I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
7
,
No
.
4
,
A
u
g
u
s
t
201
7
,
p
p
.
2
2
1
5
~
2
2
2
2
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v7
i
4
.
p
p
2
2
1
5
-
2222
2215
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
JE
C
E
Lea
rning
f
ro
m
a
Cla
ss
I
m
ba
la
nced
P
ublic
Hea
lth
Da
tas
et
:
a
Co
st
-
b
a
sed Co
m
p
a
riso
n
o
f
Cla
ss
ifie
r
Per
for
m
a
nce
Ro
hin
i R
.
Ra
o
1
,
K
rish
na
m
o
o
rt
hi M
a
kk
it
ha
y
a
2
1
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter A
p
p
li
c
a
ti
o
n
s,
M
a
n
i
p
a
l
In
stit
u
te
o
f
T
e
c
h
n
o
lo
g
y
(M
IT
),
M
a
n
ip
a
l,
I
n
d
ia
2
De
p
a
rt
m
e
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
&
En
g
in
e
e
rin
g
,
M
a
n
i
p
a
l
I
n
stit
u
t
e
o
f
T
e
c
h
n
o
lo
g
y
(M
IT
),
M
a
n
ip
a
l,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
1
4
,
2
0
1
7
R
ev
i
s
ed
Ma
y
3
1
,
2017
A
cc
ep
ted
A
u
g
2
8
,
2017
P
u
b
l
ic
h
e
a
lt
h
c
a
re
sy
ste
m
s
ro
u
t
in
e
ly
c
o
ll
e
c
t
h
e
a
lt
h
-
re
late
d
d
a
ta
f
ro
m
th
e
p
o
p
u
lati
o
n
.
T
h
is
d
a
ta
c
a
n
b
e
a
n
a
ly
z
e
d
u
sin
g
d
a
ta
m
in
in
g
tec
h
n
iq
u
e
s
to
f
in
d
n
o
v
e
l,
in
tere
stin
g
p
a
tt
e
rn
s,
w
h
ich
c
o
u
ld
h
e
lp
f
o
rm
u
late
e
ffe
c
ti
v
e
p
u
b
l
ic h
e
a
lt
h
p
o
li
c
ies
a
n
d
in
terv
e
n
ti
o
n
s
.
T
h
e
o
c
c
u
rre
n
c
e
o
f
c
h
ro
n
ic
i
ll
n
e
ss
is
ra
re
in
th
e
p
o
p
u
lati
o
n
a
n
d
th
e
e
ff
e
c
t
o
f
th
is
c
las
s
i
m
b
a
lan
c
e
,
o
n
th
e
p
e
rf
o
rm
a
n
c
e
o
f
v
a
rio
u
s
c
las
si
f
iers
w
a
s
stu
d
ied
.
T
h
e
o
b
j
ec
tiv
e
o
f
th
is
w
o
r
k
is
t
o
id
en
tif
y
th
e
b
est
class
i
f
ier
s
f
o
r
class
im
b
ala
n
ce
d
h
ea
lt
h
d
atasets
th
r
o
u
g
h
a
co
s
t
-
b
ased
co
m
p
ar
i
s
o
n
o
f
cla
s
s
i
f
ier
p
er
f
o
r
m
an
ce
.
T
h
e
p
o
p
u
lar,
o
p
e
n
-
so
u
rc
e
d
a
ta
m
in
in
g
to
o
l
W
EKA
,
w
a
s
u
se
d
to
b
u
il
d
a
v
a
riety
o
f
c
o
re
c
las
si
f
iers
a
s
w
e
ll
a
s
c
las
si
f
ie
r
e
n
se
m
b
les
,
to
e
v
a
lu
a
te
th
e
c
las
si
f
iers
‟
pe
rf
o
r
m
a
n
c
e
.
T
h
e
u
n
e
q
u
a
l
m
isc
la
ss
if
ic
a
ti
o
n
c
o
sts
w
e
re
re
p
re
se
n
ted
in
a
c
o
st
m
a
tri
x
,
a
n
d
c
o
st
-
b
e
n
e
f
it
a
n
a
l
y
sis
w
a
s
a
lso
p
e
rf
o
r
m
e
d
.
In
a
n
o
t
h
e
r
e
x
p
e
rime
n
t,
v
a
rio
u
s
sa
m
p
li
n
g
m
e
th
o
d
s
su
c
h
a
s
u
n
d
e
r
-
sa
m
p
li
n
g
,
o
v
e
r
-
s
a
m
p
li
n
g
,
a
n
d
S
M
OT
E
w
a
s
p
e
r
f
o
r
m
e
d
to
b
a
la
n
c
e
th
e
c
las
s
d
istri
b
u
ti
o
n
in
th
e
d
a
tas
e
t
,
a
n
d
th
e
c
o
sts
w
e
re
c
o
m
p
a
re
d
.
T
h
e
Ba
y
e
sia
n
c
la
ss
i
f
iers
p
e
r
f
o
r
m
e
d
w
e
ll
w
it
h
a
h
ig
h
re
c
a
ll
,
l
o
w
n
u
m
b
e
r
o
f
f
a
lse
n
e
g
a
ti
v
e
s
a
n
d
w
e
re
n
o
t
a
ff
e
c
ted
b
y
th
e
c
las
s
im
b
a
lan
c
e
.
Re
su
lt
s
c
o
n
f
ir
m
th
a
t
to
tal
c
o
st
o
f
Ba
y
e
sia
n
c
las
si
f
i
e
rs
c
a
n
b
e
f
u
rth
e
r
re
d
u
c
e
d
u
sin
g
c
o
st
-
se
n
siti
v
e
lea
rn
in
g
m
e
th
o
d
s.
C
las
sif
iers
b
u
il
t
u
sin
g
th
e
ra
n
d
o
m
u
n
d
e
r
-
sa
m
p
led
d
a
tas
e
t
s
h
o
w
e
d
a
d
ra
m
a
ti
c
d
ro
p
in
c
o
st
s
a
n
d
h
ig
h
c
las
si
f
ica
ti
o
n
a
c
c
u
ra
c
y
.
K
ey
w
o
r
d
:
C
las
s
i
m
b
ala
n
ce
C
las
s
i
f
i
er
ac
cu
r
ac
y
C
o
s
t b
en
e
f
it a
n
al
y
s
is
Data
m
i
n
i
n
g
Hea
lth
ca
r
e
Co
p
y
rig
h
t
©
2017
In
s
t
it
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
R
o
h
i
n
i R.
R
ao
,
Dep
ar
t
m
en
t
o
f
C
o
m
p
u
ter
A
p
p
licatio
n
s
,
Ma
n
ip
al
I
n
s
tit
u
te
o
f
T
ec
h
n
o
lo
g
y
,
Ma
n
ip
al
--
5
7
6
1
0
4
,
I
n
d
ia
.
E
m
ail:
r
o
h
i
n
i.r
ao
@
m
a
n
ip
al.
ed
u
1.
I
NT
RO
D
UCT
I
O
N
A
ll
o
v
er
t
h
e
w
o
r
ld
,
p
u
b
lic
h
ea
lth
o
r
g
a
n
izatio
n
s
ar
e
c
u
r
r
en
tl
y
f
ac
i
n
g
th
e
c
h
alle
n
g
e
o
f
tack
li
n
g
c
h
r
o
n
i
c
d
is
ea
s
es.
P
u
b
lic
h
ea
lt
h
ag
e
n
ci
es
n
ee
d
to
r
esp
o
n
d
w
it
h
co
s
t
-
e
f
f
ec
tiv
e,
e
v
id
en
ce
-
b
ased
i
n
ter
v
en
t
io
n
s
to
p
r
o
m
o
te
h
ea
lt
h
y
l
iv
i
n
g
an
d
p
r
ev
e
n
t
ch
r
o
n
ic
d
is
ea
s
es
.
P
u
b
lic
Hea
lth
Or
g
an
izatio
n
s
r
o
u
ti
n
el
y
co
llect
d
ata
o
n
d
em
o
g
r
ap
h
ics,
s
o
cio
-
ec
o
n
o
m
i
c
s
tatu
s
,
li
v
in
g
co
n
d
itio
n
s
,
a
n
d
u
n
d
er
l
y
i
n
g
h
ea
l
th
co
n
d
itio
n
s
in
t
h
e
p
o
p
u
latio
n
.
Data
m
i
n
i
n
g
tec
h
n
iq
u
es
ca
n
b
e
ap
p
lied
to
th
is
p
o
p
u
latio
n
-
b
a
s
ed
d
ata,
to
g
ai
n
n
e
w
i
n
s
i
g
h
ts
i
n
to
t
h
e
u
n
d
er
l
y
in
g
h
ea
lt
h
p
r
o
b
le
m
s
.
I
n
t
h
e
m
ed
i
ca
l
d
iag
n
o
s
is
d
o
m
ai
n
,
clas
s
i
f
i
er
s
h
a
v
e
b
ee
n
b
u
il
t
to
p
r
ed
ict
d
is
ea
s
e
s
s
u
c
h
as
b
r
ea
s
t
ca
n
ce
r
,
i
n
s
o
m
n
ia,
t
h
y
r
o
id
d
is
ea
s
e,
P
ar
k
in
s
o
n
‟
s
d
i
s
ea
s
e
a
n
d
e
v
en
r
ec
o
m
m
e
n
d
m
ed
icatio
n
[
1
-
6
]
.
P
o
llettin
i
et
al.
[
7
]
p
r
o
p
o
s
e
a
class
i
f
ier
w
h
ich
a
u
to
m
atica
ll
y
class
i
f
ies
p
atie
n
t
s
i
n
to
s
u
r
v
ei
llan
ce
le
v
els
b
ased
o
n
ass
o
ciatio
n
s
a
m
o
n
g
p
atien
t
f
ea
tu
r
e
s
an
d
h
ea
lt
h
.
C
las
s
i
f
ie
r
s
h
av
e
a
ls
o
b
ee
n
d
esi
g
n
ed
to
p
r
ed
ict
th
e
co
s
t
o
f
h
ea
lt
h
ca
r
e
s
er
v
ice
s
,
to
p
r
ed
ict
in
te
n
s
i
v
e
ca
r
e
u
n
it
r
ea
d
m
i
s
s
io
n
,
m
o
r
talit
y
r
ate
an
d
li
f
e
e
x
p
ec
tan
c
y
r
ate
[
1
]
,
[
8
]
.
Sen
s
o
r
b
ased
,
u
n
o
b
tr
u
s
i
v
e,
co
n
ti
n
u
o
u
s
h
o
m
e
m
o
n
ito
r
in
g
s
y
s
te
m
s
h
a
v
e
b
ee
n
d
ep
lo
y
ed
,
an
d
h
u
m
a
n
ac
tiv
it
y
i
s
b
ein
g
ass
e
s
s
ed
u
s
in
g
cla
s
s
i
f
ie
r
s
[
3
]
.
I
n
t
h
e
p
u
b
lic
h
ea
l
th
d
o
m
a
in
,
clas
s
if
icatio
n
tech
n
iq
u
es
ca
n
b
e
u
s
ed
to
an
al
y
ze
t
h
e
ef
f
ec
t
o
f
v
ar
io
u
s
s
o
cial
an
d
en
v
ir
o
n
m
e
n
tal
f
ac
to
r
s
,
s
u
ch
a
s
w
o
r
k
en
v
ir
o
n
m
en
t,
liv
in
g
co
n
d
itio
n
s
,
ed
u
ca
tio
n
o
n
t
h
e
h
ea
l
th
o
f
t
h
e
p
o
p
u
latio
n
.
T
h
e
r
elatio
n
s
h
ip
b
et
w
ee
n
p
atien
t
f
ea
t
u
r
es
an
d
d
is
ea
s
es
co
u
ld
h
elp
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
4
,
A
u
g
u
s
t
2
0
1
7
:
2
2
1
5
–
2
2
2
2
2216
f
o
r
m
u
late
ef
f
ec
ti
v
e
p
u
b
lic
h
e
al
th
in
ter
v
e
n
tio
n
s
[
8
]
.
Fo
r
in
s
tan
ce
,
T
ak
ed
a
et
al.
[
9
]
u
tili
z
ed
m
u
ltip
le
lo
g
i
s
tic
an
al
y
s
es
to
r
ep
o
r
t
th
e
s
ig
n
i
f
ic
an
t
ass
o
ciatio
n
s
b
et
w
ee
n
m
en
tal
h
ea
lth
a
n
d
p
s
y
ch
o
s
o
cial
s
t
r
ess
o
r
s
lik
e
f
a
m
il
y
r
elatio
n
s
h
ip
,
p
r
eg
n
a
n
c
y
,
a
n
d
in
co
m
e.
Ho
w
e
v
er
,
th
ese
p
u
b
lic
h
ea
lt
h
d
atasets
o
f
te
n
s
u
f
f
er
f
r
o
m
“r
ar
e
ca
s
es”
o
r
“
r
ar
e
class
es”
p
r
o
b
lem
,
w
h
ic
h
r
esu
l
t
in
i
m
b
ala
n
c
ed
class
es
i
n
t
h
e
tr
ain
i
n
g
d
atasets
[
1
0
]
.
Fo
r
ex
am
p
le,
m
o
s
t
h
ea
th
d
atasets
u
s
u
all
y
h
av
e
v
er
y
f
e
w
ca
s
e
s
o
f
th
e
ta
r
g
et
d
is
ea
s
e,
w
h
e
n
co
m
p
ar
ed
to
th
e
n
u
m
b
er
o
f
h
ea
l
th
y
p
ati
en
ts
i
n
th
e
d
ataset
[
1
1
-
1
4
]
.
I
n
th
e
b
in
ar
y
cla
s
s
i
f
icatio
n
f
o
r
m
ed
ical
d
iag
n
o
s
i
s
,
th
e
r
ar
e
m
i
n
o
r
it
y
cla
s
s
r
e
f
e
r
s
to
th
e
p
o
s
iti
v
e
in
s
ta
n
ce
s
o
r
th
e
tar
g
et
clas
s
,
wh
er
ea
s
t
h
e
m
aj
o
r
it
y
clas
s
i
s
r
e
p
r
esen
ted
b
y
t
h
e
n
e
g
ati
v
e
i
n
s
t
an
ce
s
in
th
e
d
ataset.
C
las
s
i
m
b
ala
n
ce
ca
n
al
s
o
o
cc
u
r
w
h
e
n
th
e
d
ata
co
llectio
n
p
r
o
ce
s
s
is
li
m
ited
,
r
esu
ltin
g
i
n
a
r
tif
icial
i
m
b
ala
n
ce
s
.
T
o
b
e
class
if
ied
as a
clas
s
-
i
m
b
alan
ce
d
d
ataset,
r
ar
it
y
s
h
o
u
ld
b
e
b
etw
ee
n
0
.
