I
A
E
S
I
n
t
e
r
n
at
io
n
al
Jou
r
n
al
of
A
r
t
if
ic
ia
l
I
n
t
e
ll
ig
e
n
c
e
(
I
J
-
AI
)
V
ol
.
14
, N
o.
3
,
J
une
2025
, pp.
1727
~
1742
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
3
.pp
1727
-
1742
1727
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
R
e
vi
e
w
on
c
l
ass
i
m
b
al
an
c
e
t
e
c
h
n
i
q
u
e
s t
o st
r
e
n
gt
h
e
n
m
od
e
l
p
r
e
d
i
c
t
i
on
P
u
t
t
a H
e
m
al
at
h
a, G
e
e
t
h
a M
ar
y A
m
al
an
at
h
an
S
c
hool
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng, V
e
l
l
or
e
I
ns
t
i
t
ut
e
of
T
e
c
hnol
ogy,
V
e
l
l
or
e
, I
ndi
a
A
r
t
ic
le
I
n
f
o
A
B
S
T
R
A
C
T
A
r
ti
c
le
h
is
to
r
y
:
R
e
c
e
iv
e
d
F
e
b
13
,
2024
R
e
vi
s
e
d
F
e
b
28
,
2025
A
c
c
e
pt
e
d
M
a
r
15
,
2025
Data
is
a
fundamental
component
in
various
fields,
including
s
cience,
business,
health
care,
and
technology.
It
is
often
processed,
store
d,
and
analyzed
using
computer
systems
and
software
applicati
ons.
The
imp
ortance
of
data
lies
in
its
ability
to
provide
valuable
insights,
drive
innovation,
and
improve
decision
-
making
processes.
However,
it’s
essential
to
hand
le
and
manage
data
responsibly
to
address
privacy
and
ethical
consideration
s.
Data
mining
(DM)
involves
discovering
patterns,
trends,
correlations,
or
useful
information
from
large
datasets.
Data
dredging
or
DM
and
machine
le
arning
(ML)
are
closely
related
fields
that
both
involve
the
analysis
of
data
to
discover
patterns
and
make
predictions.
DM
focuses
on
ext
racting
knowledge
from
data;
ML
emphasizes
the
development
of
algorithms
that
can
do
analysis.
The
two
fields
are
interconnected,
and
the
technique
s
from
one
state
of
art
integrated
into
the
processes
of
the
other.
In
ML
th
e
class
imbalance
problem
occurs
due
to
the
class
distribution
in
the
trai
ning
data
is
not
equal.
Imbalanced
classification
refers
to
a
condition
wh
ere
a
par
ticular
class
(minority
class)
is
under
represented
parallelled
to
another
class
(majority
class)
in
a
dataset.
This
paper
furthermore
emphasizes
on
the
synthetic
minority
oversampling
technique
(SMOTE)
variants
employed
by
the rese
archer
s, and highligh
ts the limitations th
e work.
K
e
y
w
o
r
d
s
:
C
la
s
s
i
m
ba
la
n
c
e
D
a
ta
m
in
in
g
M
a
c
hi
ne
l
e
a
r
ni
ng
M
ode
l
pe
r
f
or
m
a
nc
e
S
M
O
T
E
t
e
c
hni
que
s
This is an
open
acce
ss artic
le unde
r the
CC BY
-
SA
license.
C
or
r
e
s
pon
di
n
g A
u
th
or
:
G
e
e
th
a
M
a
r
y A
m
a
la
na
th
a
n
S
c
hool
of
C
om
put
e
r
S
c
ie
nc
e
a
nd E
ngi
ne
e
r
in
g, V
e
ll
or
e
I
ns
ti
tu
te
of
T
e
c
hnol
ogy
V
e
ll
or
e
C
a
m
pus
,
T
ir
uva
la
m
R
d, K
a
tp
a
di
, V
e
ll
or
e
,
T
a
m
il
N
a
du
632014, I
ndi
a
E
m
a
il
:
ge
e
th
a
m
a
r
y.a
@
gm
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
I
n
m
a
c
hi
ne
le
a
r
ni
ng
(
M
L
)
,
c
la
s
s
im
ba
la
nc
e
is
a
s
c
e
na
r
io
w
he
r
e
w
e
s
e
e
hi
ghl
y
s
ke
w
e
d
di
s
tr
ib
ut
io
n
of
c
la
s
s
e
s
in
a
p
a
r
ti
c
ul
a
r
da
ta
s
e
t,
w
it
h
one
or
m
or
e
c
la
s
s
e
s
be
in
g
m
uc
h
m
or
e
c
om
m
on
th
a
n
ot
he
r
s
.
C
la
s
s
if
ie
r
s
m
a
y
be
c
om
e
bi
a
s
e
d
to
w
a
r
ds
th
e
m
a
jo
r
it
y
c
la
s
s
a
s
a
r
e
s
ul
t,
w
hi
c
h
e
m
pha
s
is
on
poor
pe
r
f
or
m
a
nc
e
on
th
e
m
in
or
it
y
c
la
s
s
e
s
.
I
n
a
r
e
a
l
-
w
or
ld
a
ppl
ic
a
ti
on,
w
he
r
e
th
e
m
in
or
it
y
c
la
s
s
r
e
pr
e
s
e
nt
s
s
ig
ni
f
ic
a
nt
or
c
r
it
ic
a
l
in
s
ta
nc
e
s
, s
uc
h
a
s
m
e
di
c
a
l
di
a
gnos
is
,
f
r
a
ud
de
te
c
ti
on,
a
nd
a
no
m
a
ly
id
e
nt
if
ic
a
ti
on,
a
ddr
e
s
s
in
g
c
la
s
s
im
ba
la
nc
e
is
e
s
s
e
nt
ia
l.
M
e
th
od
s
l
ik
e
c
os
t
-
s
e
n
s
it
iv
e
l
e
a
r
ni
ng, e
ns
e
m
bl
e
a
pp
r
oa
c
he
s
, a
nd r
e
s
a
m
pl
in
g a
r
e
f
r
e
que
nt
ly
us
e
d t
o
le
s
s
e
n
th
e
im
pa
c
t
of
c
la
s
s
im
ba
l
a
nc
e
a
nd
e
nha
nc
e
M
L
m
ode
l
p
e
r
f
or
m
a
nc
e
.
