I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
pu
t
er
E
ng
ineering
(
I
J
E
CE
)
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
,
p
p
.
2998
~
3
0
0
9
I
SS
N:
2088
-
8
7
0
8
,
DOI
:
1
0
.
1
1
5
9
1
/ijece.
v
15
i
3
.
pp
2
9
9
8
-
3
0
0
9
2998
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m
Deep learni
ng
f
o
r
predic
ting drug
-
rela
ted
p
ro
blems
in dia
betes
pa
tient
s
F
a
t
im
a
M
.
Sm
a
di,
Q
a
s
em
A.
Al
-
Ra
da
ideh
D
e
p
a
r
t
me
n
t
o
f
I
n
f
o
r
m
a
t
i
o
n
S
y
st
e
ms,
Y
a
r
mo
u
k
U
n
i
v
e
r
si
t
y
,
I
r
b
i
d
,
J
o
r
d
a
n
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
2
8
,
2
0
2
4
R
ev
is
ed
J
an
3
,
2
0
2
5
Acc
ep
ted
J
an
1
6
,
2
0
2
5
M
a
c
h
in
e
lea
rn
i
n
g
a
n
d
d
e
e
p
lea
rn
in
g
h
a
v
e
m
a
d
e
a
d
v
a
n
c
e
s
in
th
e
h
e
a
lt
h
c
a
re
d
o
m
a
in
.
In
th
is
stu
d
y
,
we
a
im
to
a
p
p
l
y
d
e
e
p
lea
rn
i
n
g
m
o
d
e
ls
to
p
re
d
ict
th
e
d
ru
g
-
re
late
d
p
r
o
b
lem
s
(DRP
s
)
sta
tu
s
fo
r
d
iab
e
tes
p
a
ti
e
n
ts
.
Also
,
t
o
d
e
term
in
e
th
e
a
p
p
r
o
p
riate
m
o
d
e
l
to
u
se
fo
r
c
las
sifica
ti
o
n
u
si
n
g
d
e
e
p
lea
rn
in
g
a
lg
o
rit
h
m
s
o
r
m
a
c
h
in
e
lea
rn
in
g
m
e
th
o
d
s
to
i
n
v
e
sti
g
a
te
wh
ich
o
n
e
p
e
rfo
rm
e
d
b
e
tt
e
r
re
su
lt
s
f
o
r
tab
u
lar
d
a
ta
b
y
c
o
m
p
a
ri
n
g
t
h
e
a
c
h
iev
e
d
d
e
e
p
lea
rn
in
g
re
su
lt
s
with
th
e
m
a
c
h
in
e
lea
rn
i
n
g
m
e
th
o
d
s
t
o
f
ig
u
re
o
u
t
w
h
ich
o
n
e
g
iv
e
s
b
e
tt
e
r
re
su
lt
s.
T
o
a
p
p
ly
th
e
d
e
e
p
l
e
a
rn
in
g
m
o
d
e
ls,
t
h
e
sa
m
e
c
rit
e
ria
th
a
t
we
re
a
p
p
li
e
d
in
th
e
p
re
v
i
o
u
s
st
u
d
y
h
a
v
e
b
e
e
n
imp
lem
e
n
ted
i
n
t
h
is
i
n
v
e
stig
a
ti
o
n
,
a
n
d
t
h
e
sa
m
e
d
a
tas
e
t
wa
s
u
se
d
.
Th
e
re
su
lt
s
sh
o
w
th
a
t
th
e
m
a
c
h
in
e
lea
rn
in
g
a
lg
o
rit
h
m
s
e
sp
e
c
ially
th
e
ra
n
d
o
m
fo
re
sts
f
o
r
p
re
d
icti
n
g
th
e
D
RP
s
sta
tu
s
o
u
t
p
e
rfo
rm
t
h
e
d
e
e
p
lea
rn
i
n
g
m
o
d
e
ls.
F
o
r
c
las
sifica
ti
o
n
tas
k
s
in
h
e
a
lt
h
c
a
re
fo
r
tab
u
lar
d
a
ta,
t
h
e
fi
n
d
i
n
g
s
o
f
th
is
stu
d
y
sh
o
w
th
a
t
m
a
c
h
in
e
lea
rn
in
g
m
e
th
o
d
s
a
re
th
e
a
p
p
ro
p
riate
m
o
d
e
l
in
ste
a
d
o
f
u
sin
g
d
e
e
p
lea
rn
in
g
to
a
p
p
l
y
c
las
sifica
ti
o
n
.
K
ey
w
o
r
d
s
:
Dee
p
lear
n
in
g
Dee
p
n
eu
r
al
n
etwo
r
k
s
Hy
p
er
p
ar
a
m
eter
o
p
tim
izatio
n
L
o
n
g
-
s
h
o
r
t te
r
m
m
e
m
o
r
y
n
etwo
r
k
s
Ma
ch
in
e
lear
n
in
g
T
ab
u
lar
d
ata
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
:
Fatim
a
M.
Sm
ad
i
Dep
ar
tm
en
t o
f
I
n
f
o
r
m
atio
n
Sy
s
tem
s
,
Yar
m
o
u
k
Un
iv
er
s
ity
I
r
b
id
2
1
1
6
3
,
J
o
r
d
an
E
m
ail: f
atim
a.
s
m
ad
i1
3
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
R
ec
en
tly
,
with
lar
g
e
am
o
u
n
ts
o
f
d
ata
b
ec
o
m
in
g
av
ailab
le
a
n
d
th
e
im
p
r
o
v
em
e
n
t
in
m
em
o
r
y
h
an
d
lin
g
ca
p
ab
ilit
ies o
f
th
e
co
m
p
u
ter
s
m
ak
in
g
th
e
p
r
o
ce
s
s
o
f
h
an
d
lin
g
h
u
g
e
am
o
u
n
ts
o
f
d
ata
ea
s
ier
,
an
d
th
e
p
r
o
g
r
ess
o
f
m
ac
h
in
e
lear
n
in
g
a
n
d
d
ee
p
lea
r
n
in
g
alg
o
r
ith
m
s
[
1
]
,
[
2
]
.
Ho
w
ev
er
,
tab
u
lar
d
ata
is
th
e
m
o
s
t
p
r
ev
alen
t
d
ata
ty
p
e
em
p
lo
y
ed
in
m
a
n
y
r
ea
l
-
wo
r
ld
ap
p
licatio
n
s
.
I
t
is
u
s
ed
in
v
ar
io
u
s
f
ield
s
,
in
clu
d
in
g
m
ed
icin
e,
f
r
au
d
d
etec
tio
n
,
co
m
p
u
ter
v
is
io
n
,
an
d
d
is
ea
s
e
d
iag
n
o
s
tic
[
3
]
.
I
n
r
ec
e
n
t
y
ea
r
s
,
d
ee
p
lear
n
i
n
g
h
as
g
ain
ed
a
lo
t
o
f
tr
ac
tio
n
in
d
ata
m
in
in
g
d
u
r
in
g
th
e
last
s
ix
y
ea
r
s
,
th
e
ter
m
“d
ee
p
lear
n
in
g
”
was
co
i
n
ed
in
2
0
0
6
[
4
]
.
I
t
is
u
s
ed
to
in
v
esti
g
ate
h
ea
lth
ca
r
e
to
p
ics
i
n
o
r
d
er
to
im
p
r
o
v
e
p
atien
t
ca
r
e
[
5
]
.
Ho
wev
er
,
d
ee
p
lear
n
in
g
is
a
c
o
m
p
u
tatio
n
al
m
eth
o
d
t
h
at
allo
ws
m
ac
h
in
es
t
o
an
aly
ze
d
ata
lik
e
th
e
h
u
m
an
b
r
ain
.
T
h
u
s
,
it
is
a
s
p
ec
ial
ty
p
e
o
f
m
ac
h
in
e
le
ar
n
in
g
th
at
in
v
o
lv
es
a
d
ee
p
e
r
lev
el.
Als
o
,
d
ee
p
lear
n
in
g
d
if
f
er
s
f
r
o
m
a
n
eu
r
al
n
etwo
r
k
in
th
at
it
h
as
m
a
n
y
l
ay
er
s
an
d
n
e
u
r
o
n
s
in
b
ig
n
u
m
b
er
s
.
On
th
e
o
th
e
r
h
an
d
,
th
e
u
s
e
o
f
d
ee
p
lear
n
i
n
g
h
as
ev
o
l
v
ed
in
m
a
n
y
ar
e
as
in
th
e
f
o
llo
win
g
ap
p
licatio
n
s
s
u
ch
as
s
p
ee
ch
r
ec
o
g
n
itio
n
,
m
ed
ical
r
esear
ch
in
ca
n
ce
r
r
esear
c
h
to
a
u
to
m
ati
ca
lly
d
etec
t
ca
n
ce
r
ce
lls
,
m
e
d
ical
im
ag
in
g
,
d
r
u
g
s
d
is
co
v
er
y
,
a
n
d
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
[
1
]
–
[
5
]
.
Dee
p
n
eu
r
al
n
etwo
r
k
s
(
DNN)
tr
ain
ee
th
e
n
etwo
r
k
th
r
o
u
g
h
f
ee
d
-
f
o
r
war
d
n
eu
r
al
n
etwo
r
k
s
an
d
u
s
e
b
ac
k
p
r
o
p
ag
atio
n
.
T
h
e
s
tr
u
ct
u
r
al
b
u
ild
in
g
b
lo
ck
o
f
d
ee
p
lea
r
n
in
g
is
th
e
p
er
ce
p
tr
o
n
(
n
e
u
r
o
n
)
.
Dee
p
lea
r
n
in
g
co
n
tain
s
th
e
in
p
u
t
lay
e
r
,
h
id
d
e
n
lay
er
s
,
an
d
o
u
tp
u
t
la
y
er
[
4
]
.
I
n
ad
d
itio
n
,
th
e
DNN
is
ch
ar
ac
ter
ized
th
at
it
h
as
m
o
r
e
th
an
o
n
e
h
id
d
en
lay
er
,
t
h
e
d
ir
ec
tio
n
o
f
th
e
in
f
o
r
m
atio
n
in
th
e
n
etwo
r
k
is
f
o
r
war
d
(
t
h
er
e
ar
e
n
o
cir
cles)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Dee
p
lea
r
n
in
g
fo
r
p
r
ed
ictin
g
d
r
u
g
-
r
ela
ted
p
r
o
b
lems in
d
ia
b
etes p
a
tien
ts
(
F
a
tima
M.
S
ma
d
i
)
2999
f
r
o
m
in
p
u
t
to
th
e
o
u
tp
u
t
t
h
r
o
u
g
h
h
id
d
en
lay
e
r
s
,
also
k
n
o
wn
as
m
u
lti
-
lay
e
r
p
er
c
ep
tr
o
n
(
ML
P)
[
6
]
.
Fu
r
th
er
m
o
r
e
,
th
e
lo
n
g
-
s
h
o
r
t
ter
m
m
em
o
r
y
n
etwo
r
k
s
(
L
STM
)
im
p
lem
en
t
a
m
o
r
e
co
m
p
lex
r
ec
u
r
r
en
t
u
n
it
with
g
ates
to
co
n
tr
o
l
w
h
at
in
f
o
r
m
atio
n
is
p
ass
ed
th
r
o
u
g
h
.
Als
o
,
it
is
c
o
n
s
id
er
ed
a
s
tate
-
of
-
th
e
-
ar
t
m
o
d
el
[
7
]
.
Ho
wev
er
,
t
h
e
L
STM
c
o
n
tain
s
m
em
o
r
y
b
l
o
ck
s
th
at
co
n
tr
o
l
th
e
in
f
o
r
m
atio
n
f
lo
w,
t
h
e
k
ey
b
u
ild
in
g
b
lo
c
k
b
eh
in
d
L
STM
is
th
e
g
ates,
th
e
m
em
o
r
y
ce
lls
in
L
STM
lay
er
s
in
ter
ac
t
with
ea
ch
o
th
er
v
ia
g
ates,
an
d
th
e
g
ates
ar
e
r
esp
o
n
s
ib
le
f
o
r
en
a
b
lin
g
t
h
e
L
STM
to
b
e
ab
le
t
o
ad
d
o
r
r
em
o
v
e
i
n
f
o
r
m
atio
n
to
its
ce
l
l
s
tate.
I
n
ad
d
itio
n
,
T
h
e
L
STM
f
o
llo
ws
th
e
m
ain
f
o
u
r
s
tep
s
:
f
o
r
g
ettin
g
ir
r
elev
a
n
t
h
is
to
r
y
,
th
en
p
er
f
o
r
m
in
g
c
o
m
p
u
tatio
n
s
to
s
to
r
e
r
elev
an
t
in
f
o
r
m
atio
n
,
u
p
d
atin
g
th
e
in
ter
n
al
s
tate,
an
d
f
in
all
y
g
e
n
er
atin
g
th
e
o
u
tp
u
t
to
b
e
an
in
p
u
t
to
th
e
n
ex
t
L
STM
lay
er
.
T
r
ain
in
g
L
STM
n
etwo
r
k
ap
p
lied
th
e
b
ac
k
p
r
o
p
ag
atio
n
m
eth
o
d
[
8
]
,
[
9
]
.
As
f
o
r
u
s
in
g
d
ee
p
lear
n
in
g
te
ch
n
iq
u
es
i
n
th
e
Hea
lth
ca
r
e
ar
ea
.