1
to
1
0
%.
T
h
is
p
ap
er
co
n
s
id
er
s
a
h
ea
lth
b
ased
d
ataset
w
h
ic
h
r
ec
o
r
d
s
th
e
p
r
e
s
en
ce
o
f
c
h
r
o
n
ic
d
is
ea
s
es
i.e
.
d
iab
etes,
h
ea
r
t
d
is
ea
s
e
an
d
h
y
p
er
te
n
s
io
n
in
t
h
e
p
o
p
u
latio
n
.
P
atien
t
d
em
o
g
r
ap
h
ics,
li
v
in
g
co
n
d
itio
n
s
,
s
o
cio
-
ec
o
n
o
m
ic
s
tatu
s
ar
e
also
r
ec
o
r
d
ed
in
th
e
d
ataset.
T
h
e
au
th
o
r
s
atte
m
p
t
to
b
u
ild
cla
s
s
i
f
ier
s
to
p
r
ed
ict
th
e
o
cc
u
r
r
en
ce
o
f
an
y
o
f
th
e
t
h
r
ee
ch
r
o
n
ic
d
is
ea
s
es
i
n
th
e
p
o
p
u
l
atio
n
.
L
iter
at
u
r
e
r
ev
ie
w
i
n
d
icate
s
th
at
t
h
er
e
is
n
o
s
in
g
le
clas
s
i
f
ier
m
et
h
o
d
w
h
ic
h
y
ield
s
t
h
e
b
est
r
esu
lt
f
o
r
all
t
y
p
es
o
f
class
i
m
b
ala
n
ce
d
tr
ain
i
n
g
d
ataset
s
[
1
]
.
T
h
e
am
o
u
n
t
o
f
t
h
e
cla
s
s
-
i
m
b
a
lan
ce
b
ias
d
ep
en
d
s
o
n
f
ac
to
r
s
s
u
c
h
as
t
h
e
cla
s
s
i
f
icat
io
n
m
et
h
o
d
,
th
e
n
u
m
b
er
o
f
attr
ib
u
tes
i
n
t
h
e
d
atase
t
an
d
t
h
e
s
a
m
p
le
s
ize
[
1
5
]
.
T
h
e
m
o
ti
v
atio
n
f
o
r
th
i
s
w
o
r
k
i
s
to
s
t
u
d
y
th
e
e
f
f
ec
t
o
f
c
las
s
i
m
b
alan
ce
o
n
t
h
e
v
ar
io
u
s
clas
s
if
ier
s
f
o
r
th
e
h
ea
lt
h
d
ataset.
T
h
e
o
b
j
ec
tiv
e
o
f
t
h
i
s
w
o
r
k
i
s
to
id
en
ti
f
y
th
e
b
es
t
class
i
f
ier
s
f
o
r
th
e
class
i
m
b
al
an
ce
d
h
ea
lth
d
ataset
f
r
o
m
a
co
s
t
-
b
ased
co
m
p
ar
is
o
n
o
f
clas
s
if
ier
p
er
f
o
r
m
a
n
ce
.
T
h
is
w
o
r
k
is
r
ele
v
an
t
to
p
u
b
lic
h
ea
lth
p
o
lic
y
m
ak
er
s
,
w
h
o
ca
n
u
s
e
t
h
e
clas
s
i
f
ier
s
to
au
g
m
e
n
t
m
ed
ica
l
p
r
o
g
n
o
s
is
i
n
t
h
e
p
o
p
u
latio
n
a
n
d
also
id
en
ti
f
y
t
h
e
u
n
d
er
l
y
i
n
g
p
atien
t
f
ea
t
u
r
es
t
h
at
ar
e
co
r
r
elate
d
w
it
h
c
h
r
o
n
ic
d
is
ea
s
es.
1
.
1
.
Cla
s
s
I
m
ba
la
nce
P
ro
ble
m
.
T
h
is
class
i
m
b
ala
n
ce
p
r
o
b
lem
is
a
ch
allen
g
e
to
class
i
f
ier
tec
h
n
iq
u
es
b
ec
au
s
e
a
n
o
r
m
al
clas
s
if
ier
ai
m
s
to
i
m
p
r
o
v
e
o
v
er
a
ll c
lass
if
ier
a
cc
u
r
ac
y
.
C
o
n
s
id
er
a
t
w
o
-
cla
s
s
b
in
ar
y
c
lass
if
ier
f
o
r
a
h
ea
lth
d
ataset,
i
n
w
h
ic
h
th
e
o
u
tco
m
es
ar
e
lab
eled
as
p
o
s
it
iv
e
(
P
)
o
r
n
eg
at
iv
e
(
N)
.
T
h
e
class
i
f
ier
ac
c
u
r
ac
y
ca
n
b
e
co
m
p
u
ted
b
y
ap
p
l
y
i
n
g
th
e
class
i
f
ier
to
a
test
d
ataset
an
d
co
m
p
ar
in
g
th
e
c
las
s
i
f
i
er
r
esu
lt
w
it
h
ac
tu
a
l
class
lab
els.
T
h
er
e
a
r
e
f
o
u
r
p
o
s
s
ib
le
o
u
tco
m
es,
i
f
t
h
e
p
r
ed
icted
v
al
u
e
is
P
a
n
d
t
h
e
ac
t
u
al
v
al
u
e
is
a
ls
o
P
,
t
h
en
it
is
a
tr
u
e
p
o
s
iti
v
e
(
T
P
)
.
I
f
th
e
ac
t
u
al
v
al
u
e
is
N
a
n
d
it
is
p
r
ed
icted
as P
,
th
en
it r
es
u
lt
s
i
n
a
f
alse p
o
s
i
tiv
e
(
FP
)
.
C
o
n
v
er
s
el
y
,
w
h
e
n
b
o
th
t
h
e
p
r
ed
ictio
n
o
u
tco
m
e
an
d
th
e
a
ctu
al
v
al
u
e
ar
e
N,
it
in
d
icate
s
a
tr
u
e
n
eg
at
iv
e
(
T
N)
.
F
alse
n
eg
ati
v
e
(
FN
)
o
cc
u
r
s
w
h
e
n
t
h
e
p
r
ed
ictio
n
o
u
tco
m
e
is
N
w
h
ile
th
e
ac
t
u
al
v
alu
e
is
P
.
T
h
e
class
i
m
b
ala
n
ce
p
r
o
b
lem
a
f
f
ec
ts
d
i
f
f
er
en
t
class
i
f
ier
s
in
a
v
ar
iet
y
o
f
wa
y
s
,
f
o
r
i
n
s
ta
n
ce
i
n
d
ec
is
io
n
tr
ee
in
d
u
ctio
n
al
g
o
r
ith
m
s
,
it
r
esu
l
ts
in
s
m
aller
d
is
j
u
n
cts
[
1
1
]
.
T
h
e
class
i
f
ier
s
tr
ai
n
ed
o
n
c
lass
-
i
m
b
ala
n
ce
d
d
ata
ar
e
u
s
u
al
l
y
b
i
ased
to
w
ar
d
s
th
e
m
aj
o
r
it
y
n
eg
at
iv
e
clas
s
,
an
d
th
e
ac
cu
r
a
c
y
o
f
p
r
ed
ictio
n
s
f
o
r
th
e
m
in
o
r
it
y
tar
g
et
clas
s
is
v
er
y
p
o
o
r
.
T
h
e
class
-
i
m
b
alan
ce
p
h
en
o
m
e
n
o
n
o
f
ten
p
r
o
d
u
ce
s
class
i
f
ier
s
t
h
at
h
a
v
e
a
p
o
o
r
p
r
e
d
ictiv
e
r
ec
all,
p
ar
ticu
lar
l
y
w
h
e
n
th
e
p
o
s
itiv
e
lab
el
is
th
e
m
i
n
o
r
it
y
tar
g
et
cla
s
s
[
4
]
.
T
h
e
p
r
o
b
lem
o
f
i
m
b
ala
n
ce
d
d
ata
is
also
as
s
o
ciate
d
w
it
h
a
s
y
m
m
etr
ic
co
s
t
s
o
f
m
is
c
lass
if
y
i
n
g
ele
m
en
t
s
o
f
d
if
f
er
en
t
c
lass
e
s
.
Fo
r
e
x
a
m
p
le
,
co
n
s
id
er
a
b
in
ar
y
cla
s
s
i
f
ier
b
u
ilt
f
o
r
m
ed
ica
l
d
iag
n
o
s
i
s
,
t
h
e
co
s
t
o
f
m
is
d
ia
g
n
o
s
i
n
g
a
h
ea
lt
h
p
atie
n
t
a
s
h
av
in
g
a
h
ea
lt
h
co
n
d
itio
n
(
f
al
s
e
p
o
s
itiv
e)
is
le
s
s
t
h
an
th
e
co
s
t
o
f
f
al
s
el
y
d
ia
g
n
o
s
i
n
g
a
s
ick
p
atien
t
as
a
h
ea
l
th
y
p
er
s
o
n
(
f
alse
n
eg
at
iv
e)
.
T
h
e
FP
ca
s
e
co
u
ld
lead
to
m
o
r
e
d
iag
n
o
s
tic
test
s
u
n
til
th
e
p
atien
t
is
d
iag
n
o
s
ed
as
h
ea
l
th
y
.
T
h
e
co
s
t
o
f
FN
er
r
o
r
co
u
ld
r
esu
lt
i
n
d
ela
y
ed
d
iag
n
o
s
i
s
,
an
d
u
lti
m
ate
l
y
,
t
h
e
lo
s
s
o
f
lif
e.
T
h
er
ef
o
r
e,
i
n
m
ed
ical
d
iag
n
o
s
is
b
ased
cla
s
s
i
f
ier
s
,
th
e
co
s
t
o
f
FN
i
s
m
o
r
e
s
i
g
n
i
f
ica
n
t
t
h
at
t
h
e
FP
co
s
t.
W
h
ile
t
h
e
FP
co
s
t
ca
n
b
e
ca
lcu
lated
as
t
h
e
ex
p
e
n
s
e
s
in
c
u
r
r
ed
in
f
u
r
t
h
er
test
i
n
g
,
t
h
e
co
s
t o
f
FN
‟
s
h
ar
d
t
o
q
u
an
ti
f
y
.
1
.
2
.
L
ea
rning
f
ro
m
cla
s
s
I
m
ba
la
nced
D
a
t
a
s
et
s
T
h
er
e
ar
e
t
w
o
b
r
o
ad
ap
p
r
o
ac
h
es
to
f
i
n
d
in
g
e
f
f
ec
ti
v
e
clas
s
i
f
i
er
s
in
t
h
e
clas
s
-
i
m
b
ala
n
ce
d
d
a
tasets
:
th
e
alg
o
r
ith
m
s
p
ec
i
f
ic
ap
p
r
o
ac
h
an
d
th
e
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
.
I
n
th
e
alg
o
r
ith
m
s
p
ec
if
ic
ap
p
r
o
ac
h
,
th
e
class
i
f
ier
m
et
h
o
d
s
th
at
ar
e
k
n
o
w
n
to
w
o
r
k
e
f
f
ec
tiv
e
l
y
in
t
h
e
class
-
i
m
b
alan
ce
d
d
ataset
s
c
an
b
e
u
s
ed
w
it
h
n
o
m
o
d
i
f
icat
io
n
o
f
d
atasets
.
Fo
r
ex
a
m
p
le,
W
ei
s
s
[
12]
ad
v
o
ca
tes
th
e
u
s
e
o
f
in
s
ta
n
ce
-
b
ased
lea
r
n
in
g
m
et
h
o
d
s
li
k
e
k
-
Nea
r
est
N
eig
h
b
o
r
s
o
r
Su
p
p
o
r
t
Vec
to
r
Ma
ch
in
es,
to
p
r
ed
i
ct
th
e
m
i
n
o
r
it
y
clas
s
ef
f
ec
tiv
e
l
y
.
T
h
ey
f
o
u
n
d
th
a
t
in
d
ep
en
d
en
t
o
f
th
e
tr
ain
i
n
g
s
iz
e
,
lin
ea
r
l
y
s
ep
ar
ab
le
d
o
m
ai
n
s
ar
e
n
o
t
s
en
s
iti
v
e
to
i
m
b
ala
n
ce
.
He
also
co
n
clu
d
e
s
th
at
n
o
n
-
g
r
ee
d
y
s
ea
r
c
h
tec
h
n
i
q
u
es
u
s
ed
i
n
t
h
e
Ge
n
etic
alg
o
r
ith
m
m
ak
e
it
m
o
r
e
s
u
itab
le
f
o
r
d
ea
lin
g
w
i
th
th
e
class
i
m
b
ala
n
ce
.
I
n
t
h
e
d
ec
is
i
o
n
tr
ee
alg
o
r
it
h
m
,
h
e
s
u
g
g
es
t
s
th
at
s
p
litt
i
n
g
r
u
le
s
ca
n
b
e
m
o
d
if
ied
to
en
s
u
r
e
t
h
at
b
o
th
class
e
s
ar
e
a
d
d
r
ess
ed
.
J
ap
k
o
w
icz
[
13]
co
n
clu
d
ed
th
a
t
th
e
M
u
lti
L
a
y
er
P
er
ce
p
tr
o
n
b
ased
class
i
f
ier
s
ar
e
n
o
t
s
e
n
s
iti
v
e
to
cla
s
s
i
m
b
ala
n
ce
.
Ker
n
el
-
b
ased
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
i
n
e
clas
s
i
f
ier
s
,
cl
u
s
ter
in
g
a
n
d
u
tili
zi
n
g
d
en
s
itie
s
to
e
s
ti
m
ate
tar
g
et
c
l
ass
m
e
m
b
er
s
h
ip
ar
e
k
n
o
w
n
to
w
o
r
k
w
ell
i
n
t
h
e
cla
s
s
-
i
m
b
a
l
an
ce
d
d
atasets
[
16]
.
Si
m
u
latio
n
s
t
u
d
ies
s
h
o
w
t
h
a
t
th
e
class
-
i
m
b
alan
ce
a
n
d
h
ig
h
d
i
m
en
s
io
n
al
it
y
i
m
p
ac
t
th
e
p
er
f
o
r
m
an
ce
o
f
class
i
f
ier
s
li
k
e
a
k
-
n
ea
r
est
n
ei
g
h
b
o
r
,
d
iag
o
n
al
lin
ea
r
d
is
cr
i
m
in
an
t
an
al
y
s
i
s
,
r
an
d
o
m
f
o
r
ests
an
d
s
u
p
p
o
r
t
v
ec
to
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Lea
r
n
in
g
fr
o
m
a
C
la
s
s
I
mb
a
la
n
ce
d
P
u
b
lic
Hea
lth
Da
ta
s
et:
a
C
o
s
t
-
b
a
s
ed
C
o
mp
a
r
is
o
n
o
f …
.