T
hi
s
w
or
k
di
s
c
us
s
e
s
th
e
di
f
f
e
r
e
nt
te
c
hni
que
s
to
ha
ndl
e
th
e
c
la
s
s
im
ba
la
nc
e
pr
obl
e
m
,
di
f
f
e
r
e
nt
s
ynt
he
ti
c
m
in
or
it
y
ove
r
s
a
m
pl
in
g
te
c
hni
que
(
S
M
O
T
E
)
ba
s
e
d
m
e
th
ods
,
di
f
f
e
r
e
nt
c
la
s
s
if
ic
a
ti
on
m
ode
ls
,
v
a
r
io
us
e
va
lu
a
ti
on
m
e
tr
ic
s
w
hi
c
h
is
us
e
d
to
m
e
a
s
ur
e
th
e
m
ode
l
pe
r
f
or
m
a
nc
e
,
a
nd
c
h
a
ll
e
nge
s
a
s
s
oc
ia
te
d
w
it
h
a
c
la
s
s
im
ba
la
n
c
e
in
th
e
f
or
m
of
a
li
te
r
a
tu
r
e
r
e
vi
e
w
.
T
he
a
na
ly
s
is
of
e
xi
s
ti
ng
di
f
f
e
r
e
nt
c
a
te
gor
iz
a
ti
on
a
ppr
oa
c
he
s
he
lp
s
us
to
be
tt
e
r
unde
r
s
ta
nd
th
e
di
f
f
ic
ul
ti
e
s
th
a
t
th
e
im
ba
la
nc
e
da
t
a
s
e
t
pr
e
s
e
nt
s
.
M
L
te
c
hni
que
s
ha
ve
b
e
e
n
w
id
e
ly
u
s
e
d
in
a
va
r
ie
ty
of
a
ppl
ic
a
ti
ons
,
in
c
lu
di
ng
f
in
a
nc
e
,
c
li
ni
c
a
l
di
a
gnos
ti
c
s
,
m
a
r
ke
ti
n
g,
ba
nki
ng,
e
pi
de
m
io
lo
gy,
a
nd
bi
oi
nf
or
m
a
ti
c
s
[
1]
. M
os
t
of
th
e
da
ta
s
e
ts
i
n t
he
r
e
a
l
-
w
or
ld
s
e
e
m
t
o
be
i
m
ba
la
nc
e
d. T
he
na
tu
r
e
of
i
m
ba
la
nc
e
a
nd i
m
por
ta
nc
e
o
f
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
1727
-
1742
1728
th
e
m
in
or
it
y
s
a
m
pl
e
s
a
nd
it
s
w
id
e
r
a
nge
of
da
ta
m
in
in
g
(
D
M
)
a
ppl
ic
a
ti
on
f
ie
ld
s
a
r
e
di
s
c
us
s
e
d
in
de
ta
il
e
d
by
[
2]
,
w
hi
c
h
a
ls
o
de
li
ve
r
s
a
s
ynops
is
of
th
e
e
va
lu
a
ti
on
m
e
tr
i
c
s
a
nd
c
ur
r
e
nt
pr
oc
e
dur
e
s
f
or
a
s
s
e
s
s
in
g
a
nd
r
e
s
ol
vi
ng
im
ba
la
nc
e
pr
obl
e
m
s
.
O
ne
ove
r
-
s
a
m
pl
in
g
te
c
hni
que
th
a
t
a
ddr
e
s
s
e
s
th
is
is
s
ue
is
th
e
S
M
O
T
E
.
I
n
e
s
s
e
n
c
e
, a
n i
m
ba
la
nc
e
d da
ta
ba
s
e
ha
s
m
in
or
it
y a
nd ma
jo
r
it
y
i
nf
or
m
a
ti
on. DM
i
s
one
of
t
he
ke
y
a
ppr
oa
c
he
s
f
o
r
ha
ndl
in
g volum
in
ous
i
m
ba
la
nc
e
d da
ta
unf
ol
di
ng
hi
dde
n pa
tt
e
r
n
s
t
hr
ough it
s
t
e
c
hni
que
s
s
uc
h a
s
c
la
s
s
if
ic
a
ti
on,
a
s
s
oc
ia
ti
on
a
na
ly
s
is
,
c
lu
s
te
r
in
g,
a
nd
pa
tt
e
r
n
r
e
c
ogni
ti
on
[
3]
.
S
e
tt
in
g
up
a
r
e
a
l
-
ti
m
e
da
ta
pr
oc
e
s
s
in
g,
da
ta
s
e
ts
f
r
e
que
nt
ly
c
ons
is
t
of
a
gr
e
a
te
r
qu
a
nt
it
y
of
da
ta
be
lo
ngi
ng
to
one
c
la
s
s
,
known
a
s
th
e
m
a
jo
r
it
y
c
la
s
s
,
a
nd
a
le
s
s
e
r
num
be
r
be
lo
ngi
ng
to
th
e
m
in
or
it
y
c
la
s
s
.
A
lm
os
t
m
a
jo
r
i
ty
of
th
e
c
a
s
e
s
in
th
e
s
e
is
s
ue
s
a
r
e
la
be
ll
e
d
a
s
be
lo
ngi
ng
to
one
c
la
s
s
,
w
he
r
e
a
s
m
uc
h
f
e
w
e
r
in
s
ta
nc
e
s
a
r
e
la
b
e
ll
e
d
a
s
be
lo
ngi
ng
to
th
e
ot
he
r
c
la
s
s
,
w
hi
c
h
is
ty
pi
c
a
ll
y
th
e
m
or
e
s
ig
ni
f
ic
a
nt
c
la
s
s
.
I
m
ba
la
nc
e
is
a
s
it
ua
ti
on
w
h
e
r
e
th
e
r
e
a
r
e
bi
a
s
e
d
num
be
r
of
s
a
m
pl
e
s
in
on
e
c
la
s
s
w
he
n
c
om
p
a
r
e
d
to
ot
he
r
c
la
s
s
e
s
.
I
n
th
e
s
e
c
ir
c
um
s
ta
nc
e
s
,
th
e
c
la
s
s
if
ie
r
c
r
e
a
te
d
m
a
y
pr
e
di
c
t
in
f
a
vor
o
f
th
e
dom
in
a
nt
c
la
s
s
.
I
n
e
a
r
li
e
r
r
e
s
e
a
r
c
h
in
ve
s
ti
ga
ti
ons
,
s
e
ve
r
a
l
s
tr
a
te
gi
e
s
w
e
r
e
de
ve
lo
pe
d
to
a
ddr
e
s
s
th
e
i
s
s
ue
of
c
la
s
s
im
ba
l
a
nc
e
.
T
he
r
e
f
or
e
,
th
e
a
ut
hor
a
im
s
to
r
e
vi
e
w
va
r
io
us
te
c
hni
que
s
f
or
a
ddr
e
s
s
in
g
th
e
c
la
s
s
im
ba
la
nc
e
d pr
obl
e
m
.
2.