Kim
et
a
l.
[
1
0
]
u
s
ed
th
e
d
ee
p
b
elief
n
etwo
r
k
(
DB
N)
an
d
d
if
f
er
e
n
t
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
to
p
r
ed
ict
ca
r
d
io
v
ascu
lar
r
is
k
p
r
ed
ictio
n
(
lo
w
r
is
k
an
d
h
ig
h
r
is
k
)
)
.
T
h
e
a
u
th
o
r
s
e
v
alu
ated
th
e
m
o
d
els
u
s
in
g
th
r
e
e
p
er
f
o
r
m
an
ce
m
etr
ics
(
ac
cu
r
a
cy
,
s
p
ec
if
icity
,
an
d
s
en
s
itiv
ity
)
.
T
h
e
n
aïv
e
B
ay
e
s
(
NB
)
m
o
d
el
o
b
tain
e
d
(
7
9
%,
6
3
%,
8
4
%),
l
o
g
is
tic
r
eg
r
ess
io
n
(
L
R
)
m
o
d
el
o
b
tain
ed
(
8
0
%,
6
9
%,
8
2
%),
s
u
p
p
o
r
t
v
ec
t
o
r
m
ac
h
i
n
e
(
SVM)
m
o
d
el
o
b
tain
ed
(
7
1
%,
1
0
0
%,
7
1
%),
r
an
d
o
m
f
o
r
ests
(
R
F)
m
o
d
el
o
b
tain
ed
(
7
7
%,
6
1
%,
8
2
%),
d
ee
p
b
eli
ef
n
etwo
r
k
(
DB
N)
m
o
d
el
o
b
tain
ed
(
7
5
%,
8
2
%,
7
4
%),
a
n
d
a
s
tatis
tical
DB
N
m
o
d
el
o
b
tain
ed
(
8
3
,
7
3
%,
8
7
%),
th
e
r
esu
lts
s
h
o
wed
th
at
t
h
e
s
tatis
tical
DB
N
m
o
d
el
o
u
tp
er
f
o
r
m
ed
t
h
e
o
th
e
r
class
if
ier
s
.
Ko
r
o
tco
v
et
a
l.
[
1
1
]
u
s
ed
SVM,
R
F,
an
d
lo
g
is
tic
lin
ea
r
r
e
g
r
ess
io
n
to
c
o
m
p
ar
e
th
e
p
er
f
o
r
m
an
ce
o
f
d
if
f
er
en
t
m
ac
h
i
n
e
lear
n
in
g
al
g
o
r
ith
m
s
to
th
e
d
ee
p
n
eu
r
al
n
etwo
r
k
s
(
DNN)
to
p
r
ed
ict
d
r
u
g
d
is
co
v
er
y
.
T
h
e
r
esu
lts
o
f
th
e
p
er
f
o
r
m
an
ce
m
etr
ics
s
h
o
wed
ac
cu
r
ac
y
,
p
r
ec
i
s
io
n
,
an
d
r
ec
all,
r
esp
ec
tiv
ely
.
T
h
e
DNN
m
o
d
el
o
b
tain
ed
(
8
7
%,
6
8
%,
7
0
%),
SVM
m
o
d
el
o
b
tain
ed
(
8
3
%,
5
9
%,
7
4
%),
lo
g
is
tic
lin
ea
r
r
eg
r
ess
io
n
m
o
d
el
o
b
tain
ed
(
7
8
%,
5
8
%,
7
4
%),
R
F
m
o
d
el
o
b
tain
ed
(
7
9
%,
5
9
%,
7
0
%),
th
e
r
esu
lts
s
h
o
wed
th
at
th
e
DNN
m
o
d
el
o
u
tp
er
f
o
r
m
ed
th
e
m
ac
h
i
n
e
lea
r
n
in
g
alg
o
r
ith
m
s
.
Oth
er
r
esear
ch
u
s
ed
d
ee
p
lear
n
in
g
an
d
m
ac
h
in
e
lear
n
i
n
g
to
p
r
ed
ict
b
r
ain
tu
m
o
r
s
in
th
e
b
r
a
in
[
1
2
]
,
th
e
au
th
o
r
s
u
s
ed
d
ee
p
lear
n
i
n
g
an
d
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
to
p
r
ed
ict
b
r
ain
tu
m
o
r
s
in
th
e
b
r
ain
.
T
h
ey
u
s
ed
th
e
d
ee
p
n
eu
r
al
n
etwo
r
k
a
n
d
k
-
n
ea
r
est
n
eig
h
b
o
r
alg
o
r
it
h
m
s
(
k
-
NN)
in
th
eir
s
tu
d
y
.
T
h
e
r
esu
lts
o
f
t
h
e
p
er
f
o
r
m
an
ce
m
etr
ics
s
h
o
wed
ac
cu
r
ac
y
,
p
r
ec
is
io
n
,
an
d
r
ec
al
l,
r
esp
ec
tiv
ely
.
T
h
e
DNN
m
o
d
el
o
b
tain
e
d
(
9
6
%,
9
7
%,
9
7
%),
k
-
n
ea
r
est
n
eig
h
b
o
r
m
o
d
el
with
k
=1
o
b
tain
ed
(
9
5
%,
9
5
%,
9
5
%),
T
h
e
k
-
N
N
m
o
d
el
with
k
=3
o
b
tain
ed
(
8
6
%,
8
9
%,
8
6
%).
Ay
o
n
a
n
d
I
s
lam
[
1
3
]
u
s
ed
DNN
to
d
iag
n
o
s
e
d
iab
etes
(
ab
s
en
ce
an
d
p
r
esen
t
)
.
T
h
e
r
e
s
u
lts
o
f
th
e
p
er
f
o
r
m
an
ce
m
etr
ics
f
o
r
ten
-
f
o
ld
s
s
h
o
wed
ac
cu
r
ac
y
,
s
p
ec
if
icity
,
an
d
s
en
s
itiv
ity
,
r
esp
e
ctiv
ely
.
T
h
e
DNN
m
o
d
el
o
b
tain
ed
(
9
7
%,
9
6
%,
9
7
%).
M
a
s
s
a
r
o
e
t
a
l
.
[
1
4
]
u
s
e
d
L
ST
M
,
L
S
T
M
u
s
i
n
g
a
r
t
i
f
i
ci
a
l
r
ec
o
r
d
s
(
L
S
T
M
-
AR
)
,
a
n
d
M
L
P
t
o
p
r
e
d
i
c
t
d
i
a
b
e
t
e
s
b
y
c
l
as
s
i
f
y
i
n
g
d
i
a
b
e
te
s
s
ta
t
u
s
a
n
d
n
o
-
d
i
a
b
e
t
es
s
t
at
u
s
.
T
h
e
r
e
p
o
r
t
e
d
r
e
s
u
l
t
o
f
t
h
e
p
e
r
f
o
r
m
a
n
c
e
m
e
t
r
i
c
s
s
h
o
w
e
d
a
c
c
u
r
a
c
y
.
T
h
e
L
S
T
M
m
o
d
e
l
o
b
t
a
i
n
e
d
(
7
5
%
)
,
L
S
T
M
-
A
R
m
o
d
e
l
o
b
t
a
i
n
e
d
(
8
4
%
)
,
a
n
d
M
L
P
m
o
d
e
l
o
b
t
a
i
n
e
d
(
7
7
%
)
,
b
y
a
p
p
l
y
i
n
g
t
h
e
L
S
T
M
-
AR
t
h
e
r
e
w
as
a
n
im
p
r
o
v
e
m
e
n
t
i
n
r
es
u
l
ts
c
o
m
p
ar
e
d
w
i
t
h
M
L
P
a
n
d
L
S
T
M
.
T
i
g
g
a
a
n
d
G
a
r
g
[
1
5
]
u
s
e
d
L
R
,
k
-
N
N
,
SV
M
,
N
B
,
d
ec
is
i
o
n
t
r
e
e
(
D
T
)
,
a
n
d
R
F
t
o
p
r
e
d
ic
t
t
y
p
e
2
d
i
a
b
e
t
es
p
a
t
i
e
n
ts
.
T
h
e
r
es
u
l
ts
o
f
t
h
e
p
e
r
f
o
r
m
a
n
c
e
m
e
t
r
i
cs
s
h
o
w
e
d
a
c
c
u
r
a
c
y
.
T
h
e
L
R
m
o
d
e
l
o
b
t
a
i
n
e
d
(
8
5
%
)
,
k
-
N
N
m
o
d
e
l
o
b
t
a
i
n
e
d
(
7
7
%
)
,
S
V
M
m
o
d
e
l
o
b
t
a
i
n
e
d
(
8
6
%
)
,
N
B
m
o
d
e
l
o
b
t
a
i
n
e
d
(
8
0
%
)
,
D
T
m
o
d
e
l
o
b
t
a
i
n
e
d
(
8
4
%
)
,
a
n
d
t
h
e
R
F
o
b
t
ai
n
e
d
(
9
4
%
)
.
T
h
e
r
e
s
u
l
t
s
s
h
o
w
e
d
t
h
a
t
t
h
e
R
F
m
o
d
e
l
o
u
tp
e
r
f
o
r
m
e
d
t
h
e
o
t
h
e
r
a
l
g
o
r
i
t
h
m
s
.
Sh
war
tz
-
Z
iv
an
d
Ar
m
o
r
[
1
6
]
d
em
o
n
s
tr
ated
th
at
wh
en
wo
r
k
i
n
g
with
tab
u
lar
d
ata
in
class
if
i
ca
tio
n
an
d
r
eg
r
ess
io
n
,
d
ee
p
lear
n
in
g
m
o
d
els
ar
e
n
o
t
all
y
o
u
n
ee
d
.
I
n
t
h
eir
s
tu
d
y
,
th
ey
co
m
p
ar
ed
tr
e
e
en
s
em
b
le
m
o
d
els
s
u
ch
as
XGBo
o
s
t
with
d
ee
p
l
ea
r
n
in
g
m
o
d
els
to
s
ee
w
h
ich
o
f
th
e
m
g
i
v
es
b
etter
r
esu
lts
f
o
r
tab
u
lar
d
ata.
T
h
e
s
tu
d
y
f
o
u
n
d
th
at
th
e
XGBo
o
s
t
o
u
tp
er
f
o
r
m
s
th
e
d
ee
p
m
o
d
els,
it
r
eq
u
ir
es
m
u
ch
less
tu
n
in
g
.
So
,
th
e
y
r
ec
o
m
m
en
d
u
s
in
g
e
n
s
em
b
le
m
o
d
els wh
en
u
s
in
g
ta
b
u
lar
d
ata.
Hair
an
i
an
d
Priy
an
to
[
1
7
]
u
s
e
d
SVM
an
d
R
F
co
m
b
in
ed
wit
h
th
e
s
y
n
th
etic
m
in
o
r
ity
o
v
er
s
am
p
lin
g
tech
n
iq
u
e
(
SMOT
E
)
,
e
d
ited
n
ea
r
est
n
eig
h
b
o
r
s
(
E
NN)
,
an
d
h
y
b
r
id
SMOT
E
-
E
NN
m
et
h
o
d
s
.
T
h
e
r
ep
o
r
ted
r
esu
lt
o
f
th
e
p
er
f
o
r
m
an
ce
m
et
r
ics
s
h
o
wed
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
an
d
s
p
ec
if
icity
,
r
esp
ec
tiv
ely
.
T
h
e
SVM
with
SMOT
E
m
o
d
el
o
b
tain
ed
(
7
4
%,
7
0
%,
7
7
%),
SVM
with
E
NN
m
o
d
el
o
b
tain
ed
(
8
5
%,
8
5
%,
8
6
%),
SVM
with
SMOT
E
-
E
NN
m
o
d
el
o
b
tain
e
d
(
9
0
%,
9
1
%,
8
8
%),
R
F
with
SMOT
E
m
o
d
el
o
b
tain
ed
(
8
2
%,
8
6
%,
7
8
%),
R
F
with
E
NN
m
o
d
el
o
b
tain
e
d
(
8
7
%,
8
6
%,
8
7
%),
R
F
with
SMOT
E
-
E
NN
m
o
d
el
o
b
tain
e
d
(
9
5
%,
98%
,
9
2
%).
T
h
e
f
in
d
in
g
s
s
h
o
wed
th
at
th
e
R
F
with
SMOT
E
-
E
NN
m
o
d
el
o
u
tp
er
f
o
r
m
ed
SMOT
E
an
d
E
NN
in
d
iv
id
u
ally
in
ter
m
s
o
f
av
er
a
g
e
p
er
f
o
r
m
a
n
ce
.
C
h
u
et
a
l.
[
1
8
]
u
s
ed
DNN
to
d
ev
elo
p
a
ca
r
d
io
v
ascu
lar
d
is
e
ase
(
C
VD
)
r
is
k
p
r
ed
ictio
n
.
T
h
e
r
esu
lts
o
f
th
e
p
er
f
o
r
m
an
ce
m
etr
ics
s
h
o
wed
ac
cu
r
ac
y
,
s
p
ec
if
icity
,
an
d
s
en
s
itiv
ity
r
esp
ec
tiv
ely
.
T
h
e
D
NN
m
o
d
el
o
b
tain
e
d
(
8
7
.
5
0
%,
8
7
.
2
3
%,
8
8
.
0
6
%).