(
R
o
h
in
i R
.
R
a
o
)
2217
m
ac
h
in
e
s
w
i
th
li
n
ea
r
k
er
n
el
b
asis
[
1
5
]
.
C
o
s
t
-
s
e
n
s
iti
v
e
lear
n
i
n
g
m
eth
o
d
s
m
o
d
i
f
y
ex
i
s
ti
n
g
a
lg
o
r
ith
m
s
u
s
i
n
g
t
h
e
co
s
t
in
f
o
r
m
atio
n
.
Fo
r
in
s
ta
n
ce
,
in
a
tr
ee
b
ased
clas
s
i
f
ier
,
th
e
co
s
t
ca
n
b
e
u
s
ed
to
ch
o
o
s
e
t
h
e
s
p
litt
i
n
g
a
ttrib
u
te
o
r
m
a
n
ip
u
la
te
th
e
w
ei
g
h
t o
f
tr
ain
i
n
g
r
ec
o
r
d
s
[
17]
.
T
h
er
e
ar
e
also
s
p
ec
if
ic
alg
o
r
i
th
m
s
s
u
c
h
as
T
w
o
p
h
a
s
e
r
u
l
e
in
d
u
ctio
n
,
C
R
E
DOS
o
r
t
h
e
o
n
e
-
cla
s
s
class
i
f
ier
s
w
h
ic
h
ar
e
p
r
o
v
ed
to
b
e
u
s
ef
u
l
in
cla
s
s
if
y
i
n
g
r
ar
e
ca
s
es
i
n
tr
ain
i
n
g
d
atasets
[
1
1
]
,
[
1
6
]
,
[
1
8
-
2
0
]
.
T
h
e
m
ai
n
id
ea
o
f
t
h
ese
cla
s
s
i
f
icati
o
n
m
et
h
o
d
s
is
t
h
at
t
h
e
al
g
o
r
it
h
m
s
h
o
u
ld
co
n
ce
n
tr
ate
o
n
t
h
e
in
s
ta
n
ce
s
th
at
ar
e
d
if
f
ic
u
lt
to
lear
n
.
[
1
1
]
,
[
1
2
]
.
He
m
p
s
tal
k
et
al.
[
1
6
]
u
s
e
a
co
m
b
in
ed
d
en
s
it
y
esti
m
atio
n
w
it
h
clas
s
p
r
o
b
ab
ilit
y
esti
m
atio
n
f
o
r
t
h
e
p
u
r
p
o
s
e
o
f
o
n
e
-
cla
s
s
c
lass
if
ica
tio
n
.
T
h
e
P
Nr
u
le
al
g
o
r
ith
m
u
s
es
a
t
w
o
-
p
h
ase
r
u
le
i
n
d
u
c
tio
n
m
et
h
o
d
.
W
h
ile
t
h
e
f
ir
s
t
p
h
ase
f
o
cu
s
es
o
n
r
ec
all,
i
n
t
h
e
s
ec
o
n
d
p
h
ase
p
r
ec
is
io
n
i
s
o
p
ti
m
iz
ed
[
1
8
]
.
C
R
E
DOS
ef
f
ec
tiv
e
l
y
u
til
izes
“
r
ip
p
le
d
o
w
n
r
u
le
s
”
to
lear
n
co
m
p
ar
ab
le
o
r
b
etter
m
o
d
els
f
o
r
a
v
ar
iet
y
o
f
r
ar
e
class
es
[
2
0
]
.
I
n
m
o
s
t
ca
s
es,
th
e
co
s
t
o
f
m
i
s
class
if
ica
tio
n
er
r
o
r
s
ar
e
n
o
t
eq
u
al
,
an
d
a
co
s
t
-
s
e
n
s
iti
v
e
le
ar
n
in
g
ap
p
r
o
ac
h
i
s
r
eq
u
ir
ed
[
1
9
]
.
I
n
c
o
s
t
s
en
s
iti
v
e
lear
n
in
g
m
et
h
o
d
s
s
u
c
h
as
Ad
aCo
s
t
an
d
Me
taC
o
s
t,
th
e
co
s
t
is
r
ep
r
esen
ted
in
a
m
atr
i
x
,
an
d
it
is
u
t
ilized
to
g
en
er
ate
a
m
o
d
el
w
i
th
lo
w
er
co
s
t.
E
m
p
ir
icall
y
,
it
i
s
o
f
te
n
r
ep
o
r
ted
th
at
co
s
t
-
s
en
s
iti
v
e
lear
n
i
n
g
o
u
tp
er
f
o
r
m
s
r
an
d
o
m
r
e
-
s
a
m
p
li
n
g
[
1
1
]
.
T
h
e
l
iter
atu
r
e
r
e
v
ie
w
ed
al
s
o
i
n
d
icate
s
th
at
u
s
i
n
g
m
u
ltip
le
c
lass
i
f
ier
s
i
n
e
n
s
e
m
b
les
an
d
a
g
g
r
e
g
ati
n
g
th
e
p
r
ed
ict
io
n
s
o
f
m
u
ltip
le
cla
s
s
i
f
ier
s
,
te
n
d
to
b
e
m
o
r
e
ac
cu
r
ate
th
a
n
t
h
e
co
r
e
class
i
f
ier
s
[
1
9
]
.
W
an
g
et
al.
[
4
]
d
is
cu
s
s
an
i
m
p
le
m
e
n
tatio
n
o
f
an
en
s
e
m
b
le
o
f
lear
n
i
n
g
alg
o
r
ith
m
s
to
r
ec
o
m
m
e
n
d
m
ed
icatio
n
to
d
iab
etic
p
atien
ts
.
E
n
s
e
m
b
le
m
et
h
o
d
s
in
cl
u
d
e
b
ag
g
in
g
,
b
o
o
s
ti
n
g
an
d
r
an
d
o
m
f
o
r
ests
[
17]
.
B
ag
g
i
n
g
u
s
ed
a
m
aj
o
r
it
y
v
o
te
to
m
a
k
e
m
o
r
e
ac
cu
r
ate
class
i
f
icatio
n
s
u
s
in
g
m
u
lt
ip
le
class
i
f
ier
s
.
B
o
o
s
tin
g
,
o
n
th
e
o
th
er
h
a
n
d
,
u
s
es
a
n
ad
ap
tiv
e
s
a
m
p
lin
g
o
f
i
n
s
ta
n
c
es,
b
ased
o
n
t
h
e
w
ei
g
h
ts
o
f
t
h
e
i
n
s
ta
n
ce
s
to
i
m
p
r
o
v
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
class
i
f
ier
s
.
B
o
o
s
tin
g
m
et
h
o
d
s
lik
e
SMOT
E
B
o
o
s
t
an
d
A
d
aBo
o
s
t
h
av
e
b
ee
n
f
o
u
n
d
to
b
e
e
f
f
ec
tiv
e
i
n
th
e
r
ar
e
ca
s
e
s
ce
n
ar
io
[
1
9
]
.
B
len
d
in
g
is
an
en
s
e
m
b
le
m
et
h
o
d
w
h
er
e
m
u
ltip
le
alg
o
r
it
h
m
s
ar
e
p
r
ep
a
r
ed
o
n
th
e
tr
ain
in
g
d
ata.
Me
ta
class
i
f
ier
s
co
m
b
in
e
th
e
p
r
ed
ic
tio
n
s
o
f
m
u
lt
ip
le
class
i
f
ier
s
to
m
a
k
e
ac
cu
r
ate
p
r
ed
ictio
n
s
o
n
u
n
s
ee
n
d
ata.
T
h
e
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
to
th
e
clas
s
i
m
b
alan
ce
p
r
o
b
lem
w
o
u
ld
b
e
to
m
o
d
i
f
y
t
h
e
tr
ai
n
in
g
d
ataset
its
el
f
u
s
in
g
v
ar
io
u
s
s
a
m
p
lin
g
tec
h
n
iq
u
e
s
[
4
]
,
[
1
1
-
1
3
]
,
[
2
1
]
.
B
asic
s
a
m
p
li
n
g
m
et
h
o
d
s
in
cl
u
d
e
u
n
d
er
-
s
a
m
p
li
n
g
to
r
ed
u
ce
t
h
e
m
aj
o
r
it
y
c
lass
i
n
s
ta
n
ce
s
o
r
o
v
er
-
s
a
m
p
lin
g
w
h
er
ei
n
th
e
m
in
o
r
it
y
class
in
s
ta
n
ce
s
ar
e
in
cr
ea
s
ed
to
m
atc
h
t
h
e
n
u
m
b
e
r
o
f
m
aj
o
r
ity
c
lass
i
n
s
ta
n
ce
s
.
T
h
e
S
y
n
th
e
tic
Mi
n
o
r
it
y
Ov
er
-
s
a
m
p
li
n
g
T
ec
h
n
iq
u
e
(
SMOT
E
)
is
w
id
el
y
u
s
ed
i
n
t
h
e
class
i
m
b
alan
ce
p
r
o
b
le
m
.
S
MO
T
E
is
an
o
v
er
-
s
a
m
p
li
n
g
a
p
p
r
o
ac
h
th
at
cr
ea
tes
s
y
n
t
h
etic
m
i
n
o
r
it
y
cla
s
s
s
a
m
p
les
to
m
atc
h
t
h
e
n
u
m
b
er
o
f
m
aj
o
r
it
y
clas
s
i
n
s
ta
n
ce
s
.
S
MO
T
E
is
r
ep
o
r
ted
t
o
p
er
f
o
r
m
b
etter
th
a
n
s
i
m
p
le
o
v
er
-
s
a
m
p
li
n
g
.
SMOT
E
is
al
s
o
co
m
p
u
tatio
n
all
y
e
x
p
e
n
s
i
v
e
to
i
m
p
le
m
en
t
w
h
en
co
m
p
ar
ed
to
s
a
m
p
lin
g
m
et
h
o
d
s
lik
e
r
a
n
d
o
m
u
n
d
er
-
s
a
m
p
l
in
g
[
2
1
]
.
Ho
w
ev
er
,
o
th
er
e
x
p
er
im
en
ts
h
av
e
p
r
o
v
ed
th
at
s
i
m
p
le
u
n
d
er
-
s
a
m
p
li
n
g
ten
d
s
to
o
u
tp
er
f
o
r
m
SMOT
E
in
m
o
s
t
s
i
tu
atio
n
s
[
2
2
]
.
T
h
e
p
er
f
o
r
m
an
ce
o
f
class
i
f
ier
s
i
m
p
le
m
e
n
ti
n
g
SM
OT
E
h
as
b
ee
n
f
o
u
n
d
to
v
ar
y
b
ased
o
n
th
e
n
u
m
b
er
o
f
d
i
m
e
n
s
io
n
s
i
n
t
h
e
tr
ai
n
i
n
g
d
ataset
[
2
2
]
.
Sm
ar
t
re
-
s
a
m
p
li
n
g
ca
n
b
e
d
ep
lo
y
ed
i
n
s
tead
o
f
co
s
t
-
s
e
n
s
i
tiv
e
lear
n
i
n
g
a
s
t
h
e
y
ca
n
p
r
o
v
id
e
n
e
w
in
f
o
r
m
atio
n
o
r
eli
m
i
n
ate
r
ed
u
n
d
a
n
t
i
n
f
o
r
m
atio
n
f
o
r
th
e
l
ea
r
n
in
g
al
g
o
r
ith
m
[
1
1
]
.
T
h
e
d
is
ad
v
an
ta
g
es
o
f
s
a
m
p
li
n
g
ar
e,
th
e
r
a
n
d
o
m
u
n
d
er
s
a
m
p
li
n
g
m
eth
o
d
ca
n
p
o
ten
tiall
y
r
e
m
o
v
e
ce
r
tai
n
cr
i
tical
in
s
ta
n
ce
s
,
a
n
d
r
an
d
o
m
o
v
er
-
s
a
m
p
li
n
g
ca
n
l
ea
d
to
o
v
er
-
f
itti
n
g
[
1
1
]
,
[
1
2
]
.
T
h
e
th
r
e
s
h
o
ld
-
m
o
v
i
n
g
ap
p
r
o
ac
h
to
t
h
e
clas
s
i
m
b
alan
ce
p
r
o
b
le
m
d
o
es
n
o
t
in
v
o
l
v
e
a
n
y
s
a
m
p
li
n
g
.
C
er
tai
n
clas
s
i
f
ier
s
li
k
e
th
e
B
a
y
esia
n
o
r
d
ec
is
io
n
tr
ee
in
d
u
ctio
n
,
r
etu
r
n
a
p
r
o
b
ab
ilit
y
v
alu
e
alo
n
g
w
it
h
th
e
cla
s
s
lab
e
l
w
h
ic
h
ca
n
b
e
u
s
ed
to
co
m
p
u
t
e
a
n
e
w
th
r
es
h
o
ld
.
I
n
class
,
b
ala
n
ce
d
d
atasets
t
h
e
p
r
o
b
a
b
ilit
y
th
r
e
s
h
o
ld
is
0
.
5
.
I
n
ca
s
e
o
f
clas
s
i
m
b
ala
n
ce
,
th
e
r
esu
lts
o
f
th
e
class
i
f
ier
ca
n
al
s
o
b
e
w
ei
g
h
te
d
b
ased
o
n
co
s
t
s
.
I
n
g
e
n
er
al,
t
h
r
es
h
o
ld
m
o
v
in
g
m
o
v
es
th
e
t
h
r
es
h
o
ld
,
s
o
t
h
at
t
h
e
r
ar
e
class
t
u
p
les
ar
e
ea
s
ier
to
class
i
f
y
.
Th
r
esh
o
ld
m
o
v
i
n
g
te
ch
n
iq
u
e
is
k
n
o
w
n
to
r
ed
u
ce
s
t
h
e
co
s
tl
y
FN
er
r
o
r
s
in
clas
s
i
f
ier
s
u
s
ed
f
o
r
m
ed
ical
d
iag
n
o
s
i
s
.
1
.
3
.
E
v
a
lua
t
ing
Cla
s
s
if
ier
P
er
f
o
rm
a
nce
T
r
a
d
itio
n
al
class
i
f
icatio
n
ac
cu
r
ac
y
m
ea
s
u
r
es
s
u
c
h
as
ac
c
u
r
ac
y
o
r
m
is
cla
s
s
i
f
icat
io
n
r
ate
ar
e
n
o
t
g
o
o
d
in
d
icato
r
s
o
f
cla
s
s
i
f
ier
ac
c
u
r
a
c
y
i
n
c
lass
-
i
m
b
ala
n
ce
d
d
atase
ts
[
1
1
]
,
[
1
2
]
.