B
A
C
K
G
R
O
U
N
D
O
N
D
A
T
A
M
I
N
I
N
G
, M
A
C
H
I
N
E
L
E
A
R
N
I
N
G
, A
N
D
C
L
A
S
S
I
M
B
A
L
A
N
C
E
2
.1. Dat
a m
in
in
g p
la
y’
s
vi
t
al
r
ol
e
t
o
e
n
h
an
c
e
c
la
s
s
i
m
b
al
an
c
e
p
r
ob
le
m
s
C
la
s
s
im
ba
la
nc
e
is
a
ve
r
y
im
por
ta
nt
pr
obl
e
m
in
r
e
a
l
li
f
e
s
c
e
na
r
io
s
,
a
nd
s
uc
h
a
ppl
ic
a
ti
ons
in
c
lu
de
;
h
e
a
lt
h
c
a
r
e
,
f
r
a
ud
a
na
ly
s
is
,
in
f
or
m
a
ti
on
r
e
tr
ie
va
l,
a
nd
te
x
t
m
in
in
g.
T
hi
s
im
ba
la
nc
e
in
da
ta
s
e
ts
c
a
n
s
ig
ni
f
ic
a
nt
ly
r
e
duc
e
th
e
e
f
f
ic
ie
nc
y
of
c
onve
nt
io
na
l
ML
a
lg
o
r
it
hm
s
th
a
t
ha
ve
be
e
n
ba
s
e
d
on
f
a
ir
ly
ba
la
nc
e
d
da
ta
.
T
hi
s
pr
obl
e
m
is
a
ddr
e
s
s
e
d
e
f
f
e
c
ti
ve
ly
u
s
in
g
DM
by
a
ppl
yi
ng
va
r
io
us
m
e
th
ods
a
nd
te
c
hni
que
s
th
a
t
id
e
nt
if
y
c
om
pl
e
x
s
tr
uc
tu
r
e
s
a
nd
c
or
r
e
la
ti
ons
be
tw
e
e
n
c
la
s
s
e
s
in
s
por
a
di
c
a
ll
y
im
ba
la
n
c
e
d
d
a
ta
s
e
t
s
.
N
e
w
te
c
hni
que
s
in
c
lu
d
e
s
ynt
h
e
ti
c
s
a
m
pl
in
g,
c
os
t
-
s
e
ns
it
iv
e
l
e
a
r
ni
ng,
e
ns
e
m
bl
e
te
c
hni
que
s
,
a
nd
hybr
id
a
r
c
hi
te
c
tu
r
e
to
e
ns
ur
e
th
a
t
th
e
m
ode
l
s
s
ol
ve
th
e
pr
obl
e
m
of
unde
r
r
e
pr
e
s
e
nt
a
ti
on
of
th
e
m
in
or
it
y
c
la
s
s
e
s
w
hi
le
m
a
in
ta
in
in
g
th
e
m
e
r
it
s
of
ba
la
nc
e
d
m
ode
ls
.
T
hi
s
p
a
pe
r
pr
opos
e
s
th
e
us
e
o
f
DM
te
c
hni
que
s
in
bui
ld
in
g
r
e
li
a
bl
e
a
nd
f
a
ir
m
ode
ls
f
or
e
f
f
e
c
ti
ve
ha
ndl
in
g of
i
m
ba
la
nc
e
a
s
t
he
vol
um
e
a
nd
c
om
pl
e
xi
ty
of
da
ta
i
nc
r
e
a
s
e
s
.
‒
I
m
pr
ove
d
m
ode
l
pe
r
f
or
m
a
nc
e
:
w
he
n
th
e
da
ta
is
im
ba
la
nc
e
d,
i
t
c
a
n
le
a
d
to
th
e
c
r
e
a
ti
on
of
im
ba
la
nc
e
d
m
ode
ls
th
a
t
ha
ve
poor
pe
r
f
or
m
a
nc
e
w
he
n
it
c
om
e
s
to
th
e
m
in
o
r
it
y
c
la
s
s
.
P
r
e
pr
oc
e
s
s
in
g
te
c
hni
que
s
s
uc
h
a
s
r
e
s
a
m
pl
in
g
he
lp
s
to
c
or
r
e
c
t
im
ba
la
nc
e
in
th
e
num
be
r
of
in
s
ta
nc
e
s
th
a
t
be
lo
ng
s
to
th
e
m
in
or
it
y
c
la
s
s
th
e
r
e
by i
nc
r
e
a
s
in
g t
he
a
bi
li
ty
of
t
he
m
ode
l
in
c
or
r
e
c
tl
y i
de
nt
if
yi
ng t
hi
s
c
la
s
s
[
4]
.
‒
B
e
tt
e
r
ge
ne
r
a
li
z
a
ti
on:
im
ba
la
nc
e
d
da
ta
s
e
t
s
c
a
n
r
e
s
ul
t
in
m
ode
ls
th
a
t
ove
r
f
it
to
th
e
m
a
jo
r
it
y
c
la
s
s
.
D
M
m
e
th
ods
,
li
ke
f
e
a
tu
r
e
s
e
le
c
ti
on
a
nd
di
m
e
ns
io
na
li
ty
r
e
duc
ti
on,
c
a
n
he
lp
c
r
e
a
te
a
m
or
e
ba
la
nc
e
d
a
nd
r
e
pr
e
s
e
nt
a
ti
ve
da
ta
s
e
t,
a
ll
ow
in
g t
he
m
ode
l
to
ge
ne
r
a
li
z
e
b
e
tt
e
r
.
‒
A
nom
a
ly
de
te
c
ti
on:
in
s
c
e
na
r
io
s
li
ke
f
r
a
ud
de
te
c
ti
on,
id
e
nt
if
yi
ng
r
a
r
e
e
ve
nt
s
(
m
in
or
it
y
c
la
s
s
)
is
c
r
uc
ia
l.
D
M
m
e
th
ods
c
a
n h
e
lp
de
te
c
t
a
nom
a
li
e
s
or
r
a
r
e
c
a
s
e
s
e
f
f
e
c
ti
ve
ly
, e
ve
n i
n i
m
ba
la
nc
e
d da
ta
s
e
ts
.
‒
C
os
t
r
e
duc
ti
on:
in
c
e
r
ta
in
a
ppl
ic
a
ti
ons
,
m
is
c
la
s
s
if
yi
ng
in
s
ta
nc
e
s
f
r
om
th
e
m
in
or
it
y
c
la
s
s
c
a
n
ha
ve
s
ig
ni
f
ic
a
nt
c
os
ts
or
c
ons
e
que
nc
e
s
.
D
M
he
lp
s
in
bui
ld
in
g
m
od
e
ls
th
a
t
r
e
duc
e
th
e
s
e
c
o
s
ts
by
a
ddr
e
s
s
in
g
th
e
c
la
s
s
i
m
ba
l
a
nc
e
i
s
s
ue
.
