s
s
ing
ph
a
s
e
Pre
p
r
o
ce
s
s
in
g
is
th
e
p
r
im
ar
y
s
tep
f
o
r
p
r
o
ce
s
s
in
g
n
u
m
er
o
u
s
d
ata
f
o
r
class
if
icatio
n
.
I
n
th
is
p
ap
er
,
a
h
ea
lth
ca
r
e
d
ataset
r
ef
er
r
ed
t
o
as
,
wh
ich
is
u
s
ed
f
o
r
b
ig
d
at
a
class
if
icat
io
n
.
T
h
e
d
ata
is
n
o
r
m
alize
d
u
s
in
g
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
Th
e
r
o
le
o
f b
ig
d
a
ta
in
p
r
ec
is
io
n
med
icin
e
a
n
d
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
… (
Meen
a
ksh
i
S
a
n
k
a
r
a
s
u
b
r
a
ma
n
ia
n
)
3855
th
e
m
o
d
if
ied
m
in
-
m
ax
n
o
r
m
al
izatio
n
m
eth
o
d
[
2
0
]
.
T
h
is
tech
n
iq
u
e
ad
j
u
s
ts
th
e
v
alu
es
o
f
th
e
d
ataset
b
ased
o
n
its
m
in
im
u
m
a
n
d
m
ax
im
u
m
v
alu
es.
B
y
s
tan
d
ar
d
izin
g
t
h
e
s
m
allest
v
alu
es
to
0
a
n
d
t
h
e
lar
g
est
v
alu
es
to
1
,
th
is
ap
p
r
o
ac
h
e
n
s
u
r
es
th
at
all
o
th
er
d
ata
p
o
i
n
ts
ar
e
p
r
o
p
o
r
tio
n
ally
s
ca
led
in
th
e
r
an
g
e
[
0
-
1
]
.
I
n
(
1
)
is
u
s
ed
to
r
ep
r
esen
t th
e
f
o
r
m
u
la
f
o
r
m
in
-
m
ax
n
o
r
m
aliza
tio
n
.
̂
=
−
∗
(
1
)
I
n
(
1
)
,
̂
d
en
o
tes
th
e
n
o
r
m
alize
d
d
ata,
r
ep
r
esen
ts
th
e
in
p
u
t
v
alu
e,
s
ig
n
if
ies
th
e
s
m
allest
v
alu
e
in
th
e
d
ataset,
an
d
in
d
icate
s
th
e
la
r
g
est
v
alu
e
in
th
e
d
ataset.
T
h
e
r
ep
r
esen
ts
th
e
m
ed
ia
n
ab
s
o
lu
te
d
ev
iatio
n
as
(
2
)
o
f
th
e
in
p
u
t
d
ata,
wh
ich
is
less
s
en
s
iti
v
e
to
o
u
tlier
s
an
d
ap
p
lies
to
b
o
t
h
n
o
r
m
al
a
n
d
non
-
n
o
r
m
al
d
is
tr
ib
u
ted
d
ata.
=
(
|
−
̃
|
)
(
2
)
Her
e,
̃
r
ep
r
esen
ts
th
e
m
ea
n
o
f
t
h
e
in
p
u
t
d
ata.
T
h
is
n
o
r
m
aliza
tio
n
p
r
o
ce
s
s
en
s
u
r
e
s
u
n
if
o
r
m
it
y
an
d
s
tan
d
ar
d
izatio
n
o
f
th
e
d
ata,
f
a
cilitatin
g
m
o
r
e
ef
f
ec
tiv
e
an
aly
s
is
an
d
m
o
d
elin
g
.
T
h
e
p
r
e
p
r
o
c
ess
ed
d
ata
is
p
ar
titi
o
n
ed
in
to
m
u
ltip
le
d
ata
,
s
u
ch
as
(
1
,
2
,
…
,
)
d
u
r
in
g
th
is
p
r
ep
r
o
ce
s
s
in
g
p
h
ase.
E
ac
h
d
ata
is
s
u
b
jecte
d
to
th
e
Ma
p
R
ed
u
ce
f
r
a
m
ewo
r
k
.
Fig
u
r
e
1
.
B
ig
d
ata
class
if
icatio
n
p
r
o
ce
s
s
3
.
2
.
M
a
pReduce
f
ra
m
ewo
r
k f
o
r
ha
nd
lin
g
big
da
t
a
I
n
t
h
i
s
s
ec
o
n
d
p
h
a
s
e
,
M
a
p
R
e
d
u
c
e
f
r
a
m
e
wo
r
k
i
s
a
p
p
l
ie
d
to
e
f
f
i
c
i
e
n
t
ly
p
r
o
c
e
s
s
an
d
h
an
d
l
e
t
h
e
b
i
g
d
a
t
a.
T
h
e
Ma
p
R
e
d
u
ce
f
r
a
m
e
wo
r
k
i
s
s
p
l
i
t
i
n
to
t
wo
m
a
i
n
p
h
a
s
e
s
s
u
ch
a
s
m
ap
p
er
p
h
a
s
e
an
d
r
ed
u
c
er
p
h
a
s
e
.
T
h
e
m
ap
p
e
r
p
h
a
s
e
co
n
s
i
s
t
s
o
f
i
m
p
r
o
v
ed
f
ea
t
u
r
e
f
u
s
i
o
n
u
s
in
g
a
d
e
e
p
r
e
s
i
d
u
a
l
n
e
t
w
o
r
k
,
an
d
th
e
r
e
d
u
c
e
r
p
h
a
s
e
f
o
c
u
s
e
s
o
n
m
e
r
g
in
g
th
e
f
u
s
e
d
f
e
a
tu
r
e
s
to
g
e
n
er
a
t
e
th
e
f
in
a
l
f
e
a
tu
r
e
.
F
i
g
u
r
e
2
d
ep
i
c
t
s
t
h
e
a
r
c
h
i
t
ec
t
u
r
e
o
f
M
a
p
R
ed
u
c
e
f
r
a
m
e
wo
r
k
.
T
h
e
s
t
e
p
-
by
-
s
te
p
p
r
o
c
e
s
s
o
f
t
h
e
Ma
p
R
e
d
u
c
e
f
r
a
m
e
wo
r
k
i
s
s
t
r
u
c
t
u
r
e
d
a
s
f
o
l
l
o
w
s
:
i)
Step
1
: to
s
p
litt
in
g
th
e
o
r
ig
in
al
in
p
u
t d
ata,
ii)
Step
2
:
th
e
d
iv
i
d
ed
d
ata
is
in
p
u
tted
in
t
o
th
e
m
ap
p
er
p
h
ase,
h
er
e
an
im
p
r
o
v
ed
f
ea
tu
r
e
f
u
s
io
n
tech
n
iq
u
e
u
s
in
g
an
SVM
-
R
FE.
T
h
is
p
r
o
c
ess
im
p
r
o
v
es f
ea
tu
r
e
e
x
tr
ac
tio
n
f
o
r
b
etter
r
e
p
r
esen
tatio
n
.
iii)
Step
3
: su
b
s
eq
u
en
tly
,
t
h
e
s
h
u
f
f
lin
g
p
r
o
ce
s
s
is
u
s
ed
to
th
e
in
t
er
m
ed
iate
o
u
tp
u
t f
r
o
m
th
e
m
a
p
p
er
p
h
ase.
iv
)
Step
4
:
f
in
ally
,
th
e
f
u
s
ed
f
ea
t
u
r
es
ar
e
m
er
g
e
d
o
r
c
o
n
ca
ten
a
ted
in
t
h
e
r
e
d
u
ce
r
p
h
ase.
T
h
e
s
e
f
ea
tu
r
es
ar
e
s
u
b
jecte
d
to
th
e
class
if
icatio
n
p
h
ase.
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
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
8
5
2
-
3
8
6
4
3856
Fig
u
r
e
2
.
Ar
c
h
itectu
r
e
o
f
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
3
.
2
.
1
.
M
a
pp
er
ph
a
s
e
I
n
g
en
e
r
al,
th
e
Ma
p
R
ed
u
ce
a
lg
o
r
ith
m
is
ex
ce
llen
t
f
o
r
m
in
in
g
h
u
g
e
am
o
u
n
ts
o
f
d
atasets
th
at
ar
e
lar
g
er
th
an
a
p
etab
y
te
an
d
ca
n
n
o
t
b
e
p
h
y
s
ically
k
ep
t
in
m
em
o
r
y
.
I
n
o
r
d
er
to
d
etec
t
th
e
ch
ar
ac
ter
is
tics
o
f
m
ass
iv
e
co
llectio
n
s
o
f
h
ea
lth
in
f
o
r
m
atio
n
,
th
e
Ma
p
R
ed
u
ce
f
u
n
ctio
n
is
em
p
lo
y
ed
in
th
is
s
tu
d
y
.