I
f
th
e
tar
g
et
cla
s
s
i
s
v
er
y
r
ar
e,
s
a
y
0.
5
%
,
co
r
r
ec
tly
p
r
ed
ictin
g
a
ll i
n
s
ta
n
ce
s
o
f
th
e
m
aj
o
r
it
y
cla
s
s
ca
n
ac
h
iev
e
a
v
er
y
h
i
g
h
ac
c
u
r
ac
y
le
v
el
o
f
9
9
.
5
%.
T
h
e
ac
cu
r
ac
y
m
ea
s
u
r
e
o
f
p
r
ec
is
io
n
an
d
r
e
ca
ll
ar
e
m
o
r
e
r
elev
an
t
in
th
e
ca
s
e
o
f
clas
s
-
i
m
b
ala
n
ce
d
d
atasets
[
1
7
]
.
P
r
ec
is
io
n
d
en
o
tes
t
h
e
f
r
ac
tio
n
o
f
in
s
tan
ce
s
t
h
at
ar
e
T
P
s
in
th
e
s
et
o
f
al
l
i
n
s
ta
n
ce
s
p
r
ed
icted
as
P
(
T
P
+FP)
.
R
ec
all
m
ea
s
u
r
e
s
th
e
f
r
ac
tio
n
o
f
T
P
s
co
r
r
ec
tly
p
r
ed
icted
in
th
e
s
e
t
o
f
all
ac
t
u
al
P
in
s
tan
ce
s
(
T
P
+FN)
.
C
las
s
i
f
ier
s
w
i
th
h
i
g
h
r
ec
all
h
av
e
les
s
n
u
m
b
er
o
f
FNs
.
Hen
ce
f
o
r
r
ar
e
class
e
s
,
t
h
e
class
i
f
ier
s
h
o
u
ld
b
e
ev
alu
a
ted
b
ased
o
n
h
o
w
it
p
er
f
o
r
m
s
o
n
b
o
th
r
ec
all
an
d
p
r
e
cisi
o
n
.
Us
u
all
y
,
i
n
class
-
i
m
b
al
an
ce
d
d
atasets
,
th
e
tar
g
et
clas
s
h
a
s
m
u
c
h
lo
w
er
p
r
ec
is
io
n
an
d
r
ec
all
th
a
n
t
h
e
m
aj
o
r
ity
cla
s
s
.
Ma
n
y
p
r
ac
titi
o
n
er
s
h
a
v
e
o
b
s
er
v
ed
th
at
f
o
r
s
k
e
w
ed
c
lass
d
is
tr
i
b
u
tio
n
s
t
h
e
r
ec
all
o
f
th
e
m
in
o
r
it
y
clas
s
i
s
o
f
ten
0
,
w
h
ich
m
ea
n
s
th
a
t
n
o
class
i
f
icatio
n
r
u
le
s
h
a
v
e
b
ee
n
g
en
er
ated
f
o
r
t
h
e
tar
g
et
cl
as
s
.
C
o
m
m
o
n
l
y
u
s
ed
g
r
ap
h
ical
d
i
s
p
la
y
o
f
cla
s
s
i
f
ier
ac
cu
r
ac
y
i
n
clu
d
e
r
ec
eiv
er
o
p
er
atin
g
ch
ar
ac
ter
is
tic
cu
r
v
e
(
R
O
C
)
,
th
e
p
r
ec
is
io
n
-
r
ec
all
cu
r
v
e
(
P
R
C
)
a
n
d
co
s
t
cu
r
v
e
s
.
Fo
r
a
b
i
n
ar
y
c
lass
if
ie
r
,
R
OC
cu
r
v
e
i
s
a
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
4
,
A
u
g
u
s
t
2
0
1
7
:
2
2
1
5
–
2
2
2
2
2218
g
r
ap
h
ical
m
et
h
o
d
to
g
r
ap
h
ical
l
y
r
ep
r
esen
t
th
e
tr
ad
e
-
o
f
f
b
et
w
ee
n
T
P
r
ate
an
d
FP
r
ate
[
1
1
]
,
[
1
4
]
.
A
R
OC
p
lo
t
p
r
o
v
id
es
a
s
in
g
le
p
er
f
o
r
m
a
n
ce
m
ea
s
u
r
e
ca
lled
th
e
A
r
ea
u
n
d
e
r
th
e
R
O
C
cu
r
v
e
(
A
UC
)
s
co
r
e.
A
U
C
s
co
r
e
is
0
.
5
f
o
r
“
c
h
a
n
ce
”
cla
s
s
i
f
ier
s
,
w
h
ic
h
i
n
d
icate
s
th
e
lack
o
f
a
n
y
s
t
atis
tical
d
ep
en
d
e
n
ce
a
n
d
is
eq
u
iv
a
len
t
to
r
an
d
o
m
g
u
e
s
s
i
n
g
an
d
1
.
0
f
o
r
p
er
f
ec
t
class
i
f
ier
s
.
T
h
e
A
r
ea
u
n
d
er
R
OC
C
u
r
v
e
(
A
UC
)
ca
n
b
e
u
s
ed
to
co
m
p
ar
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
m
u
lt
ip
le
class
if
ier
s
,
b
u
t
th
e
y
ar
e
n
o
t
v
er
y
u
s
ef
u
l
f
o
r
class
-
i
m
b
ala
n
ce
d
d
atasets
.
P
r
ec
is
io
n
-
R
ec
all
c
u
r
v
e
s
(
P
R
C
)
ar
e
o
f
te
n
u
s
ed
in
s
tead
o
f
R
OC
p
lo
ts
to
r
ep
r
esen
t
ac
cu
r
ac
y
i
n
th
e
class
-
i
m
b
ala
n
ce
d
d
atasets
[
2
3
]
,
[
2
4
]
.
T
h
e
P
R
C
p
lo
t
s
h
o
w
s
p
r
ec
is
io
n
v
al
u
es
f
o
r
co
r
r
esp
o
n
d
in
g
r
ec
all
o
r
s
en
s
it
i
v
it
y
v
al
u
es.
W
h
i
le
th
e
b
aseli
n
e
is
f
i
x
ed
w
i
th
R
O
C
,
th
e
b
asel
i
n
e
o
f
P
R
C
is
d
et
er
m
in
ed
as
P
/
(
P
+
N
)
.
T
h
e
ar
ea
u
n
d
er
th
e
P
R
cu
r
v
e
,
d
e
n
o
ted
as
AUC
(
P
R
C
)
,
i
s
a
b
etter
in
d
icato
r
f
o
r
m
u
l
tip
le
cla
s
s
i
f
ier
co
m
p
a
r
is
o
n
s
in
t
h
e
cla
s
s
-
i
m
b
alan
ce
d
d
ataset
s
[
1
4
]
.
T
h
e
co
s
t
cu
r
v
e
(
C
C
)
is
an
a
lter
n
ati
v
e
to
th
e
R
O
C
p
lo
t,
an
d
th
e
y
an
a
l
y
ze
cla
s
s
i
f
icatio
n
p
er
f
o
r
m
a
n
ce
b
y
v
ar
y
in
g
o
p
er
atin
g
p
o
in
ts
,
w
h
ic
h
ar
e
b
ased
o
n
class
p
r
o
b
a
b
ilit
ies
an
d
m
is
c
lass
if
ica
tio
n
co
s
ts
[
1
4
]
.
T
h
e
p
r
o
b
a
b
ilit
y
co
s
t
f
u
n
ctio
n
o
r
P
C
F
r
ep
r
esen
ts
t
h
e
o
p
er
at
in
g
p
o
in
ts
o
n
t
h
e
x
-
ax
is
,
a
n
d
th
e
n
o
r
m
alize
d
ex
p
e
cted
co
s
t o
r
NE
[
C
]
acco
u
n
ts
f
o
r
th
e
class
i
f
icatio
n
p
er
f
o
r
m
a
n
ce
o
n
th
e
y
-
a
x
is
.
2
.
4
.
P
ro
po
s
ed
S
o
lutio
n
T
h
e
au
th
o
r
s
u
s
e
t
h
e
o
p
en
s
o
u
r
ce
W
E
KA
to
o
l
to
cr
ea
te
class
i
f
ier
s
u
s
in
g
b
o
th
th
e
a
lg
o
r
ith
m
ic
ap
p
r
o
ac
h
as
w
ell
as
t
h
e
s
a
m
p
lin
g
ap
p
r
o
ac
h
.
T
h
e
au
th
o
r
s
p
ick
ed
W
E
KA
,
b
ec
au
s
e
o
f
it
s
p
o
p
u
lar
it
y
a
m
o
n
g
r
esear
ch
er
s
[
2
5
]
.
W
E
KA
is
a
f
r
ee
l
y
a
v
ailab
le,
J
av
a
-
b
ased
co
llectio
n
o
f
m
an
y
d
ata
m
i
n
i
n
g
i
m
p
le
m
e
n
tatio
n
s
an
d
v
is
u
aliza
t
io
n
to
o
ls
.
I
ts
ea
s
y
to
u
s
e
GUI
in
ter
f
ac
e
is
b
et
ter
s
u
ited
f
o
r
n
o
n
-
tec
h
n
ical
u
s
er
s
lik
e
t
h
e
h
ea
lth
ca
r
e
p
o
licy
m
a
k
er
s
.
Sin
ce
t
h
e
s
o
f
t
w
ar
e
is
o
p
en
-
s
o
u
r
ce
,
an
y
r
esear
ch
er
ca
n
m
o
d
if
y
t
h
e
s
o
u
r
ce
an
d
r
ep
ea
t
ex
p
er
i
m
e
n
ts
to
co
m
p
ar
e
r
es
u
lts
.
A
co
s
t
-
b
en
e
f
it
a
n
al
y
s
is
u
s
i
n
g
W
E
K
A
w
as
d
o
n
e
,
an
d
th
e
cla
s
s
i
f
ie
r
p
er
f
o
r
m
a
n
ce
w
as
co
m
p
ar
ed
to
id
en
tify
t
h
e
b
est
class
if
ier
s
f
o
r
th
e
cu
r
r
en
t
class
i
m
b
ala
n
ce
d
h
ea
lth
d
ataset.
I
n
th
e
class
if
ier
s
d
ef
i
n
ed
f
o
r
p
r
e
d
ictio
n
o
f
ch
r
o
n
ic
h
ea
lth
co
n
d
itio
n
s
,
w
e
ar
e
s
p
ec
if
icall
y
in
ter
ested
in
r
ed
u
cin
g
th
e
f
alse
n
e
g
ati
v
es
b
ec
a
u
s
e
it
h
as
a
h
i
g
h
er
co
s
t.
T
h
e
class
i
f
i
er
s
h
o
u
ld
b
e
ab
le
to
p
r
ed
ict
a
s
ig
n
i
f
ica
n
t
n
u
m
b
er
o
f
t
h
e
m
i
n
o
r
it
y
o
r
tar
g
et
clas
s
in
s
ta
n
ce
s
.
O
n
ce
t
h
e
co
r
e
cla
s
s
if
ier
s
ar
e
s
t
u
d
ied
,
t
h
e
a
u
t
h
o
r
s
atte
m
p
t
to
i
m
p
r
o
v
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
class
if
i
er
s
u
s
i
n
g
an
e
n
s
e
m
b
le
o
f
clas
s
if
ier
s
a
n
d
also
d
ata
s
a
m
p
li
n
g
t
ec
h
n
iq
u
es.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
e
d
ata
f
o
r
th
is
ex
p
er
i
m
e
n
t
h
as
b
ee
n
p
r
o
v
id
ed
b
y
th
e
R
u
r
al
Ma
ter
n
it
y
an
d
C
h
ild
W
elf
a
r
e
Ho
m
e
s
(
R
MCW
H
)
o
r
g
a
n
izatio
n
,
w
h
i
ch
i
s
t
h
e
lar
g
e
s
t
p
r
i
v
ate
i
n
te
g
r
ated
h
ea
lth
ca
r
e
d
eliv
er
y
n
et
w
o
r
k
in
Kar
n
ata
k
a.
R
MCW
H
s
ar
e
m
an
n
ed
b
y
th
e
D
ep
ar
tm
e
n
t
o
f
C
o
m
m
u
n
it
y
Me
d
icin
e,
Kast
u
r
b
a
Me
d
ical
C
o
lleg
e,
Ma
n
ip
al,
I
n
d
ia.
T
h
e
d
ataset
h
as
a
to
tal
o
f
2
2
,
5
9
8
in
s
tan
ce
s
a
n
d
5
3
attr
ib
u
tes.
T
h
e
p
r
ed
icto
r
v
ar
iab
les
in
t
h
e
d
atase
t
r
ec
o
r
d
th
e
p
atien
t‟
s
d
e
m
o
g
r
a
p
h
ics,
f
a
m
i
l
y
d
etail
s
,
s
o
cio
ec
o
n
o
m
ic
s
ta
tu
s
,
an
d
li
v
i
n
g
co
n
d
itio
n
s
.
T
h
e
class
att
r
ib
u
te
is
a
b
in
ar
y
attr
ib
u
te
w
h
ic
h
in
d
icate
s
if
t
h
e
p
atien
t
h
as
o
n
e
o
r
m
o
r
e
o
f
th
e
f
o
llo
w
i
n
g
c
h
r
o
n
ic
d
is
ea
s
e
s
:
d
iab
etes,
h
ea
r
t d
is
ea
s
e
o
r
h
i
g
h
b
lo
o
d
p
r
ess
u
r
e.
T
h
e
class
is
i
m
b
a
lan
ce
d
w
i
th
1311
p
atien
ts
w
it
h
c
h
r
o
n
ic
ill
n
ess
an
d
20982
h
ea
lth
y
p
atien
t
s
.
T
h
is
d
atase
t
i
m
p
lie
s
a
r
ar
e
ca
s
e,
w
h
er
ein
5
.
8
%
o
f
t
h
e
to
tal
p
o
p
u
latio
n
i
s
th
e
r
ar
e
p
o
s
itiv
e
ca
s
e.
T
h
e
d
ataset
is
also
u
n
iq
u
e
b
ec
au
s
e
it
co
n
tai
n
s
3
0
5
in
s
ta
n
ce
s
w
it
h
m
is
s
in
g
class
lab
el
s
.
T
h
e
d
ataset
co
n
tai
n
s
an
a
l
m
o
s
t
e
q
u
al
n
u
m
b
er
o
f
m
ale
a
n
d
f
e
m
ale
r
ec
o
r
d
s
.
T
h
e
ch
r
o
n
ic
d
i
s
ea
s
e
w
a
s
f
o
u
n
d
in
p
atien
ts
w
h
o
ar
e
ab
o
v
e
t
h
e
a
g
e
o
f
4
0
.
