T
E
L
K
O
M
NIKA
T
elec
o
mm
un
ica
t
io
n,
Co
m
pu
t
ing
,
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
18
,
No
.
2
,
A
p
r
il
2
0
2
0
,
p
p
.
8
2
2
~
8
2
9
I
SS
N:
1
6
9
3
-
6
9
3
0
,
ac
cr
ed
ited
First Gr
ad
e
b
y
Ke
m
e
n
r
is
te
k
d
i
k
ti,
Dec
r
ee
No
: 2
1
/E/KPT
/2
0
1
8
DOI
: 1
0
.
1
2
9
2
8
/
T
E
L
KOM
NI
KA
.
v
1
8
i2
.
1
4
8
1
8
822
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
H
AR
-
M
I
m
et
ho
d
for
m
ulti
-
cla
ss
i
mba
la
nced data
sets
H
.
H
a
rt
o
no
1
,
Yeni
Ris
y
a
ni
2
,
E
ria
nto
O
ng
k
o
3
,
Da
hla
n Ab
du
lla
h
4
1
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
,
S
T
M
IK IBBI,
In
d
o
n
e
sia
1
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
,
Un
iv
e
rsitas
P
o
ten
si Ut
a
m
a
,
In
d
o
n
e
sia
2
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
S
y
ste
m
s,
S
T
M
IK IBBI,
In
d
o
n
e
sia
3
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
c
s,
Ak
a
d
e
m
i
T
e
k
n
o
lo
g
i
In
d
u
stri
Im
m
a
n
u
e
l,
In
d
o
n
e
sia
4
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
c
s,
Un
iv
e
rsitas
M
a
li
k
u
ss
a
leh
,
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
J
u
l 1
1
,
2
0
1
9
R
ev
i
s
ed
J
an
1
9
,
2
0
2
0
A
cc
ep
ted
Feb
2
1
,
2
0
2
0
Re
se
a
rc
h
o
n
m
u
lt
i
-
c
las
s
im
b
a
la
n
c
e
f
ro
m
a
n
u
m
b
e
r
o
f
re
se
a
r
c
h
e
rs
f
a
c
e
s
o
b
sta
c
les
in
t
h
e
f
o
rm
o
f
p
o
o
r
d
a
ta
d
iv
e
rsity
a
n
d
a
larg
e
n
u
m
b
e
r
o
f
c
las
si
f
iers
.
T
h
e
H
y
b
rid
A
p
p
ro
a
c
h
Re
d
e
f
in
it
io
n
-
M
u
lt
icla
ss
Im
b
a
lan
c
e
(H
A
R
-
M
I)
m
e
th
o
d
is
a
H
y
b
rid
En
se
m
b
les
m
e
th
o
d
w
h
ich
is
th
e
d
e
v
e
lo
p
m
e
n
t
o
f
t
h
e
Hy
b
rid
A
p
p
ro
a
c
h
Re
d
e
f
in
io
n
(HA
R)
m
e
th
o
d
.
T
h
is
stu
d
y
h
a
s
c
o
m
p
a
re
d
th
e
re
su
lt
s
o
b
tai
n
e
d
w
it
h
th
e
Dy
n
a
m
i
c
En
se
m
b
le
S
e
lec
ti
o
n
-
M
u
lt
icla
ss
Im
b
a
lan
c
e
(DES
-
M
I)
m
e
th
o
d
in
h
a
n
d
li
n
g
m
u
lt
icla
ss
i
m
b
a
lan
c
e
.
In
th
e
H
A
R
-
M
I
M
e
th
o
d
,
th
e
p
re
p
r
o
c
e
ss
in
g
sta
g
e
w
a
s
c
a
rried
o
u
t
u
sin
g
th
e
ra
n
d
o
m
b
a
lan
c
e
e
n
se
m
b
les
m
e
th
o
d
a
n
d
d
y
n
a
m
ic
e
n
se
m
b
le
se
lec
ti
o
n
t
o
p
ro
d
u
c
e
a
c
a
n
d
id
a
te
e
n
se
m
b
le
a
n
d
th
e
p
ro
c
e
ss
in
g
sta
g
e
s
wa
s
c
a
rried
o
u
t
u
sin
g
d
iff
e
re
n
t
c
o
n
tri
b
u
ti
o
n
sa
m
p
li
n
g
a
n
d
d
y
n
a
m
i
c
e
n
se
m
b
le
se
lec
ti
o
n
to
p
r
o
d
u
c
e
a
c
a
n
d
id
a
te
e
n
se
m
b
le.
T
h
is
re
s
e
a
r
c
h
h
a
s
b
e
e
n
c
o
n
d
u
c
te
d
b
y
u
sin
g
m
u
lt
i
-
c
las
s
im
b
a
lan
c
e
d
a
ta
se
ts so
u
rc
e
d
f
ro
m
t
h
e
KEEL
Re
p
o
sito
ry
.
T
h
e
re
su
lt
s
sh
o
w
th
a
t
th
e
HA
R
-
M
I
m
e
th
o
d
c
a
n
o
v
e
rc
o
m
e
m
u
lt
i
-
c
las
s
i
m
b
a
lan
c
e
w
it
h
b
e
tt
e
r
d
a
ta
d
iv
e
rsity
,
s
m
a
ll
e
r
n
u
m
b
e
r
o
f
c
l
a
ss
if
i
e
r
s,
a
n
d
b
e
tt
e
r
c
las
si
f
ier
p
e
rf
o
r
m
a
n
c
e
c
o
m
p
a
re
d
to
a
DES
-
M
I
m
e
th
o
d
.
T
h
e
se
re
su
lt
s
w
e
re
te
ste
d
w
it
h
a
W
il
c
o
x
o
n
sig
n
e
d
-
ra
n
k
sta
ti
stica
l
tes
t
w
h
ich
sh
o
w
e
d
th
a
t
th
e
su
p
e
rio
rit
y
o
f
th
e
HA
R
-
M
I
m
e
th
o
d
w
it
h
re
sp
e
c
t
t
o
DES
-
M
I
m
e
th
o
d
.
K
ey
w
o
r
d
s
:
C
las
s
i
f
ier
Data
d
iv
er
s
it
y
H
y
b
r
id
ap
p
r
o
ac
h
r
e
d
ef
in
itio
n
-
m
u
lticla
s
s
i
m
b
a
lan
ce
Mu
lti
-
cla
s
s
i
m
b
a
lan
ce
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r
th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
E
r
ian
to
On
g
k
o
,
Dep
ar
t
m
en
t o
f
I
n
f
o
r
m
atic
s
,
Ak
ad
e
m
i T
ek
n
o
lo
g
i I
n
d
u
s
tr
i I
m
m
an
u
el,
Me
d
an
,
I
n
d
o
n
e
s
ia
.
E
m
ail:
er
ia
n
to
o
n
g
k
o
@
g
m
ail.
c
o
m
1.
I
NT
RO
D
UCT
I
O
N
C
las
s
i
m
b
ala
n
ce
o
cc
u
r
s
i
f
a
class
o
r
s
ev
er
al
clas
s
es
b
ec
o
m
e
u
n
d
er
r
ep
r
esen
ted
s
o
it
is
also
ca
lled
a
m
in
o
r
it
y
cla
s
s
b
ec
au
s
e
it
h
a
s
in
s
ta
n
ce
s
th
at
ar
e
m
u
ch
s
m
aller
th
an
o
th
er
clas
s
es
[
1
]
.
I
n
m
ac
h
i
n
e
lear
n
i
n
g
r
esear
ch
,
class
i
m
b
ala
n
ce
p
r
o
b
lem
s
ar
e
t
h
e
m
ai
n
c
h
all
en
g
e
s
th
a
t
attr
ac
t
th
e
atte
n
t
io
n
o
f
a
n
u
m
b
er
o
f
r
esear
ch
er
s
[
2
]
.
R
esear
ch
o
n
th
i
s
is
s
u
e
is
i
n
cl
u
d
ed
in
th
e
2
0
m
a
in
r
esear
c
h
to
p
ics
th
at
ar
e
th
e
m
o
s
t
in
ter
esti
n
g
i
n
m
ac
h
i
n
e
lear
n
i
n
g
,
esp
ec
iall
y
b
ig
d
ata.
Mi
n
o
r
it
y
C
lass
is
al
s
o
ca
lled
a
p
o
s
iti
v
e
class
b
ec
au
s
e
it
is
a
class
w
i
th
in
ter
e
s
ti
n
g
p
atter
n
s
to
o
b
s
er
v
e.
Fo
r
co
m
p
ar
is
o
n
,
th
e
d
etec
tio
n
o
f
b
r
ea
s
t
ca
n
c
er
s
u
f
f
er
er
s
is
o
f
te
n
a
class
w
it
h
a
s
m
al
l
n
u
m
b
er
o
f
i
n
s
ta
n
ce
s
,
i
f
t
h
e
clas
s
i
f
icatio
n
p
r
o
ce
s
s
f
o
r
d
etec
tio
n
o
f
b
r
ea
s
t
ca
n
ce
r
s
u
f
f
er
er
s
ex
p
er
ien
ce
s
c
lass
i
m
b
ala
n
ce
p
r
o
b
lem
s
th
e
n
t
h
er
e
i
s
a
p
o
s
s
ib
i
lit
y
t
h
at
d
etec
t
io
n
o
f
p
atien
t
s
i
s
n
o
t o
b
tai
n
ed
ev
e
n
th
o
u
g
h
th
e
s
u
f
f
er
er
class
is
v
er
y
in
ter
e
s
ti
n
g
to
o
b
t
ain
[
3
]
.
T
h
er
e
ar
e
a
n
u
m
b
er
o
f
m
et
h
o
d
s
th
at
h
av
e
b
ee
n
p
r
o
p
o
s
ed
to
d
ea
l
w
it
h
cla
s
s
i
m
b
ala
n
c
e
p
r
o
b
lem
s
s
u
c
h
as
r
e
s
a
m
p
li
n
g
,
co
s
t
s
en
s
iti
v
e,
en
s
e
m
b
le
lear
n
i
n
g
,
k
er
n
el
-
b
ased
m
eth
o
d
s
,
a
n
d
ac
tiv
e
lear
n
in
g
m
et
h
o
d
s
[
4
]
.
Mu
lti
-
clas
s
i
m
b
a
lan
ce
p
r
o
b
lem
s
ar
e
f
ar
m
o
r
e
co
m
p
lica
ted
to
h
an
d
le
th
a
n
t
wo
-
class
i
m
b
ala
n
ce
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
HA
R
-
MI
meth
o
d
fo
r
mu
lti
-
cla
s
s
imb
a
la
n
ce
d
d
a
ta
s
ets (
H.
Ha
r
to
n
o
)
823
T
h
e
m
u
l
ti
-
c
lass
i
m
b
ala
n
ce
co
n
d
itio
n
w
ill
b
e
m
o
r
e
d
if
f
ic
u
lt
i
f
t
h
e
d
esire
d
r
es
u
lts
ar
e
as
ac
c
u
r
ate
as
p
o
s
s
ib
le
i
n
ac
co
r
d
an
ce
w
i
th
t
h
e
e
x
is
t
in
g
p
r
o
b
lem
.