Acc
o
r
d
i
n
g
ly
,
th
e
in
p
u
t
d
ata
is
s
p
lit
an
d
d
is
tr
ib
u
ted
to
th
e
m
a
p
p
er
s
,
wh
er
e
p
ar
allel
p
r
o
ce
s
s
in
g
is
ap
p
lied
d
u
r
in
g
th
e
f
ea
tu
r
e
ex
tr
ac
tio
n
s
tag
es.
T
h
e
b
ig
d
ata
is
h
an
d
led
u
s
in
g
th
e
Ma
p
R
ed
u
ce
f
r
a
m
ewo
r
k
.
Ma
p
R
ed
u
ce
[
2
1
]
is
a
p
r
o
g
r
am
m
in
g
ap
p
r
o
ac
h
u
s
ed
in
a
d
is
tr
ib
u
ted
s
y
s
tem
f
o
r
co
m
p
u
tin
g
[
2
2
]
to
h
an
d
le
en
o
r
m
o
u
s
am
o
u
n
ts
o
f
d
ata.
I
t'
s
em
p
lo
y
ed
f
o
r
m
an
a
g
in
g
d
ata
th
at
ca
n
n
o
t b
e
ac
co
m
m
o
d
ated
in
p
h
y
s
ical
m
em
o
r
y
.
T
h
ese
two
o
p
e
r
atio
n
s
ar
e
ca
r
r
ied
o
u
t
i
n
two
p
h
ases
,
ea
ch
o
f
wh
ich
is
f
o
llo
wed
b
y
a
d
ata
t
r
an
s
f
er
b
etwe
en
clu
s
ter
n
o
d
es.
D
ata
in
th
e
ty
p
e
o
f
"k
ey
,
v
alu
e"
p
air
s
is
p
r
o
ce
s
s
ed
in
p
ar
allel
th
r
o
u
g
h
o
u
t
th
ese
p
h
ases
.
Fro
m
th
e
d
ataset,
th
e
m
ap
f
u
n
cti
o
n
g
ets
th
e
in
p
u
t a
n
d
p
r
o
d
u
ce
s
in
ter
m
ed
iar
y
o
u
tp
u
t.
T
h
e
m
ap
f
u
n
cti
o
n
'
s
o
u
tp
u
t is o
r
g
an
ized
t
o
b
e
u
s
ed
as in
p
u
t f
o
r
th
e
r
ed
u
ce
r
f
u
n
ctio
n
.
T
h
is
in
v
o
lv
e
s
d
ata
ex
ch
an
g
es
am
o
n
g
m
a
p
s
an
d
r
ed
u
ce
s
o
p
er
atio
n
s
.
T
h
e
v
alu
es
ar
e
g
ath
er
e
d
at
ea
ch
n
o
d
e
wh
ile
th
e
r
ed
u
ct
io
n
f
u
n
ctio
n
is
ex
ec
u
ted
f
o
r
a
ce
r
tain
k
ey
.
T
h
e
r
ed
u
ce
f
u
n
c
tio
n
also
g
en
er
ates
th
e
f
in
al
r
esu
lt
in
th
e
f
o
r
m
"k
e
y
,
v
al
u
e"
[
2
1
]
.
Ma
p
R
ed
u
ce
f
u
n
ctio
n
s
ar
e
em
p
lo
y
ed
in
th
is
s
t
u
d
y
to
d
is
co
v
er
t
h
e
f
ea
tu
r
es
o
f
b
ig
d
ata
[
2
2
]
.
Acc
o
r
d
in
g
l
y
,
th
e
m
ap
p
er
p
h
ase
in
clu
d
es
th
e
ex
tr
ac
tio
n
o
f
f
ea
tu
r
es
lik
e
r
aw
f
ea
tu
r
es
an
d
s
tatis
tical
f
ea
tu
r
es.
i)
R
aw
f
ea
tu
r
es:
th
ese
f
ea
tu
r
es
r
ep
r
esen
t
th
e
o
r
ig
in
al
in
p
u
t
d
at
a
an
d
ar
e
s
y
m
b
o
lized
b
y
t
he
r
aw
d
ata
is
u
s
ed
to
an
aly
ze
th
e
o
r
i
g
in
al
p
ar
am
eter
s
o
f
th
e
p
atien
t
h
ea
lth
r
ec
o
r
d
an
d
p
r
ec
is
io
n
m
e
d
icin
e
,
wh
ich
r
elies
o
n
an
aly
s
is
o
f
lar
g
e
d
ata
s
ets
to
id
en
tify
clin
ically
-
r
elev
an
t
s
u
b
-
g
r
o
u
p
s
an
d
c
o
n
tr
ib
u
te
to
d
is
co
v
er
in
g
d
is
ea
s
e
m
ec
h
an
is
m
s
.
Ad
v
an
c
in
g
p
r
ec
is
io
n
m
e
d
icin
e
b
y
e
n
ab
lin
g
m
ea
s
u
r
e
m
en
t
o
f
h
ig
h
in
f
o
r
m
atio
n
co
n
ten
t,
clin
ically
r
elev
a
n
t
p
ar
am
eter
s
:
h
ea
r
t
r
ate,
r
esp
i
r
atio
n
r
ate,
b
lo
o
d
p
r
ess
u
r
e,
tem
p
er
atu
r
e,
m
y
o
ca
r
d
ial
in
f
ar
ctio
n
,
s
tr
o
k
e
,
an
d
b
o
d
y
m
ass
in
d
ex
(
B
MI
)
.
ii)
Statis
t
ical
f
ea
tu
r
es:
th
e
d
er
iv
ed
f
ea
tu
r
es
co
n
tain
m
ed
ian
,
m
ea
n
,
s
k
ewn
ess
,
k
u
r
to
s
is
,
an
d
s
tan
d
ar
d
d
ev
iatio
n
(
SD
)
f
ea
tu
r
es.
T
h
ese
f
ea
tu
r
es a
r
e
s
y
m
b
o
lized
b
y
.
Af
ter
d
er
iv
in
g
th
e
f
ea
tu
r
es in
th
e
m
ap
p
er
,
th
e
r
ed
u
ce
r
p
h
ase
co
m
b
in
es
t
h
e
f
ea
tu
r
es
f
r
o
m
all
th
e
m
ap
p
er
s
an
d
p
r
o
v
id
es
th
e
c
o
m
b
in
e
d
f
ea
tu
r
e
s
et
th
at
is
d
ef
in
ed
as
=
[
,
]
.
iii)
Featu
r
e
s
elec
tio
n
b
y
u
s
in
g
im
p
r
o
v
e
d
SVM
-
R
FE
:
th
e
f
ea
tu
r
e
s
elec
tio
n
p
r
o
ce
s
s
is
a
cr
u
cial
asp
ec
t
o
f
ML
m
o
d
el
d
ev
el
o
p
m
en
t,
esp
ec
ial
ly
in
s
ce
n
ar
io
s
wh
er
e
th
e
d
i
m
en
s
io
n
ality
o
f
th
e
d
ataset
is
h
ig
h
.
I
n
th
e
p
r
o
p
o
s
ed
m
eth
o
d
,
th
e
r
aw
f
ea
t
u
r
e
co
n
tain
s
th
e
attr
ib
u
tes
o
f
t
h
e
o
r
i
g
in
al
d
ata
wh
ich
is
h
ig
h
-
d
im
en
s
io
n
al
.
Her
e,
an
im
p
r
o
v
ed
SVM
-
R
FE
[
2
3
]
is
g
ain
in
g
tr
ac
tio
n
f
o
r
its
ab
ilit
y
to
id
en
tify
th
e
m
o
s
t
d
is
cr
im
in
ativ
e
f
ea
tu
r
es wh
ile
co
n
s
id
er
in
g
v
ar
io
u
s
f
ac
to
r
s
.
Step
1
:
in
itializatio
n
T
h
e
f
ea
tu
r
e
s
elec
tio
n
p
r
o
ce
s
s
b
eg
in
s
b
y
in
itializin
g
th
e
o
r
ig
in
al
f
ea
tu
r
e
s
et
co
n
tain
in
g
f
ea
tu
r
e
s
in
d
ex
ed
f
r
o
m
1
to
,
r
ep
r
esen
tin
g
th
e
to
tal
n
u
m
b
e
r
o
f
f
ea
tu
r
es
in
th
e
d
ataset.
