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.
8
,
No
.
1
,
Feb
r
u
ar
y
201
8
,
p
p
.
3
3
3
~3
4
3
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v8
i
1
.
p
p
3
3
3
-
3
4
3
333
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
co
r
e
.
co
m/
jo
u
r
n
a
ls
/in
d
ex
.
p
h
p
/
I
JE
C
E
An Influ
ence
of
M
ea
sure
m
e
nt
S
ca
le of P
redic
tor Va
ria
ble on
Lo
g
istic Regress
i
o
n Mo
deling
and
Lea
rning
Vect
o
r
Q
unti
z
a
tion
M
o
deling
f
o
r Obj
ect
Cla
ss
ifi
ca
tion
Wa
eg
o
H
a
di Nug
ro
ho
1
,
Sa
m
ing
un
H
a
nd
o
y
o
2
,
Yus
nita
J
uly
a
rni A
k
r
i
3
1
,
2
De
p
a
rtem
e
n
t
o
f
S
tatisti
c
s
,
F
a
c
u
lt
y
o
f
S
c
ien
c
e
s,
Un
iv
e
rsitas
Bra
w
i
ja
y
a
M
a
lan
g
,
In
d
o
n
e
sia
3
De
p
a
rtme
n
t
o
f
M
id
w
if
e
r
y
Ed
u
c
a
to
rs
,
T
rib
h
u
w
a
n
a
T
u
n
g
g
a
d
e
w
i
Un
iv
e
rsit
y
,
M
a
lan
g
,
In
d
o
n
e
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Sep
2
0
,
2
0
1
7
R
ev
i
s
ed
J
an
2
,
2
0
1
8
A
cc
ep
ted
J
an
1
6
,
2
0
1
8
M
u
c
h
re
a
l
w
o
rld
d
e
c
isio
n
m
a
k
in
g
is
b
a
se
d
o
n
b
i
n
a
ry
c
a
te
g
o
ries
o
f
in
f
o
rm
a
ti
o
n
th
a
t
a
g
re
e
o
r
d
isa
g
re
e
,
a
c
c
e
p
t
o
r
re
jec
t,
su
c
c
e
e
d
o
r
f
a
il
a
n
d
so
o
n
.
In
f
o
rm
a
ti
o
n
o
f
th
is
c
a
teg
o
ry
is
t
h
e
o
u
tp
u
t
o
f
a
c
las
si
f
ica
ti
o
n
m
e
t
h
o
d
th
a
t
is
th
e
d
o
m
a
in
o
f
sta
ti
stica
l
f
ield
stu
d
ies
(e
g
L
o
g
isti
c
Re
g
re
s
sio
n
m
e
th
o
d
)
a
n
d
m
a
c
h
in
e
lea
rn
in
g
(e
g
L
e
a
rn
in
g
V
e
c
to
r
Qu
a
n
ti
z
a
ti
o
n
(L
V
Q)).
T
h
e
in
p
u
t
a
rg
u
m
e
n
t
o
f
a
c
las
si
f
ica
ti
o
n
m
e
t
h
o
d
h
a
s
a
v
e
ry
c
ru
c
ial
ro
le
to
t
h
e
re
su
lt
in
g
o
u
t
p
u
t
c
o
n
d
it
io
n
.
T
h
is
p
a
p
e
r
i
n
v
e
stig
a
ted
th
e
in
f
lu
e
n
c
e
o
f
v
a
rio
u
s
t
y
p
e
s
o
f
in
p
u
t
d
a
ta
m
e
a
su
re
m
e
n
t
(in
terv
a
l
,
ra
ti
o
,
a
n
d
n
o
m
in
a
l)
to
th
e
p
e
rf
o
rm
a
n
c
e
o
f
lo
g
isti
c
re
g
re
ss
io
n
m
e
th
o
d
a
n
d
L
V
Q
in
c
las
sify
in
g
a
n
o
b
jec
t.
L
o
g
isti
c
re
g
re
ss
io
n
m
o
d
e
li
n
g
is
d
o
n
e
in
se
v
e
ra
l
sta
g
e
s
u
n
ti
l
a
m
o
d
e
l
th
a
t
m
e
e
ts
th
e
su
it
a
b
il
it
y
m
o
d
e
l
tes
t
is
o
b
tain
e
d
.
M
o
d
e
li
n
g
o
n
L
V
Q
w
a
s
tes
ted
o
n
se
v
e
ra
l
c
o
d
e
b
o
o
k
siz
e
s an
d
se
lec
ted
th
e
m
o
st o
p
ti
m
a
l
L
V
Q
m
o
d
e
l.
T
h
e
b
e
st
m
o
d
e
l
o
f
e
a
c
h
m
e
th
o
d
c
o
m
p
a
re
d
to
i
ts
p
e
r
f
o
r
m
a
n
c
e
o
n
o
b
jec
t
c
las
sif
ic
a
ti
o
n
b
a
se
d
o
n
Hit
Ra
ti
o
i
n
d
ica
to
r.
I
n
l
o
g
isti
c
re
g
re
ss
io
n
m
o
d
e
l
o
b
tai
n
e
d
2
m
o
d
e
l
s
th
a
t
m
e
e
t
th
e
m
o
d
e
l
su
it
a
b
il
it
y
te
st
is
a
m
o
d
e
l
w
it
h
p
re
d
ictiv
e
v
a
riab
les
s
c
a
l
e
d
in
terv
a
l
a
n
d
n
o
m
in
a
l,
w
h
il
e
in
L
V
Q
m
o
d
e
li
n
g
o
b
tai
n
e
d
3
p
iec
e
s
o
f
th
e
m
o
st
o
p
ti
m
a
l
m
o
d
e
l
w
it
h
a
d
iff
e
r
e
n
t
c
o
d
e
b
o
o
k
.
In
th
e
d
a
ta
w
it
h
in
terv
a
l
-
sc
a
le
p
re
d
icto
r
v
a
riab
le,
th
e
p
e
rf
o
r
m
a
n
c
e
o
f
b
o
t
h
m
e
th
o
d
s
is
th
e
sa
m
e
.
T
h
e
p
e
rfo
rm
a
n
c
e
o
f
b
o
t
h
m
o
d
e
ls
is
j
u
st
a
s
b
a
d
w
h
e
n
th
e
d
a
ta
h
a
v
e
th
e
p
re
d
icto
r
v
a
riab
les
o
f
th
e
n
o
m
in
a
l
sc
a
le.
In
th
e
d
a
ta
w
it
h
p
re
d
icto
r
v
a
riab
le
h
a
s
ra
ti
o
sc
a
le
,
th
e
L
V
Q
m
e
th
o
d
a
b
le
to
p
ro
d
u
c
e
m
o
d
e
ra
te
e
n
o
u
g
h
p
e
rf
o
rm
a
n
c
e
,
w
h
il
e
o
n
lo
g
isti
c
re
g
re
ss
io
n
m
o
d
e
li
n
g
is
n
o
t
o
b
tai
n
e
d
th
e
m
o
d
e
l
th
a
t
m
e
e
t
m
o
d
e
l
su
it
a
b
il
it
y
tes
t.
T
h
u
s
if
th
e
in
p
u
t
d
a
tas
e
t
h
a
s
in
terv
a
l
o
r
ra
ti
o
-
sc
a
le
p
re
d
icto
r
v
a
riab
les
th
a
n
it
is
p
re
f
e
ra
b
le
to
u
se
th
e
L
V
Q
m
e
th
o
d
f
o
r
m
o
d
e
li
n
g
th
e
o
b
jec
t
c
las
si
f
ica
ti
o
n
.
K
ey
w
o
r
d
:
L
o
g
i
s
tic
R
eg
r
es
s
io
n
L
VQ
Mo
d
el
p
er
f
o
r
m
an
ce
Ob
j
ec
t c
lass
if
icatio
n
P
r
ed
icto
r
v
ar
iab
le
s
ca
le
Co
p
y
rig
h
t
©
2
0
1
8
In
stit
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
:
W
ae
g
o
Had
i N
u
g
r
o
h
o
,
Dep
ar
te
m
en
t o
f
Stati
s
tics
,
Facu
lt
y
o
f
Sc
ien
ce
s
,
Un
i
v
er
s
ita
s
B
r
a
w
ij
a
y
a
,
J
ln
.
Vete
r
an
Ma
lan
g
6
5
1
4
5
,
E
ast J
av
a
I
n
d
o
n
esia
.
E
m
ail:
w
h
n
@
u
b
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
in
p
u
t
d
ata
f
o
r
th
e
cla
s
s
i
f
icatio
n
o
r
ca
teg
o
r
izatio
n
o
f
a
n
o
b
j
ec
t
ca
n
b
e
eith
er
an
i
m
ag
e
o
r
an
attr
ib
u
te.
B
ef
o
r
e
b
ein
g
u
s
ed
as
in
p
u
t,
t
h
e
i
m
a
g
e
m
u
s
t
g
o
th
r
o
u
g
h
a
p
r
ep
r
o
ce
s
s
in
g
s
t
ag
e
ca
lled
f
ea
t
u
r
e
ex
tr
ac
tio
n
.
T
h
er
e
ar
e
t
w
o
p
r
o
ce
s
s
es
a
t
t
h
e
p
h
a
s
e
o
f
f
ea
t
u
r
e
s
ex
tr
ac
tio
n
;
T
ex
t
u
r
e
f
ea
t
u
r
e
e
x
tr
ac
tio
n
a
n
d
s
h
ap
e
f
ea
t
u
r
e
ex
tr
ac
tio
n
[
1
]
,
[
2
]
.
T
h
e
m
et
h
o
d
o
f
ca
te
g
o
r
izin
g
o
r
class
i
f
y
in
g
a
n
o
b
j
ec
t
b
ased
o
n
t
h
e
f
ea
tu
r
e
s
o
r
attr
ib
u
tes
at
tach
ed
to
t
h
e
o
b
j
ec
t
is
a
s
t
u
d
y
i
n
t
h
e
f
ield
o
f
s
t
atis
tics
an
d
m
ac
h
i
n
e
lear
n
i
n
g
.
A
p
p
lied
f
r
o
m
t
h
e
m
et
h
o
d
is
v
er
y
w
id
e
in
v
ar
io
u
s
asp
ec
ts
o
f
lif
e
t
h
at
co
m
e
f
r
o
m
th
e
f
ield
o
f
ex
ac
t,
en
g
in
ee
r
i
n
g
,
an
d
s
o
cial.
T
h
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
1
,
Feb
r
u
ar
y
2018
:
3
3
3
–
3
4
3
334
m
o
s
t
cr
u
cial
th
in
g
is
th
e
i
n
f
o
r
m
atio
n
r
es
u
lti
n
g
f
r
o
m
th
e
clas
s
if
ica
tio
n
u
s
ed
as
th
e
b
asi
s
f
o
r
d
ec
is
io
n
-
m
ak
in
g
o
r
p
o
licy
.
A
lt
h
o
u
g
h
t
h
e
attr
ib
u
te
s
o
f
t
h
e
o
b
j
ec
t
ar
e
v
er
y
w
id
e
in
s
co
p
e,
th
e
y
ar
e
eit
h
er
ta
n
g
ib
le
o
r
i
n
ta
n
g
ib
le,
o
b
s
er
v
ab
le
o
r
u
n
o
b
s
er
v
ab
le.
Ho
w
ev
er
,
i
n
o
r
d
er
to
a
p
p
l
y
a
m
et
h
o
d
o
f
class
i
f
icat
io
n
,
it
is
n
ec
es
s
ar
y
to
ch
ar
ac
ter
ize
th
e
o
b
j
ec
ts
ar
r
an
g
ed
in
a
p
ar
ticu
lar
d
ata
s
tr
u
ct
u
r
e.
T
h
e
m
o
s
t
co
m
m
o
n
l
y
u
s
ed
d
ata
s
tr
u
ctu
r
e
as
an
in
p
u
t a
r
g
u
m
e
n
t o
f
a
clas
s
i
f
icat
io
n
m
et
h
o
d
is
an
i
n
p
u
t
-
o
u
tp
u
t
p
air
.
I
n
th
e
s
tati
s
tical
f
ie
ld
,
th
e
d
at
a
in
t
h
e
i
n
p
u
t
-
o
u
tp
u
t
p
air
f
o
r
m
at
s
h
o
u
ld
b
e
co
n
s
id
er
ed
as
a
ca
u
s
al
it
y
w
h
er
e
t
h
e
s
et
o
f
o
b
s
er
v
ed
v
alu
es
in
t
h
e
i
n
p
u
t
attr
ib
u
te
w
il
l
d
eter
m
in
e
th
e
o
b
s
er
v
ed
v
alu
e
o
f
th
e
o
u
tp
u
t
attr
ib
u
te
[
3
]
.
T
h
e
attr
ib
u
tes
o
f
an
o
b
j
ec
t
th
at
h
as
a
s
p
ec
ial
p
r
o
p
er
ty
th
at
h
as
a
u
n
iq
u
e
o
r
s
in
g
le
v
al
u
e
o
n
th
e
o
b
j
ec
t
ar
e
k
n
o
w
n
b
y
th
e
ter
m
v
ar
iab
le.