T
h
e
s
tu
d
y
c
o
n
tr
ib
u
tes
b
y
ap
p
ly
in
g
d
ee
p
lear
n
i
n
g
tech
n
iq
u
es
to
n
ew
d
ata
th
at
h
as
n
o
t
b
ee
n
ap
p
lied
b
ef
o
r
e
u
s
in
g
d
ee
p
lear
n
in
g
m
o
d
els,
th
e
d
ata
was
co
llected
f
r
o
m
s
ix
M
ajo
u
r
h
o
s
p
itals
i
n
J
o
r
d
an
esp
ec
ially
f
o
cu
s
in
g
o
n
d
iab
etic
p
atien
ts
t
o
p
r
ed
ict
th
e
s
tatu
s
o
f
d
r
u
g
-
r
el
ated
p
r
o
b
lem
s
.
B
y
co
m
p
ar
in
g
th
e
r
esu
lts
o
f
d
ee
p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
9
9
8
-
3
0
0
9
3000
lear
n
in
g
tech
n
iq
u
es
with
m
a
ch
in
e
lear
n
in
g
m
eth
o
d
s
,
th
e
s
tu
d
y
ac
h
iev
es
h
i
g
h
ac
cu
r
a
cy
in
DR
P
s
tatu
s
p
r
ed
ictio
n
a
n
d
o
u
tp
er
f
o
r
m
s
p
r
ev
io
u
s
s
tu
d
ies.
T
h
is
s
tu
d
y
h
ig
h
lig
h
ts
a
f
o
u
n
d
atio
n
f
o
r
f
u
tu
r
e
r
esear
c
h
in
p
r
ed
ictin
g
d
r
u
g
-
r
elate
d
p
r
o
b
le
m
s
.
M
a
n
y
r
e
s
e
a
r
c
h
e
r
s
h
a
v
e
u
s
e
d
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
a
n
d
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
e
t
h
o
d
s
i
n
h
e
a
l
t
h
c
a
r
e
s
y
s
te
m
s
s
u
c
h
as
b
r
e
a
s
t
ca
n
c
e
r
,
b
r
a
i
n
c
a
n
c
e
r
,
a
n
d
d
r
u
g
d
i
s
c
o
v
e
r
y
[
1
9
]
,
[
2
0
]
.
W
h
e
n
h
a
n
d
l
i
n
g
c
l
a
s
s
i
f
i
c
at
i
o
n
t
a
s
k
s
i
n
d
a
t
a
m
i
n
i
n
g
w
it
h
t
a
b
u
l
a
r
d
at
a
in
h
e
a
l
t
h
c
a
r
e
s
y
s
te
m
s
,
it
b
e
c
o
m
e
s
u
n
c
l
e
a
r
t
o
d
e
t
e
r
m
i
n
e
w
h
i
c
h
m
o
d
e
l
t
o
u
s
e
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
o
r
m
a
c
h
i
n
e
l
ea
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
.
D
ee
p
l
e
a
r
n
in
g
m
o
d
e
l
s
h
a
v
e
s
h
o
w
n
t
h
e
a
b
il
i
t
y
t
o
h
a
n
d
l
e
la
r
g
e
d
a
t
a
s
et
s
,
b
u
t
t
h
e
y
r
e
q
u
i
r
e
c
o
m
p
u
t
a
t
i
o
n
a
l
r
es
o
u
r
c
e
s
a
n
d
la
r
g
e
a
m
o
u
n
t
s
o
f
d
a
t
a
.
H
o
w
e
v
er
,
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
e
t
h
o
d
s
,
a
r
e
k
n
o
w
n
f
o
r
t
h
e
i
r
e
f
f
e
c
t
i
v
e
p
r
a
c
ti
c
a
l
,
a
n
d
e
as
y
im
p
l
e
m
e
n
t
a
t
i
o
n
.
T
h
is
s
t
u
d
y
a
i
m
s
t
o
i
n
v
e
s
t
i
g
a
t
e
t
h
e
b
e
s
t
s
t
r
u
c
t
u
r
e
o
f
t
h
e
u
s
e
d
d
e
e
p
l
e
a
r
n
i
n
g
t
e
c
h
n
i
q
u
e
s
(
D
N
N
a
n
d
L
S
T
M
)
t
o
p
r
e
d
i
c
t
t
h
e
s
t
at
u
s
o
f
d
r
u
g
-
r
e
l
a
t
e
d
p
r
o
b
l
e
m
s
,
a
n
d
w
h
e
t
h
e
r
t
o
a
p
p
l
y
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
o
r
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
m
e
t
h
o
d
s
w
h
e
n
d
e
a
l
i
n
g
w
i
t
h
t
a
b
u
l
a
r
d
a
t
a
f
o
r
c
l
ass
i
f
i
c
at
i
o
n
.
T
o
p
e
r
f
o
r
m
t
h
i
s
,
t
h
e
s
a
m
e
d
a
ta
s
et
u
s
e
d
i
n
t
h
e
[
2
1
]
h
a
s
b
e
e
n
u
s
ed
.
I
n
a
d
d
i
t
i
o
n
,
w
e
c
o
m
p
a
r
e
d
t
h
e
r
e
s
u
l
ts
o
b
t
a
i
n
e
d
f
r
o
m
d
e
e
p
l
e
a
r
n
i
n
g
m
o
d
e
l
s
wi
t
h
t
h
e
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
cl
a
s
s
i
f
i
e
r
s
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
e
m
ain
p
u
r
p
o
s
e
o
f
th
is
s
tu
d
y
is
to
in
v
esti
g
ate
th
e
b
est
s
tr
u
ctu
r
e
o
f
t
h
e
DNN
an
d
L
STM
to
p
r
ed
ict
th
e
s
tatu
s
o
f
d
r
u
g
-
r
elate
d
p
r
o
b
lem
s
.
Als
o
,
to
f
in
d
o
u
t
th
e
ef
f
ec
t
o
f
a
p
p
ly
in
g
d
ee
p
lear
n
in
g
co
m
p
ar
ed
with
th
e
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
.
Ad
d
itio
n
ally
,
wh
et
h
er
to
ap
p
ly
d
ee
p
lear
n
i
n
g
m
o
d
els
o
r
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
wh
en
d
ea
lin
g
with
ta
b
u
lar
d
ata
f
o
r
class
if
icatio
n
.
Als
o
,
to
f
in
d
o
u
t
w
h
ich
will
g
iv
e
h
ig
h
p
er
f
o
r
m
an
ce
f
o
r
tab
u
lar
d
ata
c
o
m
p
a
r
ed
wit
h
th
e
s
tu
d
y
in
[
2
1
]
.
T
o
p
er
f
o
r
m
th
is
s
tu
d
y
,
th
e
m
eth
o
d
o
lo
g
y
’
s
im
p
lem
en
tatio
n
is
s
u
b
d
iv
id
ed
in
t
o
s
ev
er
al
s
tep
s
.
Fig
u
r
e
1
s
u
m
m
ar
izes th
e
o
v
e
r
all
m
eth
o
d
o
lo
g
y
s
tep
s
in
d
etail.
T
h
e
d
ee
p
lear
n
i
n
g
m
o
d
els
th
at
wer
e
ex
p
er
im
e
n
te
d
in
th
is
s
tu
d
y
ar
e
b
r
ief
ly
d
esc
r
ib
ed
in
th
e
f
o
llo
win
g
s
u
b
s
ec
t
io
n
.
Dee
p
lear
n
i
n
g
o
r
n
eu
r
al
n
etwo
r
k
s
b
o
o
k
s
will p
r
o
v
id
e
a
d
etailed
d
escr
ip
ti
o
n
.
Fig
u
r
e
1
.
T
h
e
r
esear
ch
m
eth
o
d
o
lo
g
y
DNN
ar
e
a
k
in
d
o
f
n
eu
r
al
n
e
two
r
k
in
s
p
ir
ed
b
y
t
h
e
d
esig
n
o
f
th
e
h
u
m
a
n
n
er
v
o
u
s
s
y
s
te
m
an
d
t
h
e
b
r
ain
’
s
ar
ch
itectu
r
e
[
2
2
]
.
Ad
d
itio
n
ally
,
DNN
ch
ar
ac
ter
ized
th
at
it
h
as
m
o
r
e
th
an
o
n
e
h
id
d
en
lay
er
,
th
e
d
ir
ec
tio
n
o
f
th
e
i
n
f
o
r
m
atio
n
i
n
th
e
n
etwo
r
k
is
f
o
r
war
d
(
th
er
e
ar
e
n
o
cir
cles)
f
r
o
m
in
p
u
t
to
th
e
o
u
t
p
u
t
th
r
o
u
g
h
h
id
d
en
lay
e
r
s
,
it
also
k
n
o
wn
a
s
ML
P
[
6
]
–
[
1
3
]
.
Ho
wev
er
,
t
h
e
lay
er
s
o
f
th
e
DNNs
ar
e
th
e
in
p
u
t
lay
er
,
h
id
d
en
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Dee
p
lea
r
n
in
g
fo
r
p
r
ed
ictin
g
d
r
u
g
-
r
ela
ted
p
r
o
b
lems in
d
ia
b
etes p
a
tien
ts
(
F
a
tima
M.
S
ma
d
i
)
3001
lay
er
s
(
m
u
ltip
le
h
id
d
e
n
lay
er
s
)
,
an
d
o
u
tp
u
t
la
y
er
.
T
h
e
h
id
d
en
lay
er
’
s
lo
ca
tio
n
is
lo
ca
ted
b
etwe
en
th
e
in
p
u
t
lay
er
an
d
th
e
o
u
tp
u
t
lay
er
.
Al
th
o
u
g
h
t
h
e
in
p
u
t
lay
er
is
r
esp
o
n
s
ib
le
f
o
r
p
ass
in
g
th
e
in
p
u
ts
f
r
o
m
th
e
d
ataset
to
th
e
n
ex
t
lay
er
,
th
e
h
i
d
d
en
lay
er
s
ar
e
r
esp
o
n
s
ib
le
f
o
r
ap
p
ly
i
n
g
th
e
n
o
n
-
lin
ea
r
tr
a
n
s
f
o
r
m
ati
o
n
,
a
n
d
th
e
o
u
tp
u
t
lay
er
is
r
esp
o
n
s
ib
le
f
o
r
p
r
o
d
u
cin
g
th
e
o
u
tp
u
t.
Fu
r
th
er
m
o
r
e,
th
e
ess
en
tial
elem
en
ts
o
f
DNN
ar
e
n
eu
r
o
n
s
,
weig
h
ts
,
n
o
n
-
lin
ea
r
ac
tiv
atio
n
f
u
n
ctio
n
,
an
d
b
ias.
I
n
DNN
th
er
e
is
a
f
u
lly
co
n
n
ec
ted
la
y
er
th
at
is
d
ef
in
e
d
th
r
o
u
g
h
th
e
d
e
n
s
e
class
wh
il
e
ev
er
y
n
e
u
r
o
n
in
th
e
lay
e
r
is
co
n
s
id
er
ed
an
in
p
u
t
to
th
e
n
eu
r
o
n
s
in
th
e
n
ex
t
lay
er
s
,
th
e
m
o
d
el
u
s
es
s
to
ch
asti
c
g
r
ad
ien
t
d
escen
t
in
tr
ain
in
g
th
e
m
o
d
el.
As
well,
T
h
e
DNN
tr
ain
in
g
g
o
in
g
th
r
o
u
g
h
p
h
ases
th
at
s
tar
ts
w
i
th
f
o
r
war
d
p
r
o
p
a
g
atio
n
f
o
llo
wed
b
y
th
e
b
ac
k
p
r
o
p
a
g
atio
n
an
d
th
e
ad
ju
s
tm
en
t
p
r
o
ce
s
s
[
2
3
]
.
T
h
e
f
o
r
war
d
p
r
o
p
ag
atio
n
m
et
h
o
d
in
t
h
e
n
eu
r
al
n
etwo
r
k
p
ass
es
th
e
d
ata
th
r
o
u
g
h
th
e
n
etw
o
r
k
s
tar
tin
g
with
th
e
in
p
u
t
lay
er
b
y
p
ass
in
g
m
u
ltip
le
in
p
u
ts
at
o
n
ce
.
E
ac
h
in
p
u
t
v
alu
e
(
)
n
ee
d
s
to
b
e
m
u
ltip
lied
b
y
th
e
co
r
r
esp
o
n
d
in
g
weig
h
ts
(
)
an
d
th
en
ad
d
e
d
with
all
th
e
o
th
e
r
r
esu
lts
f
o
r
ea
ch
n
e
u
r
o
n
,
th
e
n
ca
lcu
latin
g
th
e
s
u
m
o
f
th
e
weig
h
ted
in
p
u
ts
to
ea
ch
n
e
u
r
o
n
,
an
d
th
e
b
ias
(
)
is
ad
d
ed
i
n
(
1
)
,
th
en
ap
p
l
y
in
g
a
n
o
n
-
lin
ea
r
ac
tiv
atio
n
f
u
n
ctio
n
.
T
h
e
r
esu
lt
o
f
th
e
ac
tiv
atio
n
f
u
n
ctio
n
is
co
n
s
id
er
ed
an
o
u
t
p
u
t
f
o
r
th
e
la
y
er
an
d
an
in
p
u
t
f
o
r
th
e
n
eu
r
o
n
s
in
th
e
n
ex
t h
id
d
en
lay
er
.