Ad
d
itio
n
ally
,
th
e
o
u
tp
u
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
Th
e
r
o
le
o
f b
ig
d
a
ta
in
p
r
ec
is
io
n
med
icin
e
a
n
d
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
… (
Meen
a
ksh
i
S
a
n
k
a
r
a
s
u
b
r
a
ma
n
ia
n
)
3857
f
ea
tu
r
e
o
r
d
er
e
d
s
et
is
in
itialized
to
e
v
en
tu
ally
co
n
tain
th
e
s
elec
ted
f
ea
tu
r
es
i
n
d
escen
d
in
g
o
r
d
er
o
f
im
p
o
r
tan
ce
.
Step
2
:
lo
o
p
in
g
p
r
o
ce
d
u
r
e
−
Ob
tain
a
tr
ain
in
g
s
et
with
th
e
ca
n
d
id
ate
f
ea
tu
r
e
s
et:
in
ea
ch
iter
atio
n
,
th
e
cu
r
r
en
t
f
ea
t
u
r
e
s
et
is
u
s
ed
to
tr
ain
th
e
SVM
class
if
i
er
,
en
s
u
r
in
g
co
n
s
id
er
atio
n
o
f
d
if
f
er
en
t f
ea
tu
r
e
s
u
b
s
ets.
−
T
r
ain
SVM
class
if
ier
to
o
b
tai
n
weig
h
t
v
ec
to
r
:
f
o
llo
win
g
tr
ain
in
g
,
a
weig
h
t
v
ec
t
o
r
is
o
b
ta
in
ed
f
r
o
m
th
e
SVM
class
if
ier
,
r
ep
r
esen
tin
g
th
e
im
p
o
r
tan
ce
o
f
ea
ch
f
ea
tu
r
e
f
o
r
th
e
class
if
icatio
n
t
ask
.
−
C
alcu
late
r
an
k
in
g
cr
iter
ia
s
co
r
e
:
in
th
e
co
n
v
en
tio
n
al
ap
p
r
o
ac
h
,
th
e
r
an
k
in
g
cr
iter
ia
s
co
r
e
is
ca
lcu
lated
b
y
u
s
in
g
(
3
)
.
W
er
e,
r
an
g
es f
r
o
m
1
to
th
e
n
u
m
b
er
o
f
f
ea
tu
r
es in
.
=
2
(
3
)
T
h
is
ap
p
r
o
ac
h
s
o
lely
r
elies
o
n
th
e
s
q
u
ar
e
o
f
ea
ch
f
ea
tu
r
e'
s
weig
h
t,
p
o
ten
tially
m
is
s
i
n
g
o
th
er
im
p
o
r
tan
t
f
ac
to
r
s
.
In
(
4
)
d
ef
in
es
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
,
h
er
e,
ad
d
itio
n
al
f
ac
to
r
s
,
s
u
c
h
as
th
e
in
ter
q
u
ar
tile
m
e
an
(
μ
)
an
d
SD
(
σ
)
,
a
r
e
in
c
o
r
p
o
r
ated
[
2
4
]
.
T
h
is
ac
co
u
n
ts
f
o
r
b
o
th
lin
ea
r
an
d
non
-
lin
ea
r
co
r
r
elatio
n
s
am
o
n
g
f
ea
tu
r
es,
o
v
e
r
co
m
in
g
th
e
lim
i
tatio
n
s
o
f
th
e
co
n
v
en
tio
n
al
m
e
th
o
d
.
=
{
2
+
[
|
|
+
1
∗
|
+
(
)
−
−
(
)
|
+
(
)
+
−
(
)
]
}
(
4
)
−
Fin
d
th
e
s
m
allest
r
an
k
in
g
cr
iter
ia
s
co
r
e
f
ea
t
u
r
e:
th
e
f
ea
tu
r
e
with
th
e
s
m
allest
r
an
k
in
g
cr
iter
ia
s
co
r
e
is
id
en
tifie
d
u
s
in
g
(
5
)
,
in
d
icatin
g
th
e
f
ea
t
u
r
e
co
n
tr
ib
u
tin
g
th
e
least to
th
e
class
if
icatio
n
task
.
(
5
)
−
Up
d
ate
f
ea
tu
r
e
s
et
:
th
e
s
elec
ted
f
ea
tu
r
e
is
ad
d
ed
to
th
e
f
e
atu
r
e
o
r
d
er
ed
s
et
b
y
tak
in
g
th
e
u
n
io
n
with
t
h
e
s
et
,
co
n
tain
in
g
th
e
f
ea
tu
r
es selecte
d
in
th
is
it
er
atio
n
.
−
R
em
o
v
e
f
ea
tu
r
e
f
r
o
m
:
f
in
ally
,
th
e
s
elec
ted
f
e
atu
r
e
is
r
em
o
v
ed
f
r
o
m
t
h
e
cu
r
r
en
t
f
ea
tu
r
e
s
et
,
=
to
en
s
u
r
e
s
u
b
s
eq
u
e
n
t
iter
atio
n
s
co
n
s
id
er
a
r
ed
u
ce
d
s
et
o
f
f
ea
tu
r
es.
T
h
u
s
,
th
is
m
eth
o
d
f
u
l
ly
ex
p
lo
its
th
e
class
if
icatio
n
p
e
r
f
o
r
m
a
n
ce
o
f
ea
ch
f
ea
tu
r
e
s
u
b
s
et,
en
s
u
r
in
g
n
o
p
o
ten
tially
u
s
ef
u
l
in
f
o
r
m
atio
n
is
o
v
e
r
lo
o
k
e
d
.
B
y
co
n
s
id
er
in
g
b
o
t
h
lin
ea
r
an
d
n
o
n
-
lin
ea
r
co
r
r
elatio
n
s
am
o
n
g
f
ea
tu
r
es,
it
p
r
o
v
id
es
a
m
o
r
e
co
m
p
r
eh
en
s
iv
e
f
ea
tu
r
e
r
an
k
i
n
g
.
T
h
r
o
u
g
h
th
e
in
co
r
p
o
r
atio
n
o
f
ad
d
itio
n
al
f
ac
to
r
s
in
to
th
e
r
an
k
i
n
g
cr
iter
ia,
it
e
f
f
ec
tiv
ely
ad
d
r
ess
es
th
e
lim
it
atio
n
s
o
f
th
e
co
n
v
en
tio
n
al
a
p
p
r
o
ac
h
,
r
esu
ltin
g
in
a
m
o
r
e
r
o
b
u
s
t
f
e
atu
r
e
s
elec
tio
n
p
r
o
ce
s
s
.
Fin
ally
,
th
e
I
m
p
r
o
v
ed
SVM
-
R
FE
alg
o
r
ith
m
o
f
f
er
s
a
r
ef
in
ed
ap
p
r
o
ac
h
t
o
f
ea
tu
r
e
s
elec
tio
n
,
wh
ic
h
i
s
r
ep
r
esen
ted
b
y
,
en
h
an
cin
g
th
e
ef
f
ec
tiv
en
ess
o
f
th
e
p
r
o
ce
s
s
in
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
ap
p
licatio
n
s
b
y
co
n
s
id
er
i
n
g
a
wid
er
r
an
g
e
o
f
f
ac
to
r
s
an
d
cr
iter
ia.
T
h
e
p
a
r
am
eter
s
o
f
p
r
o
p
o
s
ed
SVM
-
R
FE
ar
e
a
s
tep
f
u
n
ctio
n
as
0
.
1
0
a
n
d
n
_
f
ea
tu
r
e
_
to
_
s
elec
t a
s
1
0
.
3
.
2
.
2
.
Reducer
ph
a
s
e
I
n
th
is
wo
r
k
,
th
e
r
ed
u
ce
r
p
h
ase
is
a
cr
itical
s
tep
in
th
e
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
,
wh
ich
is
r
esp
o
n
s
ib
le
f
o
r
co
m
b
in
in
g
th
e
ex
t
r
ac
ted
f
ea
tu
r
es
th
r
o
u
g
h
th
e
m
ap
p
e
r
p
h
ase.
T
h
e
r
ed
u
ce
r
p
h
ase’
s
o
u
t
p
u
t
is
a
f
in
al
f
ea
tu
r
e
s
et,
wh
ich
is
s
u
b
jecte
d
to
th
e
s
u
b
s
eq
u
en
t
p
h
ase.
T
h
u
s
,
t
h
e
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
en
s
u
r
es
s
ca
lab
ilit
y
an
d
ef
f
ec
tiv
en
ess
in
h
an
d
lin
g
b
i
g
d
ata
,
wh
er
ea
s
u
s
ag
e
o
f
I
m
p
r
o
v
e
d
SVM
-
R
FE
en
h
an
ce
s
f
ea
tu
r
e
f
u
s
io
n
q
u
ality
.
3
.
2
.
3
.