W
h
en
an
o
b
j
ec
t
is
o
b
s
er
v
ed
o
n
th
e
b
asi
s
o
f
5
v
a
r
iab
les,
it
w
ill
g
et
5
o
b
s
er
v
atio
n
v
al
u
es
as
s
o
ciate
d
w
i
th
t
h
e
o
b
j
ec
t.
T
h
ese
t
y
p
e
s
o
f
o
b
s
er
v
at
io
n
al
v
al
u
e
s
h
av
e
4
t
y
p
e
s
o
f
m
ea
s
u
r
e
m
e
n
t
s
ca
les
:
n
o
m
i
n
al,
o
r
d
in
al,
in
ter
v
al,
an
d
r
atio
.
T
h
is
t
y
p
e
o
f
m
ea
s
u
r
e
m
e
n
t
s
c
ale
in
s
tati
s
tics
w
ill
g
r
ea
tl
y
i
n
f
lu
e
n
c
e
th
e
s
e
lectio
n
o
f
th
e
m
o
s
t
s
u
itab
le
an
al
y
ti
ca
l
m
eth
o
d
s
.
S
u
p
p
o
s
e
th
at
if
th
e
o
u
tp
u
t
v
ar
iab
les
(
v
ar
iab
les
a
f
f
ec
ted
b
y
t
h
e
in
p
u
t
v
ar
iab
le
s
)
ar
e
n
o
m
i
n
al
o
r
o
r
d
in
al,
th
e
n
t
h
e
s
tati
s
tical
m
o
d
elin
g
f
o
r
t
h
e
class
i
f
icatio
n
o
f
o
b
j
ec
ts
is
lo
g
is
tic
r
e
g
r
ess
io
n
[
4
]
,
[
5
]
.
On
th
e
o
th
er
h
a
n
d
,
t
h
e
m
ac
h
i
n
e
lea
r
n
in
g
m
et
h
o
d
d
o
es
n
o
t
r
eq
u
ir
e
a
ca
u
s
alit
y
b
et
w
ee
n
t
h
e
i
n
p
u
t
-
o
u
tp
u
t
v
ar
iab
les
a
n
d
also
d
o
es
n
o
t
co
n
ce
r
n
t
h
e
t
y
p
e
o
f
m
ea
s
u
r
e
m
e
n
t
s
ca
le
in
t
h
e
o
u
tp
u
t v
ar
iab
le
[
6
]
,
[
7
]
.
Han
d
a
n
d
He
n
le
y
[
8
]
h
a
v
e
r
e
v
ie
w
ed
t
h
e
m
et
h
o
d
s
u
s
ed
i
n
o
b
j
ec
t
class
if
ica
tio
n
.
T
h
e
y
co
n
clu
d
ed
th
a
t
th
e
clas
s
if
icatio
n
m
et
h
o
d
s
w
h
i
ch
ar
e
ea
s
y
to
u
n
d
er
s
tan
d
(
s
u
c
h
as
r
eg
r
es
s
io
n
,
n
ea
r
es
t
n
ei
g
h
b
o
u
r
an
d
tr
ee
-
b
ased
m
et
h
o
d
s
)
ar
e
m
u
ch
m
o
r
e
ap
p
ea
lin
g
,
b
o
th
to
u
s
er
s
a
n
d
to
clie
n
ts
,
t
h
a
n
ar
e
m
e
th
o
d
s
w
h
ic
h
a
r
e
ess
en
tiall
y
b
lack
b
o
x
es
(
s
u
c
h
as
A
r
ti
f
icial
Ne
u
r
al
Net
w
o
r
k
)
.
T
h
e
y
also
p
er
m
it
m
o
r
e
r
ea
d
y
e
x
p
lan
at
io
n
s
o
f
th
e
s
o
r
t
o
f
r
ea
s
o
n
s
w
h
y
t
h
e
m
eth
o
d
s
h
a
v
e
r
ea
ch
ed
th
eir
d
ec
is
io
n
s
.
Me
a
n
w
h
il
e
Dr
eiseitl
a
n
d
O
h
n
o
-
Ma
c
h
a
d
o
[
9
]
s
am
p
led
7
2
p
ap
er
s
co
m
p
ar
i
n
g
b
o
th
lo
g
is
ti
c
r
eg
r
ess
io
n
a
n
d
n
e
u
r
al
n
et
wo
r
k
m
o
d
els
o
n
m
ed
ical
d
ata
s
ets.
T
h
e
y
a
n
al
y
ze
d
th
ese
p
ap
er
s
w
i
th
r
esp
ec
t
to
s
e
v
er
al
cr
iter
ia,
s
u
c
h
as
th
e
s
ize
o
f
d
ata
s
et
s
,
m
o
d
el
p
ar
a
m
e
ter
,
s
elec
tio
n
s
c
h
e
m
e,
an
d
p
er
f
o
r
m
a
n
ce
m
ea
s
u
r
e
u
s
e
d
in
r
ep
o
r
tin
g
m
o
d
el
r
esu
lt
s
.
T
h
ey
s
a
id
th
at
w
h
er
e
p
er
f
o
r
m
an
ce
w
as
co
m
p
ar
ed
s
tatis
t
icall
y
,
t
h
er
e
w
as a
5
:2
r
atio
o
f
ca
s
es i
n
w
h
ic
h
it
w
a
s
n
o
t sig
n
i
f
ica
n
tl
y
b
etter
to
u
s
e
n
e
u
r
al
n
et
w
o
r
k
s
.
P
er
f
o
r
m
a
n
ce
i
m
p
r
o
v
e
m
e
n
t
o
f
lo
g
i
s
tic
r
eg
r
e
s
s
io
n
m
o
d
el
o
n
m
icr
o
ar
r
a
y
d
ata
w
it
h
th
e
B
ay
e
s
ian
ap
p
r
o
ac
h
to
g
en
e
s
elec
t
io
n
a
n
d
cl
ass
i
f
icatio
n
u
s
i
n
g
t
h
e
lo
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el.
T
h
e
m
et
h
o
d
ca
n
e
f
f
ec
ti
v
el
y
id
en
ti
f
y
i
m
p
o
r
tan
t
g
e
n
es
co
n
s
i
s
te
n
t
w
it
h
t
h
e
k
n
o
w
n
b
i
o
lo
g
ical
f
in
d
i
n
g
s
w
h
ile
th
e
ac
cu
r
ac
y
o
f
t
h
e
class
i
f
icatio
n
is
also
h
ig
h
[
1
0
]
.
I
n
ad
d
itio
n
,
th
e
p
er
f
o
r
m
an
ce
co
m
p
ar
is
o
n
b
et
w
ee
n
A
N
N
an
d
lo
g
is
t
ic
r
eg
r
ess
io
n
i
n
v
ar
io
u
s
f
ield
s
ar
e
d
o
n
e
b
y
Felici
s
i
m
o
,
et
a
l
[
1
1
]
d
id
Ma
p
p
in
g
la
n
d
s
lid
e
s
u
s
ce
p
tib
ili
t
y
,
M.
Sh
a
f
iee,
et
al
[
1
2
]
d
id
Fo
r
ec
asti
n
g
Sto
ck
R
et
u
r
n
s
in
I
r
an
Sto
ck
E
x
ch
a
n
g
e,
an
d
Ka
m
le
y
S,
et
a
l
[
1
3
]
d
id
Fo
r
ec
asti
n
g
o
f
Sh
ar
e
Ma
r
k
et.
Fr
o
m
v
ar
io
u
s
s
t
u
d
ies,
A
NN
m
eth
o
d
u
s
ed
is
b
ac
k
p
r
o
p
ag
atio
n
o
r
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e.
B
o
th
m
et
h
o
d
s
ar
e
w
id
el
y
u
s
ed
b
ec
au
s
e
it
h
as
a
to
p
o
lo
g
y
an
d
lear
n
in
g
m
et
h
o
d
s
th
at
ar
e
ea
s
y
to
u
n
d
er
s
ta
n
d
.
On
t
h
e
o
t
h
er
h
a
n
d
,
th
er
e
is
also
A
NN
m
et
h
o
d
k
n
o
w
n
as
L
ea
r
n
in
g
Vec
to
r
Qu
a
n
tizatio
n
(
L
VQ)
w
h
ic
h
is
s
till
r
ar
el
y
f
o
u
n
d
ap
p
lied
,
b
ec
au
s
e
th
i
s
m
eth
o
d
h
a
s
a
co
m
p
eti
tiv
e
la
y
er
t
h
at
w
o
r
k
s
u
s
i
n
g
th
e
p
r
i
n
cip
le
o
f
th
e
s
el
f
-
o
r
g
an
izin
g
m
ap
[
1
4
]
,
[
1
5
]
s
o
it
h
as
a
s
tr
u
ct
u
r
e
a
n
d
lear
n
i
n
g
m
eth
o
d
t
h
at
i
s
d
i
f
f
ic
u
lt
to
u
n
d
er
s
to
o
d
.
L
VQ
i
s
a
cl
ass
i
f
icatio
n
m
et
h
o
d
in
w
h
ich
ea
c
h
u
n
it
o
f
o
u
tp
u
t
r
ep
r
esen
ts
a
class
th
at
ca
n
b
e
u
s
ed
f
o
r
g
r
o
u
p
in
g
w
h
er
e
th
e
n
u
m
b
er
o
f
tar
g
et
g
r
o
u
p
s
o
r
class
es is
p
r
e
-
d
eter
m
i
n
ed
.
B
ased
o
n
th
e
ab
o
v
e
ex
p
o
s
u
r
e,
th
is
p
ap
er
w
ill
e
x
a
m
in
e
t
h
e
i
m
p
le
m
e
n
tat
io
n
o
f
lo
g
i
s
tic
r
eg
r
ess
io
n
a
n
d
L
VQ
n
et
w
o
r
k
f
o
r
o
b
j
ec
t
class
i
f
icatio
n
i
n
t
h
r
ee
d
ataset
s
w
it
h
in
p
u
t
v
ar
iab
les
(
p
r
ed
ict
o
r
s
)
w
it
h
d
if
f
er
en
t
m
ea
s
u
r
e
m
e
n
t
s
ca
les,
r
e
s
p
ec
ti
v
el
y
ar
e
in
ter
v
als,
r
atio
s
an
d
n
o
m
i
n
al
f
o
r
d
ata
1
,
d
ata
2
,
an
d
d
ata
3
.
I
n
t
h
e
lo
g
is
tic
r
e
g
r
ess
io
n
m
o
d
eli
n
g
i
s
d
o
n
e
p
a
r
am
eter
es
ti
m
atio
n
s
tag
e,
test
i
n
g
th
e
m
o
d
el
p
ar
a
m
eter
s
,
th
en
te
s
t
t
h
e
g
o
o
d
n
ess
o
f
f
it,
f
i
n
a
ll
y
o
b
tai
n
ed
a
s
u
itab
le
m
o
d
el.
I
n
L
V
Q
n
et
w
o
r
k
m
o
d
elin
g
t
h
e
4
co
d
eb
o
o
k
s
ar
e
test
ed
ie
2
,
1
0
,
3
0
,
an
d
5
0
.
T
h
e
b
est
m
o
d
els
p
r
o
d
u
ce
d
b
y
b
o
th
lo
g
is
tic
r
eg
r
ess
io
n
an
d
L
VQ
n
et
w
o
r
k
s
w
ill
b
e
ev
al
u
ated
f
o
r
class
i
f
icatio
n
u
s
in
g
Hit
R
a
tio
[
1
6
]
,
ie
th
e
p
r
o
p
o
r
tio
n
o
f
s
a
m
p
le
o
b
s
er
v
atio
n
s
th
at
ca
n
b
e
class
i
f
ied
b
y
th
e
cla
s
s
i
f
icatio
n
m
o
d
el
ap
p
r
o
p
r
ia
tel
y
.
I
m
p
le
m
e
n
tatio
n
o
f
b
o
th
m
et
h
o
d
s
u
s
in
g
s
o
f
t
w
ar
e
R
.
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
2
.
1
.
B
ina
ry
L
o
g
is
t
ic
Reg
re
s
s
io
n Ana
ly
s
is
B
in
ar
y
lo
g
i
s
tic
r
eg
r
es
s
io
n
is
a
lo
g
is
tic
r
eg
r
es
s
io
n
w
it
h
r
esp
o
n
s
e
v
ar
iab
le
s
th
at
ar
e
ca
teg
o
r
i
ca
l
v
alu
e
s
o
f
b
in
ar
y
o
r
d
ich
o
to
m
o
u
s
.