T
h
e
attr
ib
u
te
s
w
h
ic
h
ar
e
h
ig
h
l
y
co
r
r
elate
d
w
i
th
th
e
c
h
r
o
n
ic
ill
n
e
s
s
o
cc
u
r
r
en
ce
ar
e
ag
e,
g
e
n
d
er
an
d
m
ar
ital
s
tat
u
s
o
f
t
h
e
p
atie
n
t.
T
h
e
p
atien
ts
w
i
th
c
h
r
o
n
ic
d
is
ea
s
es
w
er
e
also
f
o
u
n
d
to
b
e
f
r
o
m
t
h
e
h
ig
h
er
in
co
m
e
g
r
o
u
p
.
B
ased
o
n
liter
atu
r
e
r
ev
ie
w
,
t
h
e
au
t
h
o
r
s
s
e
lecte
d
a
s
u
b
s
e
t
o
f
W
E
KA
cla
s
s
i
f
ier
s
th
a
t
ar
e
k
n
o
w
n
to
w
o
r
k
w
ell
i
n
th
e
clas
s
-
i
m
b
al
an
ce
d
d
atasets
[
2
5
]
.
C
lass
if
ie
r
s
w
h
ich
w
o
r
k
w
it
h
m
is
s
in
g
class
v
al
u
es
w
er
e
ch
o
s
en
d
u
e
to
a
lar
g
e
n
u
m
b
er
o
f
m
is
s
i
n
g
v
alu
e
s
in
t
h
e
h
ea
lt
h
d
ataset
.
I
n
th
e
ca
s
e
o
f
th
e
ch
r
o
n
ic
h
ea
lt
h
d
ataset
,
th
e
co
s
ts
o
f
FNs
i
s
m
u
ch
m
o
r
e
th
at
th
e
co
s
t
o
f
FP
s
.
A
lt
h
o
u
g
h
it
is
p
o
s
s
ib
le
to
co
m
p
u
te
th
e
co
s
t
o
f
th
e
FP
r
eg
ar
d
in
g
t
h
e
co
s
t
o
f
d
iag
n
o
s
tic
test
s
,
t
h
e
co
s
t
o
f
late
d
iag
n
o
s
is
an
d
d
ea
th
ca
n
n
o
t
b
e
ea
s
il
y
q
u
a
n
ti
f
ied
.
T
h
e
au
th
o
r
s
c
h
o
s
e
to
r
ep
r
esen
t
t
h
e
W
E
KA
co
s
t
m
atr
i
x
in
t
h
e
r
at
io
o
f
1
:
10
,
i.e
.
,
T
he
co
s
t
o
f
FN
is
ten
ti
m
es
m
o
r
e
th
an
t
h
e
co
s
t
o
f
th
e
FP
.
T
h
e
w
id
el
y
u
s
ed
s
tr
ati
f
ied
10
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
w
a
s
d
ep
lo
y
ed
f
o
r
th
e
test
in
g
o
f
t
h
e
class
i
f
ier
s
,
d
u
e
to
its
r
elati
v
el
y
lo
w
b
ias
a
n
d
v
ar
ia
n
ce
[
7
]
,
[
1
7]
.
T
h
e
co
r
e
class
if
ier
s
w
er
e
co
m
p
ar
ed
in
ter
m
s
of
to
tal
co
s
t
a
n
d
tr
u
e
p
o
s
iti
v
e
r
ate.
C
o
s
t
b
en
e
f
it
an
al
y
s
is
w
a
s
d
o
n
e
w
it
h
t
h
e
r
es
u
lts
o
f
b
asic
class
i
f
ier
s
,
a
n
d
t
h
e
co
s
t
f
u
n
ctio
n
w
as
m
i
n
i
m
ized
s
o
as
to
lo
w
er
to
tal
co
s
t
s
in
g
e
n
er
al
as
w
ell
a
s
r
ed
u
ce
th
e
to
tal
n
u
m
b
er
o
f
FN
s
in
th
e
clas
s
i
f
ier
.
Af
ter
t
h
e
b
est co
r
e
class
i
f
ier
s
h
ad
b
ee
n
id
en
ti
f
ied
,
t
h
e
au
t
h
o
r
s
co
n
d
u
cted
ex
p
er
i
m
e
n
ts
to
c
h
ec
k
i
f
co
s
t
s
en
s
iti
v
e
lear
n
i
n
g
,
f
ilter
ed
class
i
f
ier
s
,
an
d
en
s
e
m
b
le
m
et
h
o
d
s
co
u
ld
b
e
u
s
ed
to
im
p
r
o
v
e
th
e
r
esu
l
ts
.
W
E
K
A
s
u
p
p
o
r
ts
e
n
s
e
m
b
le
-
b
ased
cla
s
s
i
f
icatio
n
:
b
o
o
s
ti
n
g
,
b
a
g
g
in
g
a
n
d
b
len
d
i
n
g
.
B
o
o
s
tin
g
w
a
s
d
o
n
e
w
it
h
t
h
e
A
d
aB
o
o
s
tM1
w
it
h
d
if
f
er
e
n
t
b
ase
clas
s
i
f
ier
s
to
s
ee
i
f
t
h
eir
r
esu
lt
s
co
u
ld
b
e
i
m
p
r
o
v
ed
.
B
a
g
g
i
n
g
w
i
th
v
ar
io
u
s
b
ase
class
i
f
ier
s
w
as
p
er
f
o
r
m
e
d
to
s
ee
if
it
r
esu
lts
en
h
an
ce
d
b
y
t
h
e
s
ep
ar
atio
n
o
f
d
ata
in
to
s
a
m
p
les.
B
len
d
in
g
w
a
s
co
n
d
u
cted
u
s
i
n
g
Stac
k
i
n
g
in
W
E
KA
w
h
ic
h
is
b
ased
o
n
th
e
Stac
k
ed
Ag
g
r
eg
atio
n
m
et
h
o
d
u
s
in
g
a
d
iv
er
s
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Lea
r
n
in
g
fr
o
m
a
C
la
s
s
I
mb
a
la
n
ce
d
P
u
b
lic
Hea
lth
Da
ta
s
et:
a
C
o
s
t
-
b
a
s
ed
C
o
mp
a
r
is
o
n
o
f …
.
(
R
o
h
in
i R
.
R
a
o
)
2219
b
len
d
o
f
alg
o
r
ith
m
s
.
T
h
e
ch
o
ice
o
f
b
ase
class
if
ier
s
f
o
r
th
e
en
s
e
m
b
le
w
as
b
ased
o
n
th
e
as
s
u
m
p
tio
n
t
h
at
b
ase
class
i
f
ier
s
ar
e
in
d
ep
en
d
e
n
t o
f
ea
ch
o
th
er
an
d
t
h
at
th
e
b
ase
cl
ass
i
f
ier
s
p
er
f
o
r
m
b
etter
t
h
an
r
a
n
d
o
m
g
u
es
s
i
n
g
.
I
n
th
e
last
e
x
p
er
i
m
e
n
t,
th
e
d
ataset
s
w
er
e
m
o
d
if
ied
u
s
i
n
g
u
n
d
er
-
s
a
m
p
li
n
g
,
o
v
er
-
s
a
m
p
lin
g
a
n
d
SMOT
E
tech
n
iq
u
e
s
to
s
ee
t
h
eir
i
m
p
ac
t
o
n
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
class
i
f
ier
s
.
E
ac
h
o
f
t
h
ese
s
a
m
p
lin
g
tec
h
n
iq
u
e
s
en
s
u
r
ed
th
at
b
o
th
class
lab
els
ar
e
b
alan
ce
d
,
u
s
in
g
t
h
e
W
E
KA
f
ilter
s
“Res
a
m
p
le,
”
“Sp
r
ea
d
Su
b
Sa
m
p
le”
a
n
d
“
SM
OT
E
.
”
I
n
th
e
f
ir
s
t
s
tr
ate
g
y
,
t
h
e
“Sp
r
ea
d
Su
b
Sa
m
p
le”
f
ilter
w
h
ich
p
r
o
d
u
ce
s
a
r
an
d
o
m
s
u
b
s
a
m
p
le
o
f
a
d
ataset
w
as
u
s
ed
.
T
h
is
f
il
ter
p
er
f
o
r
m
s
u
n
d
er
-
s
a
m
p
li
n
g
to
en
s
u
r
e
a
u
n
i
f
o
r
m
d
is
tr
ib
u
tio
n
o
f
cla
s
s
e
s
,
w
h
ic
h
r
esu
lted
i
n
a
d
ataset
w
i
th
1
0
7
4
p
o
s
itiv
e
i
n
s
tan
ce
s
an
d
1311
n
eg
ati
v
e
i
n
s
ta
n
ce
s
.
I
n
t
h
e
s
ec
o
n
d
s
tr
ateg
y
,
th
e
“Res
a
m
p
le”
f
ilter
w
a
s
u
s
ed
,
w
it
h
th
e
“
b
ia
s
T
o
Un
if
o
r
m
D
is
t
r
ib
u
tio
n
”
o
p
tio
n
to
g
et
a
n
o
v
er
-
s
a
m
p
led
d
ataset
w
it
h
r
ep
lace
m
en
t.
T
h
e
o
v
er
-
s
a
m
p
led
d
ataset
r
es
u
lted
in
a
d
ataset
w
it
h
11299
p
o
s
itiv
e
c
l
ass
i
n
s
tan
ce
s
a
n
d
1
1
1
4
0
n
eg
ati
v
e
clas
s
in
s
ta
n
ce
s
.
T
h
e
SMOT
E
f
ilter
w
a
s
als
o
u
s
ed
to
r
esa
m
p
le
t
h
e
d
atase
t
u
s
i
n
g
f
iv
e
n
ea
r
est
n
eig
h
b
o
r
s
to
g
e
n
er
ate
14421
p
o
s
itiv
e
i
n
s
ta
n
ce
s
a
n
d
20982
n
eg
at
iv
e
i
n
s
tan
ce
s
.
T
h
e
th
r
ee
d
atasets
w
er
e
t
h
en
u
s
ed
to
p
r
o
d
u
ce
class
if
ier
s
u
s
i
n
g
d
i
f
f
er
en
t c
las
s
i
f
ier
m
e
th
o
d
s
,
an
d
th
e
r
esu
l
ts
w
er
e
r
an
k
ed
b
ased
o
n
co
s
t
.
3.
RE
SU
L
T
S
A
ND
D
I
SCU
SS
I
O
N
I
n
th
e
f
ir
s
t
s
tag
e,
co
r
e
class
i
f
ier
s
w
er
e
b
u
ilt
,
an
d
its
p
er
f
o
r
m
an
ce
f
o
r
th
e
class
-
i
m
b
alan
c
ed
d
ataset
w
a
s
an
al
y
ze
d
.
T
h
e
d
ata
also
co
n
tain
s
m
is
s
in
g
v
alu
e
s
,
an
d
o
n
l
y
th
o
s
e
cla
s
s
i
f
ier
s
w
h
ich
s
u
p
p
o
r
t
m
is
s
i
n
g
cla
s
s
v
alu
e
s
w
er
e
ev
a
lu
ated
f
o
r
t
h
eir
p
er
f
o
r
m
an
ce
.
T
h
e
S
u
p
p
o
r
t
Vec
to
r
Ma
c
h
in
e
b
ased
W
E
KA
i
m
p
le
m
e
n
tat
io
n
L
I
b
SVM,
p
r
o
d
u
ce
d
an
e
f
f
ec
ti
v
e
cla
s
s
i
f
ier
w
i
th
t
h
e
s
ig
m
o
id
k
er
n
e
l
w
h
ile
o
t
h
er
k
er
n
el
s
li
k
e
li
n
e
ar
a
n
d
r
ad
ial
r
esu
lted
in
“
c
h
a
n
ce”
class
i
f
ier
s
w
h
ic
h
w
er
e
eq
u
i
v
alen
t
to
a
r
an
d
o
m
g
u
es
s
.
T
h
is
also
i
m
p
li
es
th
at
t
h
e
s
o
lu
tio
n
s
p
ac
e
is
n
o
t
li
n
ea
r
l
y
s
ep
ar
ab
le.
C
h
a
n
ce
c
lass
if
ier
s
w
er
e
el
i
m
in
a
ted
,
an
d
t
h
e
r
e
m
ai
n
i
n
g
clas
s
i
f
ier
s
w
er
e
s
h
o
r
tlis
ted
an
d
r
an
k
ed
b
ased
o
n
th
e
to
tal
co
s
t
o
f
th
e
class
if
i
er
(
s
ee
T
ab
le
1
)
.
T
h
e
„
T
o
tal
C
o
s
t‟
w
as
ca
lc
u
lated
b
ased
o
n
th
e
co
s
t
m
atr
i
x
;
t
h
e
co
s
ts
w
er
e
f
u
r
th
er
r
ed
u
ce
d
b
y
p
er
f
o
r
m
in
g
a
co
s
t
-
b
e
n
ef
it
a
n
al
y
s
i
s
to
e
n
s
u
r
e
a
m
i
n
i
m
u
m
n
u
m
b
er
o
f
FNs
.
T
h
e
„
T
o
tal
C
o
s
t
(
Op
ti
m
ized
)
‟
r
ep
r
esen
ts
t
h
e
co
s
ts
af
ter
a
c
o
s
t
-
b
en
e
f
it
an
a
l
y
s
is
.
F
e
w
er
FNs
(
w
h
ic
h
r
es
u
lts
in
l
ess
co
s
t)
a
n
d
h
i
g
h
r
ec
al
l
is
d
e
s
ir
ab
le.
T
h
e
to
tal
n
u
m
b
er
o
f
F
Ns
b
ef
o
r
e
an
d
a
f
ter
co
s
t
-
b
en
e
f
it
a
n
al
y
s
i
s
w
er
e
ta
b
u
lated
.
T
h
e
A
UC
v
al
u
es
i
n
co
lu
m
n
8
p
r
o
v
e
th
at
all
t
h
es
e
class
i
f
ier
s
ar
e
n
o
t
eq
u
iv
ale
n
t to
r
an
d
o
m
g
u
e
s
s
b
u
t c
an
class
if
y
t
h
e
d
ata
in
s
p
ite
o
f
class
i
m
b
ala
n
ce
.
T
ab
le
1
.