‒
E
ns
e
m
bl
e
m
e
th
ods
:
e
ns
e
m
bl
e
le
a
r
ni
ng
te
c
hni
que
s
,
li
ke
ba
ggi
ng
a
nd
boos
ti
ng,
a
r
e
w
id
e
ly
u
s
e
d
in
D
M
to
im
pr
ove
m
ode
l
pe
r
f
or
m
a
nc
e
on
im
ba
la
nc
e
d
da
ta
s
e
ts
.
T
he
s
e
m
e
th
ods
c
om
bi
ne
m
ul
ti
pl
e
m
ode
l
s
to
m
a
ke
pr
e
di
c
ti
ons
, a
nd t
he
y c
a
n be
a
da
pt
e
d t
o gi
ve
m
or
e
w
e
ig
ht
t
o t
he
m
in
or
it
y
c
la
s
s
[
5]
.
‒
R
e
s
a
m
pl
in
g
te
c
hni
que
s
:
D
M
of
f
e
r
s
s
e
v
e
r
a
l
m
e
th
ods
of
r
e
s
a
m
p
li
ng
in
c
lu
di
ng
ove
r
s
a
m
pl
in
g
(
ge
ne
r
a
ti
ng
a
ddi
ti
ona
l
in
s
ta
nc
e
s
f
or
th
e
m
in
or
it
y
c
la
s
s
)
,
unde
r
-
s
a
m
pl
in
g
(
r
e
duc
in
g
th
e
num
be
r
of
in
s
ta
nc
e
s
,
be
lo
ngi
ng
to
th
e
m
a
jo
r
it
y
c
la
s
s
)
,
a
nd
s
ynt
he
ti
c
da
t
a
ge
ne
r
a
ti
on (
c
r
e
a
ti
ng
a
r
ti
f
ic
ia
l
s
a
m
pl
e
s
in
th
e
c
ont
e
xt
of
t
he
m
in
or
it
y c
la
s
s
)
. T
he
s
e
t
e
c
hni
que
s
c
a
n ba
la
n
c
e
t
he
da
ta
s
e
t
a
nd i
m
pr
ove
m
ode
l
tr
a
in
in
g
[
6]
.
‒
C
os
t
-
s
e
ns
it
iv
e
le
a
r
ni
ng
:
D
M
a
ll
ow
s
f
or
th
e
in
c
or
por
a
ti
on
of
c
os
t
-
s
e
ns
it
iv
e
le
a
r
ni
ng
te
c
hni
que
s
,
w
he
r
e
m
is
c
la
s
s
if
ic
a
ti
on
s
of
di
f
f
e
r
e
nt
c
la
s
s
e
s
a
r
e
a
s
s
ig
ne
d di
f
f
e
r
e
nt
c
os
ts
. T
hi
s
i
s
pa
r
ti
c
ul
a
r
ly
us
e
f
ul
i
n s
c
e
n
a
r
io
s
w
he
r
e
m
is
c
la
s
s
if
yi
ng t
he
m
in
or
it
y c
la
s
s
i
s
c
o
s
tl
ie
r
[
7]
, [
8]
.
‒
F
e
a
tu
r
e
e
ngi
ne
e
r
in
g:
f
e
a
tu
r
e
e
ngi
ne
e
r
in
g
te
c
hni
que
s
in
D
M
c
a
n
he
lp
c
r
e
a
te
in
f
or
m
a
ti
ve
f
e
a
tu
r
e
s
th
a
t
a
r
e
pa
r
ti
c
ul
a
r
ly
us
e
f
ul
f
or
di
s
ti
ngui
s
hi
ng be
twe
e
n c
la
s
s
e
s
i
n i
m
ba
la
nc
e
d da
ta
s
e
ts
.
‒
A
c
ti
ve
le
a
r
ni
ng:
D
M
te
c
hni
que
s
c
a
n
be
a
ppl
i
e
d
to
a
c
ti
ve
le
a
r
ni
ng
s
c
e
n
a
r
io
s
w
he
r
e
th
e
m
ode
l
s
e
le
c
ts
in
s
ta
nc
e
s
f
or
la
be
ll
in
g
th
a
t
a
r
e
li
ke
ly
to
im
pr
ove
it
s
pe
r
f
or
m
a
nc
e
on
th
e
m
in
or
it
y
c
la
s
s
,
th
us
opt
im
iz
in
g
th
e
l
e
a
r
ni
ng pr
oc
e
s
s
.
‒
E
va
lu
a
ti
on
m
e
tr
ic
s
:
D
M
pr
ovi
de
s
a
w
id
e
r
a
nge
of
e
va
lu
a
t
io
n
m
e
tr
ic
s
be
yond
a
c
c
ur
a
c
y,
s
uc
h
a
s
pr
e
c
is
io
n,
r
e
c
a
ll
,
F
1
-
s
c
or
e
,
a
nd
a
r
e
a
und
e
r
th
e
r
e
c
e
iv
e
r
ope
r
a
ti
ng
c
ha
r
a
c
te
r
is
ti
c
c
ur
ve
(
A
U
C
-
R
O
C
)
,
w
hi
c
h
a
r
e
m
or
e
in
f
or
m
a
ti
ve
f
or
im
ba
la
nc
e
d
da
ta
s
e
ts
.
T
he
s
e
m
e
tr
ic
s
gi
ve
a
b
e
tt
e
r
unde
r
s
ta
ndi
ng
of
a
m
ode
l’
s
pe
r
f
or
m
a
nc
e
on both m
a
jo
r
it
y a
nd mi
nor
it
y
c
la
s
s
e
s
[
9]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
v
ie
w
on c
la
s
s
i
m
bal
anc
e
t
e
c
hni
que
s
t
o s
tr
e
ngt
he
n m
ode
l
p
r
e
di
c
ti
on (
P
ut
ta
H
e
m
al
at
ha)
1729
2
.2. M
ac
h
in
e
l
e
ar
n
in
g
ap
p
r
oac
h
e
s
f
or
c
la
s
s
i
m
b
al
an
c
e
p
r
ob
le
m
s
M
L
is
a
s
ubs
e
t
of
a
r
ti
f
ic
ia
l
in
te
ll
ig
e
nc
e
(
A
I
)
th
a
t
e
na
bl
e
s
on
de
ve
lo
pi
ng
m
ode
ls
,
c
om
put
e
,
a
nd
le
a
r
n
f
r
om
e
vi
de
nc
e
s
a
nd
b
e
c
om
e
be
tt
e
r
a
t
c
e
r
ta
in
ta
s
ks
[
10]
.