T
h
e
v
ar
iab
le
r
esp
o
n
s
e
o
f
B
er
n
o
u
lli's
Y
d
is
tr
ib
u
tes
w
it
h
t
h
e
f
o
l
lo
w
i
n
g
p
r
o
b
ab
ilit
y
f
u
n
ctio
n
s
[
3
]
:
(
)
(
)
(
(
)
)
(
1
)
T
h
e
lo
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el
:
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:
2
0
8
8
-
8708
A
n
I
n
flu
e
n
ce
o
f Mea
s
u
r
eme
n
t
S
ca
le
o
f P
r
ed
icto
r
V
a
r
ia
b
le
o
n
Lo
g
is
tic
R
eg
r
ess
io
n
…
(
W
a
eg
o
Ha
d
i Nu
g
r
o
h
o
)
335
(
2
)
W
h
er
e
:
n
=
N
u
m
b
er
o
f
o
b
s
er
v
atio
n
s
p
=
Nu
m
b
er
o
f
p
r
ed
icto
r
v
ar
iab
les
=
I
n
ter
ce
p
t
=
L
o
g
i
s
tic
r
eg
r
e
s
s
io
n
co
e
f
f
icie
n
t f
r
o
m
th
e
j
th
j
th
p
r
ed
icto
r
v
ar
iab
le
=
T
h
e
v
alu
e
o
f
t
h
e
j
-
th
p
r
ed
icto
r
v
ar
iab
le
o
n
th
e
i
-
th
o
b
s
er
v
a
ti
o
n
.
W
h
ile
th
e
lo
g
it
f
o
r
m
is
:
(
)
(
∑
)
(
3
)
I
n
lo
g
is
t
ic
r
eg
r
ess
io
n
,
th
e
c
o
n
d
itio
n
al
d
is
tr
ib
u
tio
n
p
atter
n
o
f
th
e
r
esp
o
n
s
e
v
ar
iab
le
is
Y
=
π
(
x
i
)
+
ε,
w
h
ich
h
as 2
t
y
p
e
s
o
f
er
r
o
r
:
Y
=
1
th
en
ε
=
1
-
π
(
x
)
w
it
h
p
r
o
b
ab
ilit
y
π
(
x
)
Y
=
0
th
en
ε
=
-
π
(
x
)
w
it
h
a
ch
a
n
ce
o
f
1
-
π
(
x
)
So
th
e
er
r
o
r
d
is
tr
ib
u
tio
n
h
a
s
th
e
m
ea
n
eq
u
a
l
to
ze
r
o
,
v
ar
ian
ce
{π
(
x
)
(
1
-
π
(
x
)
)
}
a
n
d
f
o
llo
w
s
t
h
e
B
in
o
m
ial
d
is
tr
ib
u
tio
n
.
2
.
1
.
1
.
E
s
t
i
m
a
t
io
n o
f
M
o
del P
a
ra
m
et
er
s
T
h
e
m
eth
o
d
to
e
s
ti
m
ate
t
h
e
lo
g
is
t
ic
r
eg
r
es
s
io
n
m
o
d
el
p
ar
am
e
ter
s
i
s
Ma
x
i
m
u
m
L
i
k
eli
h
o
o
d
E
s
ti
m
a
tio
n
(
M
L
E
)
.
T
h
e
m
o
d
e
l
p
ar
am
eter
i
s
esti
m
ated
f
r
o
m
th
e
(
)
v
ec
to
r
,
th
e
v
alu
e
is
o
b
tain
ed
b
y
m
a
x
i
m
izi
n
g
th
e
lik
eli
h
o
o
d
f
u
n
ct
io
n
(
L
(
β))
th
r
o
u
g
h
d
er
iv
atio
n
o
f
its
p
ar
a
m
e
t
er
s
.
T
h
e
lik
eli
h
o
o
d
f
u
n
ctio
n
is
a
j
o
in
t
p
r
o
b
ab
ilit
y
f
u
n
ctio
n
o
f
t
h
e
v
ar
iab
les
an
d
.
T
h
e
p
r
o
b
ab
ilit
y
d
is
tr
ib
u
tio
n
f
u
n
ct
io
n
f
o
r
ea
ch
(
x
i
,y
i
)
,
is
(
)
(
)
(
(
)
)
(
4
)
w
h
er
e,
(
)
(
∑
)
*
(
∑
)
+
A
cc
o
r
d
in
g
to
Ho
s
m
er
a
n
d
L
e
m
es
h
o
w
[
4
]
,
if
in
ter
-
o
b
s
er
v
ati
o
n
s
ar
e
ass
u
m
ed
to
b
e
in
d
ep
en
d
en
t,
t
h
e
lik
eli
h
o
o
d
f
u
n
ctio
n
is
t
h
e
m
u
lt
ip
licatio
n
o
f
ea
ch
p
r
o
b
ab
ilit
y
d
is
tr
ib
u
tio
n
i
n
E
q
u
atio
n
(
4
)
.
(
)
∏
(
)
∏
(
)
(
(
)
)
(
)
[
(
)
]
∑
*
(
(
)
(
)
)
(
(
)
)
+
∑
*
(
∑
)
(
(
∑
)
)
+
∑
(
∑
)
∑
(
(
∑
)
)
Ma
x
lik
elih
o
o
d
is
o
b
tain
ed
b
y
d
er
iv
atin
g
(
)
to
β an
d
eq
u
atin
g
i
t
w
it
h
ze
r
o
.
(
)
∑
∑
[
(
∑
)
(
∑
)
]
∑
∑
(
)
∑
(
(
)
)
(
5
)
Sin
ce
E
q
u
atio
n
(
5
)
is
n
o
n
-
l
in
ea
r
,
th
e
s
o
l
u
tio
n
o
f
th
i
s
eq
u
atio
n
b
ec
o
m
es
d
i
f
f
icu
l
t
to
r
eso
lv
e
an
al
y
ticall
y
,
r
eq
u
ir
in
g
an
i
ter
ativ
e
s
o
l
u
tio
n
s
u
c
h
as t
h
e
Ne
w
t
o
n
-
R
ap
h
s
o
n
m
e
th
o
d
[
5
]
.
(
)
(
∑
)
(
∑
)
i=1
,
2
,
…, n
j=1
,
2
,
…, p
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
1
,
Feb
r
u
ar
y
2018
:
3
3
3
–
3
4
3
336
2
.
1
.
2
.
P
a
ra
m
et
er
Sig
nifica
nce
T
e
s
t
ing
a.
Si
m
u
lta
n
eo
u
s
T
esti
n
g
T
h
e
s
i
m
u
lta
n
eo
u
s
te
s
t
i
s
p
er
f
o
r
m
ed
to
e
x
a
m
i
n
e
t
h
e
r
o
le
o
f
ea
c
h
p
r
ed
icto
r
v
ar
iab
le
i
n
t
h
e
m
o
d
el
s
i
m
u
l
ta
n
eo
u
s
l
y
.
Stati
s
tical
h
y
p
o
th
eses
a
n
d
test
s
ta
tis
t
ics ar
e
as f
o
llo
w
s
[
3
]
:
H
y
p
o
th
es
is
:
H
0
:
v
er
s
u
s
H
1
: a
t
least o
n
e
;
(
(
)
(
)
*
[
(
)
(
)
]
(6
)
w
h
er
e:
w
it
h
;
j
=1
,
2
,
…,
p
w
it
h
o
u
t
;
j
=1
,
2
,
…,
p
T
h
e
G
v
alu
e
is
co
m
p
ar
ed
w
it
h
th
e
s
tati
s
tic
(
)
w
it
h
d
eg
r
ee
s
o
f
f
r
ee
d
o
m
co
r
r
esp
o
n
d
in
g
to
th
e
esti
m
ated
n
u
m
b
er
o
f
p
ar
a
m
ete
r
s
.
H
0
w
ill
b
e
ac
ce
p
ted
i
f
p
-
v
a
lu
e
i
s
g
r
ea
ter
t
h
a
n
th
e
p
r
o
b
ab
i
lit
y
o
f
d
o
in
g
t
y
p
e
I
er
r
o
r
o
f
α
.
b.
P
ar
tial
T
esti
n
g
P
ar
tial
test
in
g
is
u
s
ed
to
te
s
t
t
h
e
ef
f
ec
t
o
f
ea
ch
p
ar
a
m
eter
o
n
th
e
m
o
d
el
i
n
d
iv
id
u
all
y
o
r
s
ep
ar
atel
y
w
it
h
r
eg
ar
d
to
o
th
er
p
ar
am
ete
r
s
.
T
h
e
p
ar
tial
test
r
esu
lts
w
ill
s
h
o
w
w
h
et
h
er
a
p
r
ed
icto
r
v
ar
iab
le
is
elig
ib
le
to
en
ter
t
h
e
m
o
d
el
o
r
n
o
t.
I
f
h
y
p
o
th
esi
test
y
ield
s
s
ig
n
i
f
ican
t,
t
h
en
an
en
ter
in
m
o
d
el.
H
y
p
o
th
esis
:
H
0
:
=
0
v
er
s
u
s
H
1
:
0.
W
ald
s
tatis
tic
test
,
̂
(
̂
)
T
h
e
test
s
tati
s
tic
u
s
ed
i
s
t
h
e
W
ald
test
,
r
ej
ec
t
H
0
if
t
h
e
v
alu
e
o
f
|
|
(
)
o
r
p
-
v
alu
e
<0
.
0
5
,
s
o
it
ca
n
b
e
co
n
clu
d
ed
th
er
e
is
i
n
f
lu
en
ce
b
et
w
ee
n
p
r
ed
icto
r
v
ar
iab
les
w
i
th
r
esp
o
n
s
e
v
ar
iab
les
[
4
]
.
I
n
ad
d
itio
n
to
f
o
llo
w
in
g
t
h
e
n
o
r
m
al
d
is
tr
ib
u
t
io
n
,
th
e
W
ald
s
tati
s
tical
te
s
t
s
q
u
ar
ed
(
)
w
ill
f
o
llo
w
t
h
e
c
h
i
-
s
q
u
ar
e
w
it
h
th
e
d
eg
r
ee
o
f
f
r
ee
d
o
m
o
n
e
[
3
]
.
2
.
1
.
3
.
G
o
o
dn
es
s
o
f
F
it
T
est
Mo
d
el
f
it test
o
n
lo
g
is
tic
r
e
g
r
ess
io
n
,
ca
n
u
s
e
te
s
t sta
tis
tic
ca
ll
ed
g
o
o
d
n
ess
o
f
f
it
test
.
T
h
is
te
s
t stat
is
ti
c
is
u
s
ed
to
f
in
d
o
u
t
h
o
w
b
ig
th
e
ef
f
ec
ti
v
e
n
ess
o
f
t
h
e
m
o
d
el
f
o
r
m
ed
in
ex
p
lai
n
i
n
g
th
e
r
esp
o
n
s
e
v
ar
iab
le,
s
o
th
e
m
o
d
el
ca
n
r
ep
r
ese
n
t
t
h
e
ac
t
u
al
co
n
d
itio
n
r
ep
r
ese
n
ted
b
y
t
h
e
d
ata
u
s
ed
i
n
t
h
e
a
n
al
y
s
i
s
.
T
h
e
h
y
p
o
t
h
esi
s
an
d
te
s
t
s
tatis
t
ic
ar
e
as f
o
llo
w
s
[
3
]
:
H
y
p
o
th
es
is
: H
0
: T
h
e
m
o
d
el
ap
p
r
o
p
r
iate
v
s
H
1
: T
h
e
m
o
d
el
is
n
o
t a
p
p
r
o
p
r
iate.
T
h
e
test
s
tati
s
t
ic
u
s
ed
is
ie
∑
. H
0
is
r
ej
ec
ted
if
(
(
)
)
.
2
.
2
.
Art
if
icia
l N
eura
l N
et
w
o
rk
wit
h Co
m
pet
it
iv
e
L
a
y
er
A
r
ti
f
icial
n
e
u
r
al
n
et
w
o
r
k
s
(
ANN)
w
ith
co
m
p
etiti
v
e
la
y
er
s
h
a
v
e
t
h
r
ee
la
y
er
s
:
t
h
e
i
n
p
u
t
la
y
er
,
t
h
e
h
id
d
en
la
y
er
,
an
d
th
e
o
u
tp
u
t
la
y
er
.
I
n
th
is
ca
s
e
t
h
e
co
m
p
etitiv
e
la
y
er
lies
in
t
h
e
h
id
d
en
lay
er
.
Neu
r
o
n
s
i
n
n
et
w
o
r
k
s
w
it
h
co
m
p
etiti
v
e
la
y
er
s
co
m
p
ete
f
o
r
ac
tiv
e
r
i
g
h
ts
.
On
e
o
f
t
h
e
co
m
p
etit
iv
e
la
y
e
r
n
et
w
o
r
k
m
o
d
el
i
s
L
VQ.
T
h
er
e
ar
e
t
w
o
lear
n
i
n
g
m
et
h
o
d
s
in
A
NN
n
a
m
el
y
s
u
p
er
v
is
ed
lear
n
i
n
g
an
d
u
n
s
u
p
er
v
is
ed
lear
n
in
g
.
I
n
Su
p
er
v
i
s
ed
lear
n
in
g
,
ev
er
y
p
a
tter
n
g
i
v
en
as
i
n
p
u
t
f
o
r
A
NN,
h
as
b
ee
n
k
n
o
w
n
o
u
tp
u
t.
T
h
e
d
if
f
er
e
n
ce
b
et
w
ee
n
th
e
A
N
N
o
u
tp
u
t
a
n
d
t
h
e
d
esir
ed
o
u
tp
u
t
(
tar
g
et)
is
ca
lled
an
er
r
o
r
.
T
h
is
er
r
o
r
q
u
an
tit
y
i
s
u
s
ed
to
co
r
r
e
ct
A
NN
w
ei
g
h
t
s
o
t
h
at
A
N
N
ca
n
p
r
o
d
u
ce
o
u
tp
u
t a
s
clo
s
e
as
p
o
s
s
ib
le
to
k
n
o
w
n
tar
g
e
t p
atter
n
.