O
n
t
h
e
o
th
er
h
a
n
d
,
ap
p
ly
i
n
g
t
h
e
m
et
h
o
d
p
r
o
p
o
s
ed
to
h
a
n
d
le
t
w
o
-
clas
s
i
m
b
alan
ce
p
r
o
b
le
m
s
to
h
a
n
d
le
m
u
lti
-
clas
s
i
m
b
ala
n
ce
p
r
o
b
lem
s
d
o
es
n
o
t
g
et
t
h
e
d
esire
d
r
esu
lt
s
[
5
]
.
I
n
g
en
er
al
,
th
e
al
g
o
r
ith
m
f
o
r
h
a
n
d
lin
g
m
u
lti
-
clas
s
i
m
b
alan
ce
is
to
d
ev
elo
p
an
alg
o
r
ith
m
u
s
ed
f
o
r
h
a
n
d
lin
g
b
i
n
ar
y
cla
s
s
I
m
b
alan
ce
t
h
r
o
u
g
h
t
h
e
d
ec
o
m
p
o
s
itio
n
m
e
th
o
d
[
6
]
.
A
n
o
t
h
er
co
m
m
o
n
m
e
th
o
d
is
to
ad
o
p
t
an
en
s
e
m
b
le
-
b
ased
ap
p
r
o
ac
h
f
o
r
u
s
e
in
h
a
n
d
li
n
g
m
u
lti
-
cla
s
s
i
m
b
ala
n
ce
s
[
4
]
an
d
an
o
t
h
er
w
a
y
i
s
to
ad
ap
t
t
h
e
i
n
tr
i
g
u
e
p
r
o
ce
s
s
b
y
b
u
i
ld
in
g
d
ec
is
io
n
tr
ee
s
[
7
]
.
A
r
elativ
el
y
ea
s
y
w
a
y
to
d
o
is
to
v
ie
w
m
u
lti
-
clas
s
i
m
b
ala
n
ce
as
a
s
u
b
s
et
o
f
b
in
ar
y
p
r
o
b
lem
s
[8
,
9]
.
T
h
e
m
u
lti
-
c
las
s
i
m
b
ala
n
ce
p
r
o
b
lem
s
t
h
at
w
ill
b
e
s
o
l
v
ed
ar
e
p
r
o
b
lem
s
s
u
c
h
a
s
m
a
n
y
m
i
n
o
r
it
y
-
o
n
e
m
aj
o
r
ity
,
o
n
e
m
i
n
o
r
it
y
-
m
a
n
y
m
aj
o
r
it
y
,
an
d
m
a
n
y
m
i
n
o
r
i
t
y
-
m
a
n
y
m
aj
o
r
it
y
[1
0]
.
In
[
1
]
s
u
g
g
e
s
ted
th
at
to
o
v
er
co
m
e
th
e
p
r
o
b
le
m
o
f
i
m
b
alan
ce
clas
s
th
er
e
ar
e
2
(
t
w
o
)
th
i
n
g
s
th
at
n
ee
d
to
b
e
co
n
s
id
er
ed
,
n
a
m
el
y
t
h
o
s
e
r
elate
d
to
th
e
n
u
m
b
er
o
f
class
if
ier
s
an
d
d
iv
er
s
i
t
y
(
d
iv
er
s
i
t
y
)
o
f
d
ata.
In
[
1
1
]
p
r
o
p
o
s
e
th
e
D
y
n
a
m
ic
C
las
s
i
f
ier
Selectio
n
(
DC
S)
m
et
h
o
d
f
o
r
d
ea
lin
g
w
it
h
m
u
lt
i
-
cla
s
s
i
m
b
alan
ce
p
r
o
b
le
m
s
,
b
u
t
it
h
as
t
h
e
d
is
ad
v
an
tag
e
o
f
b
ein
g
a
lar
g
e
n
u
m
b
er
o
f
clas
s
if
ier
s
.
In
[
1
2
]
s
u
g
g
e
s
ted
th
e
D
yna
m
ic
E
n
s
e
m
b
le
Se
lectio
n
(
DE
S)
-
MI
m
et
h
o
d
w
h
ic
h
g
i
v
es
b
etter
r
esu
lts
c
o
m
p
ar
ed
to
th
e
D
y
n
a
m
ic
C
l
ass
i
f
ier
Selectio
n
(
DC
S)
m
et
h
o
d
.
T
h
e
DE
S
-
MI
m
et
h
o
d
f
o
u
n
d
h
as
a
s
m
all
cl
ass
i
f
ier
,
b
u
t
i
n
r
esear
c
h
co
n
d
u
cted
b
y
[
1
3
]
h
as
id
en
t
if
ied
th
at
d
i
v
er
s
it
y
d
ata
o
b
tain
ed
b
y
DE
S
-
MI
i
s
n
o
t
g
o
o
d
en
o
u
g
h
.
T
h
e
H
y
b
r
id
Ap
p
r
o
ac
h
R
ed
ef
in
it
io
n
(
H
A
R
)
m
et
h
o
d
w
h
ich
i
s
a
H
y
b
r
id
E
n
s
e
m
b
le
s
ap
p
r
o
ac
h
ca
n
o
v
er
co
m
e
t
h
e
p
r
o
b
lem
o
f
c
l
as
s
i
m
b
ala
n
ce
w
i
th
a
s
m
all
n
u
m
b
er
o
f
class
i
f
ier
s
a
n
d
g
o
o
d
d
ata
d
iv
er
s
it
y
,
o
n
t
w
o
-
cla
s
s
i
m
b
alan
ce
p
r
o
b
lem
s
[
1
4
,
15]
.
T
h
is
r
esear
ch
w
ill
o
p
ti
m
ize
th
e
H
AR
m
et
h
o
d
s
o
th
at
it
ca
n
b
e
u
s
ed
to
o
v
er
co
m
e
m
u
lti
-
clas
s
i
m
b
alan
ce
p
r
o
b
le
m
s
.
I
n
t
h
e
o
p
ti
m
izatio
n
p
r
o
ce
s
s
t
h
e
p
r
ep
r
o
ce
s
s
i
n
g
s
ta
g
es
w
er
e
ca
r
r
ied
o
u
t
u
s
in
g
th
e
r
an
d
o
m
b
alan
ce
en
s
e
m
b
le
m
et
h
o
d
p
r
o
p
o
s
ed
b
y
[
1
6
]
an
d
d
y
n
a
m
ic
en
s
e
m
b
le
s
e
lectio
n
s
o
t
h
at
a
ca
n
d
id
ate
en
s
e
m
b
le
o
n
m
u
lticla
s
s
p
r
o
b
lem
s
an
d
p
r
o
ce
s
s
i
n
g
s
tag
e
s
w
a
s
ca
r
r
ied
o
u
t
u
s
i
n
g
d
i
f
f
er
e
n
t
co
n
tr
ib
u
t
io
n
s
a
m
p
li
n
g
p
r
o
p
o
s
e
d
b
y
[
1
7
]
an
d
d
y
n
a
m
ic
e
n
s
e
m
b
le
s
elec
tio
n
.
T
h
is
r
esear
c
h
w
il
l
b
e
co
n
d
u
cted
u
s
i
n
g
m
u
l
ti
-
cla
s
s
i
m
b
alan
ce
d
d
atasets
s
o
u
r
ce
d
f
r
o
m
t
h
e
KE
E
L
R
ep
o
s
ito
r
y
[
1
8
]
.
T
h
e
r
esu
lts
o
f
t
h
e
s
t
u
d
y
ar
e
t
h
e
Hy
b
r
id
A
p
p
r
o
ac
h
R
ed
ef
i
n
itio
n
-
M
u
lt
iclas
s
I
m
b
al
an
ce
(
H
AR
-
MI
)
m
et
h
o
d
t
h
at
is
e
x
p
ec
ted
to
o
v
er
co
m
e
m
u
l
ti
-
cla
s
s
i
m
b
ala
n
c
e
w
it
h
b
etter
d
ata
d
iv
er
s
i
t
y
,
s
m
aller
n
u
m
b
er
o
f
clas
s
i
f
ier
s
,
an
d
b
etter
cla
s
s
i
f
ier
p
er
f
o
r
m
an
ce
co
m
p
ar
ed
to
a
DE
S
-
MI
Me
t
h
o
d
.
2.
RE
S
E
ARCH
M
E
T
H
O
D
T
h
is
r
esear
ch
w
i
ll
p
r
o
d
u
ce
th
e
HAR
-
MI
m
et
h
o
d
to
o
v
er
co
m
e
m
u
lt
i
-
c
lass
i
m
b
ala
n
ce
p
r
o
b
l
e
m
s
.
H
AR
Me
th
o
d
w
ill
b
e
ca
r
r
ied
o
u
t
a
n
o
p
ti
m
izatio
n
p
r
o
ce
s
s
w
i
th
HAR
-
MI
m
et
h
o
d
s
o
t
h
at
it
ca
n
h
an
d
le
m
u
lt
i
-
cla
s
s
i
m
b
alan
ce
p
r
o
b
le
m
s
b
y
ad
d
in
g
ca
p
ab
ilit
ies
f
r
o
m
H
AR
m
et
h
o
d
t
o
d
eter
m
i
n
e
ca
n
d
id
ate
en
s
e
m
b
les
b
y
u
s
in
g
d
y
n
a
m
ic
en
s
e
m
b
le
s
e
lectio
n
o
n
m
i
n
o
r
it
y
cla
s
s
es
a
n
d
m
aj
o
r
it
y
class
e
s
s
o
th
at
t
h
e
y
ca
n
r
ec
o
g
n
ize
ea
ch
s
u
b
s
et
o
f
m
in
o
r
it
y
a
n
d
m
aj
o
r
ity
clas
s
es
b
ased
o
n
2
-
Di
m
en
s
io
n
al
Data
s
ets
p
r
o
p
o
s
ed
b
y
Sáez
et
a
l
.
[
1
0
]
.
T
h
e
r
esu
lt
s
o
f
H
AR
-
MI
m
et
h
o
d
ar
e
ex
p
ec
ted
to
o
b
tain
b
etter
d
ata
d
iv
er
s
it
y
a
n
d
also
a
s
m
all
n
u
m
b
er
o
f
cla
s
s
i
f
i
er
s
.
T
h
e
s
tag
es o
f
r
esear
c
h
co
n
d
u
c
ted
b
y
r
esear
ch
er
s
f
r
o
m
t
h
i
s
s
t
u
d
y
ca
n
b
e
s
ee
n
i
n
Fi
g
u
r
e
1
.
Fig
u
r
e
1
.
Stag
es o
f
r
esear
ch
m
eth
o
d
s
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
18
,
No
.
2
,
A
p
r
il 2
0
2
0
:
8
2
2
-
8
2
9
824
I
n
Fi
g
u
r
e
1
,
it
ca
n
b
e
s
ee
n
t
h
a
t
th
e
p
r
o
ce
s
s
t
h
at
o
cc
u
r
s
in
th
e
d
ataset
s
elec
tio
n
an
d
p
r
ep
ar
atio
n
s
tag
e
is
d
eter
m
i
n
ed
b
y
t
h
e
i
m
b
ala
n
ce
d
atase
t
w
i
th
v
ar
y
in
g
i
m
b
alan
ce
r
atio
.