I
n
e
s
s
e
nc
e
,
DM
f
oc
u
s
e
s
on
th
e
di
s
c
ove
r
y
of
pa
tt
e
r
ns
a
nd
r
e
la
ti
ons
hi
ps
w
it
hi
n
d
a
ta
,
w
hi
le
ML
f
oc
us
e
s
on
bui
ld
in
g
a
lg
or
it
hm
s
th
a
t
c
a
n
le
a
r
n
f
r
om
da
ta
to
pe
r
f
or
m
ta
s
ks
s
u
c
h a
s
c
la
s
s
if
ic
a
ti
on, r
e
gr
e
s
s
io
n,
a
nd
c
lu
s
te
r
in
g. T
he
s
e
t
w
o f
ie
ld
s
a
r
e
of
te
n i
nt
e
r
twi
ne
d, w
it
h t
e
c
hni
que
s
a
nd
in
s
ig
ht
s
f
r
om
one
f
ie
ld
in
f
o
r
m
in
g
a
nd
e
nha
nc
in
g
th
e
ot
h
e
r
.
M
L
c
a
n
be
f
ur
th
e
r
be
e
n
c
a
te
gor
iz
e
d
in
to
s
upe
r
vi
s
e
d
l
e
a
r
ni
ng,
uns
upe
r
vi
s
e
d
le
a
r
ni
ng,
a
nd
r
e
in
f
or
c
e
m
e
nt
l
e
a
r
ni
ng
a
s
s
ke
tc
h
e
d
in
F
ig
ur
e
1.
I
n s
upe
r
vi
s
e
d
le
a
r
ni
ng,
e
a
c
h
da
ta
poi
nt
ha
s
a
ta
r
ge
t
la
be
l,
th
us
th
e
s
ys
te
m
le
a
r
ns
f
r
om
la
be
ll
e
d
da
ta
.
T
he
obj
e
c
ti
ve
of
a
ny
m
ode
l
is
to
de
ve
lo
p
a
c
or
r
e
la
ti
on
a
m
ong
th
e
in
put
a
tt
r
ib
ut
e
s
a
nd
out
put
la
be
ls
s
o
th
a
t
th
e
m
ode
l
c
a
n
e
f
f
e
c
ti
ve
ly
pr
e
di
c
t
th
e
c
l
a
s
s
a
c
c
ur
a
te
ly
f
r
om
uns
e
e
n
da
ta
.
I
n
c
ont
r
a
s
t,
th
e
m
e
th
od
of
un
s
upe
r
vi
s
e
d
le
a
r
ni
ng
us
e
s
unl
a
be
l
e
d
da
ta
[
11]
.
S
upe
r
vi
s
e
d
le
a
r
ni
ng
lo
oks
f
or
s
tr
uc
tu
r
e
s
,
c
or
r
e
la
ti
ons
,
a
nd
pa
tt
e
r
ns
in
th
e
da
ta
,
f
or
e
xa
m
pl
e
:
c
lu
s
te
r
in
g
or
di
m
e
ns
io
na
li
ty
r
e
duc
ti
on
or
gr
ou
pi
ng
c
om
pa
r
a
bl
e
da
ta
poi
nt
s
to
ge
th
e
r
.
I
n
r
e
in
f
or
c
e
m
e
nt
le
a
r
ni
ng,
a
s
ys
te
m
a
s
s
um
e
s
th
e
r
ol
e
of
a
n
a
ge
nt
a
nd
in
te
r
a
c
ts
w
it
h
it
s
s
ur
r
oundings
to
di
s
c
ove
r
how
to
m
a
xi
m
iz
e
r
e
w
a
r
d
s
ig
na
ls
.
I
t
is
f
r
e
que
nt
ly
e
m
pl
oye
d
in
s
it
ua
ti
ons
w
he
r
e
a
n
a
ge
nt
de
ve
lo
p
s
th
e
a
bi
li
ty
to
m
a
ke
a
s
e
r
ie
s
of
d
e
c
is
io
n
s
.
ML
is
a
c
ont
in
uou
s
ly
e
vol
vi
ng
di
s
c
ip
li
ne
th
a
t
ha
s
s
e
e
n
r
e
m
a
r
ka
bl
e
pr
ogr
e
s
s
a
c
r
os
s
m
ul
ti
pl
e
f
ie
ld
s
.
H
ow
e
ve
r
,
th
e
r
e
a
r
e
s
ti
ll
s
e
ve
r
a
l
r
e
s
e
a
r
c
h
c
ha
ll
e
nge
s
th
a
t
r
e
s
e
a
r
c
he
r
s
a
nd
e
xpe
r
ts
a
r
e
a
c
ti
ve
ly
w
or
ki
ng t
o a
ddr
e
s
s
. S
om
e
of
t
he
s
e
c
h
a
ll
e
nge
s
i
nc
lu
d
e
:
in
te
r
pr
e
ta
bl
e
a
nd e
xpl
a
in
a
bl
e
m
ode
ls
, bi
a
s
a
nd
f
a
ir
ne
s
s
,
ha
ndl
in
g
m
ul
ti
c
la
s
s
or
c
la
s
s
im
ba
la
nc
e
pr
obl
e
m
s
,
a
d
ve
r
s
a
r
ia
l
a
tt
a
c
ks
a
nd
de
f
e
ns
e
s
,
s
c
a
la
bi
li
ty
a
nd
e
f
f
ic
ie
nc
y, a
nd c
a
us
a
l
in
f
e
r
e
nc
e
.
F
ig
ur
e
1. G
e
ne
r
a
l
ove
r
vi
e
w
of
di
f
f
e
r
e
nt
t
ype
s
of
ML
ta
s
ks
2
.2.1.
M
u
lt
i
-
la
b
e
l
c
la
s
s
vs
.
m
u
lt
ic
la
s
s
c
la
s
s
i
f
ic
at
io
n
M
ul
ti
-
la
be
l
c
la
s
s
a
nd
m
ul
ti
c
la
s
s
c
la
s
s
if
ic
a
ti
on
a
r
e
th
e
c
om
m
on
pr
obl
e
m
s
in
ML
a
nd
a
f
f
e
c
t
th
e
f
ie
ld
of
DM
a
nd
s
ta
ti
s
ti
c
s
.
N
um
e
r
ous
a
ppr
oa
c
he
s
ha
v
e
ge
n
e
r
a
ll
y
be
e
n
put
f
or
th
to
ba
la
n
c
e
th
e
unba
la
n
c
e
d
d
a
ta
s
e
ts
[
12]
, [
13]
. H
ow
e
ve
r
, due
t
o t
he
l
a
r
ge
r
da
ta
s
iz
e
, i
t
r
e
m
a
in
s
a
c
ha
ll
e
ngi
ng t
a
s
k a
s
s
how
n i
n F
ig
ur
e
2.