C
o
s
ts
an
d
p
er
f
o
r
m
an
ce
o
f
co
r
e
class
if
ier
s
(
T
o
p
1
0
)
R
a
n
k
W
E
K
A
C
l
a
ssi
f
i
e
r
T
o
t
a
l
C
o
st
T
o
t
a
l
C
o
st
(
O
p
t
i
m
i
z
e
d
)
R
e
c
a
l
l
F
N
s
(
o
u
t
o
f
1
3
1
1
)
F
N
s
(
O
p
t
i
m
i
z
e
d
)
A
c
c
u
r
a
c
y
R
a
t
e
(
%)
A
U
C
A
U
C
(
P
R
C
)
1
B
a
y
e
si
a
n
N
e
t
5
5
0
4
4
9
5
3
0
.
7
3
3
5
7
1
7
4
8
9
.
7
2
0
.
9
2
5
0
.
3
7
9
2
N
a
ï
v
e
B
a
y
e
si
a
n
5
6
0
2
4
9
8
8
0
.
7
2
3
6
7
1
4
3
8
9
.
6
9
0
.
9
2
4
0
.
3
7
9
3
L
o
g
i
st
i
c
1
0
4
8
0
4
9
2
8
0
.
2
2
1
0
2
1
1
7
2
9
4
.
2
1
0
.
9
2
4
0
.
3
8
7
4
R
a
n
d
o
m
T
r
e
e
1
0
7
0
7
9
6
6
2
0
.
2
4
9
9
3
8
0
6
9
2
.
0
6
0
.
6
5
7
0
.
1
4
2
5
J4
8
1
1
1
3
7
7
4
6
7
0
.
1
7
1
0
9
1
5
2
8
9
4
.
0
8
0
.
8
3
6
0
.
2
8
6
6
V
o
t
e
d
P
e
r
c
e
p
t
r
o
n
1
1
4
8
8
1
1
3
2
7
0
.
1
4
1
1
2
7
1
1
0
5
9
3
.
9
7
0
.
5
7
2
0
.
1
1
7
7
JR
I
P
1
1
4
4
7
1
1
4
6
8
0
.
1
4
1
1
2
2
1
1
2
2
9
3
.
9
5
0
.
5
7
6
0
.
1
2
8
8
L
i
b
S
V
M
–
S
i
g
mo
i
d
k
e
r
n
e
l
1
1
9
4
5
1
1
9
7
1
0
.
1
2
1
1
5
3
1
3
1
1
9
2
.
9
7
0
.
5
5
0
0
.
0
8
3
9
S
i
mp
l
e
C
a
r
t
1
2
0
5
4
9
1
5
8
0
.
0
9
1
1
9
4
7
4
3
9
4
.
1
3
0
.
7
6
8
0
.
2
1
9
10
I
B
K
,
K
=
5
1
2
5
4
8
1
0
0
2
9
0
.
0
5
1
2
4
9
5
7
7
9
4
.
1
4
0
.
7
2
1
0
.
1
7
8
I
n
g
e
n
er
al,
m
o
s
t
o
f
t
h
e
to
p
class
i
f
ier
s
ex
h
ib
ited
lo
w
to
tal
co
s
t,
an
d
th
e
n
u
m
b
er
o
f
FNs
w
a
s
d
r
asti
ca
ll
y
r
ed
u
ce
d
u
s
i
n
g
th
e
co
s
t
-
b
en
e
f
it
a
n
al
y
s
i
s
.
T
h
e
B
ay
e
s
ia
n
clas
s
if
ier
s
h
ad
t
h
e
b
est
p
er
f
o
r
m
an
c
e
r
eg
ar
d
in
g
t
h
e
co
s
t
o
f
th
e
co
r
e
class
i
f
ier
s
i
n
th
e
class
i
m
b
ala
n
ce
d
h
ea
lth
d
ataset
(
T
ab
le
1
)
.
T
h
e
W
E
KA
b
ased
p
air
ed
t
-
test
p
r
o
v
ed
t
h
at
t
h
er
e
is
n
o
d
if
f
er
en
ce
in
th
e
p
er
f
o
r
m
an
ce
o
f
t
h
e
Na
ïv
e
B
a
y
esia
n
an
d
B
a
y
es
ian
n
et
class
i
f
ier
s
.
T
h
e
B
a
y
e
s
ian
n
et
class
i
f
ier
i
s
p
r
ef
er
r
ed
b
ec
au
s
e
in
itial
e
x
p
lo
r
atio
n
s
h
o
w
s
a
s
tr
o
n
g
co
r
r
elatio
n
a
m
o
n
g
t
h
e
p
atie
n
t
f
ea
tu
r
es.
B
a
y
esia
n
cla
s
s
i
f
ier
s
ar
e
k
n
o
w
n
to
w
o
r
k
w
ell
i
n
s
it
u
atio
n
s
li
k
e
m
ed
ical
d
iag
n
o
s
is
,
w
h
er
ein
t
h
e
r
elatio
n
s
h
ip
b
et
wee
n
th
e
attr
ib
u
te
s
et
an
d
class
v
ar
iab
le
is
n
o
n
-
d
eter
m
in
i
s
tic.
B
ay
e
s
ian
cla
s
s
i
f
ier
s
ar
e
also
r
o
b
u
s
t
to
n
o
is
e,
ir
r
elev
an
t
attr
ib
u
tes
a
n
d
co
n
f
o
u
n
d
i
n
g
f
ac
to
r
s
t
h
at
ar
e
n
o
t
in
c
lu
d
ed
i
n
t
h
e
class
i
f
icatio
n
.
A
ll
t
h
e
o
th
er
al
g
o
r
ith
m
s
l
ik
e
L
o
g
is
tic,
R
an
d
o
m
T
r
ee
an
d
Vo
ted
P
e
r
ce
p
tr
o
n
h
av
e
a
h
i
g
h
n
u
m
b
er
o
f
f
al
s
e
n
eg
at
iv
e
s
w
h
ich
d
r
ast
icall
y
i
n
cr
ea
s
e
s
t
h
e
co
s
t
o
f
th
e
class
i
f
ier
.
T
h
e
r
esu
l
ts
al
s
o
co
n
t
r
ad
ict
th
e
r
esu
lt
s
o
f
W
eiss
[
1
2
]
w
h
o
ad
v
o
ca
ted
th
e
u
s
e
o
f
i
n
s
tan
ce
b
ased
lear
n
er
s
f
o
r
th
e
clas
s
i
m
b
ala
n
ce
p
r
o
b
lem
.
T
h
e
r
u
le
-
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
4
,
A
u
g
u
s
t
2
0
1
7
:
2
2
1
5
–
2
2
2
2
2220
b
ased
class
if
ier
,
J
R
I
P
w
h
ic
h
is
an
i
m
p
le
m
e
n
tatio
n
o
f
a
p
r
o
p
o
s
itio
n
al
r
u
le
lear
n
er
,
is
w
ell
s
u
ited
f
o
r
h
an
d
li
n
g
class
i
m
b
ala
n
ce
s
a
n
d
ap
p
ea
r
s
in
th
e
to
p
10
class
if
ier
s
.
T
h
is
r
es
u
lt
is
i
n
lin
e
w
it
h
p
r
ev
io
u
s
r
esu
lt
s
w
h
ic
h
s
u
g
g
e
s
ted
t
h
at
k
er
n
el
b
ased
S
VM
S
w
o
r
k
b
etter
i
n
cla
s
s
i
m
b
alan
ce
p
r
o
b
le
m
s
[
1
2
]
,
[
1
3
]
,
[
1
5
]
.
E
v
en
th
o
u
g
h
th
e
class
i
f
ier
ac
cu
r
ac
y
f
o
r
all
clas
s
if
ier
s
is
h
i
g
h
,
b
et
w
ee
n
89
%
a
n
d
94
%,
o
n
l
y
th
e
B
a
y
e
s
ia
n
class
i
f
ier
s
,
h
a
v
e
a
h
ig
h
r
ec
all
(
0.
7
3
)
an
d
lo
w
n
u
m
b
er
o
f
FNs
.
I
n
th
e
s
ec
o
n
d
e
x
p
er
i
m
e
n
t,
t
h
e
ef
f
ec
t o
f
e
n
s
e
m
b
li
n
g
m
e
th
o
d
s
lik
e
R
an
d
o
m
Fo
r
est,
B
o
o
s
tin
g
,
B
ag
g
i
n
g
,
Stack
i
n
g
a
n
d
Vo
ti
n
g
o
n
t
h
e
s
e
b
aselin
e
clas
s
i
f
ier
s
w
as
s
t
u
d
ied
.
T
h
e
r
esu
lts
ar
e
tab
u
la
ted
i
n
T
ab
le
2
.
T
h
e
co
s
t
s
en
s
iti
v
e
lear
n
in
g
a
n
d
t
h
e
m
eta
co
s
t
i
m
p
le
m
en
ta
tio
n
s
i
n
W
E
K
A
w
er
e
also
e
v
al
u
a
ted
.
T
h
e
class
i
f
ier
p
er
f
o
r
m
a
n
ce
w
as
ag
ai
n
r
an
k
e
d
b
ased
o
n
to
tal
co
s
t
w
h
ich
w
a
s
f
u
r
th
er
o
p
ti
m
ized
u
s
i
n
g
c
o
s
t
-
b
en
e
f
it
a
n
al
y
s
is
an
d
t
ab
u
lated
in
T
ab
le
2
.
T
ab
le
2
.
C
o
s
ts
an
d
p
er
f
o
r
m
an
ce
o
f
E
n
s
e
m
b
le
&
C
o
s
t b
ased
C
las
s
i
f
ier
s
(
T
o
p
1
0
)
R
a
n
k
C
l
a
ssi
f
i
e
r
T
o
t
a
l
C
o
st
T
o
t
a
l
c
o
st
(
o
p
t
i
m
i
z
e
d
)
R
e
c
a
l
l
F
N
s
(
o
u
t
o
f
1
3
1
1
)
F
N
s
(
o
p
t
i
m
i
z
e
d
)
A
c
c
u
r
a
c
y
Rate
(
%)
A
U
C
A
U
C
(
P
R
C
)
1
C
o
st
S
e
n
s
i
t
i
v
e
(
B
a
y
e
sN
e
t
)
4
9
6
8
5
0
0
7
0
.
9
0
1
3
7
1
6
9
8
3
.
2
5
0
.
9
2
4
0
.
3
7
8
2
F
i
l
t
e
r
e
d
c
l
a
ss,
C
l
a
ss B
a
l
a
n
c
e
r
,
(
B
a
y
e
sN
e
t
)
5
1
5
0
5
0
0
0
0
.
9
1
1
1
7
1
6
9
8
1
.
6
2
0
.
9
2
4
0
.
3
7
8
3
C
o
st
se
n
s
i
t
i
v
e
(
JR
I
P
)
5
2
8
1
5
4
4
8
0
.
8
5
1
9
7
2
1
4
8
4
.
2
6
0
.
8
5
7
0
.
2
6
2
4
M
e
t
a
C
o
st
(
B
a
y
e
sN
e
t
)
5
4
2
6
5
1
8
5
0
.
9
3
95
1
9
2
7
9
.
4
9
0
.
9
1
9
0
.
3
5
5
5
C
o
st
se
n
s
i
t
i
v
e
(
L
o
g
i
st
i
c
)
5
5
0
7
5
6
0
7
0
.
8
5
1
9
6
2
2
3
8
3
.
2
1
0
.
9
0
4
0
.
3
1
2
6
B
a
g
g
i
n
g
(
B
a
y
e
sN
e
t
)
5
5
7
5
4
9
6
9
0
.
7
2
3
6
8
1
7
1
8
9
.
8
5
0
.
9
2
4
0
.
3
8
0
7
F
i
l
t
e
r
e
d
c
l
a
ss,
C
l
a
ss B
a
l
a
n
c
e
r
,
(
L
o
g
i
st
i
c
)
5
6
1
8
5
5
4
7
0
.
8
8
1
5
3
2
1
1
8
0
.
9
8
0
.
9
0
7
0
.
3
2
0
8
F
i
l
t
e
r
e
d
c
l
a
ss
C
l
a
ss
B
a
l
a
n
c
e
r
,
(
J4
8
)
6
2
4
0
6
2
8
7
0
.
7
0
3
9
9
4
0
3
8
8
.
1
2
0
.
7
9
7
0
.
2
3
1
9
V
o
t
e
(
B
a
y
e
sN
e
t
w
i
t
h
L
o
g
i
st
i
c
)
6
4
4
3
4
9
3
6
0
.
6
0
5
2
0
1
9
1
9
2
.
0
9
0
.
9
2
8
0
.
3
9
4
10
V
o
t
e
(
B
a
y
e
sN
e
t
w
i
t
h
r
a
n
d
o
m fo
r
e
st
)
6
5
4
3
4
8
3
6
0
.
5
9
5
3
4
1
5
4
9
2
.
2
0
0
.
9
2
7
0
.
4
0
0
T
h
e
co
s
t
s
en
s
iti
v
e
lear
n
i
n
g
i
m
p
le
m
e
n
tatio
n
w
it
h
d
if
f
er
en
t
co
r
e
clas
s
i
f
ier
s
e
x
h
ib
ite
d
th
e
b
est
p
er
f
o
r
m
a
n
ce
.
I
n
t
h
e
ca
s
e
o
f
t
h
e
J
R
I
P
an
d
L
o
g
is
t
ic
cla
s
s
i
f
i
er
s
,
th
e
co
s
t
-
s
e
n
s
i
tiv
e
lear
n
in
g
ap
p
r
o
ac
h
al
m
o
s
t
h
alv
e
s
th
e
to
tal
co
s
t
o
f
t
h
e
co
r
e
i
m
p
le
m
e
n
tatio
n
.
T
h
e
f
ilter
ed
class
i
f
ier
w
i
th
t
h
e
class
b
alan
ce
r
f
ilter
p
r
o
d
u
ce
s
g
o
o
d
r
esu
lts
w
ith
L
o
g
is
tic
an
d
J
4
8
m
eth
o
d
s
w
h
ic
h
w
er
e
p
r
ev
io
u
s
l
y
a
f
f
ec
ted
b
y
th
e
cla
s
s
i
m
b
alan
ce
.
Ho
w
e
v
er
,
t
h
e
co
s
t
-
b
en
e
f
it
a
n
al
y
s
i
s
s
o
m
eti
m
e
s
led
to
an
i
n
cr
ea
s
e
i
n
o
v
er
all
co
s
t
e
v
en
th
o
u
g
h
t
h
e
n
u
m
b
er
o
f
FNs
w
er
e
lo
w
.
T
h
e
in
cr
ea
s
e
in
to
tal
co
s
t
w
a
s
d
u
e
to
a
m
a
s
s
i
v
e
i
n
cr
ea
s
e
in
FP
s
as
a
r
esu
lt
o
f
t
h
r
esh
o
ld
m
o
v
in
g
.