A
NN
lear
n
i
n
g
al
g
o
r
ith
m
u
s
i
n
g
th
i
s
m
e
th
o
d
is
Heb
b
ian
,
P
er
ce
p
tr
o
n
,
A
d
alin
e
,
B
o
ltz
m
a
n
,
Ho
p
f
ield
,
B
ac
k
p
r
o
p
ag
atio
n
,
an
d
L
VQ.
On
e
o
f
th
e
ai
m
s
o
f
A
NN
m
o
d
eli
n
g
is
f
o
r
class
i
f
icat
io
n
.
A
cc
o
r
d
in
g
Fau
s
ett
[
6
]
,
th
e
b
asis
o
f
class
i
f
icatio
n
o
n
A
NN
i
s
to
u
s
e
th
e
o
p
ti
m
al
w
ei
g
h
t
o
f
th
e
l
ea
r
n
in
g
p
r
o
ce
s
s
.
T
h
e
w
ei
g
h
ts
ar
e
th
e
a
m
o
u
n
t
s
o
r
v
alu
e
s
t
h
at
e
x
is
t o
n
t
h
e
co
n
n
ec
tio
n
b
et
w
ee
n
n
e
u
r
o
n
s
th
a
t tr
an
s
f
er
d
ata
f
r
o
m
o
n
e
la
y
er
to
an
o
th
er
,
w
h
ic
h
s
er
v
e
s
to
r
eg
u
la
te
t
h
e
n
et
w
o
r
k
s
o
as
t
o
p
r
o
d
u
ce
th
e
d
esire
d
o
u
tp
u
t.
I
n
ad
d
itio
n
to
h
a
v
i
n
g
ad
v
a
n
ta
g
es
th
at
d
o
n
o
t
h
a
v
e
to
m
ee
t
th
e
clas
s
ical
a
s
s
u
m
p
t
io
n
s
a
n
d
v
ar
ian
h
o
m
o
g
e
n
it
y
o
f
er
r
o
r
s
,
ANN
al
s
o
h
a
s
a
wea
k
n
e
s
s
th
at
ta
k
es
lo
n
g
er
tr
ain
in
g
ti
m
e
to
p
er
f
o
r
m
ca
lcu
latio
n
s
in
t
h
e
f
o
r
m
atio
n
o
f
m
o
d
el
s
[
1
3
]
.
L
VQ
i
s
a
cla
s
s
i
f
icatio
n
m
eth
o
d
in
w
h
ic
h
ea
c
h
o
u
tp
u
t
u
n
it
p
r
esen
ts
a
cla
s
s
w
i
th
a
s
p
ec
i
f
ied
tar
g
e
t
class
.
L
VQ
u
s
es
a
s
u
p
er
v
i
s
ed
co
m
p
etiti
v
e
lear
n
i
n
g
alg
o
r
it
h
m
v
er
s
io
n
o
f
th
e
Ko
h
o
n
en
Se
lf
-
Or
g
a
n
izi
n
g
Ma
p
(
SOM)
alg
o
r
ith
m
.
L
VQ
n
e
t
w
o
r
k
ar
ch
itect
u
r
e
ac
co
r
d
in
g
to
K
ask
i
a
n
d
Ko
h
o
n
e
n
[
1
5
]
ca
n
b
e
s
ee
n
i
n
Fi
g
u
r
e
1
.
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:
2
0
8
8
-
8708
A
n
I
n
flu
e
n
ce
o
f Mea
s
u
r
eme
n
t
S
ca
le
o
f P
r
ed
icto
r
V
a
r
ia
b
le
o
n
Lo
g
is
tic
R
eg
r
ess
io
n
…
(
W
a
eg
o
Ha
d
i Nu
g
r
o
h
o
)
337
Fig
u
r
e
1
.
T
h
e
ar
ch
itectu
r
e
L
V
Q
n
et
w
o
r
k
w
h
er
e:
R
=
n
u
m
b
er
o
f
ele
m
e
n
ts
i
n
t
h
e
i
n
p
u
t
v
ec
to
r
S
1
=
th
e
n
u
m
b
er
o
f
co
m
p
etit
iv
e
n
eu
r
o
n
s
S
2
=
n
u
m
b
er
o
f
li
n
ea
r
n
e
u
r
o
n
s
B
ased
o
n
Fi
g
u
r
e
1
.
,
th
e
L
VQ
n
et
w
o
r
k
co
n
s
i
s
ts
o
f
t
h
r
ee
la
y
er
s
:
an
i
n
p
u
t
la
y
er
,
a
co
m
p
eti
tiv
e
la
y
er
,
an
d
an
o
u
tp
u
t
la
y
er
.
L
ea
r
n
i
n
g
in
th
e
co
m
p
etiti
v
e
la
y
er
ai
m
s
to
cla
s
s
i
f
y
ea
ch
i
n
p
u
t
v
ec
to
r
to
a
h
id
d
en
la
y
er
.
W
h
ile
lear
n
i
n
g
o
n
th
e
o
u
tp
u
t
lay
er
ai
m
s
at
tr
an
s
f
o
r
m
in
g
th
e
s
u
b
clas
s
o
n
th
e
co
m
p
et
itiv
e
la
y
er
i
n
to
th
e
class
i
f
icatio
n
o
f
t
h
e
tar
g
e
t
cla
s
s
t
h
at
h
a
s
b
ee
n
e
s
tab
lis
h
ed
.
L
ea
r
n
i
n
g
t
h
e
co
m
p
etiti
v
e
la
y
e
r
as
a
s
u
b
clas
s
a
n
d
lear
n
in
g
f
r
o
m
t
h
e
o
u
tp
u
t
la
y
er
as
t
h
e
tar
g
et
cla
s
s
.
A
cc
o
r
d
in
g
to
P
u
tr
i
(
2
0
1
2
)
,
th
er
e
ar
e
t
w
o
f
ac
to
r
s
th
a
t
in
f
lu
e
n
ce
t
h
e
lear
n
i
n
g
p
r
o
ce
s
s
in
L
VQ
n
a
m
el
y
i
n
itial i
n
itia
liz
atio
n
an
d
tr
ai
n
in
g
r
ate.
Settin
g
d
ata
a
s
i
n
p
u
t
o
n
L
V
Q
n
et
w
o
r
k
m
u
s
t
b
e
i
n
i
n
p
u
t
-
o
u
tp
u
t
p
air
f
o
r
m
at.
I
n
th
is
ca
s
e
th
e
o
u
tp
u
t
d
ata
w
ill
s
er
v
e
as a
tar
g
e
t in
t
h
e
lear
n
in
g
p
r
o
ce
s
s
.
Su
p
p
o
s
e
in
th
e
f
o
r
m
a
t a
s
f
o
llo
w
s
:
{
(
)
(
)
}
q
=
1
,
2
,
.
.
.
,
Q
(7
)
W
h
er
e:
(
)
=
v
ec
to
r
/in
p
u
t
m
a
tr
ix
(
)
=
th
e
o
u
tp
u
t v
ec
to
r
L
VQ
co
n
s
i
s
ts
o
f
a
co
m
p
etiti
v
e
la
y
er
th
at
i
n
cl
u
d
es a
co
m
p
etit
iv
e
s
u
b
n
et
a
n
d
a
lin
ea
r
o
u
tp
u
t
la
y
er
.
I
n
a
co
m
p
eti
tiv
e
la
y
er
,
ea
c
h
n
e
u
r
o
n
is
as
s
ig
n
ed
to
a
cla
s
s
.
Dif
f
er
en
t
n
eu
r
o
n
s
i
n
t
h
e
co
m
p
e
titi
v
e
la
y
er
,
it
is
p
o
s
s
ib
le
to
h
av
e
th
e
s
a
m
e
clas
s
.
E
ac
h
class
is
t
h
en
p
air
ed
w
it
h
o
n
e
o
f
th
e
n
e
u
r
o
n
s
i
n
th
e
o
u
tp
u
t
la
y
er
.
T
h
u
s
t
h
e
n
u
m
b
er
o
f
n
e
u
r
o
n
s
i
n
t
h
e
co
m
p
etit
iv
e
la
y
er
,
at
least
a
s
m
u
c
h
as
t
h
e
n
u
m
b
er
o
f
n
eu
r
o
n
s
i
n
th
e
l
in
e
ar
o
u
tp
u
t
la
y
er
[
1
4
]
.
T
h
e
r
elatio
n
s
h
ip
b
et
w
ee
n
th
e
in
p
u
t
v
ec
to
r
a
n
d
o
n
e
o
f
t
h
e
weig
h
t
v
ec
to
r
s
is
m
ea
s
u
r
ed
b
y
t
h
e
E
u
c
lid
d
is
ta
n
ce
.
A
s
u
b
n
e
t is
u
s
ed
to
f
i
n
d
th
e
s
m
alle
s
t e
le
m
e
n
t i
n
t
h
e
in
p
u
t d
ata.
(
)
[
‖
(
)
‖
‖
(
)
‖
‖
(
)
‖
]
(8
)
An
ele
m
en
t
g
i
v
e
n
v
a
lu
e
1
i
n
d
icate
s
th
at
t
h
e
i
n
p
u
t
v
ec
to
r
b
elo
n
g
s
to
t
h
e
i
n
ten
d
ed
clas
s
,
an
d
an
ele
m
e
n
t
is
as
s
ig
n
ed
a
v
alu
e
o
f
0
if
th
e
in
p
u
t
v
ec
to
r
is
n
o
t
in
clu
d
ed
in
t
h
e
d
esire
d
class
.
T
h
is
ca
n
b
e
r
ep
r
esen
ted
b
y
a
s
u
b
n
et
a
s
a
v
ec
to
r
w
it
h
th
e
f
o
llo
w
in
g
v
ec
to
r
f
u
n
ctio
n
s
:
a
(1)
=
co
m
p
et
(
(
)
)
(9
)
L
i
n
ea
r
o
u
tp
u
t
la
y
er
o
f
L
VQ
n
et
w
o
r
k
,
u
s
ed
to
co
m
b
i
n
e
s
u
b
class
es
i
n
to
a
s
i
n
g
le
clas
s
.
T
h
is
i
s
d
o
n
e
u
s
i
n
g
th
e
w
ei
g
h
t
m
atr
ix
(
)
,
ie
th
e
w
ei
g
h
t
m
atr
i
x
h
av
i
n
g
ele
m
e
n
ts
:
,
(
1
0
)
I
n
ad
d
itio
n
,
th
e
w
ei
g
h
t
m
atr
i
x
(
)
,
o
n
th
e
co
m
p
eti
tiv
e
la
y
er
m
u
s
t
b
e
tr
ain
ed
u
s
i
n
g
t
h
e
Ko
h
o
n
e
n
SOM
r
u
le
as
f
o
llo
w
s
:
A
t
ea
c
h
iter
atio
n
,
ea
ch
tr
ai
n
in
g
v
ec
to
r
is
e
n
ter
ed
in
to
t
h
e
n
e
t
w
o
r
k
a
s
i
n
p
u
t
x
an
d
t
h
e
E
u
cli
d
d
is
tan
c
e
f
r
o
m
th
e
i
n
p
u
t
v
ec
to
r
to
ea
ch
p
r
o
t
o
ty
p
e
v
ec
to
r
(
w
ei
g
h
ted
m
atr
i
x
co
lu
m
n
)
is
ca
lcu
la
ted
.
Neu
r
o
n
j
*
w
in
s
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
1
,
Feb
r
u
ar
y
2018
:
3
3
3
–
3
4
3
338
co
m
p
eti
tio
n
if
E
u
cl
id
's
d
is
ta
n
c
e
b
et
w
ee
n
x
a
n
d
j
*
p
r
o
to
t
y
p
e
v
ec
to
r
is
th
e
s
m
alle
s
t.
T
h
e
ac
t
iv
atio
n
v
a
lu
e
a
(1)
i
s
m
u
ltip
l
ied
b
y
n
(
)
in
its
r
ig
h
t
p
o
s
itio
n
to
o
b
tain
in
p
u
t
n
(2)
.
T
h
e
o
u
tp
u
t
a
(2)
=
n
(2)
,
as
lo
n
g
as
th
e
tr
an
s
f
er
f
u
n
ctio
n
i
n
th
e
o
u
tp
u
t
n
e
u
r
o
n
is
an
id
e
n
tit
y
f
u
n
ctio
n
.
T
h
e
a
(2)
also
h
a
s
o
n
l
y
o
n
e
v
al
u
e
i
n
t
h
e
ele
m
e
n
t
k
*
,
in
d
icati
n
g
t
h
at
t
h
e
in
p
u
t
v
ec
t
o
r
b
elo
n
g
s
to
th
e
clas
s
k
*
.
K
o
h
o
n
en
r
u
les
ar
e
u
s
ed
to
f
ix
th
e
w
ei
g
h
t
s
o
n
t
h
e
h
id
d
en
la
y
er
.