T
h
e
n
e
x
t
p
r
o
ce
s
s
is
p
r
ep
r
o
ce
s
s
in
g
.
T
h
e
p
r
o
ce
s
s
o
f
h
an
d
li
n
g
t
h
e
m
u
lti
-
clas
s
i
m
b
ala
n
ce
w
ill
b
eg
in
w
i
th
t
h
e
p
r
ep
r
o
ce
s
s
in
g
s
t
ag
e.
T
h
e
p
u
r
p
o
s
e
o
f
th
i
s
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
i
s
t
o
r
ed
u
ce
th
e
n
u
m
b
er
o
f
class
if
ier
s
.
W
h
er
e
th
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
w
i
ll
b
e
d
o
n
e
u
s
i
n
g
t
h
e
R
an
d
o
m
B
alan
ce
E
n
s
e
m
b
le
m
et
h
o
d
an
d
D
y
n
a
m
ic
E
n
s
e
m
b
le
Selectio
n
.
T
h
e
R
a
n
d
o
m
B
ala
n
ce
E
n
s
e
m
b
le
Me
t
h
o
d
w
il
l
u
s
e
R
an
d
o
m
u
n
d
er
Sa
m
p
lin
g
a
n
d
SMOT
E
B
o
o
s
t
.
T
h
e
r
esu
lts
o
f
th
e
p
r
ep
r
o
ce
s
s
in
g
s
tag
e
ar
e
in
th
e
f
o
r
m
o
f
a
p
r
ep
r
o
ce
s
s
in
g
d
ata
s
et
w
h
ic
h
w
il
l
t
h
en
p
r
o
ce
ed
to
th
e
p
r
o
ce
s
s
in
g
s
tag
e.
I
m
p
le
m
e
n
tatio
n
an
d
v
alid
ati
o
n
o
f
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
e
ac
h
ex
p
er
i
m
en
t
w
as
ca
r
r
ied
o
u
t
u
s
in
g
1
0
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
an
d
co
m
p
ar
ed
w
i
th
th
e
DE
S
-
MI
m
et
h
o
d
w
h
ic
h
i
s
v
er
y
g
o
o
d
in
d
ea
li
n
g
w
i
th
m
u
lt
i
-
c
lass
i
m
b
alan
ce
p
r
o
b
le
m
s
.
2
.
1
.
P
re
pro
ce
s
s
ing
a
nd
pro
ce
s
s
ing
s
t
a
g
e
in H
AR
-
M
I
m
et
ho
d
T
h
e
p
r
e
p
r
o
ce
s
s
in
g
s
ta
g
e
was
ca
r
r
ied
o
u
t
u
s
in
g
th
e
R
an
d
o
m
B
alan
ce
E
n
s
e
m
b
le
s
Me
th
o
d
an
d
D
y
n
a
m
ic
E
n
s
e
m
b
le
Selec
t
io
n
.
T
h
e
p
s
eu
d
o
co
d
e
o
f
th
is
s
t
ag
e
is
as
f
o
llo
w
s
.
Require
: Set
S o
f
ex
a
m
p
les(
x
1
,y
1
)
E
ns
ure
: N
e
w
s
et
S
’
o
f
ex
a
m
p
l
es
w
ith
R
a
n
d
o
m
B
a
la
n
ce
an
d
D
y
n
a
m
ic
E
n
s
e
m
b
le
Select
io
n
1
:
to
ta
lS
iz
e
←
|
S
|
2
: D
eter
m
in
e
k
as t
h
e
n
u
m
b
er
o
f
N
ea
r
es
t Neig
h
b
o
r
3
: Fo
r
A
ll
Sa
m
p
les i
n
S d
o
4
:
Dete
r
m
i
n
e
th
e
B
o
r
d
er
lin
e
o
f
P
o
s
itiv
e
o
r
Min
o
r
it
y
C
las
s
as
E
O
+
5
:
Dete
r
m
i
n
e
th
e
B
o
r
d
er
lin
e
o
f
Neg
ati
v
e
o
r
Ma
j
o
r
ity
C
la
s
s
as
E
O
−
6
: E
n
d
Fo
r
7
: Fo
r
A
ll Sa
m
p
les i
n
E
O
+
d
o
8
:
C
alcu
late
th
e
cn
(
e
)
i
as
n
eig
b
o
r
h
o
o
d
v
alu
e
f
o
r
ea
ch
s
a
m
p
le
9
:
Or
d
er
A
s
ce
n
d
i
n
g
t
h
e
s
a
m
p
le
ac
co
r
d
in
g
to
th
e
cn
(
e)
i
1
0
: E
n
d
Fo
r
1
1
: B
u
ild
in
g
a
ca
n
d
id
ate
en
s
e
m
b
le
f
o
r
S
a
fe
,
B
o
r
d
erli
n
e
,
R
a
r
e
,
d
an
Ou
tlier
ac
co
r
d
in
g
to
k
v
alu
e
1
2
:
T
ak
e
a
ca
n
d
id
ate
en
s
e
m
b
l
e
o
f
S
a
fe
,
B
o
r
d
erli
n
e
,
R
a
r
e
,
d
a
n
Ou
tlier
to
S
P
1
3
: Fo
r
A
ll Sa
m
p
le
s
in
E
O
−
do
1
4
:
T
ak
e
a
ca
n
d
id
ate
en
s
e
m
b
le
to
SN
1
5
: E
n
d
Fo
r
1
6
: A
d
d
I
n
s
tan
ce
f
r
o
m
w
i
th
S
|
y
i
=+
1
to
S
P
1
7
: A
d
d
I
n
s
tan
ce
f
r
o
m
w
i
th
S
|
y
i
=
-
1
to
S
N
1
8
: Calcu
late
t
h
e
s
ize
o
f
Ma
j
o
r
it
y
C
la
s
s
f
r
o
m
S
N
1
9
: Calcu
late
t
h
e
s
ize
o
f
Ma
j
o
r
it
y
C
la
s
s
f
r
o
m
S
P
2
0
:
n
ewM
a
jo
r
ityS
iz
e
←R
a
n
d
o
m
in
teg
er
b
et
w
ee
n
2
an
d
to
ta
lS
iz
e
-
2
2
1
:
n
ewM
in
o
r
ityS
iz
e
←
to
ta
lS
iz
e
–
n
ewM
a
jo
r
ityS
iz
e
8
: if
n
e
wM
aj
o
r
ity
Size
2
2
:
if
n
ewM
a
jo
r
ityS
iz
e
<
ma
jo
r
ityS
iz
e
t
hen
2
3
:
S
’
←
S
P
2
4
:
S
’
w
ill f
ill
w
it
h
a
r
an
d
o
m
in
s
ta
n
ce
f
r
o
m
S
N
2
5
:
C
r
ea
te
n
ewM
in
o
r
ityS
iz
e
–
min
o
r
ityS
iz
e
ar
tif
icial
2
6
:
else
2
7
:
S
’
←
S
N
2
8
:
S
’
w
ill f
ill
w
it
h
a
r
an
d
o
m
in
s
ta
n
ce
f
r
o
m
S
P
2
9
:
c
r
ea
te
n
ewM
a
jo
r
ityS
iz
e
–
ma
jo
r
ityS
iz
e
ar
tif
icial
3
0
:
end if
3
1
:
re
t
urn
S
’
B
ased
o
n
t
h
e
p
s
e
u
d
o
co
d
e
ab
o
v
e,
it
ca
n
b
e
s
ee
n
t
h
at
in
th
e
p
r
ep
r
o
ce
s
s
in
g
s
ta
g
e
w
a
s
ca
r
r
ied
o
u
t
u
s
in
g
R
an
d
o
m
U
n
d
er
S
a
m
p
li
n
g
a
n
d
SMOT
E
B
o
o
s
t.
I
n
th
e
R
an
d
o
m
U
n
d
er
Sa
m
p
l
i
n
g
p
r
o
ce
s
s
th
e
D
y
n
a
m
i
c
E
n
s
e
m
b
le
Selec
tio
n
p
r
o
ce
s
s
w
il
l
ta
k
e
t
h
e
f
o
r
m
o
f
b
o
r
d
er
lin
e
d
eter
m
in
atio
n
f
o
r
m
in
o
r
it
y
an
d
m
aj
o
r
ity
clas
s
.
T
h
en
f
o
r
s
a
m
p
les
th
at
ar
e
in
t
h
e
b
o
r
d
er
lin
e
m
i
n
o
r
it
y
clas
s
E
O
+
,
th
e
n
eig
h
b
o
r
h
o
o
d
v
alu
e
ca
lcu
latio
n
p
r
o
ce
s
s
cn
(
e)
w
ill
b
e
p
e
r
f
o
r
m
ed
,
t
h
e
n
it
w
i
ll
b
e
s
o
r
ted
asce
n
d
in
g
to
d
eter
m
in
e
t
h
e
ca
n
d
id
ate
en
s
e
m
b
le
f
o
r
Saf
e,
B
o
r
d
e
r
lin
e,
R
ar
e,
an
d
O
u
tlier
,
th
e
n
t
h
e
ca
n
d
id
ate
e
n
s
e
m
b
le
w
il
l
b
e
in
cl
u
d
ed
in
th
e
SP
.
N
ex
t
f
o
r
th
e
s
a
m
p
le
th
at
is
i
n
t
h
e
b
o
r
d
er
lin
e
th
e
m
a
j
o
r
class
w
ill b
e
en
ter
ed
in
t
o
SN
.
Af
ter
t
h
at,
t
h
e
p
r
o
ce
s
s
w
i
ll
c
o
n
tin
u
e
w
i
th
th
e
R
a
n
d
o
m
B
a
lan
ce
E
n
s
e
m
b
le
Me
t
h
o
d
,
w
h
i
ch
w
ill
b
e
b
ased
o
n
th
e
r
esu
lt
s
o
f
th
e
D
y
n
a
m
ic
E
n
s
e
m
b
le
Selectio
n
.
T
h
e
p
r
o
ce
s
s
s
tar
ts
w
it
h
t
h
e
d
eter
m
i
n
atio
n
o
f
Ma
j
o
r
ity
an
d
Min
o
r
it
y
Size.
T
h
en
b
ased
o
n
th
e
d
eter
m
in
a
tio
n
o
f
th
e
s
ize,
an
i
m
b
ala
n
ce
cla
s
s
w
ill
b
e
h
an
d
led
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
HA
R
-
MI
meth
o
d
fo
r
mu
lti
-
cla
s
s
imb
a
la
n
ce
d
d
a
ta
s
ets (
H.
Ha
r
to
n
o
)
825
I
f
th
e
s
ize
o
f
t
h
e
n
e
w
Ma
j
o
r
i
t
y
C
las
s
is
g
r
ea
ter
t
h
an
th
e
n
e
w
Ma
j
o
r
it
y
C
las
s
,
th
is
m
ea
n
s
th
a
t
th
e
Min
o
r
it
y
C
las
s
is
lar
g
er
th
a
n
th
e
Ma
j
o
r
it
y
C
la
s
s
an
d
p
ar
t
o
f
th
e
Mi
n
o
r
it
y
C
la
s
s
i
n
s
ta
n
ce
w
ill
b
e
tak
en
t
o
m
o
v
e
to
th
e
Ma
j
o
r
ity
C
la
s
s
a
n
d
v
ice
v
er
s
a.