F
ig
ur
e
2. D
if
f
e
r
e
nt
t
ype
s
of
c
la
s
s
l
e
a
r
ne
r
c
ha
ll
e
nge
s
i
n
ML
2
.2.2. M
u
lt
ic
la
s
s
d
at
a
M
ul
ti
c
la
s
s
da
ta
r
e
f
e
r
s
to
a
f
or
m
of
da
t
a
s
e
t
f
r
e
que
nt
ly
ut
il
iz
e
d
i
n
M
L
a
nd
da
t
a
a
na
ly
s
i
s
a
ppl
ic
a
ti
on
s
.
E
a
c
h d
a
ta
poi
nt
i
n a
m
ul
ti
-
c
l
a
s
s
d
a
ta
s
e
t
c
or
r
e
s
p
ond
s
t
o o
ne
of
s
e
ve
r
a
l
pr
e
s
e
t
c
la
s
s
e
s
or
gr
oup
s
.
T
o
put
i
t
a
not
he
r
w
a
y,
e
a
c
h
da
ta
poi
nt
c
a
n be
c
l
a
s
s
if
ie
d
in
t
o
o
ne
c
l
a
s
s
out
of
th
r
e
e
or
m
or
e
c
la
s
s
e
s
. T
h
e
m
ul
ti
c
l
a
s
s
m
e
th
od pa
v
e
s
a
pa
t
h t
o d
e
a
l
w
it
h
da
t
a
poi
nt
s
t
ha
t
be
lo
ng t
o
nume
r
ous
c
l
a
s
s
e
s
a
nd pe
r
f
or
m
M
L
a
nd d
a
ta
a
na
ly
s
is
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
1727
-
1742
1730
2
.2.3.
A
p
p
li
c
at
io
n
s
of
m
u
lt
ic
la
s
s
d
at
a
T
he
r
e
a
r
e
s
e
ve
r
a
l
u
s
e
s
f
or
m
ul
ti
c
la
s
s
da
ta
in
num
e
r
ou
s
di
s
c
ip
li
ne
s
a
nd
s
e
c
to
r
s
.
F
ol
lo
w
in
g
w
il
l
be
a
f
e
w
e
xa
m
pl
e
s
of
how
m
ul
ti
c
la
s
s
da
ta
is
a
ppl
ie
d
in
va
r
io
us
f
ie
ld
s
[
14]
:
i)
im
a
ge
pr
oc
e
s
s
in
g:
c
ons
id
e
r
a
da
ta
s
e
t
of
im
a
ge
s
w
he
r
e
th
e
a
im
is
to
c
a
te
gor
iz
e
th
e
im
a
ge
s
in
to
m
u
lt
ip
le
obj
e
c
t
c
la
s
s
e
s
,
f
or
e
xa
m
pl
e
c
a
ts
,
dogs
,
bi
r
ds
,
a
nd
c
a
r
s
.
T
h
e
c
ha
ll
e
nge
i
s
one
of
m
ul
ti
c
la
s
s
c
a
te
gor
iz
a
ti
o
n
be
c
a
us
e
e
a
c
h
im
a
ge
b
e
lo
ngs
to
one
of
th
e
s
e
va
r
io
us
c
la
s
s
e
s
;
ii
)
m
e
di
c
a
l
di
a
gnos
is
:
m
ul
ti
c
la
s
s
c
l
a
s
s
if
ic
a
ti
on
r
e
f
e
r
s
to
th
e
pr
oc
e
s
s
of
c
la
s
s
if
yi
ng
di
f
f
e
r
e
nt
di
s
e
a
s
e
s
ba
s
e
d
on
in
f
or
m
a
ti
on
a
bout
th
e
pa
ti
e
nt
,
s
u
c
h
a
s
s
ym
pt
om
s
a
nd
te
s
t
f
in
di
ngs
;
ii
i)
te
xt
c
a
te
gor
iz
a
ti
on:
in
na
tu
r
a
l
la
ngua
ge
pr
oc
e
s
s
in
g,
you
m
ig
ht
ha
v
e
a
da
ta
s
e
t
of
te
xt
doc
um
e
nt
s
th
a
t
ne
e
d
to
be
c
a
te
gor
iz
e
d
in
to
va
r
io
us
to
pi
c
s
or
th
e
m
e
s
.
E
a
c
h
doc
um
e
nt
c
a
n
be
lo
ng
to
one
of
m
a
ny
pos
s
ib
le
c
a
te
gor
ie
s
,
s
uc
h
a
s
s
por
ts
,
pol
it
ic
s
,
e
nt
e
r
ta
in
m
e
nt
,
a
nd
te
c
hnol
ogy
;
iv
)
s
p
e
e
c
h
r
e
c
ogni
ti
on:
i
t
is
th
e
pr
oc
e
s
s
of
id
e
nt
if
yi
ng
w
or
ds
us
e
d
in
s
pe
e
c
h,
phr
a
s
e
s
,
or
c
om
m
a
nds
a
nd
c
a
te
gor
iz
in
g
th
e
m
,
a
ll
ow
in
g
voi
c
e
a
s
s
is
ta
nt
s
to
c
om
pr
e
he
nd
va
r
io
us
u
s
e
r
c
om
m
a
nds
;
v)
m
a
nuf
a
c
tu
r
in
g
a
nd
qua
li
ty
c
ont
r
ol
:
i
t
in
vol
ve
s
in
s
pe
c
ti
ng
a
nd
c
a
t
e
gor
iz
in
g
pr
oduc
ts
on
a
n
a
s
s
e
m
bl
y
li
ne
a
c
c
or
di
ng
to
th
e
ir
qua
li
ty
o
r
f
la
w
s
;
a
nd
vi
)
f
i
na
nc
ia
l
da
ta
a
na
ly
s
is
:
f
in
a
nc
ia
l
da
ta
a
na
ly
s
i
s
in
vol
ve
s
e
xa
m
in
in
g
a
nd
in
te
r
pr
e
ti
ng
da
ta
r
e
la
te
d
to
f
in
a
n
c
ia
l
tr
a
ns
a
c
ti
ons
,
in
ve
s
tm
e
nt
s
,
m
a
r
ke
ts
,
a
nd
e
c
onomi
c
i
ndi
c
a
to
r
s
t
o ga
in
i
ns
ig
ht
s
a
nd ma
ke
i
nf
or
m
e
d de
c
is
io
ns
.
2
.3.
C
la
s
s
im
b
al
an
c
e
A
c
c
ur
a
te
ly
e
s
ti
m
a
ti
ng
th
e
c
la
s
s
di
s
tr
ib
ut
io
n
be
c
om
e
s
a
n
is
s
ue
w
he
n
de
a
li
ng
w
it
h
da
ta
w
he
r
e
c
e
r
ta
in
c
la
s
s
e
s
ha
ve
f
e
w
s
a
m
pl
e
s
th
a
n
ot
he
r
s
,
a
ls
o
known
a
s
c
la
s
s
i
m
ba
la
nc
e
.