I
f
x
is
co
r
r
ec
tl
y
class
i
f
ied
,
th
e
w
eig
h
t
v
ec
to
r
(
)
is
th
e
w
i
n
n
er
s
o
t
h
at
t
h
e
h
id
d
en
n
e
u
r
o
n
i
s
m
o
v
ed
clo
s
er
to
x
.
(
)
(
(
)
)
if
(
)
(
1
1
)
B
u
t
if
x
i
s
class
i
f
ied
in
co
r
r
ec
tl
y
,
it
is
o
b
v
io
u
s
t
h
at
th
e
w
r
o
n
g
h
id
d
en
n
eu
r
o
n
s
w
i
n
th
e
co
m
p
etitio
n
.
I
n
th
is
ca
s
e,
th
e
w
ei
g
h
t is
m
o
v
ed
a
w
a
y
f
r
o
m
th
e
x
.
(
)
(
(
)
)
if
(
)
(
1
2
)
Af
ter
t
h
e
tr
ain
i
n
g
,
th
e
f
i
n
al
w
ei
g
h
ts
(
w
)
w
ill
b
e
u
s
ed
f
o
r
th
e
n
e
x
t
s
i
m
u
latio
n
,
test
o
r
class
i
f
icatio
n
[
1
5
]
.
2
.
3
.
Acc
ura
cy
o
f
Cla
s
s
if
ica
t
io
n
A
cc
o
r
d
in
g
to
Dia
n
iati
(
2
0
1
3
)
,
p
r
io
r
to
class
i
f
icat
io
n
,
t
h
e
d
ata
is
d
iv
id
ed
i
n
to
t
w
o
d
atasets
.
T
h
e
f
ir
s
t
p
ar
t
is
th
e
tr
ain
in
g
d
ataset
u
s
e
d
to
f
o
r
m
th
e
o
p
ti
m
al
m
o
d
el
o
f
ar
tif
icia
l
n
eu
r
al
n
et
w
o
r
k
s
,
wh
ile
th
e
s
ec
o
n
d
p
ar
t
is
t
h
e
test
in
g
d
ataset
to
test
t
h
e
o
p
ti
m
al
m
o
d
el
o
b
tai
n
ed
f
r
o
m
t
h
e
tr
ai
n
in
g
d
atase
t.
Hair
,
e
t
a
l.
[
5
]
ex
p
lain
s
t
h
e
p
r
in
cip
le
o
f
s
h
ar
i
n
g
t
h
e
m
o
s
t
p
o
p
u
lar
p
r
o
p
o
r
tio
n
o
f
tr
ai
n
i
n
g
-
test
i
n
g
d
ataset
s
i
s
5
0
-
5
0
,
b
u
t
m
o
s
t
r
e
s
ea
r
ch
er
s
also
u
s
e
t
h
e
6
0
-
4
0
o
r
7
5
-
2
5
d
iv
is
io
n
p
r
in
cip
le,
s
in
ce
th
er
e
is
n
o
s
ta
n
d
ar
d
r
u
le
ab
o
u
t
d
i
v
id
i
n
g
th
e
d
ataset.
T
h
e
p
r
ec
is
io
n
in
cla
s
s
i
f
icat
io
n
ca
n
b
e
d
eter
m
in
ed
b
y
ca
lc
u
lat
i
n
g
t
h
e
v
al
u
e
o
f
Hit
R
atio
,
i
e
th
e
p
r
o
p
o
r
tio
n
o
f
o
b
s
er
v
atio
n
al
s
a
m
p
les
t
h
at
ca
n
b
e
class
i
f
ied
b
y
th
e
cla
s
s
i
f
ic
atio
n
f
u
n
ctio
n
[
1
6
]
.
T
h
e
v
alu
e
o
f
Hit
R
atio
ca
n
b
e
ca
lcu
lated
u
s
i
n
g
t
h
e
f
o
llo
w
i
n
g
f
o
r
m
u
la:
(
1
3
)
T
h
e
Hit
R
atio
v
al
u
e
ca
lc
u
late
d
ac
co
r
d
in
g
to
th
e
E
q
u
atio
n
(
1
3
)
s
h
o
w
s
p
er
f
o
r
m
a
n
ce
o
f
cla
s
s
i
f
icatio
n
f
u
n
ctio
n
.
A
b
ig
er
Hit Ra
tio
v
al
u
e
in
d
icate
s
a
b
etter
class
i
f
icat
io
n
m
et
h
o
d
.
3.
RE
S
E
ARCH
M
E
T
H
O
D
3
.
1
.
Da
t
a
Cha
ra
ct
er
is
t
ics
T
h
e
d
ata
u
s
ed
ar
e
s
ec
o
n
d
ar
y
d
ata
o
b
tain
ed
f
r
o
m
t
h
r
ee
p
r
ev
io
u
s
s
t
u
d
ies.
T
h
e
th
r
ee
d
at
asets
h
av
e
d
if
f
er
e
n
t r
esp
o
n
s
e
a
n
d
p
r
ed
icto
r
v
ar
ia
b
les.
T
h
e
ex
p
lan
atio
n
o
f
th
e
d
ata
u
s
ed
is
s
h
o
w
n
i
n
T
ab
le
1
.
T
ab
le
1
.
P
r
ed
icto
r
an
d
R
esp
o
n
s
e
Var
iab
le
o
f
th
e
Data
D
a
t
a
se
t
D
a
t
a
so
u
r
c
e
s
P
r
e
d
i
c
t
o
r
v
a
r
i
a
b
l
e
s
P
r
e
d
i
c
t
o
r
sca
l
e
C
a
t
e
g
o
r
i
c
a
l
r
e
sp
o
n
v
a
r
i
a
b
l
e
r
e
c
o
r
d
D
i
e
t
o
n
t
o
d
d
l
e
r
s
S
a
r
t
i
k
a
[
1
7
]
=
C
a
r
b
o
h
y
d
r
a
t
e
s
=
V
e
g
e
t
a
b
l
e
s
=
S
i
d
e
d
i
s
h
=
F
r
u
i
t
=
M
i
l
k
I
n
t
e
r
v
a
l
B
a
d
d
i
e
t
0
1
0
8
D
i
e
t
i
s e
n
o
u
g
h
1
T
h
e
g
r
a
n
t
i
n
g
o
f
c
r
e
d
i
t
f
o
r
se
a
w
e
e
d
b
u
s
i
n
e
ss
S
u
n
a
d
j
i
[
1
8
]
=
Ex
p
e
r
i
e
n
c
e
=
D
u
r
a
t
i
o
n
o
f
e
d
u
c
a
t
i
o
n
=
Emp
l
o
y
me
n
t
I
n
t
e
n
si
t
y
=
A
g
e
=
S
e
a
w
e
e
d
c
l
e
a
n
l
i
n
e
ss l
e
v
e
l
=
=
S
e
a
w
e
e
d
W
a
t
e
r
C
o
n
t
e
n
t
R
a
si
o
D
o
n
o
t
a
c
c
e
p
t
c
r
e
d
i
t
0
1
3
8
A
c
c
e
p
t
i
n
g
c
r
e
d
i
t
1
F
a
c
t
o
r
s t
h
a
t
i
n
f
l
u
e
n
c
e
t
h
e
i
n
c
i
d
e
n
c
e
o
f
l
o
w
b
i
r
t
h
w
e
i
g
h
t
i
n
f
a
n
t
s
(
L
B
W
)
P
a
n
d
i
n
[
1
9
]
=
M
o
t
h
e
r
A
g
e
=
P
a
r
i
t
y
=
B
i
r
t
h
d
i
s
t
a
n
c
e
=
A
n
e
mi
a
=
N
u
t
r
i
t
i
o
n
S
t
a
t
u
s
=
Ed
u
c
a
t
i
o
n
N
o
mi
n
a
l
C
a
se
o
f
L
B
W
1
96
N
o
t
a
c
a
se
o
f
L
B
W
2
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:
2
0
8
8
-
8708
A
n
I
n
flu
e
n
ce
o
f Mea
s
u
r
eme
n
t
S
ca
le
o
f P
r
ed
icto
r
V
a
r
ia
b
le
o
n
Lo
g
is
tic
R
eg
r
ess
io
n
…
(
W
a
eg
o
Ha
d
i Nu
g
r
o
h
o
)
339
3
.
2
.
Resea
rc
h St
a
g
es
I
n
th
is
s
t
u
d
y
,
t
h
e
f
ir
s
t
s
tep
is
to
d
iv
id
e
ea
ch
d
ata
in
to
2
p
ar
t
s
,
7
0
%
f
o
r
tr
ai
n
i
n
g
a
n
d
3
0
%
f
o
r
test
in
g
d
ataset
,
th
e
n
co
n
t
in
u
ed
lo
g
i
s
ti
c
r
eg
r
ess
io
n
an
a
l
y
s
is
a
n
d
L
V
Q
an
al
y
s
i
s
.
B
in
ar
y
lo
g
i
s
tic
r
e
g
r
ess
io
n
a
n
al
y
s
i
s
w
as
p
er
f
o
r
m
ed
u
s
i
n
g
s
o
f
t
w
ar
e
R
.
T
h
e
s
tep
s
in
b
in
ar
y
lo
g
is
tic
r
eg
r
ess
io
n
an
al
y
s
is
w
er
e
a.
C
h
ec
k
f
o
r
t
h
e
p
r
esen
ce
o
r
ab
s
en
ce
o
f
m
u
l
tico
llin
ea
r
i
t
y
b
et
wee
n
in
d
ep
en
d
e
n
t v
ar
iab
les X
b.
P
ar
am
eter
esti
m
atio
n
c.
T
esti
n
g
p
ar
a
m
eter
s
s
i
m
u
l
tan
eo
u
s
l
y
d.
P
ar
tial p
ar
am
eter
test
e.
T
h
e
estab
lis
h
m
e
n
t o
f
a
lo
g
i
s
tic
r
eg
r
ess
io
n
m
o
d
el
f.
C
h
ec
k
t
h
e
s
u
itab
ilit
y
o
f
t
h
e
m
o
d
el
g.
A
p
p
l
y
b
i
n
ar
y
lo
g
is
tic
r
e
g
r
ess
i
o
n
m
o
d
el
o
b
tain
ed
f
r
o
m
tr
ai
n
i
n
g
d
ata
to
test
i
n
g
d
ata
h.
Fo
r
m
in
g
a
tab
le
o
f
p
r
ec
is
io
n
c
lass
i
f
icatio
n
o
f
tr
ai
n
in
g
a
n
d
test
in
g
m
o
d
el
s
.
W
h
ile
t
h
e
an
al
y
s
is
o
f
L
VQ
i
s
d
o
n
e
u
s
i
n
g
pa
c
k
a
g
e
cla
s
s
o
n
s
o
f
t
w
ar
e
R
.
T
h
e
s
tep
s
i
n
L
V
Q
an
al
y
s
i
s
ar
e
:
a.
Fo
r
m
in
g
a
n
in
p
u
t
m
a
tr
ix
o
n
tr
ain
i
n
g
d
ata
an
d
f
o
r
m
i
n
g
a
v
ec
t
o
r
o
r
class
if
icatio
n
f
ac
to
r
f
o
r
tr
ain
in
g
d
ata
b.
I
n
itializatio
n
w
ei
g
h
t
s
o
f
t
h
e
L
VQ
n
et
w
o
r
k
c.
Dete
r
m
i
n
e
th
e
lear
n
i
n
g
r
ate
(
α
)
d.
R
en
e
w
w
e
ig
h
t o
n
t
h
e
co
m
p
etit
iv
e
la
y
er
to
o
b
tain
o
p
ti
m
u
m
weig
h
t
e.
Fo
r
m
an
o
p
ti
m
a
l a
r
ch
itect
u
r
e
f.
Re
-
cla
s
s
i
f
ied
t
h
e
test
i
n
g
d
ata
s
et
b
ased
o
n
th
e
b
est
ar
ch
ite
ctu
r
e
o
f
th
e
L
VQ
m
et
h
o
d
f
o
r
m
ed
f
r
o
m
t
h
e
tr
ain
i
n
g
d
ata
s
et.
Af
ter
t
h
e
co
m
p
le
tio
n
o
f
t
h
e
L
VQ
an
al
y
s
is
d
o
n
e
th
e
n
ca
lc
u
l
ated
in
d
icato
r
o
f
cla
s
s
i
f
icatio
n
ac
c
u
r
ac
y
th
at
is
H
it R
atio
a
n
d
n
ex
t to
co
m
p
ar
e
v
al
u
e
o
f
Hit
R
atio
f
r
o
m
b
o
th
m
eth
o
d
s.
4.
RE
SU
L
T
S
A
ND
AN
AL
Y
SI
S
4
.
1
.
L
o
g
is
t
ic
Reg
re
s
s
io
n M
o
delin
g
T
h
e
p
r
o
ce
s
s
o
f
d
ata
an
al
y
s
i
s
b
eg
in
s
w
ith
m
u
ltico
li
n
ea
r
it
y
t
esti
n
g
a
m
o
n
g
t
h
e
p
r
ed
icto
r
v
ar
iab
les
o
n
ea
ch
d
ataset.
I
n
th
e
t
h
r
ee
d
atasets
u
s
ed
in
th
is
s
t
u
d
y
,
t
h
er
e
is
n
o
m
u
ltico
lli
n
ea
r
it
y
t
h
at
is
i
n
d
icate
d
b
y
t
h
e
VI
F
v
alu
e
g
r
ea
ter
th
a
n
1
0
.