Da
t
a
cla
s
s
if
ica
t
io
n
Af
ter
p
er
f
o
r
m
in
g
f
ea
tu
r
e
f
u
s
io
n
,
th
e
m
er
g
e
d
o
u
tp
u
t
is
g
iv
en
as
an
in
p
u
t
to
t
h
e
p
h
ase
o
f
class
if
icatio
n
.
Her
e,
th
e
DQN
[
2
5
]
class
if
ier
is
u
s
ed
f
o
r
d
etec
tio
n
.
T
h
e
m
ain
ad
v
a
n
tag
e
o
f
u
s
in
g
th
e
DQN
class
if
ier
is
th
at
it
ca
n
ef
f
ec
t
iv
ely
d
etec
t
d
ia
b
etes
b
y
i
n
cr
ea
s
in
g
th
e
s
p
ee
d
o
f
tr
ain
in
g
,
th
er
eb
y
ac
h
iev
i
n
g
p
o
wer
f
u
l
r
esu
lts
.
DQN
[
2
5
]
is
a
Q
-
lear
n
in
g
m
ec
h
an
is
m
th
at
u
s
es
C
NN
to
m
ea
s
u
r
e
th
e
ac
t
io
n
v
alu
e
f
u
n
ctio
n
(Q
-
f
u
n
ctio
n
)
as
well
as
b
ein
g
th
e
b
asis
f
o
r
th
e
DQN
ar
ch
ite
ctu
r
e
an
d
t
r
ain
in
g
p
r
o
ce
d
u
r
es.
T
h
e
ad
v
a
n
tag
e
o
f
ex
p
lo
itin
g
DQN
is
th
at
it
u
tili
ze
s
th
e
r
ep
la
y
h
y
p
o
th
esis
to
m
itig
ate
th
e
is
s
u
es
b
ased
o
n
d
ata
d
is
tr
ib
u
tio
n
th
at
o
cc
u
r
in
th
e
n
etwo
r
k
.
T
h
e
a
r
ch
itectu
r
e
in
clu
d
es
v
ar
i
o
u
s
lay
er
s
,
s
u
ch
as
in
p
u
t
lay
er
s
,
co
n
v
o
l
u
tio
n
lay
er
s
,
d
r
o
p
o
u
t
lay
er
s
,
p
o
o
lin
g
lay
er
s
,
f
latten
in
g
lay
er
s
,
d
en
s
e
lay
er
s
,
an
d
o
u
tp
u
t
lay
er
s
.
T
h
e
d
ata
au
g
m
en
tatio
n
o
u
tco
m
e
is
tak
en
b
y
th
e
in
p
u
t
lay
er
a
n
d
t
h
e
o
u
tp
u
t
is
g
en
e
r
ated
in
t
h
e
d
im
e
n
s
io
n
o
f
[
1
0
×
1
]
.
I
n
ad
d
itio
n
,
two
d
if
f
er
en
t
la
y
er
s
o
f
c
o
n
v
o
lu
tio
n
ar
e
e
x
p
lo
ited
,
a
n
d
th
e
r
esu
lt
g
en
er
ated
b
y
th
e
in
p
u
t
l
ay
er
is
p
r
esen
ted
to
th
e
f
ir
s
t
lay
er
o
f
c
o
n
v
o
lu
tio
n
an
d
th
er
ea
f
ter
th
e
o
u
tp
u
t
is
ac
h
iev
ed
with
d
im
e
n
s
io
n
o
f
[
8
×
3
2
]
.
Als
o
,
th
e
s
ec
o
n
d
lay
er
r
ec
eiv
es
th
e
o
u
tp
u
t
o
f
th
e
p
r
i
o
r
co
n
v
o
lu
tio
n
lay
er
as
in
p
u
t
an
d
p
r
o
d
u
ce
s
o
u
tp
u
t
with
s
ize
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
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
8
5
2
-
3
8
6
4
3858
[
6
×
6
4
]
.
T
h
e
n
ex
t
lay
e
r
is
a
d
r
o
p
o
u
t
lay
er
th
at
d
ec
r
ea
s
es
th
e
p
r
o
b
lem
o
f
o
v
e
r
f
itti
n
g
.
B
esid
es,
th
e
d
r
o
p
o
u
t
lay
er
tak
es
th
e
o
u
tp
u
t
p
r
o
d
u
ce
d
b
y
th
e
s
ec
o
n
d
co
n
v
o
l
u
tio
n
l
ay
er
as
in
p
u
t
an
d
th
e
r
esu
lt
is
g
en
er
ated
with
s
ize
of
[
6
×
6
4
]
.
Fu
r
th
e
r
m
o
r
e,
th
e
o
u
tp
u
t
o
f
th
e
d
r
o
p
o
u
t
lay
e
r
is
p
r
esen
ted
to
th
e
m
ax
im
u
m
p
o
o
lin
g
lay
er
a
n
d
th
e
o
u
tco
m
e
attain
ed
is
in
th
e
d
i
m
en
s
io
n
o
f
[
6
×
6
4
]
,
wh
ich
is
f
ed
to
th
e
fl
atten
in
g
lay
e
r
.
T
h
e
lay
er
r
esu
lt
g
en
er
ated
f
r
o
m
f
latten
in
g
lay
e
r
h
as
th
e
d
im
en
s
io
n
o
f
[
1
×
3
8
4
]
.
I
n
ad
d
itio
n
,
th
e
Q
-
lear
n
in
g
s
tr
u
ctu
r
e
with
th
e
lo
s
s
f
u
n
ctio
n
is
(
6
)
.
(
)
=
[
(
+
′
̂
(
′
,
′
;
−
)
−
(
,
;
)
)
2
]
(
6
)
Her
e,
s
y
m
b
o
lizes
r
ewa
r
d
,
i
n
d
icate
s
d
is
co
u
n
t
f
ac
to
r
,
s
p
e
cif
ies
th
e
n
etwo
r
k
p
ar
am
eter
at
iter
atio
n
,
d
en
o
tes
ac
tio
n
,
an
d
im
p
lies
th
e
n
etwo
r
k
s
tate.
T
h
e
o
u
tco
m
e
ac
h
iev
e
d
b
y
th
e
f
latten
ed
lay
er
is
g
i
v
en
to
d
en
s
e
lay
er
.
He
r
e,
th
e
f
ir
s
t
an
d
s
ec
o
n
d
d
en
s
e
lay
e
r
s
p
r
o
d
u
ce
a
r
esu
lt
with
a
s
ize
[
1
×
1
2
8
]
.
T
h
u
s
,
th
e
o
u
tco
m
e
o
f
th
e
s
ec
o
n
d
d
en
s
e
lay
er
is
g
iv
en
to
th
ir
d
d
e
n
s
e
lay
er
.
T
h
is
lay
er
m
ea
s
u
r
es
th
e
o
u
tp
u
t
with
d
im
en
s
io
n
[
1
×
2
]
.
T
h
u
s
,
th
e
o
u
t
p
u
t p
r
o
d
u
ce
d
b
y
th
e
DQN
class
if
ier
is
r
ep
r
esen
ted
as
.
I
n
th
is
class
if
ier
,
th
e
co
n
v
o
lu
tio
n
lay
er
em
p
l
o
y
s
a
s
et
o
f
k
er
n
els
o
r
f
ea
tu
r
e
m
a
p
to
e
x
tr
ac
t
v
ital
f
ea
tu
r
e
f
r
o
m
d
ata
.
Po
o
lin
g
lay
er
s
r
ed
u
ce
th
e
f
ea
tu
r
e’
s
s
p
atial
d
im
en
s
io
n
,
wh
ich
en
s
u
r
es
co
m
p
u
tatio
n
a
l
ef
f
ec
tiv
en
ess
,
f
o
llo
wed
b
y
f
u
lly
co
n
n
ec
te
d
lay
er
s
.
Fin
ally
,
a
s
o
f
tm
ax
ac
tiv
atio
n
is
em
p
l
o
y
ed
i
n
th
e
o
u
tp
u
t
lay
er
an
d
th
e
p
r
ed
icted
o
u
tp
u
t
is
0
an
d
1
.
Acc
o
r
d
in
g
to
th
e
m
ed
ical
d
ataset,
th
e
v
al
u
e
o
f
“0
”
s
p
ec
if
ies
“n
o
r
m
al”
an
d
“1
”
s
p
ec
if
ies
“a
b
n
o
r
m
al”.
T
h
e
h
y
p
er
p
ar
am
eter
s
o
f
th
e
p
r
o
p
o
s
ed
DQN
m
o
d
el
is
g
iv
en
in
T
ab
le
1
.