Dete
r
m
in
a
tio
n
o
f
th
e
s
a
m
p
le
w
ill
b
e
d
o
n
e
b
y
R
an
d
o
m
Un
d
er
Sa
m
p
li
n
g
an
d
th
e
d
eter
m
i
n
atio
n
o
f
th
e
in
s
tan
ce
t
h
at
w
il
l
b
e
tr
an
s
f
er
r
ed
w
ill
b
e
d
o
n
e
w
ith
SMOT
E
B
o
o
s
t.
T
h
e
Pro
ce
s
s
in
g
s
tag
e
s
w
as
c
ar
r
ied
o
u
t
u
s
in
g
th
e
D
i
f
f
er
en
t
C
o
n
tr
ib
u
tio
n
Sa
m
p
l
in
g
an
d
D
y
n
a
m
ic
E
n
s
e
m
b
le
Selectio
n
.
T
h
e
p
s
eu
d
o
co
d
e
o
f
th
is
s
tag
e
i
s
as
f
o
llo
w
s
.
1
:
I
np
ut
:
S
:
Tr
a
in
in
g
S
et;
T:
Nu
m
b
er
o
f
I
ter
atio
n
s
;
n
: B
o
o
t
s
tr
ap
Size
;
k
:
n
eig
h
b
o
r
s
2
:
O
utput
:
B
ag
g
ed
C
las
s
i
f
ier
:
H
(
x
)
=
s
ig
n
(
∑
ℎ
(
)
=
1
)
w
h
er
e
h
t
[
-
1
,
1
]
ar
e
th
e
in
d
u
ce
d
clas
s
i
f
ier
s
3
:
P
ro
ce
s
s
:
4
:
Fo
r
A
ll Sa
m
p
les i
n
S d
o
5
:
Dete
r
m
i
n
e
th
e
B
o
r
d
er
lin
e
o
f
P
o
s
itiv
e
o
r
Min
o
r
it
y
C
las
s
as
E
O
+
6
:
Dete
r
m
i
n
e
th
e
B
o
r
d
er
lin
e
o
f
Neg
ati
v
e
o
r
Ma
j
o
r
ity
C
la
s
s
as
E
O
−
7
: E
n
d
Fo
r
8
: Fo
r
A
ll Sa
m
p
les i
n
E
O
+
d
o
9
:
C
alcu
late
th
e
cn
(
e
)
i
as
n
eig
b
o
r
h
o
o
d
v
alu
e
f
o
r
ea
ch
s
a
m
p
le
1
0
:
Or
d
er
A
s
ce
n
d
i
n
g
t
h
e
s
a
m
p
le
ac
co
r
d
in
g
to
th
e
cn
(
e)
i
1
1
: E
n
d
Fo
r
1
2
: B
u
ild
in
g
a
ca
n
d
id
ate
en
s
e
m
b
le
f
o
r
S
a
fe
,
B
o
r
d
erli
n
e
,
R
a
r
e
,
d
an
Ou
tlier
ac
co
r
d
in
g
to
k
v
alu
e
1
3
:
T
ak
e
a
ca
n
d
id
ate
en
s
e
m
b
l
e
o
f
S
a
fe
,
B
o
r
d
erli
n
e
,
R
a
r
e
,
d
a
n
Ou
tlier
to
SP
1
4
: Fo
r
A
ll Sa
m
p
le
s
in
E
O
−
do
1
5
:
T
ak
e
a
ca
n
d
id
ate
en
s
e
m
b
le
to
SN
1
6
: E
n
d
Fo
r
1
7
:
f
o
r
i
=
1
to
Nu
m
b
er
o
f
I
n
s
tan
ce
in
P
r
ep
r
o
ce
s
s
ed
Data
s
et
do
1
8
:
A
d
d
P
r
ep
r
o
ce
s
s
ed
Data
s
et
to
S
i
1
9
B
-
SVM
w
ill d
o
f
o
r
class
if
y
i
n
g
S
i
2
0
:
Dete
r
m
in
e
t
h
e
Ma
j
o
r
it
y
C
las
s
2
1
:
Dete
r
m
in
e
t
h
e
Mi
n
o
r
it
y
C
las
s
2
2
:
Fo
r
A
ll I
n
s
tan
ce
i
n
Ma
j
o
r
ity
C
las
s
d
o
2
3
:
New
SV
Sets
[
]
w
ill
f
o
r
m
b
y
c
h
ec
k
in
g
a
n
d
d
elete
th
e
n
o
is
e
in
SV Set
s
2
4
:
New
NS
VSet
s
[
]
w
i
ll
f
o
r
m
b
y
m
u
l
tip
le
R
U
S
2
5
:
en
d
w
h
ile
2
6
:
Fo
r
A
ll in
s
ta
n
ce
f
r
o
m
n
e
w
S
V
Sets
a
n
d
NSV
d
o
2
7
:
C
r
ea
te
an
in
s
ta
n
ce
f
o
r
Ma
jo
r
ity
C
la
s
s
2
8
:
E
n
d
Fo
r
2
9
:
Fo
r
A
ll I
n
s
tan
ce
i
n
Mi
n
o
r
ity
C
las
s
d
o
3
0
:
SMO
T
E
B
o
o
s
t P
r
o
ce
s
s
f
o
r
SV Se
ts
a
n
d
cr
ea
te
SMOT
E
Sets
3
1
:
en
d
w
h
ile
3
2
:
Fo
r
A
ll SMOT
E
Sets
an
d
Ne
w
NS
VSet
s
d
o
3
3
:
New
P
o
s
itiv
e
Sa
m
p
l
eSets
3
4
:
E
n
d
Fo
r
3
5
:
Fo
r
A
ll Ne
w
Ne
g
at
i
v
eS
a
m
p
leSet
s
an
d
Ne
w
P
o
s
iti
v
eSa
m
p
leSet
s
d
o
3
6
:
R
esu
ltData
Set
3
7
:
E
n
d
Fo
r
3
8
: E
n
d
Fo
r
Af
ter
th
e
p
r
ep
r
o
ce
s
s
in
g
d
ataset
is
g
en
er
ated
,
th
e
D
y
n
a
m
ic
E
n
s
e
m
b
le
Selectio
n
p
r
o
ce
s
s
w
il
l
o
cc
u
r
at
th
e
i
n
itial
s
tag
e
f
o
r
b
o
r
d
er
lin
e
d
eter
m
i
n
atio
n
o
f
m
i
n
o
r
it
y
an
d
m
aj
o
r
it
y
clas
s
.
T
h
en
th
e
n
ex
t
s
tep
w
ill
b
e
th
e
Di
f
f
er
e
n
tial
C
o
n
tr
ib
u
tio
n
Sa
m
p
li
n
g
p
r
o
ce
s
s
w
h
er
e
b
o
t
h
m
aj
o
r
it
y
class
e
s
an
d
m
i
n
o
r
it
y
clas
s
es
w
ill
b
e
d
iv
id
ed
in
to
SV
Sets
an
d
NS
V
Sets
.
NSV
Sets
i
n
th
e
Ne
g
a
tiv
e
Sa
m
p
le
w
ill
u
n
d
er
g
o
a
Mu
ltip
le
R
U
S
p
r
o
ce
s
s
,
w
h
ile
S
V
Sets
i
n
t
h
e
P
o
s
itiv
e
Sa
m
p
le
w
ill e
x
p
er
ien
ce
a
SM
OT
E
B
o
o
s
t.
2
.
2
.
Da
t
a
div
er
s
it
y
I
n
th
e
e
n
s
e
m
b
le
lear
n
i
n
g
p
r
o
ce
s
s
,
in
r
ea
li
t
y
i
f
t
h
er
e
is
a
clas
s
if
ier
t
h
at
ca
n
g
u
ar
an
tee
th
at
t
h
er
e
is
n
o
m
is
c
lass
if
ica
tio
n
,
an
e
n
s
e
m
b
l
e
p
r
o
ce
s
s
is
n
o
t
n
ee
d
ed
o
n
th
e
class
i
f
ier
.
T
h
e
en
s
e
m
b
le
p
r
o
ce
s
s
in
t
h
e
clas
s
i
f
ier
o
cc
u
r
s
in
t
h
e
h
o
p
e
th
at
b
et
ter
r
esu
lts
ca
n
b
e
o
b
tain
ed
.
A
s
s
u
m
i
n
g
th
a
t
if
t
h
er
e
is
a
m
is
cla
s
s
i
f
icat
io
n
o
f
th
e
cla
s
s
i
f
ier
i
n
a
p
ar
t
it
ca
n
b
e
co
v
er
ed
b
y
m
er
g
i
n
g
w
i
th
o
th
er
clas
s
i
f
ier
s
t
h
at
also
m
is
c
lass
if
i
ca
tio
n
i
n
o
th
er
p
ar
ts
[
1
9
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
18
,
No
.
2
,
A
p
r
il 2
0
2
0
:
8
2
2
-
8
2
9
826
A
cc
o
r
d
in
g
to
Díez
-
P
asto
r
,
R
o
d
r
íg
u
ez
,
Gar
cía
-
Oso
r
io
,
an
d
Ku
n
ch
e
v
a
[
1
6
]
it
is
i
m
p
o
r
tan
t
to
p
a
y
atten
tio
n
to
th
e
d
i
v
er
s
it
y
o
f
d
ata
in
h
an
d
li
n
g
i
m
b
ala
n
ce
cla
s
s
es.
T
h
is
m
ea
n
s
t
h
at
atte
m
p
t
ed
m
is
cla
s
s
i
f
icatio
n
p
r
o
d
u
ce
d
b
y
ea
c
h
cla
s
s
i
f
ier
is
as
s
m
a
ll
a
s
p
o
s
s
ib
le
a
n
d
i
f
th
er
e
is
m
is
c
lass
if
icatio
n
it
i
s
e
x
p
ec
ted
to
o
cc
u
r
o
n
d
if
f
er
e
n
t
o
b
j
ec
ts
o
r
p
ar
ts
[
2
0
]
.
Su
p
p
o
s
e
th
at
Z
=
{
1
,
.
.
.
,
}
w
h
ic
h
is
a
d
ataset
th
at
i
s
i
n
th
e
d
ec
i
s
io
n
r
eg
io
n
ℜ
,
s
o
th
at
∈
ℜ
it
is
an
in
s
tan
ce
i
n
v
o
lv
ed
in
th
e
clas
s
i
f
icatio
n
p
r
o
b
le
m
.
T
h
en
th
e
o
u
tp
u
t
o
f
th
e
cl
ass
i
f
ier
as a
class
i
f
ier
p
air
ed
co
m
p
ar
i
s
o
n
m
a
tr
ix
(
r
elatio
n
s
h
ip
p
air
w
is
e
class
if
ier
)
ca
n
b
e
s
ee
n
i
n
T
ab
le
1
.
T
ab
le
1
.