T
h
is
p
r
o
ce
s
s
will b
e
ap
p
lied
in
ea
ch
lay
er
u
n
til we
r
ea
ch
th
e
o
u
tp
u
t la
y
e
r
to
g
et
th
e
p
r
ed
ictio
n
(
)
[
8
]
.
=
∑
+
(1
)
T
h
e
b
ac
k
p
r
o
p
ag
atio
n
m
eth
o
d
is
u
s
ed
to
m
in
im
ize
th
e
er
r
o
r
i
n
th
e
n
etwo
r
k
,
wh
e
n
a
n
e
u
r
al
n
etwo
r
k
is
tr
ain
ed
u
s
in
g
f
o
r
war
d
p
r
o
p
a
g
a
tio
n
b
y
r
a
n
d
o
m
ly
i
n
itializin
g
t
h
e
weig
h
ts
o
f
all
n
eu
r
o
n
s
t
o
a
p
p
ly
t
h
e
p
r
ed
ictio
n
,
i
f
th
e
p
r
ed
ictio
n
f
o
r
th
e
m
o
d
el
is
in
co
r
r
ec
t
(
er
r
o
r
in
th
e
p
r
ed
ictio
n
)
,
th
e
b
ac
k
p
r
o
p
ag
ati
o
n
m
u
s
t
b
e
u
s
ed
to
m
in
im
ize
th
e
er
r
o
r
,
wh
ich
is
ca
lcu
lated
b
y
th
e
d
if
f
er
en
ce
b
etwe
en
th
e
o
u
tp
u
t
f
r
o
m
th
e
f
o
r
war
d
p
r
o
p
ag
atio
n
(
p
r
ed
ictio
n
o
u
tp
u
t)
an
d
th
e
ex
p
ec
ted
o
u
tp
u
ts
(
d
esire
d
o
u
tp
u
t)
.
T
h
e
g
o
al
is
to
let
th
e
er
r
o
r
c
lo
s
e
to
ze
r
o
.
W
h
en
th
e
p
r
ed
ictio
n
er
r
o
r
is
g
en
er
at
ed
th
r
o
u
g
h
th
e
f
o
r
war
d
p
r
o
p
a
g
atio
n
,
it will b
e
in
a
h
ig
h
n
u
m
b
er
,
b
y
ap
p
ly
in
g
th
e
b
ac
k
war
d
p
r
o
p
ag
atio
n
f
r
o
m
th
e
o
u
tp
u
t
lay
er
an
d
u
p
d
atin
g
t
h
e
weig
h
ts
to
ac
h
ie
v
e
th
e
f
ir
s
t
l
ay
er
,
th
e
n
ap
p
ly
in
g
th
e
f
o
r
war
d
p
r
o
p
ag
atio
n
f
o
r
t
h
e
s
ec
o
n
d
tim
e
an
d
g
en
er
ati
n
g
th
e
p
r
e
d
ictio
n
er
r
o
r
f
o
r
th
e
s
ec
o
n
d
tim
e,
th
e
m
o
d
el
k
ee
p
s
d
o
in
g
th
ese
s
tep
s
f
o
r
all
s
am
p
les
in
th
e
tr
ain
in
g
d
ata
u
n
til
th
e
s
p
ec
if
ied
ep
o
c
h
s
ar
e
r
ea
ch
ed
an
d
th
e
er
r
o
r
is
m
in
im
ized
[
8
]
.
T
h
e
L
STM
n
etwo
r
k
s
is
co
n
s
i
d
er
ed
a
s
p
ec
ial
ty
p
e
o
f
r
ec
u
r
r
en
t
n
e
u
r
al
n
etwo
r
k
s
(
R
NN)
.
I
t
h
as
a
n
in
p
u
t
lay
er
,
L
STM
h
id
d
en
la
y
er
s
,
an
d
an
o
u
tp
u
t
lay
e
r
.
Acc
o
r
d
in
g
t
o
T
ak
eu
c
h
i
et
a
l.
[
2
4
]
,
th
e
R
NN
h
as
a
d
if
f
icu
lty
in
tr
ain
in
g
b
ec
au
s
e
o
f
th
e
v
an
is
h
in
g
g
r
ad
ien
t
p
r
o
b
lem
,
th
u
s
th
e
L
STM
o
v
er
co
m
es
th
is
p
r
o
b
lem
b
y
co
m
p
o
s
in
g
an
in
p
u
t
g
ate,
an
o
u
tp
u
t
g
ate,
an
d
a
f
o
r
g
et
g
ate.
Ho
wev
er
,
th
e
L
STM
ar
ch
itectu
r
e
co
n
tain
s
a
s
et
o
f
m
em
o
r
y
b
lo
ck
s
,
ea
c
h
b
l
o
ck
c
o
n
tain
in
g
r
ec
u
r
r
en
tly
c
o
n
n
ec
t
ed
m
em
o
r
y
ce
lls
th
at
ar
e
co
n
n
ec
ted
v
ia
g
ates
th
at
allo
w
th
e
L
STM
to
a
d
d
o
r
r
e
m
o
v
e
i
n
f
o
r
m
atio
n
f
r
o
m
th
e
c
ell.
T
h
e
m
em
o
r
y
ce
ll’s
ac
tiv
a
tio
n
f
u
n
ctio
n
allo
ws
s
to
r
in
g
a
s
tate
f
o
r
eith
er
a
s
h
o
r
t
m
o
m
e
n
t
o
r
an
e
x
ten
d
e
d
a
m
o
u
n
t
o
f
tim
e.
Als
o
,
ea
ch
m
em
o
r
y
ce
ll
ap
p
lies
s
ev
er
al
s
tep
s
to
m
o
v
e
th
e
s
tate
to
t
h
e
n
ex
t
L
STM
h
id
d
en
lay
er
a
n
d
i
m
p
lem
en
t
th
e
s
am
e
p
r
o
ce
s
s
to
r
ea
ch
th
e
o
u
tp
u
t
lay
er
.
T
h
e
L
STM
is
tr
ain
ed
th
r
o
u
g
h
f
o
r
war
d
p
r
o
p
a
g
atio
n
an
d
b
ac
k
p
r
o
p
a
g
atio
n
m
eth
o
d
s
[
2
5
]
.
T
h
e
r
e
ar
e
m
an
y
ap
p
licatio
n
s
o
f
L
ST
M
s
u
ch
as
Au
to
m
atic
im
ag
e
c
ap
tio
n
g
en
er
atio
n
an
d
au
to
m
atic
tr
an
s
latio
n
o
f
th
e
tex
t.
Als
o
,
th
e
L
STM
h
as b
ec
o
m
e
in
cr
ea
s
in
g
ly
u
s
ed
in
h
ea
lth
ca
r
e
in
r
ec
e
n
t y
ea
r
s
.
2
.
1
.
E
x
perim
ent
s
d
esig
n
I
n
th
is
s
tu
d
y
,
we
u
s
ed
th
e
s
am
e
m
eth
o
d
th
at
was
u
s
ed
in
th
e
p
r
ev
io
u
s
s
tu
d
y
[
2
1
]
,
th
e
cr
o
s
s
-
v
alid
atio
n
m
eth
o
d
with
1
0
f
o
l
d
s
h
as b
ee
n
ap
p
lied
to
b
u
ild
th
e
d
ee
p
lear
n
in
g
p
r
ed
ictiv
e
m
o
d
els
. T
h
e
f
o
llo
win
g
s
tep
s
wer
e
f
o
llo
wed
to
p
er
f
o
r
m
th
is
s
tu
d
y
:
a.
Usi
n
g
th
e
cr
o
s
s
-
v
alid
atio
n
wit
h
1
0
f
o
ld
s
in
d
ee
p
lear
n
i
n
g
m
o
d
els.
b.
T
u
n
in
g
th
e
h
y
p
er
-
p
ar
am
eter
s
to
g
et
th
e
b
est
s
tr
u
ctu
r
e
f
o
r
th
e
DNN
an
d
L
STM
th
r
o
u
g
h
b
u
ild
in
g
an
d
tr
ain
in
g
th
e
m
o
d
els b
y
th
e
tr
ai
n
in
g
d
ata.
c.
Usi
n
g
th
e
test
in
g
d
ata
to
ap
p
ly
th
e
p
r
ed
ictio
n
.
d.
Usi
n
g
th
e
co
n
f
u
s
io
n
m
atr
ix
m
etr
ics to
ev
alu
ate
th
e
d
ee
p
lear
n
in
g
m
o
d
els.
e.
C
o
m
p
ar
in
g
th
e
ev
alu
ated
m
o
d
el
r
esu
lts
with
th
e
p
r
ev
io
u
s
s
tu
d
y
[
2
1
]
.
2
.
2
.
P
er
f
o
r
m
a
nce
m
e
a
s
ures t
o
ev
a
lua
t
e
t
he
mo
dels
T
o
ev
alu
ate
th
e
d
ee
p
lear
n
i
n
g
m
o
d
els,
th
e
s
am
e
cr
iter
ia
th
at
wer
e
ap
p
lied
f
o
r
th
e
p
r
ev
io
u
s
s
tu
d
y
[
2
1
]
h
av
e
b
ee
n
im
p
lem
en
ted
in
th
i
s
in
v
esti
g
atio
n
.
Mo
r
eo
v
er
,
th
e
o
v
er
all
p
er
f
o
r
m
an
ce
m
etr
ics
wer
e
g
en
er
ated
v
ia
th
e
co
n
f
u
s
io
n
m
etr
ics
to
d
is
c
o
v
er
th
e
m
o
d
el’
s
p
er
f
o
r
m
an
c
e
in
clu
d
i
n
g
ac
cu
r
ac
y
,
s
en
s
itiv
ity
,
an
d
s
p
ec
if
icity
.
T
h
e
co
n
f
u
s
io
n
m
etr
ics we
r
e
ca
lcu
lated
u
s
in
g
(
2
)
-
(
4
)
:
Acc
u
r
ac
y
=
+
+
+
+
(
2
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
9
9
8
-
3
0
0
9
3002
Sen
s
itiv
ity
=
+
(
3
)
Sp
ec
if
icity
=
+
(
4
)
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
W
e
attem
p
ted
to
tu
n
e
th
e
b
est
n
etwo
r
k
s
tr
u
ctu
r
e
t
o
ac
h
iev
e
th
e
b
est
p
er
f
o
r
m
a
n
ce
.
T
h
e
r
e
wer
e
m
an
y
d
if
f
er
en
t
v
alu
es
f
r
o
m
h
y
p
e
r
-
p
ar
am
eter
s
to
tr
ain
th
e
DNN
a
n
d
L
STM
m
o
d
els
to
e
n
h
an
ce
it
an
d
g
et
th
e
b
est
p
er
f
o
r
m
an
ce
o
f
th
e
DNN
an
d
L
STM
m
o
d
els
b
ased
o
n
m
an
y
ex
p
er
im
en
ts
to
ch
o
o
s
e
th
e
b
est
s
tr
u
ctu
r
e
o
f
th
e
DNN
an
d
L
STM
.
Ap
p
ly
in
g
t
h
e
p
r
o
ce
s
s
o
f
tr
ain
in
g
th
e
m
o
d
el
with
th
e
wr
ite
h
y
p
er
-
p
ar
a
m
eter
s
is
b
ased
o
n
ex
p
er
im
en
ts
an
d
r
esu
lts
b
y
tr
i
al
an
d
er
r
o
r
.
I
t
d
ep
e
n
d
s
o
n
th
e
ex
p
er
ien
ce
with
m
an
y
ex
p
e
r
im
en
ts
b
y
tu
n
in
g
th
e
class
if
ier
m
o
d
el,
it
tak
es
tim
e
an
d
ef
f
o
r
t
to
tu
n
e
th
e
h
y
p
e
r
-
p
ar
am
eter
s
,
wh
en
tu
n
i
n
g
h
y
p
er
-
p
ar
a
m
eter
s
,
we
h
av
e
to
s
elec
t
s
o
m
e
p
ar
am
eter
s
with
th
e
co
r
r
ec
t
co
n
f
ig
u
r
atio
n
f
o
r
ea
ch
p
ar
am
ete
r
o
n
e
at
a
tim
e
an
d
th
en
co
n
tin
u
e
to
co
n
f
i
g
u
r
e
t
h
e
n
ex
t
p
ar
am
eter
an
d
s
o
o
n
to
g
et
t
h
e
b
est p
er
f
o
r
m
an
ce
y
o
u
a
r
e
lo
o
k
in
g
f
o
r
[
2
3
]
.
3
.
1
.
Dee
p
neura
l net
wo
rk
s
e
x
perim
ent
s
a
nd
re
s
ults
T
h
e
DNN
alg
o
r
ith
m
was
u
s
ed
to
cr
ea
te
th
e
class
if
ier
m
o
d
el,
we
ap
p
lied
th
e
cr
o
s
s
-
v
alid
atio
n
m
eth
o
d
.
B
ased
o
n
s
ev
er
al
ex
p
e
r
im
en
ts
,
we
s
tar
ted
b
y
d
ef
in
in
g
th
e
b
aselin
e
ar
ch
itectu
r
e
to
s
tar
t
b
u
ild
in
g
th
e
m
o
d
el
o
f
th
e
d
ee
p
n
eu
r
al
n
etwo
r
k
s
m
o
d
el
with
th
e
h
y
p
er
-
p
ar
am
eter
s
s
p
ec
if
ied
.