W
ith
th
ese
h
y
p
er
p
ar
a
m
eter
s
,
m
o
d
el’
s
g
e
n
er
aliza
tio
n
an
d
r
o
b
u
s
tn
ess
ca
n
b
e
in
cr
ea
s
ed
.
T
ab
le
1
.
Hy
p
er
p
ar
a
m
eter
s
o
f
DQN
C
a
t
e
g
o
r
y
H
y
p
e
r
p
a
r
a
me
t
e
r
V
a
l
u
e
Q
-
n
e
t
w
o
r
k
p
a
r
a
me
t
e
r
s
H
i
d
d
e
n
l
a
y
e
r
2
Tr
a
i
n
i
n
g
p
a
r
a
m
e
t
e
r
s
N
o
d
e
s
p
e
r
l
a
y
e
r
50
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
S
o
f
t
ma
x
O
p
t
i
mi
z
e
r
A
d
a
ms
Le
a
r
n
i
n
g
r
a
t
e
0
.
0
0
1
O
t
h
e
r
p
a
r
a
me
t
e
r
s
B
a
t
c
h
si
z
e
32
E
p
o
c
h
s
50
4.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
4
.
1
.
Sim
ula
t
i
o
n
pro
ce
du
r
e
T
h
e
p
r
o
p
o
s
ed
b
ig
d
ata
class
if
icatio
n
was
s
im
u
lated
u
s
in
g
PYTH
ON,
s
p
ec
if
ically
“Ve
r
s
io
n
3
.
7
”
.
T
h
e
p
r
o
ce
s
s
o
r
u
tili
ze
d
was
1
1
th
Gen
I
n
tel(
R
)
C
o
r
e
(
T
M)
i5
-
1
1
3
5
G7
@
2
.
4
0
GHz
2
.
4
2
GH
z
,
an
d
th
e
in
s
talled
R
AM
s
ize
wa
s
1
6
.
0
GB
.
Fu
r
t
h
er
,
th
e
b
ig
d
ata
class
if
icatio
n
was
an
aly
ze
d
u
s
in
g
th
e
d
ia
b
e
tes
h
ea
lth
in
d
icato
r
s
d
atase
t
[
2
6
]
.
4
.
2
.
Da
t
a
s
et
des
cr
iptio
n
T
h
is
d
ataset
was
d
er
iv
ed
f
r
o
m
th
e
“CDC
's
B
R
F
SS
2
0
1
5
”
s
u
r
v
ey
.
Als
o
,
it
en
co
m
p
ass
es
3
f
iles
th
at
wer
e
d
escr
ib
ed
:
i)
“Dia
b
etes
_
0
1
2
_
h
ea
lth
_
in
d
icato
r
s
_
B
R
FS
S2
0
1
5
.
csv
:”
th
is
d
ataset
co
n
s
is
ts
o
f
2
5
3
,
6
8
0
s
u
r
v
ey
r
esp
o
n
s
es
,
an
d
it
u
tili
ze
s
th
r
ee
d
is
tin
ct
class
es
:
“0
in
d
icate
s
n
o
d
ia
b
etes
o
r
o
n
ly
d
u
r
in
g
p
r
e
g
n
an
cy
,
1
f
o
r
p
r
ed
iab
etes,
an
d
2
f
o
r
d
iab
etes
.
”
I
t
was im
b
alan
ce
d
a
n
d
co
m
p
r
is
es 2
1
f
ea
tu
r
e
v
ar
iab
les.
ii)
“d
iab
etes
_
b
in
a
r
y
_
5
0
5
0
s
p
lit
_
h
ea
lth
_
in
d
icato
r
s
_
B
R
FS
S2
0
1
5
.
csv
:”
th
is
d
a
taset
co
n
tain
s
7
0
,
6
9
2
s
u
r
v
ey
r
esp
o
n
s
es.
I
t
h
ad
an
e
q
u
iv
alen
t
5
0
-
5
0
s
p
lit
o
f
r
esp
o
n
d
en
ts
with
“
n
o
d
ia
b
etes
an
d
with
eith
er
p
r
ed
iab
etes
o
r
d
ia
b
ete
s
.
”
I
t
in
co
r
p
o
r
ates
2
class
es,
in
clu
d
in
g
“0
f
o
r
n
o
d
iab
et
es,
an
d
1
f
o
r
p
r
ed
iab
etes o
r
d
iab
etes
.
”
I
t
wa
s
a
b
alan
ce
d
d
ataset
an
d
co
n
s
i
s
ted
o
f
2
1
f
ea
tu
r
e
v
ar
iab
les.
iii)
“d
iab
etes
_
b
in
ar
y
_
h
ea
lth
_
in
d
icato
r
s
_
B
R
F
SS
2
0
1
5
.
c
s
v
:”
th
is
d
ataset
co
m
p
r
is
es
2
5
3
,
6
8
0
s
u
r
v
ey
r
esp
o
n
s
es.
I
t
in
clu
d
es
2
class
es
lik
e
“0
f
o
r
n
o
d
iab
etes,
an
d
1
f
o
r
p
r
ed
iab
etes
o
r
d
iab
etes.”
T
h
is
d
ataset
h
as 2
1
f
ea
tu
r
e
v
ar
ia
b
les an
d
it
was a
class
im
b
alan
ce
d
ataset.
I
n
th
is
in
v
esti
g
atio
n
,
t
h
e
to
tal
am
o
u
n
t
o
f
d
ata
u
tili
ze
d
was
4
8
,
0
0
0
an
d
it
co
n
s
is
ts
o
f
two
d
i
s
tin
ct
class
e
s
,
s
u
ch
as
n
o
r
m
al
(
lab
el
0
)
an
d
ab
n
o
r
m
al
(
lab
el
1
)
.
T
h
e
n
o
r
m
al
class
in
clu
d
es
2
3
,
9
0
0
d
ata
an
d
ab
n
o
r
m
al
class
co
n
tain
s
2
4
,
1
0
0
d
ata
.
Her
e
,
t
h
e
d
ataset
is
b
alan
ce
d
to
av
o
id
o
v
er
f
itti
n
g
a
n
d
g
en
er
aliza
tio
n
is
s
u
es.
4
.
2
.
1
.
P
er
f
o
r
m
a
nce
a
na
l
y
s
is
Fu
r
th
er
,
th
e
p
r
o
p
o
s
ed
b
ig
d
ata
class
if
icat
io
n
m
o
d
el
was
co
m
p
ar
ed
with
th
e
co
n
v
en
tio
n
al
tech
n
iq
u
es,
s
u
ch
as
KNN
[
6
]
,
o
p
tim
al
r
ec
u
r
r
en
t
n
eu
r
al
n
etwo
r
k
(
R
NN
)
[
2
7
]
,
lin
k
n
etwo
r
k
(
L
in
k
Net
)
,
b
id
ir
ec
tio
n
al
lo
n
g
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
Th
e
r
o
le
o
f b
ig
d
a
ta
in
p
r
ec
is
io
n
med
icin
e
a
n
d
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
… (
Meen
a
ksh
i
S
a
n
k
a
r
a
s
u
b
r
a
ma
n
ia
n
)
3859
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
Bi
-
L
ST
M
)
,
C
NN
,
an
d
L
eNe
t
,
r
esp
ec
tiv
ely
.
Mo
r
eo
v
er
,
it
was
ex
am
in
ed
with
r
esp
ec
t
to
s
p
ec
if
ic
ity
,
f
alse
n
eg
ativ
e
r
a
te
(
FNR
)
,
ac
cu
r
ac
y
,
Ma
tth
e
w’
s
co
r
r
elatio
n
c
o
ef
f
icien
t
(
MCC
)
,
an
d
o
th
e
r
p
er
f
o
r
m
an
ce
m
etr
ics.
Ad
d
iti
o
n
ally
,
th
e
ab
latio
n
an
d
s
tat
is
tical
ev
alu
atio
n
an
d
co
m
p
u
tatio
n
al
tim
e
wer
e
an
aly
ze
d
.
I
n
itially
,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
is
an
aly
ze
d
with
v
ar
io
u
s
lear
n
in
g
r
ate
with
8
0
%
tr
ain
in
g
d
ata
wh
ic
h
is
g
iv
en
in
Fig
u
r
e
3
.
B
y
o
b
s
er
v
in
g
t
h
e
f
ig
u
r
e,
t
h
e
ac
cu
r
a
cy
,
f
-
m
ea
s
u
r
e,
p
r
ec
is
io
n
,
a
n
d
MCC
is
h
ig
h
f
o
r
lear
n
in
g
r
ate
0
.
0
0
1
wh
e
n
co
m
p
ar
ed
to
o
t
h
er
lear
n
in
g
r
ate
s
.