R
elatio
n
s
h
ip
p
air
w
is
e
class
i
f
ier
m
atr
i
x
[
2
0
]
D
k
C
o
rre
c
t
(
1
)
D
k
Wr
o
n
g
(
0
)
D
i
C
o
rre
c
t
(
1
)
N
11
N
10
D
i
Wr
o
n
g
(
0
)
N
01
N
10
Div
er
s
it
y
d
ata
ca
n
b
e
ca
lcu
late
d
u
s
in
g
Q
-
Sta
tis
tic
s
[
2
1
]
.
Q
i
,k
=
11
00
−
01
10
11
00
+
01
10
(1
)
2
.
3
.
Cla
s
s
if
ier
C
las
s
i
f
ier
s
ca
n
g
en
er
all
y
b
e
d
ef
i
n
ed
as
Dec
i
s
io
n
R
eg
io
n
ℜ
th
at
p
lace
an
o
b
j
ec
t
in
to
a
s
e
t
c
la
s
s
Ω
,
w
h
er
e
Ω
co
n
s
i
s
ts
o
f
cla
s
s
1
,
2
,
u
n
ti
l
.
T
h
is
ca
n
b
e
s
ee
n
in
(
9
)
[
2
0
]
.
:
ℜ
→
(
2
)
W
h
er
e
D
is
th
e
cla
s
s
i
f
ier
an
d
i
s
th
e
s
et
o
f
ea
ch
p
o
in
t i
n
t
h
e
d
ec
is
io
n
r
eg
io
n
ℜ
w
h
ic
h
i
s
in
te
n
d
ed
f
o
r
cla
s
s
.
2
.
4
.
Cla
s
s
if
ier
perf
o
r
m
a
nce
R
OC
C
u
r
v
e
i
s
o
n
e
s
tatis
t
i
ca
l
m
e
th
o
d
t
h
at
is
o
f
te
n
u
s
ed
to
d
eter
m
i
n
e
t
h
e
p
er
f
o
r
m
a
n
ce
o
f
a
class
i
f
ier
.
T
h
is
cu
r
v
e
is
g
en
er
ated
b
y
p
lo
tti
n
g
th
e
tr
u
e
p
o
s
itiv
e
f
r
ac
tio
n
o
f
a
p
o
s
iti
v
e
s
a
m
p
le
in
t
h
e
Y
ax
is
w
it
h
th
e
f
alse
p
o
s
iti
v
e
f
r
ac
tio
n
o
f
a
n
e
g
a
ti
v
e
s
a
m
p
le
(
Fals
e
P
o
s
itiv
e
R
ate)
in
t
h
e
X
ax
i
s
[
2
2
]
.
T
h
e
co
n
ce
p
t
s
o
f
T
r
u
e
Po
s
itiv
e
an
d
Fal
s
e
P
o
s
itiv
e
ca
n
b
e
s
ee
n
i
n
th
e
C
o
n
f
u
s
io
n
Ma
tr
ix
as c
a
n
b
e
s
ee
n
i
n
T
ab
le
2
[
2
3
]
.
T
ab
le
2
.
C
o
n
f
u
s
io
n
m
atr
i
x
[
2
4
]
C
l
a
ssi
f
i
e
d
a
s
p
o
si
t
i
v
e
C
l
a
ssi
f
i
e
d
a
s
n
e
g
a
t
i
v
e
Po
si
t
i
v
e
s
a
m
p
l
e
s
T
ru
e
Po
s
i
t
i
v
e
(
T
P
)
Fa
l
s
e
N
e
g
a
t
i
v
e
(
F
N
)
N
e
g
a
t
i
v
e
s
a
m
p
l
e
s
Fa
l
s
e
P
o
s
i
t
i
v
e
(
F
P
)
T
ru
e
N
e
g
a
t
i
v
e
(
T
N
)
T
h
e
n
u
m
b
er
o
f
p
er
f
o
r
m
a
n
ce
cl
ass
i
f
ier
m
ea
s
u
r
e
m
e
n
t p
ar
a
m
et
er
s
in
th
e
t
w
o
cla
s
s
p
r
o
b
le
m
s
a
r
e
as f
o
llo
w
s
[
2
5
]
.
T
P
r
a
te
=
+
(3
)
F
Pr
a
te
=
+
(
4
)
TN
r
a
te
=
+
(
5
)
R
e
c
a
l
l
=
TP
r
a
te
(
6
)
P
r
e
c
isi
on
=
P
PV
a
l
ue
=
+
(
7
)
F
-
M
e
a
s
ure
=
2
+
(
8
)
G
-
M
e
a
n
=
√
.
(
9
)
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
HA
R
-
MI
meth
o
d
fo
r
mu
lti
-
cla
s
s
imb
a
la
n
ce
d
d
a
ta
s
ets (
H.
Ha
r
to
n
o
)
827
T
r
u
e
P
o
s
itiv
e
R
ate
(
T
P
r
ate)
is
s
tated
a
s
a
r
ec
all
w
h
ic
h
s
tates
t
h
e
p
er
ce
n
tag
e
o
f
d
ata
ca
p
tu
r
ed
is
r
elev
a
n
t
d
ata.
P
o
s
itiv
e
P
r
ed
ictiv
e
Val
u
e
(
P
P
Valu
e)
is
s
tated
as
P
r
ec
is
io
n
w
h
ic
h
s
ta
t
es
th
e
p
er
ce
n
ta
g
e
o
f
r
elev
a
n
t
d
ata
id
en
ti
f
ied
t
o
b
e
tak
en
.
F
-
Me
a
s
u
r
e
s
tate
s
th
e
h
ar
m
o
n
ic
av
er
a
g
e
v
al
u
e
b
et
w
ee
n
r
ec
all
an
d
p
r
ec
is
io
n
.
T
h
e
F
-
Me
a
s
u
r
e
v
alu
e
is
u
s
u
all
y
s
m
aller
t
h
a
n
2
,
th
e
h
i
g
h
er
t
h
e
v
al
u
e
o
f
F
-
Me
asu
r
e
s
tates
t
h
at
b
o
th
r
ec
all
an
d
p
r
ec
is
io
n
ar
e
q
u
ite
h
ig
h
.
G
-
Me
an
s
o
n
th
e
o
th
er
h
an
d
s
tates
t
h
e
b
alan
c
e
b
et
w
ee
n
p
o
s
iti
v
e
s
a
m
p
les
a
n
d
n
eg
a
tiv
e
s
a
m
p
l
es
[
2
3
]
.
P
e
r
f
o
r
m
a
n
ce
m
ea
s
u
r
e
m
en
t
i
n
m
u
lti
cla
s
s
i
m
b
a
l
an
ce
is
b
asicall
y
a
m
o
d
if
ica
tio
n
o
f
t
w
o
clas
s
p
r
o
b
lem
s
,
a
n
d
i
n
g
en
er
al
t
h
e
r
e
ar
e
2
(
t
w
o
)
p
ar
a
m
eter
s
u
s
ed
,
n
a
m
el
y
:
MAv
A
an
d
MFM
[
2
6
]
.
=
∑
=
1
(
1
0
)
w
h
er
e
m
is
th
e
n
u
m
b
er
o
f
class
es
a
n
d
s
tan
d
s
f
o
r
th
e
ac
cu
r
ac
y
r
ate
f
o
r
th
e
class
I
an
d
MAv
A
is
th
e
a
v
er
ag
e
v
alu
e
o
f
ac
cu
r
a
c
y
.
=
−
(
1
1
)
w
h
er
e
MFM
i
s
th
e
m
u
lti
-
clas
s
F
-
Me
a
s
u
r
e.
3.
RE
SU
L
T
S AN
D
AN
AL
Y
SI
S
3
.
1
.
Da
t
a
s
et
des
cr
iptio
n
T
h
is
s
tu
d
y
u
s
e
s
a
m
u
lti
-
clas
s
i
m
b
ala
n
ce
d
d
ataset
th
at
is
s
o
u
r
ce
d
f
r
o
m
th
e
KE
E
L
R
ep
o
s
ito
r
y
.
T
h
e
d
ataset
s
elec
ted
in
th
is
s
t
u
d
y
h
a
s
r
ep
r
esen
ted
a
lo
w
,
m
e
d
iu
m
a
n
d
h
i
g
h
i
m
b
alan
ce
r
ati
o
.
Fo
r
d
atasets
w
it
h
a
lo
w
i
m
b
ala
n
ce
r
atio
ar
e
B
al
an
ce
Scale
d
atasets
,
d
atasets
w
it
h
m
o
d
er
ate
i
m
b
ala
n
ce
r
ati
o
ar
e
C
ar
E
v
alu
at
i
o
n
d
atasets
,
a
n
d
d
ataset
w
it
h
h
i
g
h
i
m
b
a
lan
ce
r
atio
ar
e
R
ed
W
in
e
Q
u
alit
y
d
atase
ts
,
E
co
li
,
an
d
P
ag
eb
lo
ck
s
.
Data
s
et
d
escr
ip
tio
n
ca
n
b
e
s
ee
n
in
T
ab
l
e
3
[
1
8
]
.
T
ab
le
3
.
Data
s
et
d
escr
ip
tio
n
[
1
8
]
D
a
t
a
se
t
#
Ex
#
A
t
t
s
D
i
st
r
i
b
u
t
i
o
n
o
f
c
l
a
ss
IR
B
a
l
a
n
c
e
s
c
a
l
e
6
2
5
4
2
8
8
/
4
9
/
2
8
8
5
.
8
8
C
a
r
e
v
a
l
u
a
t
i
o
n
1
7
2
8
6
3
8
4
/
6
9
/
1
2
1
0
/
6
5
1
8
.
6
2
R
e
d
w
i
n
e
q
u
a
l
i
t
y
1
5
9
9
11
1
0
/
5
3
/
6
8
1
/
6
3
8
/
1
9
9
/
1
8
6
8
.
1
Ec
o
l
i
3
3
6
7
2
/
2
/
5
/
2
0
/
3
5
/
5
2
/
7
7
/
1
4
3
7
1
.
5
P
a
g
e
b
l
o
c
k
s
5
4
8
10
3
/
8
/
1
2
/
3
3
/
4
9
2
1
6
4
3
.
2
.
T
esting
re
s
ult
T
h
e
f
ir
s
t
te
s
t
i
s
to
o
b
tain
a
co
m
p
ar
is
o
n
o
f
th
e
n
u
m
b
er
o
f
clas
s
i
f
ier
an
d
d
i
v
er
s
it
y
d
at
a
o
b
tain
ed
b
y
u
s
i
n
g
H
A
R
-
MI
an
d
DE
S
-
MI
m
e
th
o
d
.
T
esti
n
g
o
f
ea
c
h
m
et
h
o
d
w
ill
b
e
ca
r
r
ied
o
u
t
as
m
an
y
as
1
0
test
in
g
f
o
r
ea
ch
d
ataset.
T
h
e
av
er
ag
e
t
est r
esu
l
ts
ca
n
b
e
s
ee
n
i
n
T
ab
l
e
4
.
T
ab
le
4
.