T
h
e
b
aselin
e
ar
c
h
itectu
r
e
th
at
was
u
s
ed
to
s
tar
t
b
u
ild
in
g
th
e
DNN
m
o
d
el:
a
f
u
lly
co
n
n
ec
ted
n
eu
r
al
n
etwo
r
k
with
th
r
ee
h
id
d
e
n
lay
er
s
with
3
1
n
eu
r
o
n
s
in
ea
ch
h
id
d
en
lay
er
,
r
esp
ec
ti
v
ely
,
th
e
ac
tiv
atio
n
f
u
n
ctio
n
f
o
r
ea
c
h
lay
e
r
was
r
ec
tifie
d
lin
ea
r
u
n
it
(
R
eL
U
)
,
th
e
d
r
o
p
o
u
t
was
0
.
5
,
f
o
r
th
e
o
p
ti
m
izer
was
Ad
am
,
th
e
ep
o
c
h
s
was
5
0
,
an
d
th
e
b
atch
s
ize
was
3
2
.
T
h
e
r
esu
lts
f
o
r
th
e
b
aselin
e
wer
e
ac
cu
r
ac
y
:
8
8
.
5
4
,
s
en
s
itiv
ity
:
8
7
.
5
0
,
an
d
s
p
ec
if
icity
:
8
9
.
5
8
.
T
h
e
f
o
llo
wi
n
g
ex
p
er
im
e
n
ts
wer
e
p
er
f
o
r
m
ed
to
g
et
th
e
b
est s
tr
u
c
tu
r
e
f
o
r
th
e
DNN
m
o
d
el.
3
.
1
.
1
.
T
un
e
t
he
nu
m
ber
o
f
hid
den ne
uro
ns
in ea
ch
h
idd
e
n la
y
er
T
h
e
f
o
llo
win
g
e
x
p
er
im
e
n
ts
wer
e
u
s
ed
to
tu
n
e
t
h
e
n
u
m
b
er
o
f
h
id
d
e
n
n
e
u
r
o
n
s
in
ea
c
h
h
i
d
d
en
lay
er
.
T
ab
le
1
d
is
p
lay
s
th
e
ex
p
er
im
en
t
r
esu
lts
o
n
d
if
f
er
en
t
n
u
m
b
er
s
o
f
h
id
d
en
n
eu
r
o
n
s
.
W
h
en
u
s
in
g
7
8
,
1
0
9
,
1
2
4
,
1
5
5
,
an
d
1
7
1
h
id
d
e
n
n
eu
r
o
n
s
in
th
e
th
r
ee
h
id
d
e
n
lay
er
s
,
we
test
ed
th
ese
n
eu
r
o
n
s
with
ep
o
ch
s
(
5
0
)
,
an
d
b
atch
s
ize
(
3
2
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
ed
wh
en
we
u
s
ed
th
e
1
2
4
h
id
d
en
n
eu
r
o
n
s
,
th
er
e
was
an
i
m
p
r
o
v
e
m
en
t
in
th
e
r
esu
lts
u
n
til
we
r
ea
ch
ed
th
e
1
2
4
h
id
d
en
n
eu
r
o
n
s
u
s
ed
in
t
h
e
DNN
m
o
d
el,
af
ter
(
1
2
4
)
we
n
o
ted
th
at
th
e
r
e
was
n
o
im
p
r
o
v
em
e
n
t in
th
e
r
esu
lts
,
we
co
n
clu
d
e
d
th
at
(
1
2
4
)
h
id
d
en
n
eu
r
o
n
s
u
s
ed
in
th
e
DNN
m
o
d
el
was th
e
b
est.
T
ab
le
1
.
T
h
e
h
id
d
en
n
e
u
r
o
n
s
ex
p
er
im
en
t
f
o
r
DNN
N
u
mb
e
r
o
f
h
i
d
d
e
n
n
e
u
r
o
n
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
78
9
0
.
1
0
9
1
.
6
6
8
8
.
5
4
1
0
9
8
9
.
0
6
8
8
.
5
4
8
9
.
5
8
1
2
4
9
1
.
1
4
8
8
.
5
4
9
3
.
7
5
1
5
5
8
8
.
5
4
9
2
.
7
0
8
4
.
3
7
1
7
1
8
7
.
5
0
9
4
.
7
9
8
0
.
2
0
3
.
1
.
2
.
T
un
e
t
he
nu
m
ber
o
f
hid
den la
y
er
s
T
h
e
f
o
llo
win
g
e
x
p
er
im
en
ts
w
er
e
u
s
ed
t
o
tu
n
e
th
e
n
u
m
b
er
o
f
h
i
d
d
en
lay
er
s
.
T
a
b
le
2
d
is
p
lay
s
th
e
ex
p
er
im
en
t
r
esu
lts
o
n
d
if
f
e
r
e
n
t
n
u
m
b
er
s
o
f
h
id
d
en
lay
er
s
.
W
h
en
u
s
in
g
3
,
4
,
5
,
6
,
a
n
d
7
h
id
d
en
lay
e
r
s
,
we
test
ed
th
ese
h
id
d
en
la
y
er
s
with
(
1
2
4
)
h
id
d
en
n
eu
r
o
n
s
in
e
ac
h
h
id
d
en
lay
e
r
r
esp
ec
tiv
ely
,
ep
o
ch
s
(
5
0
)
,
an
d
b
atch
s
ize
(
3
2
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
e
d
wh
e
n
we
u
s
ed
t
h
r
ee
h
id
d
en
lay
er
s
,
we
c
o
n
clu
d
e
d
th
at
th
e
n
u
m
b
er
o
f
h
id
d
en
lay
e
r
s
u
s
ed
in
th
e
D
NN
m
o
d
el
was
th
e
b
est,
as
th
er
e
was
n
o
im
p
r
o
v
em
en
t
i
n
th
e
r
esu
lts
wh
en
in
cr
ea
s
in
g
th
e
n
u
m
b
er
o
f
h
id
d
en
lay
er
s
.
T
ab
le
2
.
Nu
m
b
er
o
f
h
id
d
en
la
y
er
s
ex
p
er
im
e
n
t f
o
r
DNN
N
u
mb
e
r
o
f
h
i
d
d
e
n
l
a
y
e
r
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
3
9
1
.
1
4
8
8
.
5
4
9
3
.
7
5
4
8
6
.
9
7
8
9
.
5
8
8
4
.
3
7
5
8
5
.
4
1
9
0
.
6
2
8
0
.
2
0
6
7
9
.
6
8
9
6
.
8
7
6
2
.
5
0
7
7
2
.
9
1
9
5
.
8
3
5
0
.
0
0
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Dee
p
lea
r
n
in
g
fo
r
p
r
ed
ictin
g
d
r
u
g
-
r
ela
ted
p
r
o
b
lems in
d
ia
b
etes p
a
tien
ts
(
F
a
tima
M.
S
ma
d
i
)
3003
3
.
1
.
3
.
T
un
e
t
he
E
po
ch
s
f
o
r
DNN
T
h
e
f
o
llo
win
g
e
x
p
er
im
en
ts
w
er
e
u
s
ed
to
tu
n
e
th
e
ep
o
ch
s
.
T
ab
le
3
d
is
p
lay
s
th
e
ex
p
er
im
en
t
r
esu
lts
o
n
d
if
f
er
en
t
E
p
o
ch
s
.
W
h
en
u
s
in
g
5
0
,
2
0
0
,
3
0
0
,
4
0
0
,
an
d
5
0
0
e
p
o
ch
s
,
we
test
ed
th
ese
ep
o
ch
s
with
(
1
2
4
)
h
id
d
e
n
n
eu
r
o
n
s
in
ea
ch
h
id
d
en
lay
e
r
r
esp
ec
tiv
ely
,
with
th
r
ee
h
i
d
d
en
lay
er
s
,
an
d
b
atch
s
ize
(
3
2
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
e
d
wh
en
we
u
s
ed
4
0
0
e
p
o
ch
s
,
th
er
e
was
an
im
p
r
o
v
em
en
t
in
th
e
r
esu
lts
u
n
til
we
r
ea
c
h
ed
th
e
4
0
0
ep
o
c
h
s
in
th
e
DNN
m
o
d
el,
af
ter
4
0
0
ep
o
ch
s
,
we
n
o
ted
t
h
at
th
er
e
was
n
o
im
p
r
o
v
em
e
n
t
in
t
h
e
r
es
u
lts
,
we
co
n
clu
d
e
d
th
at
(
4
0
0
)
ep
o
ch
s
u
s
ed
in
t
h
e
DNN
m
o
d
el
wer
e
th
e
b
est.
T
ab
le
3
.
Nu
m
b
er
o
f
E
p
o
ch
s
ex
p
er
im
en
t
f
o
r
DNN
N
u
mb
e
r
o
f
E
p
o
c
h
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
50
9
1
.
1
4
8
8
.
5
4
9
3
.
7
5
2
0
0
9
5
.
3
1
9
7
.
9
1
9
2
.
7
0
3
0
0
9
4
.
7
9
9
4
.
7
9
9
4
.
7
9
4
0
0
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
5
0
0
9
2
.
7
0
9
5
.
8
3
8
9
.
5
8
3
.
1
.
4
.
T
un
e
t
he
ba
t
ch
s
ize
f
o
r
DNN
T
h
e
f
o
llo
win
g
ex
p
er
im
e
n
ts
w
er
e
u
s
ed
to
t
u
n
e
th
e
b
atch
s
ize
.
T
ab
le
4
d
is
p
lay
s
th
e
ex
p
e
r
im
en
t
r
esu
lts
o
n
d
i
f
f
er
en
t
b
atch
s
izes
.
W
h
en
u
s
in
g
1
6
,
3
2
,
6
4
,
1
2
8
,
a
n
d
2
5
6
b
atch
s
ize,
we
test
ed
th
e
s
e
b
atch
s
izes
with
(
1
2
4
)
h
i
d
d
en
n
e
u
r
o
n
s
in
ea
ch
h
id
d
en
lay
er
r
esp
ec
tiv
ely
,
wit
h
th
r
ee
h
i
d
d
en
lay
er
s
,
an
d
ep
o
ch
s
(
4
0
0
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
e
d
wh
en
we
u
s
ed
3
2
b
atch
s
ize
,
we
c
o
n
clu
d
e
d
t
h
at
th
e
3
2
-
b
atch
s
ize
u
s
ed
in
th
e
DNN
m
o
d
el
was
th
e
b
est,
as th
er
e
was n
o
im
p
r
o
v
em
en
t in
th
e
r
esu
lts
wh
en
te
s
tin
g
d
if
f
er
en
t e
p
o
ch
s
.
T
ab
le
4
.
Nu
m
b
er
o
f
b
atch
s
ize
ex
p
er
im
en
t
f
o
r
DNN
N
u
mb
e
r
o
f
b
a
t
c
h
si
z
e
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
16
9
4
.
7
9
9
4
.
7
9
9
4
.
7
9
32
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
64
9
5
.
3
1
9
8
.
9
5
9
1
.
6
6
1
2
8
9
6
.
8
7
9
8
.
9
5
9
4
.
7
9
2
5
6
9
2
.
7
0
9
6
.
8
7
8
8
.
5
4
3
.
1
.
5
.
T
un
e
t
he
dro
po
ut
r
a
t
e
f
o
r
DNN
T
h
e
f
o
l
l
o
w
i
n
g
e
x
p
e
r
i
m
e
n
t
s
we
r
e
u
s
e
d
t
o
t
u
n
e
t
h
e
d
r
o
p
o
u
t
.
T
a
b
l
e
5
d
i
s
p
l
a
y
s
t
h
e
e
x
p
e
r
i
m
e
n
t
r
e
s
u
l
ts
o
n
d
i
f
f
e
r
e
n
t
d
r
o
p
o
u
t
r
a
t
es
.
W
h
e
n
u
s
i
n
g
2
0
%
,
3
0
%
,
4
0
%
,
a
n
d
5
0
%
d
r
o
p
o
u
t
r
a
t
es
,
w
e
t
e
s
t
e
d
t
h
e
s
e
d
r
o
p
o
u
t
s
wi
t
h
(
1
2
4
)
h
i
d
d
e
n
n
e
u
r
o
n
s
i
n
e
a
c
h
h
i
d
d
e
n
l
a
y
e
r
r
e
s
p
e
ct
i
v
e
l
y
,
w
i
t
h
t
h
r
e
e
h
i
d
d
e
n
l
a
y
e
r
s
,
a
n
d
e
p
o
ch
s
(
4
0
0
)
,
a
n
d
b
a
t
c
h
s
i
z
e
(
3
2
)
,
t
h
e
b
e
s
t
r
e
s
u
lt
o
c
c
u
r
r
e
d
w
h
e
n
w
e
u
s
e
d
5
0
%
(
0
.
5
)
d
r
o
p
o
u
t
r
a
t
e
,
w
e
c
o
n
c
l
u
d
e
d
t
h
a
t
t
h
e
0
.
5
d
r
o
p
o
u
t
u
s
e
d
i
n
t
h
e
D
N
N
m
o
d
e
l
w
as
t
h
e
b
e
s
t
,
as
t
h
e
r
e
w
a
s
n
o
im
p
r
o
v
e
m
e
n
t
i
n
t
h
e
r
es
u
l
ts
wh
e
n
t
e
s
ti
n
g
d
i
f
f
e
r
e
n
t
d
r
o
p
o
u
t
s
.
T
ab
le
5
.