T
h
e
m
o
d
elin
g
p
r
o
ce
s
s
in
lo
g
i
s
tic
r
e
g
r
ess
io
n
o
f
t
h
e
t
h
r
ee
d
atase
ts
c
an
b
e
co
n
t
in
u
ed
b
y
esti
m
ati
n
g
t
h
e
m
o
d
el
p
ar
a
m
et
er
s
o
f
ea
ch
d
ata
s
et.
B
elo
w
is
th
e
p
ar
a
m
eter
m
o
d
el
esti
m
at
o
r
o
b
tain
ed
b
y
t
h
e
m
ax
i
m
u
m
li
k
eli
h
o
o
d
m
e
th
o
d
,
p
r
esen
ted
in
T
ab
le
2
.
T
ab
le
2
.
T
h
e
P
a
r
am
eter
E
s
t
i
m
atio
n
R
es
u
lt
s
o
n
A
ll Da
taset
s
P
r
e
d
i
c
t
o
r
s
V
a
l
u
e
s o
f
e
st
i
mat
e
s
p
a
r
a
me
t
e
r
D
a
t
a
se
t
1
D
a
t
a
se
t
2
D
a
t
a
se
t
3
5
.
7
6
0
0
.
2
3
4
0
.
6
1
9
2
.
1
8
4
0
.
0
0
2
-
0
.
1
2
0
2
.
2
7
9
0
.
0
3
5
-
0
.
1
9
1
4
.
4
1
5
-
0
.
0
8
9
-
1
.
3
8
2
4
.
9
5
5
0
.
3
7
0
1
.
0
9
2
X
6
-
2
.
9
7
6
2
.
3
3
2
C
o
n
st
a
n
t
-
1
0
0
.
7
5
3
4
5
.
2
8
9
2
.
1
5
3
T
h
e
f
u
ll
m
o
d
el
f
o
r
m
ed
f
o
r
d
ataset
1
,
d
ataset
2
,
an
d
d
ataset
3
r
esp
ec
tiv
el
y
ar
e
(
)
.
(
)
(
)
.
T
ests
o
n
p
ar
a
m
e
ter
est
i
m
a
to
r
s
ar
e
s
i
m
u
l
tan
eo
u
s
l
y
p
er
f
o
r
m
ed
to
d
eter
m
i
n
e
t
h
e
e
f
f
ec
t
o
f
p
r
ed
icto
r
v
ar
iab
les
co
n
ta
in
ed
i
n
t
h
e
lo
g
i
s
tic
r
eg
r
es
s
io
n
m
o
d
el
to
t
h
e
r
e
s
p
o
n
s
e
v
ar
iab
le
as
a
w
h
o
le
o
r
to
g
eth
er
.
T
h
is
te
s
t
is
b
ased
o
n
th
e
r
atio
test
s
tatis
t
ic
(
G)
lik
elih
o
o
d
r
atio
test
.
Her
e
is
th
e
h
y
p
o
th
e
s
is
u
s
ed
:
v
er
s
us
j
=
1
,
2
,
.
.
.
,
p
T
h
e
test
r
esu
lts
a
g
ain
s
t
p
ar
a
m
eter
esti
m
ato
r
s
s
i
m
u
lta
n
eo
u
s
l
y
o
n
th
e
d
atasets
1
,
2
,
an
d
3
as
s
h
o
w
n
i
n
Fig
u
r
e
3
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
1
,
Feb
r
u
ar
y
2018
:
3
3
3
–
3
4
3
340
T
ab
le
3
.
Sim
u
lta
n
eo
u
s
ly
P
ar
am
eter
T
esti
n
g
o
f
A
ll
Data
s
ets
D
a
t
a
se
t
-
2
l
o
g
l
i
k
e
l
i
h
o
o
d
L
i
k
e
l
i
h
o
o
d
R
a
t
i
o
T
e
st
p
-
v
a
l
u
e
M
o
d
e
l
i
n
t
e
r
se
p
M
o
d
e
l
p
a
r
a
me
t
e
r
1
-
4
3
.
8
0
1
-
9
.
7
5
0
0
7
6
6
8
.
1
0
3
<
0
.
0
0
0
2
-
6
1
.
5
0
8
-
7
.
1
0
5
1
0
8
.
8
0
5
0
.
0
0
0
3
-
4
6
.
3
7
4
-
3
7
.
8
5
4
1
7
.
0
3
9
0
.
0
0
9
B
ased
o
n
th
e
te
s
ti
n
g
o
f
p
ar
a
m
eter
s
s
i
m
u
ltan
eo
u
s
l
y
i
n
T
ab
le
2
.
f
o
r
d
ataset
1
it
ca
n
b
e
s
ee
n
th
at
t
h
e
p
-
v
alu
e
o
f
th
e
li
k
eli
h
o
o
d
r
atio
test
i
s
<
0
.
0
0
0
.
T
h
e
v
al
u
e
i
s
le
s
s
t
h
an
th
e
le
v
el
o
f
s
i
g
n
i
f
ica
n
t
α
i
s
0
.
0
5
,
s
o
it
w
a
s
d
ec
id
ed
to
r
e
j
ec
t
H
0
w
h
ic
h
m
ea
n
s
t
h
at
ca
r
b
o
h
y
d
r
ates,
v
eg
etab
les,
s
id
e
d
is
h
es,
f
r
u
it
s
,
an
d
m
il
k
to
g
et
h
er
s
ig
n
i
f
ica
n
tl
y
a
f
f
ec
t
t
h
e
d
iet
s
ta
tu
s
o
f
c
h
ild
r
en
u
n
d
er
f
iv
e.
T
h
e
r
esu
lt
o
f
s
tati
s
tical
test
o
n
d
ataset
2
ca
n
b
e
s
ee
n
th
at
p
-
v
al
u
e
o
f
lik
el
ih
o
o
d
r
atio
test
is
0
.
0
0
0
.
T
h
e
v
alu
e
is
l
ess
th
a
n
th
e
s
ig
n
i
f
ica
n
t
lev
el
o
f
α
is
0
.
0
5
,
s
o
it
is
d
ec
id
ed
to
r
e
j
ec
t
H
0
w
h
ich
m
e
an
s
t
h
at
ex
p
er
ie
n
ce
,
d
u
r
atio
n
o
f
ed
u
ca
tio
n
,
lab
o
r
o
u
tp
o
u
r
,
ag
e,
lev
el
o
f
s
ea
w
ee
d
clea
n
li
n
es
s
,
an
d
s
ea
w
ee
d
co
n
ten
ts
s
ig
n
i
f
ica
n
tl
y
af
f
ec
t
f
ar
m
i
n
g
cr
ed
it
to
f
ar
m
er
s
s
ea
wee
d
.
W
h
ile
th
e
test
r
esu
lt
s
i
n
d
ata
3
ca
n
b
e
s
ee
n
t
h
at
t
h
e
p
-
v
al
u
e
o
f
t
h
e
l
ik
el
ih
o
o
d
r
atio
test
is
0
.
0
0
9
.
T
h
e
v
al
u
e
i
s
les
s
t
h
a
n
α
i
s
0
.
0
5
,
s
o
it
is
d
ec
id
e
d
to
r
e
ject
H
0
w
h
ic
h
m
ea
n
s
th
at
m
a
ter
n
al
ag
e,
p
ar
it
y
,
g
estat
io
n
al
d
is
tan
ce
,
an
e
m
ia,
n
u
tr
i
tio
n
al
s
tatu
s
,
an
d
ed
u
ca
tio
n
to
g
et
h
er
h
a
v
e
a
s
i
g
n
i
f
ica
n
t e
f
f
ec
t o
n
th
e
i
n
cid
e
n
ce
o
f
lo
w
b
ir
th
w
ei
g
h
t b
ab
ies.
T
o
d
eter
m
i
n
e
t
h
e
p
r
ed
icto
r
v
ar
iab
les
t
h
at
s
i
g
n
i
f
ican
tl
y
i
n
f
lu
e
n
ce
t
h
e
r
esp
o
n
s
e
v
ar
ia
b
les,
it
is
n
ec
es
s
ar
y
to
test
th
e
s
ig
n
i
f
ica
n
ce
o
f
th
e
p
ar
a
m
eter
s
in
ea
c
h
p
r
ed
icto
r
v
ar
iab
les
u
s
in
g
W
ald
(
W
j
)
test
s
tatis
tic
.
T
h
e
s
tatis
tical
h
y
p
o
th
es
is
test
e
d
is
H
0
:
β
j
=
0
v
er
s
u
s
H
1
:
β
j
≠
0
.
W
ald
test
s
tatis
tic
is
C
h
i
-
s
q
u
a
r
e
d
is
tr
ib
u
ted
w
it
h
o
n
e
d
eg
r
ee
o
f
f
r
ee
d
o
m
.
B
ased
o
n
p
-
v
al
u
e
o
n
W
ald
tes
t
s
tatis
tic,
f
o
r
th
e
f
ir
s
t
d
ata,
it
w
a
s
f
o
u
n
d
t
h
at
X
2
(
v
eg
etab
le)
an
d
X
3
(
s
id
e
d
is
h
)
v
ar
iab
les
d
id
n
o
t
s
ig
n
i
f
ican
t
l
y
af
f
ec
t
t
h
e
clas
s
i
f
icatio
n
o
f
in
f
a
n
t
d
iet.
I
n
t
h
e
s
ec
o
n
d
d
ata,
o
n
ly
p
r
ed
icto
r
X
5
(l
ev
el
o
f
s
ea
w
ee
d
clea
n
lin
e
s
s
)
h
as
a
s
ig
n
i
f
ica
n
t
ef
f
ec
t
o
n
t
h
e
d
eter
m
in
a
tio
n
o
f
cr
ed
it
f
o
r
s
ea
w
ee
d
f
ar
m
er
s
.
As
f
o
r
th
e
th
ir
d
d
ata,
it
is
k
n
o
wn
th
at
t
h
e
X
1
(
m
ater
n
al
a
g
e)
,
X
2
(
p
ar
ity
)
,
X
3
(
b
ir
th
d
is
tan
ce
)
,
a
n
d
X
5
(
n
u
tr
itio
n
a
l
s
tatu
s
)
d
id
n
o
t
s
ig
n
i
f
ica
n
tl
y
a
f
f
ec
t
th
e
clas
s
i
f
icatio
n
o
f
lo
w
b
ir
th
w
ei
g
h
t
b
ab
ies.
T
h
e
lo
g
is
tic
r
eg
r
es
s
io
n
m
o
d
el
th
at
is
f
o
r
m
ed
b
ased
o
n
s
i
g
n
if
i
ca
n
t p
r
ed
icto
r
v
ar
iab
les ar
e
as so
w
n
i
n
Fi
g
u
r
e
4
:
T
ab
le
4
.
T
h
e
L
o
g
is
tic
R
e
g
r
ess
io
n
Mo
d
el
o
f
A
l
l D
ata
s
et
s
an
d
Go
o
d
n
ess
o
f
Fit
T
est
D
a
t
a
se
t
T
h
e
f
i
n
a
l
mo
d
e
l
o
f
l
o
g
i
st
i
c
r
e
g
r
e
ssi
o
n
C
h
i
-
S
q
u
a
r
e
df
p
-
v
a
l
u
e
1
(
)
2
.
8
9
7
8
0
.
9
4
1
2
(
)
7
.
3
0
5
1
0
.
0
0
7
3
(
)
0
.
5
1
8
2
0
.
7
7
2
T
esti
n
g
t
h
e
s
u
i
tab
ilit
y
(
g
o
o
d
n
ess
)
m
o
d
el
u
s
ed
to
d
eter
m
i
n
e
w
h
eth
er
th
e
r
e
s
u
l
tin
g
m
o
d
el
i
s
ap
p
r
o
p
r
iate
(
f
ea
s
ib
le)
.
T
h
e
s
tatis
tical
h
y
p
o
th
e
s
is
u
s
e
d
in
t
h
is
te
s
t
i
s
:
̂
(
o
b
s
er
v
atio
n
f
r
eq
u
en
c
y
=
e
x
p
ec
ted
f
r
eq
u
e
n
c
y
)
v
er
s
u
s
̂
(
o
b
s
er
v
atio
n
f
r
eq
u
en
c
y
≠
e
x
p
ec
ted
f
r
eq
u
e
n
c
y
)
.
B
ased
o
n
T
ab
le
3
.
p
-
v
alu
e
f
o
r
th
e
1
s
t
an
d
3
r
d
d
ata
h
as
a
v
alu
e
o
f
m
o
r
e
th
an
0
.
0
5
s
o
th
e
d
ec
is
io
n
is
to
r
ec
eiv
e
H
0
an
d
it
ca
n
b
e
co
n
cl
u
d
ed
th
at
t
h
e
b
i
n
ar
y
lo
g
is
t
ic
r
eg
r
es
s
io
n
m
o
d
el
g
e
n
er
ated
is
g
o
o
d
(
ap
p
r
o
p
r
i
ate)
,
th
e
m
o
d
el
h
as
b
ee
n
s
u
f
f
icien
t
to
e
x
p
lain
t
h
e
f
ir
s
t
d
ata
(c
las
s
i
f
icatio
n
o
f
in
f
an
t
d
iet)
,
a
n
d
t
h
e
3
r
d
d
ata
(
cl
ass
i
f
icatio
n
o
f
lo
w
b
ir
th
w
ei
g
h
t
i
n
f
a
n
t
s
)
.