T
esti
n
g
r
es
u
lt f
o
r
n
u
m
b
er
o
f
cla
s
s
i
f
ier
a
n
d
d
ata
d
iv
er
s
it
y
f
o
r
ea
ch
m
et
h
o
d
D
a
t
a
se
t
H
A
R
-
M
I
m
e
t
h
o
d
D
ES
-
M
I
m
e
t
h
o
d
N
u
m
b
e
r
o
f
C
l
a
ssi
f
i
e
r
D
a
t
a
D
i
v
e
rs
i
t
y
(Q
-
S
t
a
t
i
s
t
i
c
s)
N
u
m
b
e
r
o
f
C
l
a
ssi
f
i
e
r
D
a
t
a
D
i
v
e
rs
i
t
y
(Q
-
S
t
a
t
i
s
t
i
c
s)
B
a
l
a
n
c
e
s
c
a
l
e
1
9
1
.
6
0
.
3
9
7
1
9
7
.
2
0
.
4
2
1
C
a
r
e
v
a
l
u
a
t
i
o
n
4
7
1
.
6
0
.
4
5
7
4
8
7
.
9
0
.
4
6
1
R
e
d
w
i
n
e
q
u
a
l
i
t
y
3
9
7
.
8
0
.
4
3
1
3
9
5
.
3
0
.
4
1
1
Ec
o
l
i
9
1
.
1
0
.
3
9
7
1
2
1
.
2
0
.
4
1
3
P
a
g
e
b
l
o
c
k
s
1
1
7
.
8
0
.
4
4
1
1
1
9
.
6
0
.
4
5
6
B
ased
o
n
th
e
r
es
u
lts
in
T
ab
le
4
,
it
ca
n
b
e
s
ee
n
t
h
at
H
AR
-
MI
Me
th
o
d
g
i
v
es
b
etter
r
es
u
l
ts
o
n
b
etter
d
ata
d
iv
er
s
it
y
i
n
th
e
t
h
r
ee
d
atasets
w
h
en
co
m
p
ar
ed
w
i
th
D
E
S
-
MI
Me
th
o
d
.
T
h
e
test
r
esu
lts
f
o
r
th
e
H
AR
-
MI
m
et
h
o
d
class
i
f
ier
ar
e
b
etter
in
th
e
B
alan
ce
Scale,
C
ar
E
v
a
lu
a
tio
n
,
E
co
li,
an
d
P
ag
eb
lo
ck
s
d
atasets
.
Fo
r
th
e
R
ed
W
in
e
Qu
al
it
y
d
ata
s
et,
DE
S
-
M
I
is
s
li
g
h
t
l
y
s
u
p
er
io
r
co
m
p
ar
ed
to
HA
R
-
MI
.
T
h
er
e
is
a
ten
d
en
c
y
i
f
t
h
e
n
u
m
b
er
o
f
attr
ib
u
tes
i
n
cr
ea
s
e
s
,
th
e
s
a
m
p
lin
g
p
r
o
ce
s
s
,
esp
ec
iall
y
o
n
R
an
d
o
m
Un
d
er
Sa
m
p
l
in
g
,
r
eq
u
ir
es
a
l
ar
g
er
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6930
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
,
Vo
l.
18
,
No
.
2
,
A
p
r
il 2
0
2
0
:
8
2
2
-
8
2
9
828
class
i
f
ier
.
Ho
w
e
v
er
,
th
e
d
i
f
f
er
en
ce
in
t
h
e
n
u
m
b
er
o
f
class
if
i
er
s
is
n
o
t
v
er
y
s
i
g
n
if
ican
t.
T
h
e
r
esu
lts
o
f
test
i
n
g
MAv
A
an
d
MFM
ca
n
b
e
s
ee
n
i
n
T
ab
le
5
.
T
ab
le
5
.
T
esti
n
g
R
e
s
u
lt f
o
r
M
A
vA
an
d
MFM
f
o
r
E
ac
h
Me
t
h
o
d
D
a
t
a
se
t
H
A
R
-
M
I
m
e
t
h
o
d
D
ES
-
M
I
m
e
t
h
o
d
MA
v
A
MF
M
MA
v
A
MF
M
B
a
l
a
n
c
e
s
c
a
l
e
6
6
.
7
1
0
.
7
1
6
1
.
2
9
0
.
6
1
C
a
r
e
v
a
l
u
a
t
i
o
n
9
7
.
6
8
0
.
9
7
9
4
.
2
7
0
.
9
4
5
R
e
d
w
i
n
e
q
u
a
l
i
t
y
4
5
.
2
4
0
.
4
3
4
1
.
8
1
0
.
3
9
5
Ec
o
l
i
5
7
.
3
1
0
.
5
8
4
9
.
6
7
0
.
5
1
P
a
g
e
b
l
o
c
k
s
4
7
.
8
1
0
.
4
9
4
5
.
9
2
0
.
4
4
I
n
T
ab
le
5
it
ca
n
b
e
s
ee
n
t
h
at
H
AR
-
MI
Me
t
h
o
d
g
i
v
es
b
e
tter
r
esu
lts
f
o
r
MAv
A
a
n
d
M
F
M
w
h
en
co
m
p
ar
ed
to
DE
S
-
MI
Me
th
o
d
.
B
o
th
m
et
h
o
d
s
h
a
v
e
p
r
o
v
id
ed
ex
ce
lle
n
t
MAv
A
a
n
d
MFM
v
a
lu
es.
A
g
o
o
d
MAv
A
m
ea
n
s
th
at
t
h
e
ac
cu
r
ac
y
o
f
t
h
e
class
i
f
icat
io
n
h
a
s
b
ee
n
v
e
r
y
g
o
o
d
,
w
h
er
e
th
e
m
is
cla
s
s
i
f
icatio
n
th
at
o
cc
u
r
s
is
v
er
y
m
i
n
i
m
al.
T
h
is
m
ea
n
s
th
at
th
e
i
n
s
ta
n
ce
o
f
t
h
e
m
i
n
o
r
it
y
clas
s
h
as
b
ee
n
c
lass
i
f
ied
co
r
r
ec
tl
y
an
d
also
th
e
m
aj
o
r
it
y
cla
s
s
i
n
s
ta
n
ce
s
t
h
at
ar
e
i
n
co
r
r
ec
tl
y
class
i
f
ied
a
s
m
i
n
o
r
it
y
c
lass
e
s
ar
e
also
m
i
n
i
m
al.
T
h
is
is
b
ec
a
u
s
e
F
-
Me
as
u
r
e
s
t
ates
h
o
w
m
a
n
y
i
n
s
tan
ce
s
i
n
t
h
e
m
in
o
r
it
y
clas
s
ar
e
co
r
r
ec
tly
d
ef
i
n
ed
an
d
also
m
ea
s
u
r
es
h
o
w
m
a
n
y
i
n
s
tan
ce
s
in
th
e
m
aj
o
r
it
y
class
ar
e
i
n
co
r
r
ec
tl
y
clas
s
if
ied
as
m
i
n
o
r
it
y
cl
ass
es.
3
.
2
.
T
esting
re
s
ult
T
h
e
s
tatis
tical
te
s
t
is
p
er
f
o
r
m
e
d
u
s
i
n
g
t
h
e
W
ilco
x
o
n
s
i
g
n
ed
-
r
an
k
te
s
t
w
h
ic
h
is
a
s
ta
tis
t
ical
p
r
o
ce
d
u
r
e
to
m
ea
s
u
r
e
p
er
f
o
r
m
a
n
ce
b
ased
o
n
p
air
w
i
s
e
co
m
p
ar
is
o
n
[
2
7
]
.
W
ilco
x
o
n
test
s
ar
e
ca
r
r
ied
o
u
t
to
c
o
m
p
ar
e
th
e
p
er
f
o
r
m
a
n
ce
o
f
t
h
e
H
AR
-
MI
m
eth
o
d
w
i
th
th
e
DE
S
-
MI
m
et
h
o
d
u
s
in
g
MAv
A
an
d
MFM
.
T
h
e
r
es
u
lts
o
b
tain
ed
ca
n
b
e
s
ee
n
i
n
T
ab
le
6
.
T
ab
le
6
.
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
f
o
r
co
m
p
ar
in
g
p
er
f
o
r
m
a
n
ce
m
ea
s
u
r
e
m
e
n
ts
u
s
i
n
g
MAv
A
an
d
MFM
P
e
r
f
o
r
man
c
e
m
e
a
su
r
e
me
n
t
P
-
V
a
l
u
e
H
y
p
o
t
h
e
si
s
MA
v
A
0
.
0
4
3
1
1
4
H
0
(
n
o
si
g
n
i
f
i
c
a
n
t
sc
o
re
d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n
H
AR
-
MI a
n
d
D
E
S
-
MI)
i
s re
j
e
c
t
e
d
a
n
d
t
h
i
s
m
e
a
n
s H
1
(
t
h
e
re
i
s
a
s
i
g
n
i
f
i
c
a
n
t
d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n
H
AR
-
MI a
n
d
D
E
S
-
MI
i
n
sc
o
r
e
)
i
s
Ac
c
e
p
t
e
d
b
e
c
a
u
se
t
h
e
p
-
v
a
l
u
e
<
0
.
0
5
MF
M
0
.
0
4
3
1
1
4
H
0
(
n
o
si
g
n
i
f
i
c
a
n
t
sc
o
re
d
i
f
f
e
r
e
n
c
e
b
e
t
w
e
e
n
H
AR
-
MI a
n
d
D
E
S
-
MI)
r
e
j
e
c
t
e
d
a
n
d
t
h
i
s
m
e
a
n
s
H
1
(
t
h
e
re
i
s
a
s
i
g
n
i
f
i
c
a
n
t
d
i
f
f
e
re
n
c
e
b
e
t
w
e
e
n
H
AR
-
MI a
n
d
D
E
S
-
MI
i
n
sc
o
re)
Ac
c
e
p
t
e
d
b
e
c
a
u
se
t
h
e
p
-
v
a
l
u
e
<
0
.
0
5
B
ased
o
n
th
e
r
esu
lt
s
o
f
tes
ti
n
g
w
i
th
t
h
e
W
ilco
x
o
n
s
ig
n
ed
-
r
an
k
test
t
h
at
ca
n
b
e
s
ee
n
i
n
T
ab
le
6
,
th
er
e
is
a
s
i
g
n
i
f
ica
n
t
d
if
f
er
e
n
ce
b
et
w
ee
n
H
AR
-
MI
a
n
d
DE
S
-
MI
an
d
t
h
is
i
n
d
icate
s
t
h
at
th
e
s
u
p
er
io
r
it
y
o
f
th
e
H
AR
-
MI
m
et
h
o
d
.
4.
CO
NCLU
SI
O
N
B
ased
o
n
th
e
te
s
t
r
es
u
lt
s
it
ca
n
b
e
s
ee
n
t
h
at
H
AR
-
MI
m
eth
o
d
g
i
v
es
b
etter
r
es
u
lt
s
co
m
p
ar
ed
to
DE
S
-
MI
m
e
th
o
d
f
o
r
b
o
th
th
e
n
u
m
b
er
o
f
clas
s
i
f
ier
,
d
ata
d
iv
er
s
it
y
,
an
d
also
th
e
p
er
f
o
r
m
a
n
ce
class
if
ier
.