As
in
d
icate
d
in
th
e
liter
at
u
r
e
,
th
e
en
s
e
m
b
le
m
eth
o
d
s
li
k
e
Vo
ti
n
g
a
n
d
A
D
A
B
o
o
s
tM1
s
ig
n
i
f
i
ca
n
tl
y
i
n
cr
ea
s
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
cla
s
s
i
f
ier
s
i
n
i
m
b
a
lan
ce
d
clas
s
d
ataset
s
i
n
co
m
p
ar
i
s
o
n
to
co
r
e
class
i
f
ier
s
[
1
9
]
.
I
n
th
e
t
h
ir
d
ex
p
er
i
m
e
n
t,
th
e
e
f
f
ec
t
o
f
s
a
m
p
li
n
g
to
b
alan
ce
th
e
class
es
w
a
s
d
o
n
e
u
s
in
g
tec
h
n
iq
u
e
s
lik
e
u
n
d
er
-
sa
m
p
lin
g
,
o
v
er
s
a
m
p
li
n
g
a
n
d
SMOT
E
(
T
ab
le
3
)
.
T
h
e
r
es
u
lt
s
w
er
e
r
an
k
ed
b
ased
o
n
t
o
tal
co
s
t,
an
d
a
co
s
t
-
b
en
e
f
it
a
n
al
y
s
i
s
w
as
p
er
f
o
r
m
ed
to
s
ee
if
co
s
ts
co
u
ld
b
e
r
ed
u
ce
d
.
I
n
g
en
er
al,
as
in
d
ica
ted
in
th
e
liter
at
u
r
e
,
u
n
d
er
-
s
a
m
p
lin
g
s
ee
m
s
to
w
o
r
k
b
etter
th
an
o
v
er
-
s
a
m
p
li
n
g
an
d
SMOT
E
[
2
2
]
.
T
h
e
au
th
o
r
s
r
ec
o
m
m
e
n
d
th
e
u
s
a
g
e
o
f
r
an
d
o
m
u
n
d
er
-
s
a
m
p
li
n
g
a
s
a
s
o
lu
tio
n
f
o
r
class
i
m
b
ala
n
ce
d
d
atasets
b
ec
au
s
e
it
is
al
s
o
co
m
p
u
tatio
n
all
y
le
s
s
e
x
p
en
s
iv
e
to
im
p
le
m
e
n
t
th
a
n
SMOT
E
o
r
o
v
er
-
s
a
m
p
li
n
g
.
I
t
also
r
ed
u
ce
s
th
e
s
ize
o
f
t
h
e
d
ataset,
w
h
ich
w
il
l
i
m
p
r
o
v
e
ti
m
e
co
m
p
lex
i
t
y
w
it
h
o
u
t
s
ac
r
if
i
ci
n
g
cla
s
s
i
f
icatio
n
p
er
f
o
r
m
an
c
e.
I
n
th
e
ca
s
e
o
f
th
e
J4
8
an
d
I
B
K,
it
w
a
s
o
b
s
er
v
ed
th
at
all
th
r
ee
s
a
m
p
li
n
g
s
tr
ate
g
i
es
i
m
p
r
o
v
ed
th
e
co
s
t
d
r
a
m
atic
all
y
.
T
h
e
s
a
m
p
lin
g
r
esu
lt
s
in
d
icate
th
a
t
J
4
8
an
d
I
B
K
w
o
r
k
b
etter
in
class
b
alan
ce
d
d
ataset
s
.
T
h
is
r
esu
lt
co
n
tr
ad
icts
s
o
m
e
p
r
ev
io
u
s
e
m
p
ir
ical
r
esu
lts
; t
h
e
s
a
m
p
li
n
g
m
et
h
o
d
s
o
u
tp
er
f
o
r
m
ed
th
e
c
o
s
t
-
s
e
n
s
iti
v
e
m
et
h
o
d
s
[
1
1
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Lea
r
n
in
g
fr
o
m
a
C
la
s
s
I
mb
a
la
n
ce
d
P
u
b
lic
Hea
lth
Da
ta
s
et:
a
C
o
s
t
-
b
a
s
ed
C
o
mp
a
r
is
o
n
o
f …
.
(
R
o
h
in
i R
.
R
a
o
)
2221
T
ab
le
3
.
C
o
s
ts
an
d
p
er
f
o
r
m
an
ce
o
f
C
las
s
i
f
ier
s
u
s
i
n
g
s
a
m
p
li
n
g
tech
n
iq
u
e
s
(
T
o
p
1
0
)
Ra
nk
C
las
s
i
f
ier
T
o
tal
C
o
s
t
T
o
tal
c
o
s
t
(
o
p
tim
ize
d)
R
ec
all
A
cc
u
r
ac
y
R
ate
(
%)
FNs
FNs
(
o
p
tim
ized
)
A
U
C
A
U
C
(
P
R
C
)
1
Ov
er
s
a
m
p
lin
g
R
an
d
o
m
T
r
ee
766
530
1
.
0
0
9
6
.
5
8
0
/1
1
2
9
9
5
0
.
9
7
9
0
.
9
5
9
2
Un
d
er
Sa
m
p
li
n
g
J
R
I
P
1198
1349
0
.
9
2
8
4
.
1
0
9
1
/1
0
7
4
68
0
.
8
4
1
0
.
7
1
1
3
Un
d
er
Sa
m
p
li
n
g
B
ay
e
s
n
et
1215
756
0
.
9
1
8
6
.
0
4
9
8
/1
0
7
4
11
0
.
8
8
6
0
.
7
7
7
4
Ov
er
Sa
m
p
lin
g
J
4
8
1265
1292
0
.
9
9
9
5
.
2
0
2
1
/1
0
7
4
14
0
.
9
6
3
0
.
9
3
4
5
Un
d
er
Sa
m
p
li
n
g
J
4
8
1455
1170
0
.
8
9
8
3
.
8
9
1
1
9
/1
0
7
4
65
0
.
8
3
8
0
.
6
9
3
Un
d
er
Sa
m
p
li
n
g
Vo
ted
P
er
ce
p
tr
o
n
1533
1039
0
.
8
8
8
5
.
5
3
1
3
2
/1
0
7
4
56
0
.
8
4
7
0
.
7
0
9
6
Un
d
er
Sa
m
p
li
n
g
L
o
g
i
s
tic
1682
804
0
.
8
6
8
4
.
1
9
1
4
5
/1
0
7
4
11
0
.
8
7
5
0
.
7
4
8
7
Un
d
er
Sa
m
p
li
n
g
I
B
K
k
=5
2205
1096
0
.
8
2
7
8
.
4
9
1
8
8
/1
0
7
4
9
0
.
8
3
1
0
.
7
0
9
8
Un
d
er
Sa
m
p
li
n
g
R
an
d
o
m
T
r
ee
3271
1548
0
.
7
2
7
5
.
3
0
2
9
8
/1
0
7
4
0
0
.
7
5
0
0
.
6
1
1
9
Un
d
er
Sa
m
p
li
n
g
L
I
b
SVM
(
s
ig
m
o
id
k
er
n
el)
3528
1431
0
.
7
1
7
0
.
5
7
3
1
4
/1
0
7
4
105
0
.
6
8
7
0
.
5
4
1
10
Ov
er
s
a
m
p
lin
g
I
B
K
k
5
5033
3991
0
.
9
9
8
3
.
9
5
1
5
9
/1
1
2
9
9
956
0
.
9
6
2
0
.
9
3
8
4.
CO
NCLU
SI
O
N
T
h
is
w
o
r
k
is
r
elev
a
n
t
to
p
u
b
lic
h
ea
lth
p
o
lic
y
m
ak
er
s
,
w
h
o
ca
n
u
s
e
th
e
class
if
ier
s
to
p
r
ed
ict
th
e
o
cc
u
r
r
en
ce
o
f
ch
r
o
n
ic
d
i
s
ea
s
e
in
th
e
p
o
p
u
latio
n
an
d
also
id
en
ti
f
y
t
h
e
f
ac
to
r
s
th
at
ar
e
co
r
r
elate
d
w
it
h
ch
r
o
n
ic
d
is
ea
s
es.
T
h
e
class
if
ier
s
w
ill
h
elp
h
ea
lth
ca
r
e
p
r
o
v
id
er
s
in
im
p
r
o
v
in
g
th
e
ir
p
r
o
g
n
o
s
is
,
d
iag
n
o
s
is
an
d
tr
ea
t
m
e
n
t
p
lan
s
.
E
x
p
er
i
m
en
ts
w
er
e
co
n
d
u
cted
b
ased
o
n
v
ar
io
u
s
ap
p
r
o
ac
h
es
s
u
g
g
e
s
ted
in
t
h
e
liter
a
tu
r
e
,
to
tack
le
t
h
e
class
i
m
b
alan
ce
p
r
o
b
le
m
.
T
h
e
W
E
KA
b
ased
class
if
ier
s
w
er
e
u
s
ed
to
r
ec
o
r
d
an
d
an
al
y
ze
t
h
e
class
if
ier
p
er
f
o
r
m
a
n
ce
in
ter
m
s
o
f
co
s
t
.
T
h
e
B
ay
e
s
ia
n
clas
s
i
f
ier
s
w
e
r
e
id
en
ti
f
ied
as
th
e
b
est
cla
s
s
if
ier
s
f
o
r
t
h
e
cla
s
s
i
m
b
alan
ce
d
d
atase
t.
T
h
e
au
t
h
o
r
s
r
ec
o
m
m
e
n
d
th
e
B
a
y
esia
n
Net
clas
s
i
f
ier
b
ec
au
s
e
o
f
u
n
d
er
ly
i
n
g
co
r
r
elatio
n
a
m
o
n
g
p
atien
t
f
ea
tu
r
es.
T
h
e
c
o
s
t
s
en
s
i
ti
v
e
i
m
p
le
m
en
ta
tio
n
s
an
d
co
s
t
-
b
e
n
ef
it
an
a
l
y
s
is
ca
n
f
u
r
t
h
er
r
ed
u
ce
t
h
e
to
tal
co
s
t
w
h
ile
m
ai
n
tai
n
in
g
t
h
e
ac
cu
r
ac
y
.
Ho
w
e
v
er
,
th
e
e
n
s
e
m
b
le
m
et
h
o
d
s
is
a
co
m
p
le
x
s
o
l
u
tio
n
w
h
er
ei
n
th
er
e
ar
e
a
h
u
g
e
n
u
m
b
er
o
f
s
o
lu
tio
n
s
t
h
at
s
ti
ll
n
ee
d
s
to
b
e
ex
p
lo
r
ed
.
Un
d
er
-
s
a
m
p
l
in
g
is
a
n
ef
f
icie
n
t
d
ata
p
r
e
-
p
r
o
ce
s
s
in
g
ap
p
r
o
ac
h
w
it
h
lo
w
co
m
p
u
ta
tio
n
co
s
ts
,
a
n
d
i
t
is
r
ec
o
m
m
e
n
d
ed
f
o
r
b
u
ild
in
g
co
s
t
e
f
f
ec
ti
v
e
class
i
f
ier
s
.
T
h
e
u
n
d
er
-
s
a
m
p
li
n
g
ca
n
d
r
a
m
at
icall
y
i
m
p
r
o
v
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
m
e
th
o
d
s
li
k
e
J
4
8
,
I
B
K
w
h
ic
h
ar
e
af
f
ec
ted
b
y
th
e
clas
s
i
m
b
al
an
ce
.
T
h
e
cu
r
r
en
t
w
o
r
k
as
s
u
m
es t
h
at
th
e
co
s
t o
f
FNs
i
s
te
n
ti
m
e
s
m
o
r
e
t
h
a
n
th
e
co
s
t o
f
T
P
s
.
T
h
e
w
o
r
k
ca
n
b
e
i
m
p
r
o
v
ed
b
y
ac
t
u
all
y
q
u
an
t
if
y
i
n
g
th
e
ac
t
u
al
co
s
t
s
i
n
g
en
er
ati
n
g
clas
s
i
f
ier
er
r
o
r
s
.
I
n
f
u
t
u
r
e
w
o
r
k
,
t
h
e
e
f
f
ec
t
o
f
f
ea
tu
r
e
s
e
lectio
n
o
n
t
h
e
c
lass
if
ie
r
co
s
t
w
ill
b
e
s
t
u
d
ied
.
T
h
o
u
g
h
ir
r
elev
a
n
t
f
ea
t
u
r
es
ar
e
n
o
t k
n
o
w
n
to
i
m
p
r
o
v
e
clas
s
if
ica
tio
n
p
er
f
o
r
m
a
n
ce
s
i
g
n
i
f
ic
an
tl
y
,
t
h
e
y
ca
n
s
lo
w
d
o
w
n
th
e
class
i
f
ier
p
r
o
ce
s
s
.
ACK
NO
WL
E
D
G
E
M
E
NT
W
e
th
a
n
k
Dr
.
Har
is
h
c
h
an
d
r
a
Heb
b
ar
,
P
r
o
f
ess
o
r
,
Sc
h
o
o
l
o
f
I
n
f
o
r
m
a
tio
n
Scie
n
ce
s
,
Ma
n
ip
al
an
d
th
e
D
ep
ar
t
m
en
t
o
f
C
o
m
m
u
n
it
y
M
ed
icin
e,
KM
C
,
Ma
n
ip
al
f
o
r
s
h
ar
in
g
w
it
h
u
s
v
al
u
ab
le
d
ata.
W
e
th
an
k
Dr
.
Vee
n
a
Ka
m
at
h
,
P
r
o
f
ess
o
r
,
Dep
ar
t
m
en
t
o
f
C
o
m
m
u
n
it
y
Me
d
ici
n
e
,
KM
C
,
Ma
n
ip
al,
f
o
r
ex
ten
d
in
g
u
s
h
er
s
u
b
j
ec
t
ex
p
er
tis
e.
RE
F
E
R
E
NC
E
S
[1
]
T
o
m
a
r
D,
Ag
a
r
wa
l
S
.
“
A
su
rv
e
y
o
n
Da
ta
M
in
i
n
g
a
p
p
ro
a
c
h
e
s
f
o
r
He
a
lt
h
c
a
re
”,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Bi
o
-
S
c
ien
c
e
a
n
d
Bi
o
-
T
e
c
h
n
o
lo
g
y
.
2
0
1
3
;
5
(
5
):
2
4
1
-
2
6
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
7
,
No
.
4
,
A
u
g
u
s
t
2
0
1
7
:
2
2
1
5
–
2
2
2
2
2222
[2
]
Diss
a
n
a
y
a
k
a
C,
A
b
d
u
ll
a
h
H,
A
h
m
e
d
B,
P
e
n
z
e
l
T
,
Cv
e
tk
o
v
ic
D.