Fo
r
th
e
2
n
d
d
ata
b
ec
au
s
e
p
-
v
al
u
e
h
as
a
v
alu
e
les
s
th
a
n
0
.
0
5
it
ca
n
b
e
co
n
clu
d
ed
th
at
th
e
b
in
ar
y
lo
g
is
tic
r
e
g
r
ess
io
n
m
o
d
el
g
en
er
ated
i
n
t
h
e
2
n
d
d
ata
h
as
n
o
t
b
ee
n
s
u
i
tab
le
o
r
n
o
t
en
o
u
g
h
to
ex
p
lai
n
th
e
d
ata.
B
ased
o
n
t
h
ese
r
esu
lts
,
s
tatis
t
icall
y
,
lo
g
i
s
tic
r
eg
r
es
s
io
n
m
o
d
els
o
f
t
h
e
1
s
t
a
n
d
3
r
d
d
ata
ca
n
b
e
u
s
ed
f
o
r
o
b
j
ec
t c
lass
if
icatio
n
a
n
d
s
h
o
u
l
d
b
e
ab
le
t
o
p
r
o
d
u
ce
f
air
l
y
g
o
o
d
class
if
icatio
n
ac
c
u
r
ac
y
.
4
.
2
.
L
ea
rning
Vec
t
o
r
Q
ua
ntiz
a
t
i
o
n
(
L
VQ
)
O
pti
m
u
m
M
o
del
T
o
o
b
tain
an
o
p
ti
m
al
L
V
Q
n
et
w
o
r
k
m
o
d
el,
L
VQ
n
et
w
o
r
k
m
o
d
eli
n
g
p
r
o
ce
s
s
i
s
p
er
f
o
r
m
ed
o
n
a
v
ar
io
u
s
n
u
m
b
er
o
f
n
e
u
r
o
n
s
i
n
a
h
id
d
en
la
y
er
ca
lled
co
d
eb
o
o
k
.
T
h
e
a
m
o
u
n
t
o
f
co
d
eb
o
o
k
u
s
ed
is
2
,
1
0
,
3
0
,
an
d
5
0
.
T
h
e
s
ize
o
f
th
e
co
d
eb
o
o
k
w
il
l
d
eter
m
i
n
e
th
e
w
ei
g
h
t
m
a
t
r
ix
d
i
m
e
n
s
io
n
to
b
e
ca
lcu
lated
to
o
b
tain
o
p
ti
m
al
L
VQ
n
et
w
o
r
k
.
B
ased
o
n
th
e
d
i
m
e
n
s
io
n
o
f
th
is
w
ei
g
h
t
m
atr
i
x
,
th
e
n
th
e
w
ei
g
h
ts
ar
e
r
an
d
o
m
l
y
in
i
tialized
.
T
h
e
o
p
tim
a
l
w
e
ig
h
t
w
ill
b
e
o
b
tain
ed
th
r
o
u
g
h
th
e
tr
ain
i
n
g
p
r
o
ce
s
s
b
y
u
tili
zi
n
g
th
e
tr
ain
in
g
d
ata
in
p
u
t.
A
f
ter
all
th
e
co
n
n
ec
ti
n
g
w
eig
h
t
s
b
et
w
ee
n
n
o
d
es
in
d
if
f
er
e
n
t
la
y
er
s
o
f
th
e
L
VQ
n
e
t
w
o
r
k
h
a
v
e
b
ee
n
o
b
tain
ed
,
th
e
n
e
t
w
o
r
k
o
u
tp
u
t
ca
n
b
e
o
b
tain
ed
b
y
in
p
u
tti
n
g
t
h
e
i
n
p
u
t
d
ata
in
to
t
h
e
n
et
w
o
r
k
.
I
f
u
s
ed
as
an
i
n
p
u
t
a
r
g
u
m
e
n
t
is
tr
ain
in
g
d
ata
th
en
o
b
tain
ed
t
h
e
o
u
tp
u
t
o
f
tr
ain
in
g
d
ata.
Si
m
i
lar
l
y
,
i
f
u
s
ed
as
a
n
i
n
p
u
t
ar
g
u
m
en
t
is
test
in
g
d
ata
t
h
e
n
o
b
tain
ed
th
e
o
u
tp
u
t
o
f
tes
tin
g
d
ata.
T
h
e
v
alu
e
o
f
Hit
R
atio
ca
n
b
e
ca
lcu
lated
f
r
o
m
t
h
e
o
u
tp
u
t
o
b
tain
ed
f
r
o
m
th
e
n
et
w
o
r
k
,
eit
h
er
o
u
tp
u
t
o
f
tr
ain
i
n
g
o
r
te
s
ti
n
g
d
ata.
T
h
e
i
n
itialized
w
e
ig
h
t
s
o
f
t
h
e
f
o
u
r
co
d
eb
o
o
k
s
f
o
r
th
e
f
ir
s
t d
ata
ar
e
s
h
o
w
n
i
n
T
ab
le
5
.
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:
2
0
8
8
-
8708
A
n
I
n
flu
e
n
ce
o
f Mea
s
u
r
eme
n
t
S
ca
le
o
f P
r
ed
icto
r
V
a
r
ia
b
le
o
n
Lo
g
is
tic
R
eg
r
ess
io
n
…
(
W
a
eg
o
Ha
d
i Nu
g
r
o
h
o
)
341
T
ab
le
5
.
T
h
e
in
itial
w
ei
g
h
t
s
b
et
w
ee
n
i
n
p
u
t a
n
d
h
id
d
en
la
y
er
o
f
th
e
f
ir
s
t d
ataset
S
i
z
e
C
o
d
e
b
o
o
k
N
e
u
r
o
n
H
i
d
d
e
n
(
)
(
)
(
)
(
)
(
)
2
1
4
.
7
3
3
.
5
8
5
.
0
1
4
.
1
8
5
.
4
3
2
6
.
7
0
4
.
7
6
4
.
3
3
5
.
0
0
5
.
2
4
10
1
5
.
3
4
2
.
6
2
4
.
4
2
3
.
3
6
3
.
7
8
2
4
.
5
8
3
.
7
1
3
.
3
0
3
.
3
0
5
.
0
0
...
...
...
...
...
...
10
6
.
7
0
4
.
4
0
4
.
8
3
3
.
3
4
7
.
5
0
30
1
5
.
3
4
2
.
6
2
4
.
4
2
3
.
3
6
3
.
7
8
2
4
.
7
3
3
.
0
4
2
.
8
0
2
.
3
4
3
.
6
8
...
...
...
...
...
...
30
6
.
1
0
4
.
1
4
5
.
4
9
5
.
2
8
5
.
0
0
50
1
4
.
7
3
3
.
5
8
5
.
0
1
4
.
1
8
5
.
4
3
2
5
.
6
0
4
.
5
0
2
.
8
4
3
.
3
0
4
.
3
2
...
...
...
...
...
...
50
6
.
1
8
4
.
9
1
5
.
1
9
4
.
0
0
5
.
4
3
As e
x
p
lain
ed
in
t
h
e
p
r
ev
io
u
s
s
ess
io
n
t
h
at
d
ata
1
h
a
s
5
i
n
p
u
t
v
ar
iab
les co
n
s
id
er
ed
to
h
av
e
a
n
e
f
f
ec
t
o
n
th
e
r
esp
o
n
s
e
v
ar
iab
le.
I
n
co
d
eb
o
o
k
=
2
,
it
m
u
s
t
b
e
i
n
itialized
a
5
x
2
m
atr
ix
w
h
o
s
e
ele
m
e
n
ts
ar
e
th
e
co
n
n
ec
t
in
g
w
ei
g
h
ts
b
et
w
ee
n
t
h
e
i
n
p
u
t
l
a
y
er
an
d
th
e
h
id
d
en
la
y
er
,
a
n
d
a
2
x
1
-
s
ized
v
ec
to
r
w
h
o
s
e
ele
m
e
n
t
s
ar
e
t
h
e
co
n
n
ec
ti
n
g
w
eig
h
t
s
o
f
th
e
h
id
d
en
la
y
er
to
th
e
o
u
tp
u
t
la
y
er
,
w
h
er
ea
s
i
n
th
e
co
d
eb
o
o
k
=
5
0
,
T
h
e
d
im
e
n
s
io
n
s
o
f
th
e
w
ei
g
h
ted
m
atr
ices
a
n
d
v
ec
to
r
s
to
b
e
in
itial
ized
ar
e
5
x
5
0
an
d
5
0
x
1
w
h
ich
ar
e
th
e
co
n
n
ec
ti
n
g
w
ei
g
h
t
s
b
et
w
ee
n
th
e
i
n
p
u
t
la
y
er
an
d
th
e
h
id
d
en
la
y
er
,
an
d
th
e
co
n
n
ec
tin
g
w
eig
h
t
s
o
f
th
e
h
id
d
en
la
y
e
r
to
th
e
o
u
tp
u
t
la
y
er
.
Af
ter
t
h
e
tr
ain
i
n
g
p
r
o
ce
s
s
,
f
i
n
all
y
w
e
g
et
t
h
e
o
p
ti
m
al
w
ei
g
h
t
w
h
ic
h
s
o
m
e
o
f
t
h
e
ele
m
en
ts
ar
e
p
r
esen
ted
in
T
ab
le
6
as f
o
llo
w
s
:
T
ab
le
6
.
T
h
e
Fin
al
W
eig
h
ts
b
e
t
w
ee
n
I
n
p
u
t a
n
d
Hid
d
en
L
a
y
er
o
f
th
e
Firs
t D
ataset
S
i
z
e
C
o
d
e
b
o
o
k
N
e
u
r
o
n
H
i
d
d
e
n
(
)
(
)
(
)
(
)
(
)
2
1
4
.
8
8
9
3
.
8
2
7
4
.
5
6
2
3
.
9
4
9
4
.
9
5
4
2
6
.
2
2
6
4
.
7
3
5
5
.
1
1
0
4
.
9
0
2
6
.
2
9
3
10
1
5
.
1
4
1
3
.
1
1
4
4
.
3
4
5
3
.
8
1
0
3
.
9
2
0
2
4
.
9
1
2
3
.
7
6
0
3
.
0
3
6
3
.
0
4
7
4
.
4
5
7
...
...
...
...
...
...
10
5
.
9
3
9
4
.
8
3
5
4
.
2
8
5
3
.
8
9
2
7
.
1
7
0
30
1
5
.
3
4
0
2
.
6
2
0
4
.
4
2
0
3
.
3
6
0
3
.
7
8
0
2
4
.
7
3
0
3
.
0
4
0
2
.
8
0
0
2
.
3
4
0
3
.
6
8
0
...
...
...
...
...
...
30
5
.
7
5
6
4
.
4
6
9
4
.
8
9
9
5
.
5
0
9
5
.
2
5
5
50
1
4
.
9
4
7
3
.
8
4
8
5
.
3
0
3
4
.
0
8
5
5
.
1
3
0
2
5
.
6
0
0
4
.
5
0
0
2
.
8
4
0
3
.
3
0
0
4
.
3
2
0
...
...
...
...
...
...
50
6
.
3
0
6
5
.
1
3
6
5
.
3
2
7
4
.
0
7
6
5
.
7
3
9
T
ab
le
7.
Hit Ratio
f
o
r
A
ll Da
t
aset
o
f
Var
io
u
s
C
o
d
eb
o
o
k
Size
D
a
t
a
C
o
d
e
b
o
o
k
si
z
e
H
i
t
R
a
t
i
o
T
r
a
i
n
i
n
g
D
a
t
a
T
e
st
i
n
g
D
a
t
a
1
2
8
4
.
2
1
8
4
.
4
10
9
2
.
1
8
1
.
2
5
30
9
3
.
4
7
8
.
1
2
50
9
6
.
0
5
7
8
.
1
2
2
2
6
7
.
0
1
7
5
.
6
1
10
8
0
.
4
1
6
5
.
8
5
30
8
3
.
5
7
0
.
7
3
50
9
1
.
7
5
7
3
.
1
7
3
2
6
1
.
2
3
7
.
9
3
10
7
4
.
6
2
4
4
.
8
2
30
7
6
.
1
1
5
5
.
1
7
50
7
6
.
1
2
4
4
.
8
2
T
h
e
f
in
al
w
ei
g
h
t
o
f
L
V
Q
t
h
at
h
as
b
ee
n
o
b
tain
ed
f
r
o
m
L
VQ
n
et
w
o
r
k
lear
n
i
n
g
p
r
o
ce
s
s
u
s
i
n
g
tr
ai
n
in
g
d
ata
is
t
h
en
u
s
ed
as
n
et
w
o
r
k
weig
h
t.