I
t
s
h
o
u
ld
b
e
n
o
ted
th
at
f
o
r
th
e
n
u
m
b
er
o
f
clas
s
if
ier
s
,
w
h
er
e
if
th
e
d
ataset
h
a
s
m
a
n
y
attr
ib
u
t
es
s
u
c
h
as
th
e
R
ed
W
in
e
Qu
alit
y
,
th
e
n
t
h
e
H
A
R
-
MI
m
e
th
o
d
ca
n
p
r
o
d
u
ce
p
o
o
r
r
esu
lts
.
I
n
g
en
er
al,
th
e
i
m
b
al
an
ce
r
atio
d
o
es
n
o
t
h
av
e
a
s
ig
n
i
f
ica
n
t
e
f
f
ec
t
o
n
t
h
e
test
r
esu
lts
.
T
h
is
m
ea
n
s
t
h
at
b
o
th
HAR
-
MI
m
e
th
o
d
an
d
D
E
S
-
MI
m
eth
o
d
ca
n
h
an
d
le
th
e
i
m
b
a
lan
ce
p
r
o
b
lem
cla
s
s
v
er
y
w
el
l.
F
u
t
u
r
e
r
es
ea
r
ch
,
it
is
e
x
p
ec
ted
t
h
at
H
A
R
-
MI
m
e
th
o
d
ca
n
b
e
o
p
tim
ized
s
o
th
at
it
ca
n
b
e
ap
p
lied
to
d
atasets
f
o
r
a
lar
g
e
n
u
m
b
er
o
f
attr
ib
u
tes
w
it
h
o
u
t
ca
u
s
in
g
a
lar
g
e
n
u
m
b
er
o
f
clas
s
i
f
ier
s
.
T
h
e
m
ain
atte
n
tio
n
n
ee
d
s
to
b
e
g
iv
e
n
to
th
e
s
a
m
p
li
n
g
m
et
h
o
d
u
s
ed
i
n
th
e
H
A
R
-
MI
m
et
h
o
d
.
I
t is n
ec
es
s
ar
y
to
f
i
n
d
an
o
t
h
er
s
a
m
p
li
n
g
alter
n
ati
v
e
at
t
h
e
p
r
ep
r
o
ce
s
s
in
g
a
n
d
p
r
o
ce
s
s
in
g
s
ta
g
es.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
is
w
o
r
k
w
a
s
s
u
p
p
o
r
ted
b
y
t
h
e
Gr
an
t
o
f
Min
is
tr
y
o
f
R
e
s
ea
r
ch
,
T
ec
h
n
o
lo
g
y
,
a
n
d
Hig
h
er
E
d
u
ca
tio
n
(
KE
ME
NR
I
ST
E
KDI
KT
I
)
o
f
th
e
R
ep
u
b
lic
o
f
I
n
d
o
n
esia.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
K
A
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l
C
o
n
tr
o
l
HA
R
-
MI
meth
o
d
fo
r
mu
lti
-
cla
s
s
imb
a
la
n
ce
d
d
a
ta
s
ets (
H.
Ha
r
to
n
o
)
829
RE
F
E
R
E
NC
E
S
[1
]
S
.
W
a
n
g
a
n
d
X.
Ya
o
,
"
M
u
lt
icla
ss
im
b
a
lan
c
e
p
ro
b
lem
s:
A
n
a
l
y
sis
a
n
d
p
o
ten
t
ial
so
l
u
ti
o
n
s,"
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
S
y
ste
ms
,
M
a
n
,
a
n
d
Cy
b
e
rn
e
ti
c
s,
P
a
rt B
(
Cy
b
e
rn
e
ti
c
s)
,
v
o
l.
4
2
,
n
o
.
4
,
p
p
.
1
1
1
9
–
11
3
0
,
2
0
1
2
.
[2
]
B.
Kra
wc
z
y
k
,
"
L
e
a
rn
in
g
f
ro
m
i
m
b
a
lan
c
e
d
d
a
ta:
O
p
e
n
c
h
a
ll
e
n
g
e
s
a
n
d
f
u
tu
re
d
irec
ti
o
n
s
,"
Pro
g
re
ss
in
Arti
fi
c
ia
l
In
telli
g
e
n
c
e
,
v
o
l.
5
,
p
p
.
2
2
1
–
3
2
,
2
0
1
6
.
[3
]
A
.
A
li
,
S
.
M
.
S
h
a
m
su
d
d
in
,
a
n
d
A
.
Ra
les
c
u
,
"
Clas
sif
ic
a
ti
o
n
w
it
h
c
la
s
s
im
b
a
lan
c
e
p
ro
b
lem
:
A
re
v
ie
w
,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
A
d
v
a
n
c
e
s in
S
o
ft
C
o
mp
u
ti
n
g
a
n
d
Its
Ap
p
li
c
a
t
io
n
,
v
o
l.
7
,
n
o
.
3
,
p
p
.
1
7
6
–
2
0
4
,
2
0
1
5
.
[4
]
G
.
Ha
ix
ian
g
,
L
.
Yiji
n
g
,
J.
S
h
a
n
g
,
G
.
M
in
g
y
u
n
,
H.
Yu
a
n
y
u
e
,
a
n
d
G
.
Bin
g
,
"
Lea
rn
in
g
f
ro
m
c
las
s
-
i
m
b
a
lan
c
e
d
d
a
ta:
Re
v
ie
w
o
f
m
e
th
o
d
s a
n
d
a
p
p
li
c
a
ti
o
n
s
,"
Exp
e
rt S
y
ste
ms
w
it
h
A
p
p
li
c
a
ti
o
n
s
,
v
o
l
.
73
,
p
p
.
2
2
0
–
3
9
,
2
0
1
7
.
[5
]
A
.
F
e
rn
á
n
d
e
z
,
V
.
L
ó
p
e
z
,
M.
Ga
lar,
M
.
J.
d
e
l
J
e
su
s
,
a
n
d
F
.
He
rre
ra
,
"
A
n
a
l
y
sin
g
th
e
c
l
a
ss
i
f
ica
ti
o
n
o
f
i
m
b
a
lan
c
e
d
d
a
ta
-
se
ts
w
it
h
m
u
lt
ip
le
c
las
se
s:
Bin
a
riza
ti
o
n
t
e
c
h
n
iq
u
e
s
a
n
d
a
d
-
h
o
c
a
p
p
r
o
a
c
h
e
s,"
Kn
o
wl
e
d
g
e
-
Ba
se
d
S
y
ste
ms
,
v
o
l.
4
2
,
p
p
.
9
7
–
1
1
0
,
2
0
1
3
.
[6
]
J.
Bi
a
n
d
C.
Z
h
a
n
g
,
"
A
n
e
m
p
iri
c
a
l
c
o
m
p
a
riso
n
o
n
sta
te
-
of
-
th
e
-
a
rt
m
u
lt
i
-
c
las
s
i
m
b
a
lan
c
e
lea
rn
in
g
a
lg
o
rit
h
m
s
a
n
d
a
n
e
w
d
iv
e
r
sif
ied
e
n
se
m
b
le l
e
a
rn
in
g
sc
h
e
m
e
,
"
Kn
o
wled
g
e
-
B
a
se
d
S
y
ste
ms
,
v
o
l.
1
5
8
,
p
p
.
81
–
93
,
2
0
1
8
.
[7
]
T
.
R.
Ho
e
n
s,
Q.
Qia
n
,
N.
V
.
C
h
a
w
la,
a
n
d
Z
-
H.
Zh
o
u
,
"
Bu
il
d
in
g
d
e
c
isio
n
tree
s
f
o
r
th
e
m
u
lt
i
-
c
las
s
i
m
b
a
lan
c
e
p
ro
b
l
e
m
,
"
In
:
T
a
n
P
-
N,
Ch
a
w
la
S
,
Ho
CK,
Ba
il
e
y
J,
e
d
it
o
rs,
Ad
v
a
n
c
e
s
in
Kn
o
wled
g
e
Disc
o
v
e
ry
a
n
d
Da
ta
M
in
i
n
g
,
L
e
c
tu
re
No
tes
in
C
o
mp
u
ter
S
c
ien
c
e
,
S
p
rin
g
e
r
Be
rli
n
He
id
e
lb
e
rg
;
v
o
l.
7
3
0
1
,
p
p
.
1
2
2
–
3
4
,
2
0
1
2
.
[8
]
E.
L
.
A
ll
w
e
in
,
R.
E.
S
c
h
a
p
ire,
a
n
d
Y.
S
in
g
e
r
,
"
Re
d
u
c
in
g
m
u
lt
icla
ss
to
b
in
a
ry
:
A
u
n
ify
in
g
a
p
p
ro
a
c
h
f
o
r
m
a
rg
in
cl
a
ss
i
f
iers
,"
J
o
u
rn
a
l
o
f
M
a
c
h
i
n
e
L
e
a
rn
in
g
Res
e
a
rc
h
,
v
o
l
.
1
,
1
1
3
–
1
4
1
,
2
0
0
0
.
[9
]
M.
G
a
lar,
A.
F
e
rn
á
n
d
e
z
,
E.
Ba
rre
n
e
c
h
e
a
,
H.
Bu
stin
c
e
,
a
n
d
F
.
He
rre
ra
,
"
A
n
o
v
e
r
v
ie
w
o
f
e
n
se
m
b
l
e
m
e
th
o
d
s
f
o
r
b
in
a
ry
c
las
si
f
ier
s
in
m
u
lt
i
-
c
las
s
p
ro
b
lem
s:
Ex
p
e
ri
m
e
n
tal
stu
d
y
o
n
o
n
e
-
vs
-
o
n
e
a
n
d
o
n
e
-
vs
-
a
ll
sc
h
e
m
e
s,"
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
v
o
l.
4
4
,
n
o
.
8
,
p
p
.
1
7
6
1
–
1
7
7
6
,
2
0
1
1
.
[1
0
]
J.
A
.
S
á
e
z
,
B.
Kra
wc
z
y
k
,
a
n
d
M
.
W
o
ź
n
iak
,
"
A
n
a
l
y
z
in
g
th
e
o
v
e
rsa
m
p
li
n
g
o
f
d
iff
e
r
e
n
t
c
las
s
e
s a
n
d
t
y
p
e
s o
f
e
x
a
m
p
le
s
in
m
u
lt
i
-
c
las
s im
b
a
lan
c
e
d
d
a
tas
e
t
s
,"
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
v
o
l.
57
,
p
p
.
1
6
4
–
1
78
,
2
0
1
6
.
[1
1
]
Z
-
L
.
Zh
a
n
g
,
X
-
G
.
L
u
o
,
S.
Ga
rc
í
a
,
J
-
F.
T
a
n
g
,
a
n
d
F
.
He
rre
ra
,
"
E
x
p
lo
rin
g
th
e
e
ff
e
c
ti
v
e
n
e
ss
o
f
d
y
n
a
m
ic
e
n
se
m
b
le
se
lec
ti
o
n
in
th
e
o
n
e
-
v
e
rsu
s
-
o
n
e
sc
h
e
m
e
,"
Kn
o
wled
g
e
-
B
a
se
d
S
y
ste
m
s
,
v
o
l.
1
2
5
,
p
p
.
5
3
–
6
3
,
2
0
1
7
.