“
Cla
ss
if
ica
t
io
n
o
f
He
a
lt
h
y
S
u
b
je
c
ts
a
n
d
In
s
o
mn
i
a
c
Pa
ti
e
n
ts
Ba
se
d
o
n
A
u
to
m
a
ted
S
l
e
e
p
On
se
t
De
tec
ti
o
n
”
.
In
In
ter
n
a
ti
o
n
a
l
C
o
n
f
e
re
n
c
e
f
o
r
In
n
o
v
a
ti
o
n
in
Bio
m
e
d
ica
l
En
g
in
e
e
rin
g
a
n
d
L
if
e
S
c
ien
c
e
s: I
CIBEL
2
0
1
5
;
2
0
1
5
;
P
u
traja
y
a
,
M
a
la
y
sia
.
[3
]
Ch
ien
C,
P
o
tt
ie
G
J,
“
A
Un
iv
e
r
sa
l
Hy
b
rid
De
c
isio
n
T
re
e
Clas
si
f
ier
De
sig
n
f
o
r
Hu
m
a
n
”,
In
3
4
t
h
A
n
n
u
a
l
In
ter
n
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
f
th
e
IEE
E
EM
B
S
;
2
0
1
2
;
S
a
n
Die
g
o
,
USA
.
[4
]
W
a
n
g
Y
,
L
i
P
f
,
T
ian
Y,
Re
n
Jj,
L
i
Js
,
“
A
S
h
a
re
d
De
c
isio
n
M
a
k
in
g
S
y
st
e
m
f
o
r
Dia
b
e
tes
M
e
d
ica
ti
o
n
Ch
o
ice
Util
izi
n
g
El
e
c
tro
n
ic He
a
lt
h
Re
c
o
r
d
Da
ta
”
,
EE
E
J
o
u
rn
a
l
o
f
Bi
o
me
d
ica
l
a
n
d
He
a
lt
h
I
n
fo
rm
a
ti
c
s
.
2
0
1
6
;
p
p
(
9
9
):
1
-
1.
[5
]
Ko
n
d
a
S
,
Ba
lm
u
ri
KR,
Ba
sire
d
d
y
RR,
M
o
g
il
i
R,
“
H
y
b
rid
A
p
p
ro
a
c
h
f
o
r
P
re
d
icti
o
n
o
f
Ca
rd
io
v
a
sc
u
lar
Dise
a
s
e
Us
in
g
Clas
s
A
ss
o
c
iatio
n
Ru
les
a
n
d
M
L
P
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
El
e
c
trica
l
a
n
d
Co
mp
u
ter
En
g
in
e
e
rin
g
.
2
0
1
6
;
6
(4
):
1
8
0
0
.
[6
]
Bo
ris
M
il
o
v
ic
,
M
il
a
n
M
il
o
v
ic
,
“
P
re
d
ictio
n
a
n
d
De
c
isio
n
M
a
k
in
g
i
n
He
a
lt
h
Ca
re
u
sin
g
Da
ta
M
in
i
n
g
”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
P
u
b
li
c
He
a
l
th
S
c
ien
c
e
,
De
c
e
m
b
e
r
2
0
1
2
;
1
(
2
):
6
9
-
78.
[7
]
P
o
ll
e
tt
i
n
i
JT
,
P
a
n
ico
S
RG
,
Da
n
e
lu
z
z
i
JC,
T
in
ó
s
R,
Ba
ra
n
a
u
sk
a
s
J
A
,
M
a
c
e
d
o
AA
,
“
Us
in
g
M
a
c
h
in
e
L
e
a
rn
in
g
Clas
sif
ier
s to
A
ss
ist
H
e
a
lt
h
c
a
re
-
R
e
late
d
De
c
isio
n
s: Cl
a
ss
i
f
ica
ti
o
n
o
f
El
e
c
tro
n
ic P
a
ti
e
n
t
Re
c
o
rd
s
”,
J
o
u
rn
a
l
o
f
M
e
d
ica
l
S
y
ste
ms
,
2
0
1
2
;
3
6
:
3
8
6
1
-
3
8
7
4
.
[8
]
He
r
lan
d
M
,
Kh
o
sh
g
o
f
taa
r
T
M
,
W
a
ld
R,
“
A
re
v
ie
w
o
f
d
a
ta
m
in
in
g
u
sin
g
b
ig
d
a
ta
in
h
e
a
lt
h
i
n
f
o
rm
a
ti
c
s”
,
J
o
u
rn
a
l
o
f
Bi
g
Da
ta
,
2
0
1
4
;
1
:2
.
[9
]
T
a
k
e
d
a
F
,
T
a
m
i
y
a
N,
No
g
u
c
h
i
H,
M
o
n
m
a
T,
“
R
e
latio
n
b
e
tw
e
e
n
M
e
n
tal
He
a
lt
h
S
ta
tu
s
a
n
d
P
sy
c
h
o
so
c
ial
S
tres
so
rs
a
m
o
n
g
P
re
g
n
a
n
t
a
n
d
P
u
e
rp
e
ri
u
m
W
o
m
e
n
in
Ja
p
a
n
:
F
r
o
m
th
e
P
e
rsp
e
c
ti
v
e
o
f
W
o
rk
in
g
S
tatu
s”
,
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
P
u
b
li
c
He
a
lt
h
S
c
ien
c
e
,
2
0
1
2
;
1(
2
):
3
7
-
4
8
.
[1
0
]
M
il
o
v
ic
B,
M
il
o
v
ic
M
,
“
P
re
d
icti
o
n
a
n
d
De
c
isio
n
M
a
k
in
g
in
He
a
lt
h
Ca
re
u
sin
g
Da
ta
M
in
i
n
g
”
,
I
n
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
P
u
b
li
c
He
a
lt
h
S
c
ien
c
e
,
2
0
1
2
;
1
(2
):
6
9
-
7
8
.
[1
1
]
Ch
a
w
la
NV
.
Da
ta
M
in
in
g
f
o
r
I
m
b
a
lan
c
e
d
D
a
tas
e
ts:
a
n
Ov
e
rv
ie
w
.
In
Da
t
a
M
in
in
g
a
n
d
Kn
o
w
led
g
e
Disc
o
v
e
r
y
Ha
n
d
b
o
o
k
,
S
p
rin
g
e
r
US;
2
0
0
5
,
8
5
3
-
8
6
7
.
[1
2
]
W
e
iss
G
M
.
,
“
M
in
in
g
w
i
th
Ra
rit
y
:
A
Un
ify
in
g
F
ra
m
e
w
o
r
k
,
A
CM
S
IG
KD
D
Ex
p
lo
ra
ti
o
n
s
Ne
w
sle
tt
e
r
-
S
p
e
c
ial
issu
e
o
n
le
a
rn
in
g
f
ro
m
i
m
b
a
lan
c
e
d
d
a
tas
e
ts”
,
Ju
n
e
2
0
0
4
;
6
(
1
):
7
-
1
9
.
[1
3
]
Ja
p
k
o
w
ic
z
N,
“
T
h
e
Cla
ss
Imb
a
la
n
c
e
Pro
b
le
m:
S
i
g
n
i
fi
c
a
n
c
e
a
n
d
S
tra
teg
ies
”
,
I
n
th
e
2
0
0
0
In
tern
a
t
io
n
a
l
Co
n
f
e
re
n
c
e
o
n
A
rti
f
icia
l
In
telli
g
e
n
c
e
[
ICA
I]
;
2
0
0
0
;
L
a
s V
e
g
a
s,
USA
.
[1
4
]
S
a
it
o
T
,
Re
h
m
s
m
e
ier
M
,
“
T
h
e
P
re
c
isio
n
-
Re
c
a
ll
P
lo
t
Is
M
o
re
I
n
f
o
rm
a
ti
v
e
th
a
n
th
e
ROC
P
l
o
t
W
h
e
n
Ev
a
lu
a
ti
n
g
Bin
a
ry
Clas
si
f
ier
s o
n
Im
b
a
lan
c
e
d
Da
tas
e
ts”
,
P
L
o
S
ON
E
1
0
(3
):
e
0
1
1
8
4
3
2
.
d
o
i
:
1
0
.
1
3
7
1
/j
o
u
rn
a
l.
p
o
n
e
.
0
1
1
8
4
3
2
[1
5
]
L
u
sa
L
,
Bla
g
u
s
R,
“
T
h
e
c
la
ss
-
imb
a
l
a
n
c
e
p
r
o
b
lem
f
o
r
h
i
g
h
-
d
ime
n
s
io
n
a
l
c
l
a
ss
p
re
d
ictio
n
”
,
i
n
2
0
1
2
1
1
th
I
n
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
rn
in
g
a
n
d
A
p
p
li
c
a
ti
o
n
s;
2
0
1
2
.
[1
6
]
He
m
p
sta
lk
K,
F
ra
n
k
E,
W
it
ten
IH,
“
On
e
-
c
las
s
Clas
sif
ic
a
ti
o
n
b
y
Co
m
b
in
in
g
De
n
sity
a
n
d
Cl
a
ss
P
ro
b
a
b
il
it
y
Esti
m
a
ti
o
n
”
,
M
a
c
h
in
e
L
e
a
rn
in
g
a
n
d
K
n
o
w
led
g
e
Disc
o
v
e
r
y
in
Da
ta
b
a
se
s,
2
0
0
8
;
5
2
1
1
:
5
0
5
-
5
1
9
.
[1
7
]
T
a
n
P
n
,
S
tei
n
b
a
c
h
M
,
K
u
m
a
r
V
.
I
n
tro
d
u
c
ti
o
n
to
Da
ta M
i
n
i
n
g
:
P
e
a
rso
n
P
u
b
l
ica
ti
o
n
;
2
0
1
4
.
[1
8
]
Jo
sh
i
M
V
,
A
g
a
r
w
a
l
RC,
Ku
m
a
r
V
,
“
M
in
in
g
Ne
e
d
les
in
a
Ha
y
st
a
c
k
:
Cla
ss
if
y
in
g
Ra
re
Cl
a
ss
e
s
v
ia
T
wo
-
Ph
a
se
Ru
le
In
d
u
c
ti
o
n
”
, i
n
t
h
e
2
0
0
1
A
CM
S
I
G
M
OD
in
tern
a
ti
o
n
a
l
c
o
n
f
e
re
n
c
e
o
n
M
a
n
a
g
e
m
e
n
t
o
f
d
a
ta;
2
0
0
1
;
Ne
w
Yo
rk
,
US
A
.
[1
9
]
Jo
sh
i
M
V,
A
g
a
r
w
a
l
RC,
Ku
m
a
r
V
,
“
Pre
d
icti
n
g
Ra
re
Cl
a
ss
e
s:
Ca
n
B
o
o
sti
n
g
M
a
k
e
An
y
W
e
a
k
L
e
a
r
n
e
r
S
tro
n
g
?”
,
i
n
th
e
e
ig
h
th
A
CM
S
IG
KD
D
in
tern
a
ti
o
n
a
l
c
o
n
f
e
re
n
c
e
o
n
Kn
o
w
led
g
e
d
isc
o
v
e
r
y
a
n
d
d
a
ta
m
in
in
g
;
2
0
0
2
;
Ne
w
Yo
rk
,
USA
.
[2
0
]
Jo
sh
i
M
V
,
Ku
m
a
r
V
,
“
CRE
DO
S
:
Cla
ss
if
ica
ti
o
n
u
si
n
g
R
ip
p
le
d
o
wn
S
tru
c
t
u
re
[
A
Ca
se
fo
r
Ra
re
Cla
ss
e
s
]
”
,
In
t
h
e
2
0
0
4
S
IA
M
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
Da
ta M
in
i
n
g
;
2
0
0
4
;
F
l
o
ri
d
a
,
USA
.
[2
1
]
Ditt
m
a
n
DJ
,
Kh
o
sh
g
o
f
taa
r
T
M
,
Ra
n
d
a
ll
W
a
ld
,
Na
p
o
li
tan
o
A
,
“
Co
m
p
a
riso
n
o
f
Da
ta
S
a
m
p
li
n
g
A
p
p
ro
a
c
h
e
s
f
o
r
Im
b
a
lan
c
e
d
Bio
in
f
o
rm
a
ti
c
s
Da
ta
”
,
In
th
e
T
w
e
n
t
y
-
S
e
v
e
n
th
I
n
t
e
r
n
a
ti
o
n
a
l
F
l
o
ri
d
a
A
rti
f
icia
l
In
telli
g
e
n
c
e
Re
se
a
r
c
h
S
o
c
iety
Co
n
f
e
re
n
c
e
;
2
0
1
4
;
F
lo
r
id
a
.
[2
2
]
Blag
u
s
R,
L
u
sa
L
,
“
S
M
OT
E
f
o
r
h
ig
h
-
d
im
e
n
sio
n
a
l
c
las
s
-
im
b
a
lan
c
e
d
d
a
ta”
,
BM
C
Bi
o
in
f
o
rm
a
ti
c
s,
2
0
1
3
;
1
4
:
1
0
6
.
[
2
3
]
Da
v
is
J,
G
o
a
d
rich
M
,
“
T
h
e
Re
latio
n
sh
ip
b
e
tw
e
e
n
P
re
c
isi
o
n
-
Re
c
a
ll
a
n
d
ROC
Cu
rv
e
s”
,
In
2
7
rd
In
tern
a
ti
o
n
a
l
Co
n
f
e
re
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
rn
in
g
;
2
0
0
6
;
P
it
tsb
u
rg
h
,
USA
.
[2
4
]
Ja
n
e
z
De
m
sa
r,
“
S
tatisti
c
a
l
Co
m
p
a
riso
n
s
o
f
Clas
sif
iers
o
v
e
r
M
u
lt
ip
le
Da
ta
S
e
ts”
,
J
o
u
rn
a
l
o
f
M
a
c
h
in
e
L
e
a
rn
i
n
g
Res
e
a
rc
h
,
2
0
0
6
;
7
:
1
-
30.
[2
5]
F
ra
n
k
E,
Ha
ll
M
A
,
W
it
ten
IH,
“
T
h
e
W
EK
A
W
o
rk
b
e
n
c
h
.
On
li
n
e
A
p
p
e
n
d
ix
f
o
r
"
Da
ta
M
in
in
g
:
P
r
a
c
ti
c
a
l
M
a
c
h
in
e
L
e
a
rn
in
g
T
o
o
ls
a
n
d
T
e
c
h
n
iq
u
e
s”
,
M
o
rg
a
n
Ka
u
fm
a
n
n
;
2
0
1
6
[
c
it
e
d
2
0
1
6
M
a
y
0
1
,
Av
a
il
a
b
le
f
ro
m
:
h
tt
p
:
//
ww
w
.
c
s.
w
a
ik
a
to
.
a
c
.
n
z
/m
l/
w
e
k
a
/.
Evaluation Warning : The document was created with Spire.PDF for Python.