Nu
m
b
er
o
f
d
r
o
p
o
u
t e
x
p
er
im
en
ts
f
o
r
DNN
N
u
mb
e
r
o
f
d
r
o
p
o
u
t
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
0
.
2
9
5
.
3
1
9
8
.
9
5
9
1
.
6
6
0
.
3
9
6
.
3
5
9
8
.
9
5
9
3
.
7
5
0
.
4
9
3
.
7
5
9
7
.
9
1
8
9
.
5
8
0
.
5
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
3
.
1
.
6
.
T
un
e
t
he
a
ct
iv
a
t
io
n f
u
nct
io
n
f
o
r
DNN
T
h
e
f
o
llo
win
g
ex
p
e
r
im
en
ts
wer
e
u
s
ed
to
tu
n
e
th
e
ac
ti
v
atio
n
f
u
n
ctio
n
.
T
ab
le
6
d
i
s
p
lay
s
th
e
ex
p
er
im
en
t
r
esu
lts
o
n
d
if
f
er
en
t
ac
tiv
atio
n
f
u
n
ctio
n
s
.
W
h
en
u
s
in
g
R
eL
U,
T
a
n
h
,
a
n
d
Sig
m
o
id
ac
tiv
atio
n
f
u
n
ctio
n
s
,
we
test
ed
th
ese
ac
ti
v
atio
n
f
u
n
ctio
n
s
with
(
1
2
4
)
h
i
d
d
en
n
eu
r
o
n
s
in
ea
ch
h
id
d
e
n
lay
er
r
esp
ec
tiv
ely
,
with
th
r
ee
h
id
d
e
n
lay
er
s
,
an
d
ep
o
ch
s
(
4
0
0
)
,
d
r
o
p
o
u
t
(
0
.
5
)
,
a
n
d
b
atch
s
ize
(
3
2
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
ed
wh
e
n
we
u
s
ed
R
e
L
U
ac
tiv
atio
n
f
u
n
ctio
n
,
we
co
n
cl
u
d
ed
th
at
th
e
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
u
s
e
d
in
th
e
DNN
m
o
d
el
was
th
e
b
est,
a
s
th
er
e
was
n
o
im
p
r
o
v
em
en
t
in
th
e
r
esu
lts
wh
en
test
in
g
d
if
f
er
en
t
ac
tiv
atio
n
f
u
n
ctio
n
s
.
Acc
o
r
d
in
g
ly
,
Fig
u
r
e
2
s
u
m
m
a
r
izes th
e
s
tr
u
ctu
r
e
o
f
DNN.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
9
9
8
-
3
0
0
9
3004
T
ab
le
6
.
T
h
e
ac
tiv
atio
n
f
u
n
cti
o
n
s
ex
p
er
im
e
n
t f
o
r
DNN
Th
e
a
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
R
e
L
U
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
Ta
n
h
9
4
.
2
7
9
5
.
8
3
9
2
.
7
0
S
i
g
m
o
i
d
9
3
.
7
5
9
4
.
7
9
9
2
.
7
0
Fig
u
r
e
2
.
T
h
e
s
tr
u
ctu
r
e
o
f
d
ee
p
n
eu
r
al
n
etwo
r
k
s
3
.
2
.
L
ST
M
net
wo
rk
s
ex
perim
ent
s
a
nd
re
s
ults
B
ased
o
n
s
ev
er
al
ex
p
er
im
en
ts
,
we
s
tar
ted
b
y
d
ef
in
in
g
th
e
b
aselin
e
ar
ch
itectu
r
e
to
s
tar
t
b
u
ild
in
g
th
e
m
o
d
el
o
f
t
h
e
L
STM
with
th
e
h
y
p
er
-
p
ar
am
eter
s
s
p
ec
if
ied
.
T
h
e
r
esu
lts
f
o
r
th
e
b
aselin
e
ar
ch
itectu
r
e
wer
e
Acc
u
r
ac
y
:
8
6
.
4
5
,
Sen
s
itiv
ity
:
8
4
.
3
7
,
an
d
Sp
ec
i
f
icity
:
8
8
.
5
4
.
T
h
e
n
u
m
b
er
o
f
L
STM
lay
er
s
was
(
3
)
,
th
e
h
id
d
en
n
eu
r
o
n
s
wer
e
(
4
7
)
in
ea
ch
L
STM
lay
er
,
th
e
ep
o
ch
s
wer
e
(
5
0
)
,
th
e
b
atch
s
ize
was
(
1
6
)
,
th
e
d
r
o
p
o
u
t
was
(
0
.
3
)
,
an
d
th
e
o
p
tim
izer
was
Ad
am
.
T
h
e
f
o
llo
win
g
ex
p
er
im
en
ts
wer
e
p
er
f
o
r
m
e
d
to
g
et
th
e
b
e
s
t
s
tr
u
ctu
r
e
f
o
r
t
h
e
L
STM
m
o
d
el.
3
.
2
.
1
.
T
un
e
t
he
nu
m
ber
o
f
hid
den ne
uro
ns
in ea
ch
L
S
T
M
la
y
er
s
T
h
e
f
o
llo
win
g
e
x
p
er
im
e
n
ts
wer
e
u
s
ed
to
tu
n
e
t
h
e
n
u
m
b
er
o
f
h
id
d
e
n
n
e
u
r
o
n
s
in
ea
c
h
h
i
d
d
en
lay
er
.
T
ab
le
7
,
d
is
p
lay
s
th
e
ex
p
e
r
im
e
n
t
r
esu
lts
o
n
d
if
f
e
r
en
t n
u
m
b
e
r
s
o
f
h
i
d
d
en
n
eu
r
o
n
s
.
W
h
en
u
s
i
n
g
3
1
,
6
2
,
7
8
,
1
0
9
,
an
d
1
2
4
h
i
d
d
en
n
eu
r
o
n
s
in
th
e
th
r
ee
h
id
d
en
lay
e
r
s
,
we
test
ed
th
ese
n
eu
r
o
n
s
with
ep
o
c
h
s
(
5
0
)
,
b
atch
s
ize
(
1
6
)
,
an
d
d
r
o
p
o
u
t (
0
.
3
)
,
th
e
b
est r
es
u
lt o
cc
u
r
r
e
d
wh
en
we
u
s
ed
th
e
3
1
h
id
d
en
n
e
u
r
o
n
s
,
th
e
r
e
was a
n
im
p
r
o
v
em
en
t in
th
e
r
esu
lts
wh
en
test
ed
3
1
h
i
d
d
en
n
eu
r
o
n
s
u
s
ed
i
n
th
e
L
STM
m
o
d
el,
af
ter
(
3
1
)
we
n
o
ted
th
at
th
er
e
was
n
o
im
p
r
o
v
em
e
n
t in
th
e
r
esu
lts
,
we
co
n
clu
d
e
d
th
at
(
3
1
)
h
id
d
e
n
n
eu
r
o
n
s
u
s
ed
i
n
th
e
L
STM
m
o
d
el
was th
e
b
est.
T
ab
le
7
.
T
h
e
h
id
d
en
n
e
u
r
o
n
s
ex
p
er
im
en
t
f
o
r
L
STM
N
u
mb
e
r
o
f
h
i
d
d
e
n
n
e
u
r
o
n
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
31
8
8
.
0
2
8
4
.
3
7
9
1
.
6
6
62
8
4
.
8
9
8
8
.
5
4
8
1
.
2
5
78
8
2
.
8
1
9
1
.
6
6
7
3
.
9
5
1
0
9
8
8
.
0
2
8
5
.
4
1
9
0
.
6
2
1
2
4
8
7
.
5
0
8
4
.
3
7
9
0
.
6
2
3
.
2
.
2
.
T
un
e
t
he
nu
m
ber
o
f
L
ST
M
la
y
er
s
T
h
e
f
o
llo
win
g
e
x
p
er
im
e
n
ts
wer
e
u
s
ed
t
o
tu
n
e
th
e
n
u
m
b
er
o
f
L
STM
h
id
d
en
lay
er
s
.
T
a
b
le
8
d
is
p
lay
s
th
e
ex
p
er
im
en
t
r
esu
lts
o
n
d
if
f
er
en
t
n
u
m
b
er
s
o
f
L
STM
h
id
d
en
lay
er
s
.
W
h
en
u
s
in
g
3
,
4
,
5
,
6
,
an
d
7
L
STM
h
id
d
en
la
y
er
s
,
we
test
ed
th
ese
h
id
d
en
lay
er
s
with
(
3
1
)
h
id
d
en
n
eu
r
o
n
s
in
ea
ch
L
STM
h
id
d
en
la
y
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Dee
p
lea
r
n
in
g
fo
r
p
r
ed
ictin
g
d
r
u
g
-
r
ela
ted
p
r
o
b
lems in
d
ia
b
etes p
a
tien
ts
(
F
a
tima
M.
S
ma
d
i
)
3005
r
esp
ec
tiv
ely
,
ep
o
c
h
s
(
5
0
)
,
an
d
b
atch
s
ize
(
1
6
)
,
a
n
d
d
r
o
p
o
u
t
(
0
.
3
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
ed
wh
en
we
u
s
ed
f
iv
e
L
STM
h
id
d
en
lay
er
s
,
th
er
e
w
as
an
im
p
r
o
v
em
en
t
in
th
e
r
esu
lts
u
n
til
we
r
ea
ch
ed
th
e
f
iv
e
L
STM
lay
er
s
u
s
ed
in
th
e
L
STM
m
o
d
el,
af
te
r
f
iv
e
L
STM
lay
er
s
we
n
o
ted
th
a
t
th
er
e
was
n
o
im
p
r
o
v
em
en
t
in
th
e
r
esu
lts
,
we
co
n
clu
d
e
d
th
at
f
iv
e
L
STM
lay
er
s
u
s
ed
in
th
e
L
STM
m
o
d
el
wer
e
th
e
b
est.
T
ab
le
8
.
Nu
m
b
er
o
f
h
id
d
en
la
y
er
s
ex
p
er
im
e
n
t f
o
r
L
STM
N
u
mb
e
r
o
f
LST
M
l
a
y
e
r
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
3
8
8
.
0
2
8
4
.
3
7
9
1
.
6
6
4
8
6
.
4
5
9
1
.
6
6
8
1
.
2
5
5
8
8
.
0
2
8
2
.
2
9
9
3
.
7
5
6
8
6
.
9
7
8
3
.
3
3
9
0
.
6
2
7
8
6
.
4
5
8
6
.
4
5
8
6
.
4
5
3
.
2
.
3
.
T
un
e
t
he
epo
chs
f
o
r
L
ST
M
T
h
e
f
o
llo
win
g
e
x
p
er
im
en
ts
w
er
e
u
s
ed
to
tu
n
e
th
e
ep
o
ch
s
.
T
ab
le
9
d
is
p
lay
s
th
e
ex
p
er
im
en
t
r
esu
lts
o
n
d
if
f
er
en
t
E
p
o
ch
s
.
W
h
en
u
s
in
g
5
0
,
1
0
0
,
2
0
0
,
3
0
0
,
an
d
4
0
0
e
p
o
ch
s
,
we
test
ed
th
ese
ep
o
c
h
s
with
(
3
1
)
h
id
d
en
n
eu
r
o
n
s
in
ea
ch
L
STM
h
id
d
e
n
lay
er
s
r
esp
ec
tiv
ely
,
with
f
iv
e
L
STM
h
id
d
en
lay
er
s
,
a
n
d
b
atch
s
ize
(
1
6
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
e
d
wh
en
we
u
s
ed
5
0
ep
o
c
h
s
,
we
co
n
clu
d
ed
th
at
th
e
5
0
ep
o
ch
s
u
s
ed
i
n
th
e
L
STM
m
o
d
el
wer
e
th
e
b
est,
as th
er
e
was n
o
im
p
r
o
v
em
e
n
t in
th
e
r
esu
lts
wh
en
test
in
g
d
if
f
er
e
n
t e
p
o
c
h
s
.
T
ab
le
9
.
E
p
o
ch
s
ex
p
er
im
en
t f
o
r
L
STM
N
u
mb
e
r
o
f
E
p
o
c
h
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
50
8
8
.
0
2
8
2
.
2
9
9
3
.
7
5
1
0
0
8
8
.
0
2
8
5
.
4
1
9
0
.
6
2
2
0
0
9
0
.
6
2
9
0
.
6
2
9
0
.
6
2
3
0
0
9
0
.
1
0
9
7
.
9
1
8
2
.
2
9
4
0
0
9
2
.
7
0
9
7
.
9
1
8
7
.
5
0
3
.
2
.
4
.
T
un
e
t
he
ba
t
ch
s
ize
f
o
r
L
ST
M
T
h
e
f
o
llo
win
g
ex
p
er
im
en
ts
wer
e
u
s
ed
to
tu
n
e
t
h
e
b
atc
h
s
ize.
T
ab
le
1
0
d
is
p
lay
s
th
e
ex
p
er
im
e
n
t
r
esu
lts
o
n
d
if
f
er
e
n
t
b
atch
s
ize
s
.
W
h
en
u
s
in
g
1
6
,
3
2
,
6
4
,
1
2
8
,
an
d
2
5
6
b
atch
s
ize
,
we
test
ed
th
ese
b
atch
s
izes
with
(
3
1
)
h
i
d
d
en
n
eu
r
o
n
s
in
e
ac
h
L
STM
h
id
d
en
la
y
er
s
r
esp
ec
tiv
ely
,
with
f
iv
e
L
STM
h
id
d
en
lay
er
s
,
d
r
o
p
o
u
t
(
0
.