T
h
e
r
e
is
a
te
nd
e
nc
y
to
in
c
r
e
a
s
e
th
e
de
pe
nde
nc
e
on
m
ode
ls
tr
a
in
e
d
on
s
uc
h
da
ta
a
nd
th
e
y
m
a
y
be
a
ls
o
pr
e
ju
di
c
e
d
to
th
e
m
a
jo
r
it
y
c
la
s
s
,
le
a
di
ng
to
in
a
c
c
ur
a
te
pr
e
di
c
ti
ons
a
nd
m
i
s
c
la
s
s
if
ic
a
ti
on.
T
hi
s
is
s
ue
f
r
e
que
nt
ly
a
r
is
e
s
in
f
ie
ld
s
li
ke
a
nom
a
ly
d
e
te
c
ti
on,
r
is
k
a
s
s
e
s
s
m
e
nt
,
c
r
im
e
de
te
c
ti
on,
m
e
di
c
a
l
a
na
ly
s
i
s
,
a
nd
s
pa
m
f
il
te
r
in
g,
w
he
r
e
one
c
la
s
s
a
ppe
a
r
s
s
ig
ni
f
ic
a
nt
ly
le
s
s
of
te
n t
ha
n t
he
ot
he
r
s
.
‒
T
he
c
ha
ll
e
nge
s
of
c
la
s
s
im
ba
la
nc
e
a
r
e
f
ol
lo
w
s
:
i)
b
ia
s
to
w
a
r
ds
m
a
jo
r
it
y
c
la
s
s
:
m
ode
ls
te
nd
to
pe
r
f
or
m
w
e
ll
i
n t
he
m
a
jo
r
it
y
c
la
s
s
w
hi
le
s
tr
uggl
in
g t
o pr
e
di
c
t
th
e
m
in
o
r
i
ty
c
la
s
s
due
t
o t
he
a
bunda
nc
e
of
m
a
jo
r
it
y
c
la
s
s
s
a
m
pl
e
s
;
ii
)
r
e
duc
e
d
ge
ne
r
a
li
z
a
ti
on:
th
e
m
ode
l
m
ig
ht
ge
n
e
r
a
li
z
e
poor
ly
to
ne
w
da
ta
,
e
s
pe
c
ia
ll
y
f
or
th
e
m
in
or
it
y
c
la
s
s
,
le
a
di
ng
to
poor
pe
r
f
or
m
a
nc
e
on
uns
e
e
n
da
ta
;
a
nd
ii
i)
bi
a
s
e
d
e
va
lu
a
ti
on:
tr
a
di
ti
ona
l
a
c
c
ur
a
c
y
m
ig
ht
be
m
is
le
a
di
ng
s
in
c
e
a
na
iv
e
m
ode
l
th
a
t
pr
e
di
c
t
s
th
e
m
a
jo
r
it
y
c
la
s
s
f
or
e
ve
r
yt
hi
ng
c
oul
d
s
ti
ll
a
c
hi
e
ve
hi
gh a
c
c
ur
a
c
y i
n i
m
ba
la
nc
e
d da
t
a
s
e
t
s
.
‒
S
tr
a
te
gi
e
s
to
a
ddr
e
s
s
c
la
s
s
im
ba
la
nc
e
a
r
e
f
ol
lo
w
s
:
i)
r
e
s
a
m
pl
in
g
te
c
hni
que
s
:
i
n
ML
,
r
e
s
a
m
pl
in
g
i
s
a
ty
pi
c
a
l
m
e
th
od
f
or
a
ddr
e
s
s
in
g
c
l
a
s
s
im
ba
la
nc
e
.
B
y
c
hoo
s
in
g
in
s
ta
nc
e
s
f
r
om
th
e
or
ig
in
a
l
da
ta
s
e
t,
a
ne
w
tr
a
in
in
g
da
ta
s
e
t
w
it
h
a
di
f
f
e
r
e
nt
c
la
s
s
di
s
tr
ib
ut
io
n
w
il
l
b
e
c
r
e
a
t
e
d.
R
e
s
a
m
pl
in
g
te
c
hni
que
s
c
a
n
be
di
vi
de
d
in
to
th
e
f
ol
lo
w
in
g
c
a
te
gor
ie
s
;
ii
)
ove
r
s
a
m
pl
in
g
:
in
c
r
e
a
s
in
g
th
e
num
be
r
of
in
s
ta
nc
e
s
in
th
e
m
in
or
it
y
c
la
s
s
by
dupl
ic
a
ti
ng
or
ge
ne
r
a
ti
ng
s
ynt
he
ti
c
da
ta
poi
nt
s
[
15]
;
ii
i)
unde
r
s
a
m
pl
in
g:
r
e
duc
in
g
th
e
num
be
r
of
in
s
ta
nc
e
s
in
th
e
m
a
jo
r
it
y
c
la
s
s
by
r
a
ndoml
y
r
e
m
ovi
ng
s
a
m
pl
e
s
[
16]
;
iv
)
S
M
O
T
E
:
it
s
e
nge
nde
r
s
f
a
br
ic
a
te
d
or
s
ynt
he
ti
c
s
a
m
pl
e
s
f
or
th
e
m
in
or
it
y
c
la
s
s
ba
s
e
d
o
n
in
c
or
por
a
ti
ng
a
m
ong
th
e
e
xi
s
ti
ng
da
ta
poi
nt
s
[
17]
;
v)
c
os
t
-
s
e
ns
it
iv
e
le
a
r
ni
ng:
t
hi
s
a
ppr
oa
c
h
in
ML
a
c
c
ount
s
f
or
th
e
va
r
yi
ng
c
o
s
ts
r
e
la
t
e
d
w
it
h
m
ul
ti
pl
e
ki
nds
of
m
is
c
la
s
s
if
ic
a
ti
on
e
r
r
or
s
in
a
c
la
s
s
if
ic
a
ti
on
m
ode
l.
C
onve
nt
io
na
l
ML
m
ode
l
s
a
im
to
de
c
r
e
a
s
e
ove
r
a
ll
e
r
r
or
, but
t
he
y do not a
c
c
ount
f
or
t
he
pa
r
t
ic
ul
a
r
c
os
ts
c
onne
c
te
d w
it
h f
a
ls
e
po
s
it
iv
e
s
a
nd
f
a
ls
e
ne
ga
ti
ve
s
,
in
s
c
e
n
a
r
io
s
w
he
r
e
m
is
c
la
s
s
if
ic
a
ti
on
c
os
ts
a
r
e
i
m
ba
la
nc
e
.