T
h
u
s
th
e
L
VQ
Ne
t
w
o
r
k
alr
ea
d
y
h
a
s
t
h
e
co
n
n
ec
ti
n
g
weig
h
ts
o
f
ea
c
h
n
o
d
e
b
et
w
ee
n
th
e
la
y
er
s
,
s
o
th
at
th
e
L
VQ
n
et
w
o
r
k
is
r
ea
d
y
to
b
e
u
s
ed
as
a
m
o
d
el
f
o
r
o
b
j
ec
t
cla
s
s
i
f
icatio
n
.
T
ab
le
8
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec
&
C
o
m
p
E
n
g
,
Vo
l.
8
,
No
.
1
,
Feb
r
u
ar
y
2018
:
3
3
3
–
3
4
3
342
is
an
in
d
icato
r
o
f
L
VQ
n
et
w
o
r
k
p
er
f
o
r
m
a
n
ce
th
a
t
is
H
it
R
atio
v
al
u
e
f
r
o
m
t
h
r
ee
d
atasets
a
n
d
v
ar
io
u
s
co
d
eb
o
o
k
.
B
ase
d
o
n
th
e
v
al
u
e
o
f
Hi
t
R
a
tio
in
T
ab
le
7
w
e
g
e
t
th
e
b
est
L
VQ
n
et
w
o
r
k
m
o
d
el
f
o
r
d
ataset
1
an
d
d
ataset
2
is
t
h
e
m
o
d
el
w
i
th
co
d
eb
o
o
k
=
2
,
w
h
er
ea
s
,
i
n
d
ataset
3
,
th
e
b
est
m
o
d
el
i
s
co
d
eb
o
o
k
=
3
0
.
T
h
e
m
o
d
el
s
ar
e
s
aid
to
b
e
th
e
b
est
m
o
d
els
b
ec
au
s
e
th
e
y
h
a
v
e
t
h
e
g
r
ea
test
Hit
R
atio
v
al
u
e
in
t
h
e
d
a
ta
test
in
g
.
I
f
T
ab
le
7
is
o
b
s
er
v
ed
f
u
r
th
er
,
it
ca
n
b
e
s
aid
th
at
th
e
in
cr
ea
s
e
in
t
h
e
n
u
m
b
er
o
f
co
d
eb
o
o
k
s
is
also
f
o
llo
w
ed
b
y
th
e
i
n
cr
ea
s
e
o
f
Hit
R
atio
v
al
u
e
i
n
tr
ai
n
i
n
g
d
ata,
b
u
t
th
e
i
n
cid
en
ce
d
o
es
n
o
t
ap
p
ly
to
t
h
e
v
a
lu
e
o
f
Hit
r
a
tio
in
d
ata
test
in
g
.
T
h
is
p
h
en
o
m
e
n
o
n
i
s
k
n
o
w
n
a
s
o
v
er
f
itti
n
g
.
4
.
3
.
T
he
Co
m
pa
ri
s
o
n o
f
Acc
ura
cy
Cla
s
s
if
ica
t
io
n bet
w
ee
n
L
o
g
is
t
ic
R
eg
re
s
s
io
n a
nd
L
V
Q
T
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
t
w
o
m
et
h
o
d
s
o
f
class
i
f
y
i
n
g
o
b
j
ec
ts
is
co
m
p
ar
ed
b
y
t
h
e
v
al
u
e
o
f
Hit
R
at
io
ca
lcu
lated
o
n
b
o
th
tr
ain
i
n
g
a
n
d
test
i
n
g
d
ataset
s
o
f
all
th
e
b
est
m
o
d
els.
T
h
e
r
esu
lt
s
o
f
th
e
class
i
f
icatio
n
ac
cu
r
ac
y
o
f
all
d
ata
s
ets ar
e
p
r
esen
ted
i
n
T
ab
le
8
b
elo
w
:
T
ab
le
8
.
T
h
e
A
cu
r
ac
y
o
f
th
e
B
est M
o
d
el
f
r
o
m
B
o
th
L
o
g
is
tic
R
eg
r
es
s
io
n
a
n
d
L
V
Q
D
a
t
a
T
r
a
i
n
i
n
g
T
e
st
i
n
g
L
o
g
i
st
i
c
R
e
g
r
e
ssi
o
n
L
V
Q
L
o
g
i
st
i
c
R
e
g
r
e
ssi
o
n
L
V
Q
9
0
.
8
8
4
.
2
1
8
4
.
4
8
4
.
4
-
6
7
.
0
1
-
7
5
.
6
1
6
5
.
7
7
6
.
1
1
5
5
.
1
7
5
5
.
1
7
B
ased
o
n
th
e
p
er
ce
n
ta
g
e
o
f
a
cc
u
r
ac
y
o
f
clas
s
if
icatio
n
r
es
u
l
ts
in
T
ab
le
8
.
it
ca
n
b
e
s
ee
n
t
h
at
i
n
t
h
e
s
ec
o
n
d
d
ataset
(
g
r
a
n
ti
n
g
o
f
s
e
a
w
ee
d
f
ar
m
i
n
g
)
,
lo
g
i
s
tic
r
eg
r
e
s
s
io
n
m
et
h
o
d
ca
n
n
o
t
b
e
ca
lc
u
lated
class
i
f
icatio
n
ac
cu
r
ac
y
b
ec
a
u
s
e
t
h
e
m
o
d
el
o
b
tain
ed
f
r
o
m
th
e
d
ata
d
o
es
n
o
t
m
ee
t
t
h
e
m
o
d
el
f
it
te
s
t.
T
h
is
is
s
u
p
p
o
r
ted
b
y
p
ar
tial
p
ar
am
ete
r
test
r
es
u
lt
o
n
l
y
o
b
tain
ed
o
n
e
p
r
ed
icto
r
v
a
r
iab
le
(
lev
el
o
f
s
ea
w
ee
d
clea
n
lin
es
s
)
w
h
ich
h
a
v
e
s
ig
n
i
f
ica
n
t
e
f
f
ec
t
to
r
esp
o
n
s
e
v
ar
iab
le
(
cr
ed
it
ap
p
r
o
v
al
d
ec
is
io
n
to
s
ea
w
ee
d
f
ar
m
er
)
.
Kee
p
in
m
in
d
t
h
at
th
e
s
ec
o
n
d
d
ataset
h
as
p
r
ed
icto
r
v
ar
iab
les
th
at
ar
e
all
s
ca
lab
le
r
atio
s
.
Di
f
f
er
e
n
t
ac
c
u
r
ac
y
r
e
s
u
l
ts
ar
e
f
o
u
n
d
in
th
e
L
VQ
m
o
d
el
t
h
at
i
s
i
n
t
h
e
s
ec
o
n
d
d
ataset
s
till
o
b
tai
n
ed
th
e
v
alu
e
o
f
Hit
R
atio
,
b
o
th
i
n
th
e
tr
ain
i
n
g
a
n
d
te
s
ti
n
g
d
ataset
o
f
m
o
d
er
ate
en
o
u
g
h
s
i
ze
o
f
6
7
% a
n
d
7
5
% r
esp
ec
tiv
el
y
.
Mo
d
elin
g
o
n
t
h
e
f
ir
s
t
d
at
a
s
e
t
is
a
ca
s
e
t
h
at
id
ea
ll
y
d
e
m
o
n
s
tr
ates
t
h
at
lo
g
is
t
ic
r
eg
r
e
s
s
io
n
m
o
d
el
p
er
f
o
r
m
s
eq
u
a
ll
y
w
ell
co
m
p
a
r
ed
to
L
VQ
m
o
d
el
b
ased
o
n
Hit
r
atio
o
n
d
atase
t
tes
tin
g
=
8
4
%.
Hav
i
n
g
s
t
u
d
ied
m
o
r
e
d
ee
p
l
y
to
th
is
lo
g
i
s
tic
r
eg
r
ess
io
n
m
o
d
el,
it
tu
r
n
s
o
u
t
in
t
h
e
p
r
o
ce
s
s
o
f
d
iag
n
o
s
tic
e
x
a
m
in
at
io
n
o
f
t
h
e
er
r
o
r
o
b
tain
ed
th
e
r
es
u
lt
th
a
t
th
e
er
r
o
r
o
f
t
h
is
m
o
d
el
is
ab
le
to
m
ee
t
a
ll
t
h
e
as
s
u
m
p
tio
n
s
in
t
h
e
r
e
g
r
ess
io
n
m
o
d
eli
n
g
.
T
h
e
as
s
u
m
p
tio
n
s
a
r
e
er
r
o
r
in
d
ep
en
d
en
l
y
ea
ch
o
th
er
s
,
er
r
o
r
h
as
a
co
n
s
ta
n
t
v
ar
ian
,
a
n
d
er
r
o
r
h
a
s
n
o
r
m
al
d
is
tr
ib
u
t
io
n
.
T
h
e
m
o
s
t
d
if
f
ic
u
lt
t
h
i
n
g
to
b
e
m
et
w
it
h
r
eg
r
ess
io
n
a
n
al
y
s
is
i
s
th
e
n
o
r
m
alit
y
as
s
u
m
p
tio
n
o
f
th
e
d
is
tr
ib
u
tio
n
o
f
er
r
o
r
.
I
n
th
e
t
h
ir
d
d
ataset
o
b
tain
ed
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
b
o
th
m
o
d
e
ls
ar
e
j
u
s
t
as
b
ad
th
at
i
s
s
h
o
w
n
b
y
t
h
e
v
alu
e
o
f
Hit
r
atio
o
n
d
ataset
test
i
ng
=
5
5
%.
I
t
s
h
o
u
ld
b
e
n
o
ted
ca
r
ef
u
l
l
y
th
at
t
h
e
p
r
ed
icto
r
v
ar
iab
les
ar
e
all
n
o
m
i
n
al
s
ca
le
m
ea
s
u
r
e
m
en
t
s
.
I
n
th
e
lo
g
i
s
tic
r
eg
r
es
s
io
n
m
o
d
el,
o
n
l
y
t
w
o
ca
teg
o
r
ies
o
f
p
r
ed
icto
r
s
(
ie,
an
em
ia
an
d
lo
w
ed
u
ca
ted
m
o
th
er
s
)
h
ad
a
s
ig
n
i
f
ica
n
t
e
f
f
ec
t
o
n
lo
w
b
ir
th
w
eig
h
t
(
L
B
W
)
in
f
a
n
t
s
.
T
h
is
i
m
p
lie
s
th
at
lo
g
is
tic
r
e
g
r
ess
io
n
m
o
d
eli
n
g
a
n
d
L
VQ
n
et
w
o
r
k
o
n
p
r
ed
ictiv
e
v
ar
iab
les
o
f
n
o
m
i
n
al
s
ca
le
r
eq
u
ir
e
m
o
r
e
f
ac
to
r
s
th
at
i
n
f
lu
e
n
ce
t
h
e
r
esp
o
n
s
e
v
ar
iab
le.
A
lt
h
o
u
g
h
th
e
L
V
Q
m
o
d
el
o
n
th
e
tr
ai
n
i
n
g
d
ata
s
et
h
a
s
Hit
R
atio
=
7
6
%,
th
e
m
o
d
el
al
s
o
r
e
m
ai
n
s
u
n
ab
le
to
class
i
f
y
th
e
te
s
t d
ataset
s
at
is
f
a
cto
r
il
y
.
5.
CO
NCLU
SI
O
N
T
h
e
m
ea
s
u
r
e
m
e
n
t scale
o
f
t
h
e
p
r
ed
icto
r
v
ar
iab
le
is
v
er
y
i
n
f
l
u
en
tial o
n
th
e
m
o
d
elin
g
a
n
d
p
er
f
o
r
m
a
n
ce
o
f
th
e
clas
s
if
icatio
n
m
o
d
el,
b
o
th
lo
g
is
t
ic
r
eg
r
ess
io
n
m
o
d
el,
an
d
L
VQ
m
o
d
el.
I
n
th
e
i
n
ter
v
al
-
s
ca
le
p
r
ed
icto
r
v
ar
iab
le,
th
e
b
es
t
m
o
d
el
o
f
b
o
th
m
et
h
o
d
s
p
r
o
d
u
ce
s
a
n
e
q
u
all
y
h
i
g
h
ac
cu
r
ac
y
o
f
Hit
R
atio
=
8
4
%.
I
n
t
h
e
n
o
m
i
n
al
-
s
ca
le
p
r
ed
icto
r
v
ar
ia
b
le,
th
e
class
if
icatio
n
ac
c
u
r
ac
y
o
f
t
h
e
b
est
m
o
d
el
in
b
o
th
m
et
h
o
d
s
is
s
i
m
ilar
l
y
lo
w
:
Hit
R
atio
=
55%
.
W
h
ile
o
n
R
atio
-
s
ca
le
p
r
ed
icto
r
v
ar
iab
le,
lo
g
i
s
tic
r
e
g
r
ess
io
n
m
o
d
eli
n
g
d
id
n
o
t
p
r
o
d
u
ce
th
e
b
es
t
m
o
d
el,
B
u
t
th
e
r
es
u
lti
n
g
L
VQ
m
o
d
el
h
as
a
n
ac
c
u
r
ac
y
f
o
r
f
air
l
y
m
o
d
er
ate
o
b
j
ec
t
class
i
f
icatio
n
,
ie
Hi
t
R
atio
=
7
5
%.
Evaluation Warning : The document was created with Spire.PDF for Python.