[1
2
]
S
.
G
a
rc
í
a
,
Z.
-
L
.
Zh
a
n
g
,
A
.
A
lt
a
l
h
i,
S
.
A
lsh
o
m
ra
n
i,
a
n
d
F
.
He
rre
ra
.
"
D
y
n
a
m
ic
e
n
se
m
b
le
se
lec
ti
o
n
f
o
r
m
u
lt
i
-
c
las
s
im
b
a
lan
c
e
d
d
a
tas
e
ts
,"
In
fo
rm
a
ti
o
n
S
c
ien
c
e
s
,
v
o
l.
4
4
5
–
4
4
6
,
p
p
.
2
2
–
37,
2
0
1
8
,
[1
3
]
P.
P
é
re
z
-
G
á
ll
e
g
o
,
A
.
Ca
sta
ñ
o
,
J.
R.
Qu
e
v
e
d
o
,
a
n
d
J.
J
.
d
e
l
Co
z
,
"
D
y
n
a
m
i
c
e
n
se
m
b
le
se
l
e
c
ti
o
n
f
o
r
q
u
a
n
t
if
ica
ti
o
n
tas
k
s,"
In
fo
rm
a
ti
o
n
Fu
si
o
n
,
v
o
l.
4
5
,
p
p
.
1
–
1
5
,
2
0
1
9
.
[1
4
]
Ha
rto
n
o
,
E.
O
n
g
k
o
,
O.
P
.
S
i
to
m
p
u
l,
T
u
l
u
s,
E.
B.
Na
b
a
b
a
n
,
a
n
d
D.
A
b
d
u
ll
a
h
,
"
Hy
b
rid
a
p
p
r
o
a
c
h
re
d
e
f
i
n
it
io
n
(HA
R)
m
e
th
o
d
w
it
h
lo
ss
f
a
c
to
rs
in
h
a
n
d
li
n
g
c
las
s
i
m
b
a
lan
c
e
p
ro
b
lem
,
"
In
In
ter
n
a
ti
o
n
a
l
S
y
mp
o
si
u
m
o
n
Ad
v
a
n
c
e
d
In
telli
g
e
n
t
I
n
fo
rm
a
t
ics
(
S
AIN)
,
p
p
.
5
6
–
6
1
,
2
0
1
8
.
[1
5
]
Ha
rto
n
o
,
E.
On
g
k
o
,
E
.
B.
Na
b
a
b
a
n
,
T
u
lu
s,
D.
A
b
d
u
ll
a
h
,
a
n
d
A
.
S
A
h
m
a
r,
"
A
n
e
w
d
iv
e
r
sity
tec
h
n
iq
u
e
f
o
r
i
m
b
a
lan
c
e
lea
rn
in
g
e
n
se
m
b
les
,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
o
f
En
g
i
n
e
e
rin
g
&
T
e
c
h
n
o
lo
g
y
,
v
o
l.
7
,
n
o
.
2
,
p
p
.
4
7
8
–
4
8
3
,
2
0
1
8
.
[1
6
]
j.
F
.
Díe
z
-
P
a
st
o
r,
J.
J.
R
o
d
ríg
u
e
z
,
C.
I.
G
a
rc
ía
-
Os
o
rio
,
a
n
d
L
.
I.
Ku
n
c
h
e
v
a
,
"
Div
e
rsit
y
tec
h
n
iq
u
e
s
im
p
ro
v
e
th
e
p
e
rf
o
rm
a
n
c
e
o
f
th
e
b
e
s
t
im
b
a
lan
c
e
lea
rn
in
g
e
n
se
m
b
les
,
"
In
fo
rm
a
ti
o
n
S
c
ien
c
e
s
,
v
o
l.
3
2
5
,
p
p
.
9
8
–
1
1
7
,
2
0
1
5
.
[1
7
]
C.
Jia
n
,
J.
G
a
o
,
a
n
d
Y.
Ao
,
"
A
n
e
w
sa
m
p
li
n
g
m
e
th
o
d
f
o
r
c
las
sif
y
in
g
i
m
b
a
la
n
c
e
d
d
a
ta
b
a
se
d
o
n
su
p
p
o
rt
v
e
c
to
r
m
a
c
h
in
e
e
n
se
m
b
le,"
N
e
u
ro
c
o
mp
u
ti
n
g
,
v
o
l.
1
9
3
,
p
p
.
1
1
5
–
1
2
2
,
2
0
1
6
.
[1
8
]
J.
A
lca
lá
-
F
d
e
z
e
t
a
l
.
,
"
KEEL
:
A
so
f
t
w
a
re
to
o
l
to
a
ss
e
ss
e
v
o
lu
ti
o
n
a
r
y
a
lg
o
rit
h
m
s
f
o
r
d
a
ta
m
in
in
g
p
ro
b
lem
s
,"
S
o
ft
Co
mp
u
t
.
,
v
o
l
.
1
3
,
n
o
.
3
,
p
p
.
3
0
7
–
3
1
8
,
2
0
0
9
.
[1
9
]
K
.
Na
p
iera
la
,
A
n
d
J.
S
tef
a
n
o
w
s
k
i,
"
Id
e
n
ti
f
ica
ti
o
n
o
f
d
if
f
e
re
n
t
t
y
p
e
s
o
f
m
in
o
rit
y
c
las
s
e
x
a
m
p
les
in
imb
a
lan
c
e
d
d
a
t
a
,
"
In
Co
rc
h
a
d
o
E.
,
S
n
á
še
l
V
.
,
A
b
r
a
h
a
m
A
.
,
W
o
ź
n
iak
M
.
,
G
ra
ñ
a
M
.
,
C
h
o
S
B.
(e
d
s
)
Hy
b
ri
d
Arti
f
icia
l
In
telli
g
e
n
t
S
y
ste
ms
,
S
p
rin
g
e
r,
Be
rli
n
,
He
id
e
l
b
e
rg
,
pp
.
1
3
9
-
1
5
0
,
2
0
1
2
[2
0
]
L
.
I.
Ku
n
c
h
e
v
a
,
"
Co
m
b
in
in
g
p
a
tt
e
rn
c
la
ss
if
iers
,"
Jo
h
n
W
il
e
y
&
S
o
n
s,
p
p
.
2
9
5
–
3
2
7
,
2
0
0
4
.
[2
1
]
L
.
I.
Ku
n
c
h
e
v
a
a
n
d
C.
J.
W
h
it
a
k
e
r
,
"
M
e
a
su
re
s
o
f
d
iv
e
rsit
y
in
c
las
sif
ier
e
n
se
m
b
les
a
n
d
th
e
ir
re
l
a
ti
o
n
sh
i
p
w
it
h
th
e
e
n
se
m
b
le ac
c
u
ra
cy
,"
M
a
c
h
in
e
L
e
a
rn
i
n
g
,
v
o
l.
51
,
n
o
.
2
,
p
p
.
1
8
1
–
2
0
7
,
2
0
0
3
.
[2
2
]
C.
G
ig
li
a
ra
n
o
,
S.
F
ig
in
i
,
a
n
d
P
.
M
u
li
e
re
,
"
M
a
k
in
g
c
las
sif
ier
p
e
r
f
o
rm
a
n
c
e
c
o
m
p
a
riso
n
s
w
h
e
n
ROC
c
u
rv
e
s
in
ters
e
c
t
,"
Co
mp
u
t
a
ti
o
n
a
l
S
ta
t
isti
c
s
&
Da
ta
An
a
lys
is
,
v
o
l
.
7
,
p
p
.
3
0
0
–
3
12
,
2
0
1
4
.
[2
3
]
Y.
S
u
n
,
M
.
S
.
Ka
m
e
l,
A
.
K.
C.
W
o
n
g
,
a
n
d
Y.
Wan
g
,
"
Co
st
-
se
n
s
it
iv
e
b
o
o
stin
g
f
o
r
c
las
si
f
ica
ti
o
n
o
f
i
m
b
a
lan
c
e
d
d
a
ta
,"
Pa
tt
e
rn
Rec
o
g
n
i
ti
o
n
,
v
o
l
.
4
0
,
n
o
.
1
2
,
p
p
.
3
3
5
8
–
3
3
7
8
,
2
0
0
7
.
[2
4
]
L
.
Zh
a
n
g
,
H.
Ya
n
g
,
a
n
d
Z.
Jia
n
g
,
"
Im
b
a
lan
c
e
d
b
io
m
e
d
ica
l
d
a
ta
c
las
sif
i
c
a
ti
o
n
u
si
n
g
se
l
f
-
a
d
a
p
ti
v
e
m
u
lt
il
a
y
e
r
E
L
M
c
o
m
b
in
e
d
w
it
h
d
y
n
a
m
ic
GA
N,
"
Bi
o
me
d
E
n
g
.
On
li
n
e
,
v
o
l.
1
7
,
p
p
.
1
-
2
1
,
2
0
1
8
.
[2
5
]
A
.
T
h
a
r
w
a
t
,
"
Cl
a
s
sif
i
c
a
ti
o
n
a
ss
e
ss
m
e
n
t
m
e
th
o
d
s,"
Ap
p
l
ied
Co
m
p
u
ti
n
g
a
n
d
I
n
f
o
rm
a
ti
c
s
[
In
tern
e
t
].
2
0
1
8
A
u
g
2
1
[
c
it
e
d
2
0
1
9
Oc
t
1
9
]
;
A
v
a
il
a
b
le f
ro
m
:
h
tt
p
:/
/w
ww
.
sc
ien
c
e
d
irec
t.
c
o
m
/
sc
ien
c
e
/article
/p
ii
/S
2
2
1
0
8
3
2
7
1
8
3
0
1
5
4
6
[2
6
]
R.
A
lejo
,
J.
A
.
A
n
to
n
io
,
R
.
M
.
V
a
ld
o
v
in
o
s,
a
n
d
J.
H.
P
a
c
h
e
c
o
-
S
á
n
c
h
e
z
,
"
A
s
se
ss
m
e
n
ts
m
e
tri
c
s
f
o
r
m
u
lt
i
-
c
las
s
im
b
a
lan
c
e
l
e
a
r
n
in
g
:
A
p
re
li
m
in
a
r
y
stu
d
y
,
"
In
a
rra
s
c
o
-
Oc
h
o
a
J.A
.
,
M
a
rtí
n
e
z
-
T
rin
id
a
d
J.F
.
,
Ro
d
r
íg
u
e
z
J.S
.
,
d
i
Ba
ja
G
.
S
.
(e
d
s)
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
L
e
c
tu
re
No
tes
in
Co
mp
u
ter
S
c
ien
c
e
,
S
p
rin
g
e
r
Be
rli
n
He
id
e
lb
e
rg
,
v
o
l.
7
9
1
4
,
p
p
.
3
3
5
–
3
4
3
,
2
0
1
3
.
[2
7
]
F
.
W
il
c
o
x
o
n
,
“
In
d
iv
i
d
u
a
l
c
o
m
p
a
riso
n
s b
y
ra
n
k
in
g
m
e
th
o
d
s
,”
Bi
o
me
trics
Bu
ll
e
ti
n
,
v
o
l.
1
,
n
o
.
6
,
p
p
.
80
-
83
,
1
9
4
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.