T
h
e
p
r
o
p
o
s
ed
DQN
is
ev
a
lu
ated
with
v
ar
io
u
s
lear
n
in
g
r
ate
s
u
ch
as
0
.
1
,
0
.
0
1
,
0
.
0
5
,
a
n
d
0
.
0
0
1
.
T
h
e
h
y
p
er
p
ar
am
eter
tu
n
in
g
is
d
o
n
e
f
o
r
a
lear
n
in
g
r
ate
0
.
0
0
1
wh
ich
g
iv
es b
etter
ac
c
u
r
ac
y
a
n
d
p
r
ec
is
io
n
.
Fig
u
r
e
3
.
Per
f
o
r
m
an
c
e
an
aly
s
i
s
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
v
ar
io
u
s
lear
n
i
n
g
r
ates
4
.
2
.
2
.
Co
m
pa
r
a
t
iv
e
a
na
ly
s
is
o
n c
la
s
s
if
ica
t
io
n per
f
o
rm
a
nc
e
wit
h v
a
rio
us
t
ra
ini
ng
da
t
a
T
h
e
co
m
p
ar
ativ
e
ev
alu
atio
n
o
f
th
e
DQN
ap
p
r
o
ac
h
in
th
e
r
ea
lm
o
f
b
ig
d
ata
class
if
icatio
n
ex
p
o
s
es
f
ascin
atin
g
co
n
tr
ad
ictio
n
s
wh
e
n
co
n
tr
asted
ag
ain
s
t
ex
is
tin
g
s
tr
ateg
ies
,
KNN
[
6
]
,
o
p
tim
al
R
NN
[
2
7
]
,
L
in
k
Net
,
Bi
-
L
STM
,
C
NN
,
an
d
L
eNe
t.
T
h
e
ass
ess
m
en
ts
ar
e
co
n
d
u
cted
b
y
v
ar
y
in
g
th
e
tr
ain
in
g
d
ata
to
6
0
%,
7
0
%,
8
0
%
,
an
d
9
0
%.
W
ith
in
t
h
is
co
m
p
a
r
ativ
e
s
tu
d
y
,
th
e
ef
f
icien
cy
o
f
b
ig
d
ata
class
if
icatio
n
d
ep
e
n
d
s
u
p
o
n
its
ca
p
a
b
ilit
y
to
ac
h
iev
e
g
r
ea
ter
p
o
s
itiv
e
an
d
n
e
u
tr
al
r
atin
g
s
wh
ile
m
in
i
m
izin
g
n
e
g
ativ
e
m
ea
s
u
r
e
s
co
r
es.
Fo
r
6
0
%
o
f
th
e
tr
ain
in
g
d
ata,
am
o
n
g
th
e
m
o
d
e
ls
,
th
e
B
i
-
L
STM
,
C
NN,
L
in
k
Net
,
an
d
o
p
tim
al
R
NN
s
co
r
ed
less
er
ac
cu
r
ac
y
r
ate
o
f
0
.
9
0
,
f
o
llo
wed
b
y
L
eNe
t
an
d
KNN
,
wh
ich
ac
co
m
p
lis
h
ed
th
e
ac
cu
r
ac
y
o
f
0
.
9
1
.
Nev
er
th
eless
,
th
e
p
r
o
p
o
s
ed
DQN
s
u
r
p
ass
e
s
th
e
tr
ad
itio
n
al
m
eth
o
d
o
lo
g
ies.
W
ith
an
ac
cu
r
ac
y
s
co
r
e
o
f
0
.
9
3
,
th
e
p
r
o
p
o
s
ed
s
ch
em
e
estab
lis
h
es
s
ig
n
if
ican
t
en
h
an
c
em
en
ts
,
p
ar
ticu
lar
ly
with
lar
g
er
tr
ain
in
g
d
ata.
R
eg
ar
d
in
g
6
0
%
tr
ain
in
g
d
ata,
th
e
p
r
o
p
o
s
ed
DQN
s
ch
em
e
o
u
tp
e
r
f
o
r
m
s
all
o
th
e
r
c
o
n
v
e
n
tio
n
al
m
eth
o
d
o
l
o
g
ies,
ex
h
ib
itin
g
th
e
g
r
ea
test
f
-
m
ea
s
u
r
e
o
f
0
.
7
9
.
T
h
ese
o
b
s
er
v
atio
n
s
ar
e
d
escr
ib
ed
in
Fig
u
r
e
4
.
Fig
u
r
e
4
.
C
o
m
p
a
r
ativ
e
an
aly
s
i
s
o
f
p
r
o
p
o
s
ed
m
eth
o
d
with
v
ar
io
u
s
ex
is
tin
g
alg
o
r
ith
m
s
with
6
0
% tr
ain
in
g
d
ata
0.000
0.200
0.40
0
0.600
0.800
1.000
A
c
c
u
racy
Pre
c
i
s
i
o
n
F
_
me
as
u
re
MCC
V
al
u
e
s
Per
f
o
r
ma
n
c
e
L
r=
0.1
L
r=
0.01
L
r
=
0.05
L
r=
0.001
0
0.2
0.4
0.6
0.8
1
V
al
u
e
s
Pe
r
f
o
r
man
c
e
Pro
p
o
s
e
d
KN
N
O
p
ti
mal R
N
N
L
i
n
kN
e
t
Bi-L
S
T
M
CN
N
L
e
N
e
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
.
4
,
Au
g
u
s
t 2
0
2
6
:
3
8
5
2
-
3
8
6
4
3860
Mo
r
e
p
ar
ticu
lar
ly
,
th
e
p
r
o
p
o
s
ed
DQN
s
ch
em
e
ex
h
ib
its
th
e
l
ea
s
t
f
alse
p
o
s
itiv
e
r
ate
(
F
PR
)
at
0
.
0
8
1
at
7
0
% tr
ain
in
g
d
ata
in
Fig
u
r
e
5
.
I
n
co
n
tr
ast,
th
e
ex
is
tin
g
ap
p
r
o
ac
h
es o
b
tain
ed
h
ig
h
er
FP
R
r
at
es with
KNN
[
6
]
at
0
.
1
1
,
o
p
tim
al
R
NN
[
2
7
]
at
0
.
1
3
,
L
in
k
Net
at
0
.
1
,
B
i
-
L
STM
at
0
.
1
1
,
C
NN
at
0
.
1
1
,
an
d
L
eNe
t
at
0
.
1
2
,
r
esp
ec
tiv
ely
.
Ad
d
itio
n
ally
,
th
e
p
r
o
p
o
s
ed
DQN
s
ch
em
e
s
u
r
p
ass
es
th
e
tr
ad
itio
n
al
m
eth
o
d
o
lo
g
ies
in
ter
m
s
o
f
n
eu
tr
al
m
etr
ics
ac
r
o
s
s
all
tr
ai
n
in
g
d
ata.
I
n
p
a
r
ticu
lar
,
th
e
p
r
ec
is
io
n
ac
h
iev
ed
u
s
in
g
th
e
p
r
o
p
o
s
ed
m
eth
o
d
is
0
.
8
3
3
in
tr
ain
i
n
g
d
ata
7
0
%,
wh
ich
is
ex
tr
em
ely
h
ig
h
er
th
an
KNN
[
1
]
(
0
.
7
6
2
)
,
o
p
tim
a
l
R
NN
[
2
]
(
0
.
7
2
6
)
,
L
in
k
Net
(
0
.
7
8
)
,
Bi
-
L
STM
(
0
.
7
6
6
)
,
C
NN
(
0
.
7
7
)
,
an
d
L
e
Net
(
0
.
7
6
1
)
,
r
esp
ec
tiv
ely
.
T
h
e
em
p
lo
y
m
en
t
o
f
im
p
r
o
v
e
d
SVM
-
R
FE
ca
n
ac
ce
ler
ate
tr
ain
in
g
p
r
o
ce
s
s
an
d
r
e
d
u
ce
tr
ain
in
g
tim
e
as
well
as
co
m
p
u
tatio
n
al
co
s
t.
I
t c
an
h
elp
p
r
eser
v
e
im
p
o
r
tan
t
f
ea
tu
r
es a
n
d
r
ed
u
ce
i
n
f
o
r
m
atio
n
lo
s
s
,
lead
in
g
to
b
etter
f
ea
tu
r
e
r
ep
r
esen
tatio
n
s
.
Similar
ly
,
th
e
co
m
p
ar
ativ
e
an
aly
s
is
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
v
ar
io
u
s
ex
is
tin
g
m
eth
o
d
s
is
g
iv
en
in
Fig
u
r
es
6
an
d
7
.
W
h
en
t
h
e
tr
ain
in
g
d
ata
i
n
cr
ea
s
es,
th
e
p
er
f
o
r
m
a
n
ce
also
in
c
r
ea
s
es.