3
)
,
an
d
e
p
o
ch
s
(
5
0
)
,
th
e
b
es
t
r
esu
lt
o
cc
u
r
r
ed
wh
en
we
u
s
ed
3
2
b
atc
h
s
ize,
th
e
r
e
was
a
n
i
m
p
r
o
v
e
m
en
t
in
th
e
r
esu
lts
wh
en
test
ed
3
2
b
atc
h
s
ize
u
s
ed
in
th
e
L
STM
m
o
d
el,
af
ter
(
3
2
)
we
n
o
ted
th
at
th
er
e
was
n
o
im
p
r
o
v
em
e
n
t in
th
e
r
esu
lts
,
we
co
n
clu
d
e
d
th
at
(
3
2
)
b
atch
s
ize
u
s
ed
in
th
e
L
STM
m
o
d
el
wa
s
th
e
b
est.
T
ab
le
10.
B
atch
s
ize
ex
p
e
r
im
e
n
t f
o
r
L
STM
N
u
mb
e
r
o
f
b
a
t
c
h
si
z
e
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
16
8
8
.
0
2
8
2
.
2
9
9
3
.
7
5
32
8
6
.
9
7
7
9
.
1
6
9
4
.
7
9
64
8
5
.
9
3
8
0
.
2
0
9
1
.
6
6
1
2
8
8
4
.
8
9
8
5
.
4
1
8
4
.
3
7
2
5
6
8
4
.
8
9
8
3
.
3
3
8
6
.
4
5
3
.
2
.
5
.
T
un
e
t
he
dro
po
ut
ra
t
e
f
o
r
L
ST
M
T
h
e
f
o
llo
win
g
ex
p
er
im
en
ts
w
er
e
u
s
ed
to
tu
n
e
th
e
d
r
o
p
o
u
t.
T
ab
le
1
1
d
is
p
lay
s
th
e
ex
p
er
i
m
en
t
r
esu
lts
o
n
d
i
f
f
er
en
t
d
r
o
p
o
u
t
r
ates.
W
h
en
u
s
in
g
2
0
%,
3
0
%,
4
0
%,
a
n
d
5
0
%
d
r
o
p
o
u
t
r
ates,
we
test
ed
th
ese
d
r
o
p
o
u
ts
with
(
3
1
)
h
id
d
e
n
n
eu
r
o
n
s
in
ea
ch
L
STM
h
id
d
en
lay
er
s
r
e
s
p
ec
tiv
ely
,
with
f
iv
e
L
STM
h
id
d
en
lay
er
s
,
an
d
ep
o
ch
s
(
5
0
)
,
an
d
b
atch
s
ize
(
3
2
)
,
th
e
b
est
r
esu
lt
o
cc
u
r
r
ed
wh
en
we
u
s
ed
4
0
%
(
0
.
4
)
d
r
o
p
o
u
t
r
ate,
t
h
er
e
was
a
n
im
p
r
o
v
em
e
n
t
in
th
e
r
esu
lts
wh
en
test
ed
0
.
4
d
r
o
p
o
u
ts
u
s
ed
in
th
e
L
STM
m
o
d
el,
af
te
r
(
0
.
4
)
we
n
o
ted
th
at
th
e
r
e
was
n
o
im
p
r
o
v
em
e
n
t
in
th
e
r
e
s
u
lts
,
we
co
n
clu
d
ed
th
at
(
0
.
4
)
d
r
o
p
o
u
t
u
s
ed
in
th
e
L
STM
m
o
d
el
was
th
e
b
est.
Acc
o
r
d
in
g
ly
,
Fig
u
r
e
3
s
u
m
m
a
r
izes th
e
s
tr
u
ctu
r
e
o
f
L
STM
.
Af
ter
tr
y
in
g
m
an
y
ex
p
er
im
e
n
t
s
to
f
in
d
o
u
t
th
e
b
est
s
tr
u
ctu
r
e,
th
e
tr
ials
s
h
o
wed
th
at
th
e
b
est
s
tr
u
ctu
r
e
f
o
r
th
e
DNN
alg
o
r
ith
m
was
w
h
en
u
s
in
g
1
2
4
h
id
d
en
n
e
u
r
o
n
s
in
th
r
ee
h
id
d
en
la
y
er
s
,
with
a
d
r
o
p
o
u
t
r
ate
o
f
0
.
5
,
ep
o
ch
s
(
4
0
0
)
,
b
atch
s
ize
(
3
2
)
,
an
d
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
.
Als
o
,
th
e
tr
ials
s
h
o
wed
th
at
th
e
b
est s
tr
u
ctu
r
e
f
o
r
th
e
L
STM
alg
o
r
ith
m
was
wh
en
u
s
in
g
3
1
h
i
d
d
en
n
eu
r
o
n
s
in
f
iv
e
L
STM
h
id
d
e
n
lay
er
s
,
with
a
d
r
o
p
o
u
t
r
ate
0
.
4
,
ep
o
ch
s
(
5
0
)
,
an
d
b
atch
s
ize
(
3
2
)
.
A
s
im
ilar
s
p
ec
if
icity
ac
h
iev
ed
b
y
th
e
DNN
a
n
d
L
STM
m
o
d
els
is
o
b
s
er
v
ed
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8
7
0
8
I
n
t J E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
15
,
No
.
3
,
J
u
n
e
20
25
:
2
9
9
8
-
3
0
0
9
3006
W
h
er
ea
s
th
e
ac
cu
r
ac
y
an
d
t
h
e
s
en
s
itiv
ity
f
o
r
th
e
DNN
m
o
d
el
wer
e
h
ig
h
er
th
a
n
th
e
L
ST
M
m
o
d
el.
T
ab
le
1
2
d
em
o
n
s
tr
ates th
e
r
esu
lts
f
o
r
t
h
e
DNN
an
d
L
STM
to
g
et
th
e
b
est s
tr
u
ctu
r
e.
T
ab
le
1
1
.
N
u
m
b
er
d
r
o
p
o
u
t e
x
p
er
im
en
t f
o
r
L
STM
N
u
mb
e
r
o
f
d
r
o
p
o
u
t
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
0
.
2
8
6
.
9
7
8
1
.
2
5
9
2
.
7
0
0
.
3
8
6
.
9
7
7
9
.
1
6
9
4
.
7
9
0
.
4
8
7
.
5
0
7
9
.
1
6
9
5
.
8
3
0
.
5
8
6
.
4
5
7
9
.
1
6
9
3
.
7
5
Fig
u
r
e
3
.
T
h
e
s
tr
u
ctu
r
e
o
f
lo
n
g
-
s
h
o
r
t te
r
m
m
e
m
o
r
y
n
etwo
r
k
s
T
ab
le
1
2
.
Su
m
m
ar
y
o
f
th
e
r
esu
lts
o
b
tain
ed
b
y
DNN
an
d
L
S
T
M
b
est s
tr
u
ctu
r
e
C
l
a
s
si
f
i
e
r
s
H
y
p
e
r
p
a
r
a
me
t
e
r
s
O
p
t
i
o
n
s
A
c
c
u
r
a
c
y
%
S
e
n
s
i
t
i
v
i
t
y
%
S
p
e
c
i
f
i
c
i
t
y
%
DNN
N
o
.
o
f
h
i
d
d
e
n
n
e
u
r
o
n
s
[
7
8
,
1
0
9
,
124
,
1
5
5
,
1
7
1
]
9
1
.
1
4
8
8
.
5
4
9
3
.
7
5
N
o
.
o
f
h
i
d
d
e
n
l
a
y
e
r
s
[
3
,
4
,
5
,
6
,
7
]
9
1
.
1
4
8
8
.
5
4
9
3
.
7
5
Ep
o
c
h
s
[
5
0
,
2
0
0
,
3
0
0
,
4
0
0
,
5
0
0
]
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
B
a
t
c
h
si
z
e
[
1
6
,
32
,
6
4
,
1
2
8
,
2
5
6
]
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
D
r
o
p
o
u
t
[
0
.
2
,
0
.
3
,
0
.
4
,
0
.
5
]
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
A
c
t
i
v
a
t
i
o
n
f
u
n
c
t
i
o
n
[
R
e
L
U
,
T
a
n
h
,
S
i
g
m
o
i
d
]
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
DNN
b
e
st
st
r
u
c
t
u
r
e
[
1
2
4
,
3
,
4
0
0
,
3
2
,
0
.
5
]
9
5
.
5
7
9
5
.
3
1
9
5
.
8
3
LSTM
N
o
.
o
f
h
i
d
d
e
n
n
e
u
r
o
n
s
[
31
,
6
2
,
7
8
,
1
0
9
,
1
2
4
]
8
8
.
0
2
8
4
.
3
7
9
1
.
6
6
N
o
.
o
f
LST
M
l
a
y
e
r
s
[
3
,
4
,
5
,
6
,
7
]
8
8
.
0
2
8
2
.
2
9
9
3
.
7
5
Ep
o
c
h
s
[
50
,
1
0
0
,
2
0
0
,
3
0
0
,
4
0
0
]
8
8
.
0
2
8
2
.
2
9
9
3
.
7
5
B
a
t
c
h
si
z
e
[
1
6
,
32
,
6
4
,
1
2
8
,
2
5
6
]
8
6
.
9
7
7
9
.
1
6
9
4
.
7
9
D
r
o
p
o
u
t
[
0
.
2
,
0
.
3
,
0
.
4
,
0
.
5
]
8
7
.
5
0
7
9
.
1
6
9
5
.
8
3
L
S
T
M
b
e
st
st
r
u
c
t
u
r
e
[
3
1
,
5
,
5
0
,
3
2
,
0
.
4
]
8
7
.
5
0
7
9
.
1
6
9
5
.
8
3
T
h
e
f
in
d
in
g
s
o
f
t
h
is
s
tu
d
y
in
d
icate
th
at
th
e
o
b
tain
ed
r
esu
lts
s
h
o
wed
th
at
th
e
b
est
s
tr
u
ctu
r
e
f
o
r
t
h
e
DNN
alg
o
r
ith
m
was w
h
en
u
s
in
g
1
2
4
h
id
d
en
n
e
u
r
o
n
s
in
th
r
e
e
h
id
d
en
lay
e
r
s
,
with
a
d
r
o
p
o
u
t r
ate
o
f
0
.
5
,
ep
o
c
h
s
(
4
0
0
)
,
b
atch
s
ize
(
3
2
)
,
a
n
d
R
eL
U
ac
tiv
atio
n
f
u
n
ctio
n
.
Als
o
,
th
e
b
est
s
tr
u
ctu
r
e
f
o
r
th
e
L
STM
alg
o
r
ith
m
was
wh
en
u
s
in
g
3
1
h
id
d
en
n
e
u
r
o
n
s
in
f
iv
e
L
STM
h
id
d
en
lay
er
s
,
with
a
d
r
o
p
o
u
t
r
ate
(
0
.
4
)
,
ep
o
ch
s
(
5
0
)
,
a
n
d
b
atch
s
ize
(
3
2
)
.
Ad
d
itio
n
ally
,
t
h
e
ex
p
er
im
en
t
r
esu
lts
s
h
o
wed
t
h
at
th
e
DNN
o
b
tain
e
d
a
n
a
cc
u
r
ac
y
o
f
9
5
.
5
7
%,
s
en
s
itiv
ity
o
f
9
5
.
3
1
%,
an
d
s
p
ec
if
icity
o
f
9
5
.
8
3
%.
On
th
e
o
th
er
h
an
d
,
th
e
L
STM
o
b
tain
ed
an
ac
cu
r
ac
y
o
f
8
7
.
5
0
%,
s
en
s
itiv
ity
o
f
7
9
.
1
6
%,
an
d
s
p
ec
if
icity
o
f
9
5
.
8
3
%.
Af
ter
co
m
p
ar
in
g
th
e
r
esu
lts
f
o
r
th
e
d
ee
p
lear
n
in
g
m
o
d
els
an
d
th
e
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
s
p
ar
ticu
lar
ly
th
e
r
an
d
o
m
f
o
r
ests
f
o
r
p
r
ed
ict
in
g
th
e
d
r
u
g
-
r
elate
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J E
lec
&
C
o
m
p
E
n
g
I
SS
N:
2088
-
8
7
0
8
Dee
p
lea
r
n
in
g
fo
r
p
r
ed
ictin
g
d
r
u
g
-
r
ela
ted
p
r
o
b
lems in
d
ia
b
etes p
a
tien
ts
(
F
a
tima
M.
S
ma
d
i
)
3007
p
r
o
b
lem
s
(
DR
Ps
)
s
tatu
s
ap
p
lie
d
in
[
2
1
]
,
t
h
e
r
a
n
d
o
m
f
o
r
ests
alg
o
r
ith
m
o
u
tp
er
f
o
r
m
e
d
th
e
d
e
ep
lear
n
in
g
m
o
d
els
in
ter
m
s
o
f
ac
cu
r
ac
y
a
n
d
s
en
s
itiv
ity
wh
en
wo
r
k
in
g
with
tab
u
lar
d
ata
in
class
if
icatio
n
task
s
.
I
n
th
e
h
ea
lth
ca
r
e
f
ield
,
ac
cu
r
ac
y
is
ess
en
tial.
C
h
o
o
s
in
g
th
e
ap
p
r
o
p
r
iate
m
o
d
e
l
f
o
r
th
e
d
ata
ca
n
h
av
e
a
m
ajo
r
ef
f
ec
t
o
n
p
atien
t
o
u
tco
m
es.