Fro
m
th
e
f
ig
u
r
es,
it
is
o
b
s
er
v
ed
th
at
th
e
ac
cu
r
ac
y
in
cr
ea
s
es
f
r
o
m
0
.
9
3
1
to
0
.
9
6
6
wh
en
th
e
tr
ain
in
g
d
ata
i
n
cr
ea
s
es
f
o
r
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
.
L
ik
ewise,
th
e
F
-
m
ea
s
u
r
e
an
d
p
r
ec
is
io
n
ar
e
also
b
o
o
s
ted
as
0
.
7
9
3
to
0
.
9
0
0
an
d
0
.
7
9
5
to
0
.
9
0
6
,
r
esp
ec
tiv
ely
,
wh
ich
is
h
ig
h
er
t
h
an
th
e
e
x
is
tin
g
m
eth
o
d
o
lo
g
ie
s
s
u
ch
as
KNN
[
6
]
,
o
p
tim
al
R
NN
[
2
7
]
,
L
i
n
k
Net
,
Bi
-
L
STM
,
C
NN,
an
d
L
eNe
t.
T
h
e
e
x
is
tin
g
m
eth
o
d
o
lo
g
ies
r
ea
ch
a
m
ax
im
u
m
o
f
0
.
8
7
3
p
r
ec
is
io
n
an
d
0
.
8
7
9
F
-
m
ea
s
u
r
e
f
o
r
B
i
-
L
STM
.
B
ec
au
s
e
o
f
im
p
r
o
v
ed
n
o
r
m
aliza
tio
n
,
t
h
e
p
r
o
p
o
s
ed
DQN
f
r
am
e
wo
r
k
attain
s
h
ig
h
er
p
er
f
o
r
m
an
ce
in
all
tr
ain
in
g
d
at
a
p
er
ce
n
tag
es.
T
h
e
r
o
b
u
s
t
s
ca
l
in
g
tech
n
iq
u
e
ac
cu
r
ately
id
e
n
tifie
s
th
e
o
u
tlier
s
in
ex
tr
em
e
v
alu
es
also
.
T
h
e
DQN
class
if
ier
h
as
d
eliv
er
ed
th
e
lo
wer
FNR
an
d
FP
R
r
an
g
es
to
v
er
if
y
in
g
th
at
th
eir
f
r
am
ewo
r
k
s
ar
e
d
etec
tin
g
t
h
e
d
is
ea
s
ed
an
d
n
o
n
-
d
is
ea
s
ed
ca
s
es
m
o
s
t
ac
cu
r
ately
with
o
u
t
an
y
m
is
class
if
icat
io
n
is
o
cc
u
r
r
ed
in
th
eir
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
s
y
s
tem
.
Fig
u
r
e
5
.
C
o
m
p
a
r
ativ
e
an
aly
s
i
s
o
f
p
r
o
p
o
s
ed
m
eth
o
d
with
v
ar
io
u
s
ex
is
tin
g
alg
o
r
ith
m
s
with
7
0
% tr
ain
in
g
d
ata
Fig
u
r
e
6
.
C
o
m
p
a
r
ativ
e
an
aly
s
i
s
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
with
v
ar
io
u
s
ex
is
tin
g
alg
o
r
ith
m
s
with
8
0
% tr
ain
in
g
d
ata
F
i
g
u
r
e
7
.
C
o
m
p
a
r
a
ti
v
e
a
n
a
l
y
s
is
o
f
t
h
e
p
r
o
p
o
s
e
d
m
e
t
h
o
d
w
i
t
h
v
a
r
i
o
u
s
e
x
is
t
i
n
g
a
l
g
o
r
it
h
m
s
wi
t
h
9
0
%
t
r
a
i
n
i
n
g
d
a
t
a
4.
3
.
St
a
t
is
t
ica
l
a
na
ly
s
is
o
n a
cc
ura
cy
T
h
e
s
tatis
tical
a
s
s
es
s
m
en
t
illu
s
tr
ated
in
T
ab
le
2
o
f
f
er
s
an
e
x
h
au
s
tiv
e
co
m
p
ar
is
o
n
o
f
th
e
ef
f
icac
y
o
f
th
e
p
r
o
p
o
s
ed
DQN
s
ch
em
e
a
g
ain
s
t
ex
is
tin
g
ap
p
r
o
ac
h
es,
in
clu
d
in
g
KNN
[
6
]
,
o
p
tim
al
R
NN
[
2
7
]
,
L
i
n
k
Net
,
Bi
-
L
STM
,
C
NN
,
an
d
L
eNe
t
f
o
r
a
b
ig
d
ata
class
if
icatio
n
f
r
am
ewo
r
k
.
Fo
r
t
h
e
m
a
x
im
u
m
s
tatis
t
ical
m
ea
s
u
r
e
,
th
e
p
r
o
p
o
s
ed
s
ch
em
e
ex
h
ib
ite
d
a
s
u
p
er
io
r
ac
cu
r
ac
y
o
f
0
.
9
6
7
,
s
ig
n
if
y
in
g
t
h
at
it
ac
cu
r
ately
ca
teg
o
r
izes
th
e
b
ig
d
ata.
C
o
m
p
ar
ativ
ely
,
KNN
[
6
]
,
o
p
tim
al
R
NN
[
2
7
]
,
L
in
k
Net
,
B
i
-
L
STM
,
C
NN
,
an
d
L
eNe
t
o
b
tain
ed
th
e
ac
cu
r
ac
ies
o
f
0
.
9
5
4
,
0
.
9
4
0
,
0
.
9
4
5
,
0
.
9
5
8
,
0
.
9
4
2
,
a
n
d
0
.
9
4
6
,
ex
h
ib
ited
less
er
ac
cu
r
ac
y
v
alu
es
co
m
p
ar
ed
to
th
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
.
I
n
ad
d
itio
n
,
th
e
wo
r
s
t p
er
f
o
r
m
an
ce
is
ac
co
m
p
lis
h
ed
u
s
in
g
th
e
o
p
tim
al
R
NN
[
2
7
]
o
f
0
.
9
0
6
.
T
h
e
p
r
o
p
o
s
ed
ap
p
r
o
ac
h
ca
n
b
e
m
o
r
e
r
o
b
u
s
t
to
s
h
if
ts
in
d
ata
d
is
tr
ib
u
tio
n
,
en
h
an
cin
g
th
e
m
o
d
el'
s
p
er
f
o
r
m
a
n
ce
.
Fro
m
th
e
s
tatis
tical
an
aly
s
i
s
,
i
t
is
o
b
s
er
v
ed
th
at
th
e
m
o
d
el'
s
ac
cu
r
ac
y
is
co
n
s
is
ten
tly
ab
o
v
e
9
5
%,
with
a
p
ea
k
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
Th
e
r
o
le
o
f b
ig
d
a
ta
in
p
r
ec
is
io
n
med
icin
e
a
n
d
h
ea
lth
ca
r
e
m
o
n
ito
r
in
g
… (
Meen
a
ksh
i
S
a
n
k
a
r
a
s
u
b
r
a
ma
n
ia
n
)
3861
at
9
6
.
7
%.
T
h
e
m
ea
n
an
d
m
e
d
ian
v
alu
es
a
r
e
alm
o
s
t
clo
s
e
w
h
ich
in
d
icate
s
a
s
y
m
m
etr
ic
d
i
s
tr
ib
u
tio
n
an
d
t
h
e
lo
wer
SD
s
u
g
g
ests
m
o
d
er
ate
v
ar
iab
ilit
y
ac
r
o
s
s
r
u
n
s
.
T
h
e
p
e
r
f
o
r
m
a
n
ce
an
aly
s
is
is
ex
ec
u
ted
f
o
r
1
0
r
u
n
s
u
n
d
er
9
0
% tr
ain
in
g
d
ata
(
d
if
f
er
en
t su
b
s
ets).
T
ab
le
2
.
Statis
tical
ev
alu
atio
n
o
f
ac
cu
r
ac
y
S
t
a
t
i
st
i
c
a
l
met
r
i
c
s
P
r
o
p
o
se
d
m
e
t
h
od
K
N
N
O
p
t
i
mal
R
N
N
Li
n
k
N
e
t
Bi
-
LST
M
C
N
N
Le
N
e
t
M
e
a
n
0
.
9
4
9
0
.
9
2
8
0
.
9
1
9
0
.
9
2
4
0
.
9
3
3
0
.
9
2
5
0
.
9
3
2
M
e
d
i
a
n
0
.
9
5
2
0
.
9
2
4
0
.
9
1
5
0
.
9
2
8
0
.
9
2
7
0
.
9
3
4
0
.
9
2
9
S
t
d
-
D
e
v
0
.
0
1
6
0
.
0
2
1
0
.
0
1
8
0
.
0
1
7
0
.
0
2
1
0
.
0
1
6
0
.
0
1
3
M
i
n
0
.
9
3
1
0
.
9
0
2
0
.
8
9
8
0
.
9
0
3
0
.
9
0
7
0
.
9
0
8
0
.
9
1
9
M
a
x
0
.
9
6
7
0
.
9
5
4
0
.
9
4
0
0
.
9
4
5
0
.
9
5
8
0
.
9
4
2
0
.
9
4
6
4.
4
.
Abla
t
io
n
s
t
ud
y
An
ab
latio
n
s
tu
d
y
t
h
o
r
o
u
g
h
l
y
elim
in
ates
p
ar
ts
o
f
a
tech
n
iq
u
e
to
estab
lis
h
wh
ich
p
ar
ts
ar
e
s
ig
n
if
ican
t
f
o
r
th
e
s
y
s
tem
'
s
o
u
tco
m
e.
T
a
b
le
3
p
r
esen
ts
th
e
ab
latio
n
ass
ess
m
en
t
o
n
p
r
o
p
o
s
ed
DQN
with
ex
is
tin
g
f
ea
tu
r
e
f
u
s
io
n
f
o
r
b
ig
d
ata
class
if
icatio
n
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
attain
ed
th
e
p
ea
k
n
eg
ativ
e
p
r
ed
icti
v
e
v
alu
e
(
NPV
)
o
f
0
.
9
6
2
2
,
wh
ile
t
h
e
p
r
o
p
o
s
ed
with
ex
is
tin
g
f
ea
tu
r
e
f
u
s
io
n
s
co
r
ed
th
e
p
r
ec
is
io
n
o
f
0
.
7
8
4
.
Mo
r
eo
v
e
r
,
th
e
least
FNR
is
atta
in
ed
u
s
in
g
th
e
p
r
o
p
o
s
ed
DQN
m
o
d
el
is
0
.
1
1
4
,
wh
ich
is
ex
ce
p
tio
n
ally
less
er
th
an
th
at
p
r
o
p
o
s
ed
with
ex
is
tin
g
f
ea
tu
r
e
f
u
s
io
n
.
T
h
is
s
ig
n
if
ies
th
at
th
e
p
r
o
p
o
s
ed
s
ch
em
e
r
ed
u
ce
d
th
e
er
r
o
r
v
alu
es
co
m
p
a
r
ed
t
o
co
n
v
en
tio
n
al
m
eth
o
d
o
lo
g
ies.
T
ab
le
3
.
Ab
latio
n
an
aly
s
is
M
e
t
r
i
c
s
P
r
o
p
o
se
d
m
e
t
h
o
d
w
i
t
h
o
u
t
p
r
e
p
r
o
c
e
ss
i
n
g
P
r
o
p
o
se
d
m
e
t
h
o
d
w
i
t
h
c
o
n
v
e
n
t
i
o
n
a
l
S
V
M
-
R
F
E
P
r
o
p
o
se
d
m
e
t
h
o
d
A
c
c
u
r
a
c
y
0
.
9
0
1
2
0
.
8
7
6
2
0
.
9
6
0
1
S
e
n
s
i
t
i
v
i
t
y
0
.
7
4
5
6
0
.
6
9
4
5
0
.
8
8
5
9
S
p
e
c
i
f
i
c
i
t
y
0
.
9
1
3
1
0
.
8
8
1
3
0
.
9
4
1
7
P
r
e
c
i
s
i
o
n
0
.
7
6
1
9
0
.
7
8
4
9
0
.
8
8
0
0
F
-
mea
su
r
e
0
.
7
4
5
6
0
.
6
9
4
7
0
.
8
8
2
9
M
C
C
0
.
7
6
1
4
0
.
6
6
5
5
0
.
8
4
4
1
N
P
V
0
.
9
2
3
0
0
.
8
9
8
8
0
.
9
6
2
2
FPR
0
.
0
7
6
9
0
.
1
0
1
1
0
.
0
5
8
3
F
N
R
0
.
2
1
8
6
0
.
3
0
5
5
0
.
1
1
4
1
4.
5
.
Co
m
pu
t
a
t
io
na
l
t
im
e
a
na
ly
s
is
Dete
r
m
in
in
g
a
n
d
ass
ess
in
g
th
e
am
o
u
n
t
o
f
p
r
o
ce
s
s
in
g
tim
e
r
eq
u
ir
ed
f
o
r
a
m
eth
o
d
to
s
o
lv
e
a
p
r
o
b
lem
is
k
n
o
wn
as
co
m
p
u
tatio
n
al
tim
e
ass
ess
m
en
t.
T
h
e
p
r
o
p
o
s
e
d
m
eth
o
d
h
as
a
co
m
p
u
tatio
n
al
tim
e
o
f
2
0
.
3
6
9
s
wh
en
co
m
p
a
r
ed
to
o
th
er
m
et
h
o
d
o
lo
g
ies
lik
e
KNN
[
6
]
,
o
p
tim
al
R
NN
[
2
7
]
,
L
in
k
Net
,
B
i
-
L
STM
,
C
NN
,
an
d
L
eNe
t.
T
h
e
KNN
[
1
]
h
as
th
e
wo
r
s
t
p
er
f
o
r
m
a
n
ce
o
f
h
ig
h
co
m
p
u
tatio
n
al
tim
e
o
f
2
5
.
6
5
6
s
.
Fro
m
T
ab
le
4
,
it
is
o
b
s
er
v
ed
th
at
th
e
p
r
o
p
o
s
ed
f
r
am
ewo
r
k
with
im
p
r
o
v
ed
n
o
r
m
aliza
tio
n
an
d
im
p
r
o
v
ed
f
ea
t
u
r
e
f
u
s
io
n
o
b
tain
ed
b
etter
p
er
f
o
r
m
an
ce
in
ter
m
s
o
f
co
m
p
u
tatio
n
al
tim
e.
T
ab
le
4
.
C
o
m
p
u
tatio
n
tim
e
an
aly
s
is
(
in
s
ec
ond
s)
A
l
g
o
r
i
t
h
m
C
o
m
p
u
t
a
t
i
o
n
a
l
t
i
m
e
(
i
n
s
e
c
o
n
d
s)
P
r
o
p
o
se
d
2
0
.
3
6
9
K
N
N
2
5
.
6
5
6
O
p
t
i
mal
R
N
N
2
2
.
7
6
2
Li
n
k
N
e
t
2
2
.
0
6
5
Bi
-
LST
M
2
4
.
6
5
6
C
N
N
2
4
.
8
7
9
Le
N
e
t
2
3
.
1
8
4
5.
CO
NCLU
SI
O
N
Ultim
ately
,
a
n
o
v
el
b
ig
d
ata
class
if
icatio
n
m
o
d
el
was
in
tr
o
d
u
ce
d
an
d
s
u
cc
ess
f
u
lly
u
tili
ze
d
in
th
is
p
ap
er
.
T
h
is
p
r
o
ce
s
s
s
tar
ted
with
th
e
s
u
g
g
ested
Ma
p
R
ed
u
ce
f
r
am
ewo
r
k
,
wh
ich
in
cl
u
d
es
m
ap
p
er
s
an
d
r
ed
u
ce
r
s
f
o
r
m
an
a
g
in
g
lar
g
e
d
ata
s
ets.
T
h
e
in
p
u
t
d
ata
u
n
d
e
r
wen
t
n
o
r
m
aliza
tio
n
in
th
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
an
d
was
s
u
b
s
eq
u
en
tly
in
tr
o
d
u
ce
d
in
to
t
h
e
m
ap
p
er
p
h
ase,
wh
ich
e
n
co
m
p
ass
ed
b
o
th
r
aw
f
ea
tu
r
es
an
d
s
tatis
tical
f
ea
tu
r
es
f
o
r
ex
tr
ac
tin
g
ch
ar
ac
ter
is
tics
.
T
h
e
ex
tr
ac
ted
f
ea
tu
r
es
ar
e
s
u
b
s
eq
u
en
tly
in
p
u
t
in
to
th
e
SVM
-
R
FE
f
o
r
f
ea
tu
r
e
s
elec
tio
n
,
wh
ich
in
ter
ac
ts
with
an
d
is
co
m
b
in
e
d
d
u
r
in
g
th
e
r
ed
u
ce
r
p
h
ase.
Ultim
ately
,
th
e
co
m
b
in
e
d
f
ea
tu
r
e
Evaluation Warning : The document was created with Spire.PDF for Python.