Ou
r
s
tu
d
y
r
ec
o
m
m
en
d
s
th
at
p
h
ar
m
ac
is
ts
s
h
o
u
ld
u
s
e
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
wh
en
wo
r
k
in
g
o
n
id
en
tify
in
g
t
h
e
DR
Ps
s
ta
tu
s
o
f
d
iab
etic
p
atien
ts
f
o
r
class
if
icatio
n
task
s
in
h
ea
lth
ca
r
e
to
in
c
r
ea
s
e
th
e
q
u
ality
o
f
h
ea
lth
ca
r
e
s
er
v
ices a
n
d
i
d
en
tif
y
th
e
DR
Ps
s
tatu
s
f
o
r
d
iab
etic
p
atien
ts
.
T
h
e
p
r
im
ar
y
o
b
jectiv
e
o
f
th
is
s
tu
d
y
was
to
ap
p
ly
d
ee
p
lear
n
i
n
g
m
o
d
els
to
p
r
ed
ict
th
e
DR
P
s
s
tatu
s
o
f
d
iab
etes
p
atien
ts
.
W
e
in
v
esti
g
ate
th
e
b
est
s
tr
u
ctu
r
e
o
f
th
e
DNN
an
d
L
STM
to
p
r
e
d
ict
th
e
s
tatu
s
o
f
d
r
u
g
-
r
elate
d
p
r
o
b
lem
s
.
Mo
r
e
o
v
er
,
t
o
f
in
d
o
u
t
th
e
ef
f
ec
t
o
f
ap
p
ly
i
n
g
d
ee
p
lear
n
in
g
co
m
p
a
r
ed
with
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
.
Ad
d
itio
n
ally
,
wh
eth
er
to
ap
p
ly
d
ee
p
lear
n
in
g
m
o
d
els
o
r
m
ac
h
in
e
lear
n
in
g
m
et
h
o
d
s
wh
en
d
ea
lin
g
with
tab
u
lar
d
ata
f
o
r
class
if
icatio
n
an
d
to
f
in
d
o
u
t
wh
ic
h
will
g
iv
e
h
i
g
h
p
er
f
o
r
m
an
c
e
f
o
r
ta
b
u
lar
d
ata
co
m
p
ar
ed
with
th
e
s
tu
d
y
in
[
2
1
]
.
C
h
o
o
s
in
g
th
e
r
ig
h
t m
o
d
el
f
o
r
t
h
e
tab
u
lar
d
ata
d
ep
en
d
s
o
n
o
u
r
u
n
d
er
s
tan
d
i
n
g
o
f
th
e
n
atu
r
e
o
f
th
e
d
ata,
wh
eth
er
to
u
s
e
d
ee
p
lear
n
i
n
g
o
r
m
ac
h
in
e
lear
n
in
g
.
As
d
em
o
n
s
tr
ated
in
th
e
s
tu
d
y
[
1
6
]
,
th
e
au
th
o
r
s
co
m
p
a
r
ed
tr
ee
en
s
em
b
le
m
o
d
els s
u
ch
as XG
B
o
o
s
t
with
d
ee
p
lear
n
in
g
m
o
d
els to
d
eter
m
in
e
wh
ich
p
e
r
f
o
r
m
s
b
etter
r
esu
lts
f
o
r
ta
b
u
lar
d
ata.
T
h
eir
r
esu
lts
s
h
o
wed
th
at
th
e
XGBo
o
s
t
o
u
t
p
er
f
o
r
m
s
th
e
d
ee
p
lear
n
i
n
g
m
o
d
els
an
d
r
e
q
u
ir
es
m
u
ch
less
tu
n
in
g
.
T
h
er
ef
o
r
e,
i
n
s
tead
o
f
ap
p
l
y
in
g
d
ee
p
lear
n
i
n
g
m
o
d
els
wh
en
wo
r
k
in
g
with
tab
u
lar
d
ata,
t
h
ey
r
ec
o
m
m
en
d
u
s
in
g
en
s
em
b
le
m
o
d
els.
Ad
d
itio
n
ally
,
o
u
r
r
es
u
lts
s
h
o
w
th
at
m
ac
h
in
e
lea
r
n
i
n
g
,
p
ar
ticu
lar
ly
th
e
r
an
d
o
m
f
o
r
ests
m
eth
o
d
a
p
p
lie
d
in
[
2
1
]
,
p
er
f
o
r
m
e
d
b
etter
th
an
p
r
ev
io
u
s
s
tu
d
ies
with
h
ig
h
ac
cu
r
ac
y
(
9
7
.
3
9
%),
s
p
ec
if
icity
(
9
5
.
8
3
%),
a
n
d
s
en
s
itiv
ity
(
9
8
.
9
5
%)
as
s
h
o
wn
in
T
ab
le
1
3
.
T
ab
le
1
3
,
s
u
m
m
ar
iz
es
th
e
p
er
f
o
r
m
an
ce
co
m
p
ar
is
o
n
o
f
d
ee
p
lear
n
i
n
g
a
n
d
m
ac
h
i
n
e
lear
n
in
g
with
p
r
e
v
io
u
s
s
tu
d
ies.
Acc
o
r
d
in
g
to
Al
-
R
ad
aid
eh
et
a
l.
[
2
1
]
,
th
e
r
an
d
o
m
f
o
r
ests
m
eth
o
d
ac
h
iev
ed
th
e
f
o
llo
w
in
g
r
esu
lts
ac
cu
r
ac
y
o
f
9
7
.
3
9
%,
s
en
s
itiv
ity
o
f
9
8
.
9
5
%,
an
d
s
p
ec
if
icity
o
f
9
5
.
8
3
%
)
.
I
n
t
h
is
s
tu
d
y
,
th
e
ex
p
er
im
e
n
t
r
esu
lts
s
h
o
wed
th
at
th
e
DNN
o
b
tain
ed
an
ac
cu
r
ac
y
o
f
9
5
.
5
7
%,
s
en
s
itiv
ity
o
f
9
5
.
3
1
%,
a
n
d
s
p
e
cif
icity
o
f
9
5
.
8
3
%.
Ad
d
itio
n
ally
,
th
e
L
STM
o
b
tain
ed
an
ac
cu
r
ac
y
o
f
8
7
.
5
0
%,
s
en
s
itiv
ity
o
f
7
9
.
1
6
%,
an
d
s
p
e
cif
icity
o
f
9
5
.
8
3
%.
W
h
en
co
m
p
a
r
in
g
th
e
r
esu
lts
o
f
th
e
DNN
an
d
L
STM
with
th
e
r
an
d
o
m
f
o
r
ests
r
esu
lts
,
we
n
o
ted
th
at
th
e
r
an
d
o
m
f
o
r
ests
alg
o
r
ith
m
ap
p
l
ied
in
[
2
1
]
h
as
ac
h
iev
ed
th
e
s
am
e
s
p
ec
if
icity
m
etr
ic
r
esu
lts
co
m
p
ar
ed
with
th
e
DNN
an
d
L
STM
m
o
d
els
as
s
h
o
wn
in
T
ab
le
1
3
.
Ad
d
itio
n
ally
,
we
n
o
ted
th
at
th
e
r
an
d
o
m
f
o
r
ests
o
u
tp
er
f
o
r
m
ed
th
e
r
esu
lts
o
f
th
e
DNN
i
n
ter
m
s
o
f
ac
c
u
r
ac
y
with
a
n
in
cr
e
ase
o
f
(
1
.
8
2
%)
an
d
s
en
s
itiv
ity
with
an
in
cr
ea
s
e
o
f
(
3
.
6
4
)
.
Als
o
,
it
o
u
tp
er
f
o
r
m
ed
th
e
r
esu
lts
o
f
th
e
L
STM
in
ter
m
s
o
f
ac
cu
r
ac
y
with
an
in
cr
e
ase
o
f
(
9
.
8
9
%)
an
d
s
en
s
itiv
ity
with
an
in
cr
ea
s
e
o
f
(
1
9
.
7
9
%).
As
a
r
esu
lt,
wh
en
co
m
p
ar
in
g
th
e
r
esu
lts
,
it
was
f
o
u
n
d
th
at
u
s
in
g
m
ac
h
in
e
lear
n
i
n
g
alg
o
r
ith
m
s
p
ar
ticu
lar
ly
th
e
r
an
d
o
m
f
o
r
ests
u
s
ed
in
[
2
1
]
to
p
r
e
d
ict
th
e
DR
Ps
s
tatu
s
o
b
tain
ed
th
e
b
est
r
esu
lts
co
m
p
ar
ed
to
th
e
d
ee
p
lear
n
in
g
m
o
d
els
in
ter
m
s
o
f
ac
cu
r
ac
y
an
d
s
en
s
itiv
ity
wh
en
wo
r
k
in
g
with
tab
u
lar
d
ata
in
class
if
icatio
n
task
s
.
C
o
m
p
ar
ed
to
d
ee
p
lear
n
in
g
m
o
d
els,
th
e
m
ac
h
in
e
lear
n
i
n
g
alg
o
r
ith
m
s
ap
p
lied
in
[
2
1
]
ap
p
ea
r
to
b
e
m
o
r
e
ef
f
ec
tiv
e
an
d
o
f
f
er
a
s
ig
n
if
ican
t
im
p
r
o
v
e
m
en
t
o
v
e
r
p
r
ev
io
u
s
d
ee
p
lea
r
n
in
g
m
o
d
els
in
ter
m
s
o
f
ac
cu
r
a
cy
an
d
s
en
s
itiv
ity
a
s
well
as
im
p
r
o
v
ed
o
u
tco
m
e
s
.
I
n
ad
d
itio
n
,
tu
n
in
g
th
e
d
e
ep
lear
n
in
g
h
y
p
er
-
p
ar
am
eter
s
to
ac
h
iev
e
th
e
b
est
s
tr
u
ctu
r
e
is
a
c
o
m
p
lex
p
r
o
ce
s
s
th
at
d
ep
e
n
d
s
o
n
tr
ial
an
d
er
r
o
r
a
n
d
r
eq
u
ir
es
tim
e
an
d
ef
f
o
r
t.
I
t
also
r
eq
u
ir
es
a
l
o
n
g
r
u
n
-
tim
e
to
tr
ain
th
e
m
o
d
el,
u
n
lik
e
m
ac
h
in
e
lear
n
in
g
wh
ich
r
eq
u
ir
es
less
tim
e
an
d
ef
f
o
r
t.
T
h
e
d
ata
u
s
ed
in
th
is
s
tu
d
y
was
co
llected
f
r
o
m
s
ix
m
ajo
r
h
o
s
p
itals
in
J
o
r
d
a
n
.
B
ec
au
s
e
o
f
th
is
r
eg
io
n
al
r
estrictio
n
,
t
h
e
d
ataset
m
ay
b
e
s
k
ewe
d
to
war
d
th
e
p
ar
ticu
lar
r
eg
io
n
s
wh
er
e
th
ese
h
o
s
p
itals
ar
e
lo
ca
ted
an
d
m
i
g
h
t
n
o
t
r
ep
r
esen
t
lar
g
e
p
o
p
u
latio
n
s
,
s
u
ch
as
th
o
s
e
in
t
h
e
US.
Fu
tu
r
e
r
esear
c
h
s
h
o
u
ld
ex
p
a
n
d
th
e
d
ata
to
in
clu
d
e
d
ata
f
r
o
m
o
th
er
r
eg
i
o
n
s
an
d
p
o
p
u
latio
n
s
.
T
h
is
w
ill
im
p
r
o
v
e
th
e
g
en
er
aliza
b
ili
ty
o
f
th
e
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
an
d
o
f
f
er
a
d
ee
p
er
u
n
d
er
s
tan
d
i
n
g
o
f
d
r
u
g
-
r
elate
d
p
r
o
b
lem
s
ac
r
o
s
s
d
if
f
er
en
t
g
eo
g
r
ap
h
ic
g
r
o
u
p
s
.
T
ab
le
1
3
.
A
p
er
f
o
r
m
an
ce
co
m
p
ar
is
o
n
b
etwe
en
t
h
e
ap
p
lied
d
ee
p
lear
n
in
g
m
o
d
els an
d
p
r
e
v
i
o
u
s
s
tu
d
ies
S
t
u
d
i
e
s
M
o
d
e
l
s
A
c
c
u
r
a
c
y
%
S
p
e
c
i
f
i
c
i
t
y
%
S
e
n
s
i
t
i
v
i
t
y
%
[
1
0
]
S
t
a
t
i
st
i
c
a
l
D
B
N
83
73
87
[
1
1
]
DNNs
87
-
-
[
1
2
]
DNNs
96
-
-
[
1
3
]
DNNs
97
96
97
[
1
4
]
LSTM
u
s
i
n
g
A
R
(
LST
M
-
A
R
)
84
-
-
[
1
5
]
RF
94
-
-
[
1
7
]
R
F
w
i
t
h
S
M
O
T
E
-
EN
N
95
92
98
[
1
8
]
DNNs
87
87
88
[
2
1
]
RF
97
95
98
O
u
r
st
u
d
y
(
D
N
N
)
DNNs
95
95
95
O
u
r
st
u
d
y
(
LSTM
)
LSTM
n
e
t
w
o
r
k
s
87
95
79
Evaluation Warning : The document was created with Spire.PDF for Python.