C
os
t
-
s
e
n
s
it
iv
e
le
a
r
ni
ng
c
a
n
be
e
a
s
il
y
in
te
gr
a
te
d
w
it
h
c
la
s
s
if
ic
a
ti
on
r
e
gr
e
s
s
io
n
tr
e
e
(
C
A
R
T
)
,
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
,
a
nd
na
ïv
e
B
a
ye
s
[
18]
;
vi
)
a
nom
a
ly
de
te
c
ti
on:
th
e
id
e
nt
if
ic
a
ti
on
of
pa
tt
e
r
ns
or
oc
c
ur
r
e
nc
e
s
w
it
hi
n
a
d
a
ta
s
e
t
th
a
t
de
vi
a
te
f
r
om
e
xpe
c
te
d
be
ha
vi
or
is
c
a
ll
e
d
a
nom
a
ly
de
te
c
ti
on,
s
om
e
ti
m
e
s
r
e
f
e
r
r
e
d
to
a
s
out
li
e
r
de
te
c
ti
on.
T
he
c
hoi
c
e
of
a
nom
a
ly
de
te
c
ti
on
te
c
hni
que
,
s
uc
h
a
s
s
ta
ti
s
ti
c
a
l,
de
ns
it
y
-
or
di
s
ta
nc
e
-
ba
s
e
d
m
e
th
odol
ogi
e
s
, ba
s
e
d on the
na
tu
r
e
of
t
he
da
ta
s
e
t
a
nd t
he
e
xpl
i
c
it
ne
c
e
s
s
it
ie
s
of
t
he
a
ppl
ic
a
ti
on
[
19]
;
vi
i
)
da
ta
a
ugm
e
nt
a
ti
on:
it
is
a
pr
oc
e
dur
e
of
pe
r
f
or
m
in
g
va
r
io
us
m
odi
f
ic
a
ti
ons
on
th
e
tr
a
in
in
g
da
ta
a
nd
it
s
w
id
e
ly
a
ppl
ie
d
on
c
om
put
e
r
vi
s
io
n
a
nd
ML
to
s
ynt
he
ti
c
a
ll
y
r
a
is
e
th
e
vol
um
e
of
a
tr
a
in
in
g
da
ta
s
e
t.
S
e
ve
r
a
l
w
e
ll
-
li
ke
d
m
e
th
ods
f
or
e
nha
nc
in
g
d
a
ta
in
c
lu
de
r
a
ndo
m
s
w
a
ppi
ng,
tr
a
ns
la
ti
ng,
c
r
oppi
ng,
pi
tc
h
s
hi
f
ti
ng,
ti
m
e
w
a
r
pi
ng,
s
c
a
li
ng,
a
nd
de
f
or
m
a
ti
on
[
20]
;
vi
ii
)
e
ns
e
m
bl
e
m
e
th
ods
:
e
ns
e
m
bl
e
m
e
th
ods
a
r
e
a
pow
e
r
f
ul
a
ppr
oa
c
h
th
a
t
ha
ndl
e
s
th
e
c
la
s
s
im
ba
la
nc
e
in
ML
.
S
om
e
e
ns
e
m
bl
e
m
e
th
ods
th
a
t
c
a
n
b
e
e
f
f
e
c
ti
ve
f
or
a
ddr
e
s
s
in
g
c
la
s
s
im
ba
la
nc
e
li
ke
S
M
O
T
E
ba
s
e
d
a
p
pr
oa
c
he
s
,
e
a
s
y
e
n
s
e
m
bl
e
,
A
da
B
oo
s
t,
a
nd
X
G
B
oos
t
w
it
h s
c
a
le
pos
e
w
e
ig
ht
[
21]
.
‒
S
M
O
T
E
:
I
t
is
a
n
im
por
ta
nt
te
c
hni
que
us
e
d
in
ML
a
nd
DM
to
de
a
l
w
it
h
im
ba
la
nc
e
pr
obl
e
m
i
n
c
la
s
s
if
ic
a
ti
on
pr
obl
e
m
s
[
22]
.
I
t
s
pe
c
if
ic
a
ll
y
a
ddr
e
s
s
e
s
th
is
i
s
s
u
e
s
in
c
e
it
pr
oduc
e
s
s
ynt
he
ti
c
s
a
m
pl
e
s
of
th
e
m
in
or
it
y
c
la
s
s
if
ic
a
ti
on.
T
he
c
or
e
pr
in
c
ip
le
of
S
M
O
T
E
in
vol
ve
s
c
r
e
a
ti
ng
s
ynt
he
ti
c
in
s
ta
nc
e
s
th
a
t
r
e
s
e
m
bl
e
e
xi
s
ti
ng
m
in
or
it
y
c
la
s
s
s
a
m
pl
e
s
,
th
e
r
e
by
e
nha
nc
in
g
th
e
r
e
pr
e
s
e
nt
a
ti
on
of
th
e
m
in
or
it
y
s
a
m
pl
e
s
a
nd
im
pr
ovi
ng
c
la
s
s
di
s
tr
ib
ut
io
n
ba
la
nc
e
th
r
ough
th
e
s
e
s
ynt
he
s
iz
e
d
s
a
m
pl
e
s
,
S
M
O
T
E
ope
ns
th
e
a
bi
li
ty
of
th
e
m
ode
l
to
le
a
r
n
m
o
s
t
of
th
e
c
ha
r
a
c
te
r
is
ti
c
s
of
th
e
m
in
or
i
ty
c
la
s
s
r
a
th
e
r
th
a
n
ju
s
t
f
oc
us
in
g
on
th
e
m
a
jo
r
it
y
c
la
s
s
th
us
im
pr
ovi
ng
on
th
e
c
la
s
s
if
ic
a
ti
on
of
th
e
m
od
e
l.
T
hi
s
le
a
ds
to
im
pr
ove
d
c
la
s
s
if
ic
a
ti
on
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
v
ie
w
on c
la
s
s
i
m
bal
anc
e
t
e
c
hni
que
s
t
o s
tr
e
ngt
he
n m
ode
l
p
r
e
di
c
ti
on (
P
ut
ta
H
e
m
al
at
ha)
1731
pe
r
f
or
m
a
nc
e
on
th
e
unde
r
r
e
pr
e
s
e
nt
e
d
c
la
s
s
a
nd
ov
e
r
a
ll
be
tt
e
r
ge
ne
r
a
li
z
a
ti
on
of
th
e
m
ode
l.
T
he
r
e
a
r
e
s
e
ve
r
a
l
va
r
ia
ti
ons
of
S
M
O
T
E
,
e
a
c
h
de
s
ig
ne
d
to
a
ddr
e
s
s
s
pe
c
if
ic
c
ha
ll
e
nge
s
or
s
c
e
na
r
io
s
.
T
he
di
f
f
e
r
e
nt
ty
pe
s
of
S
M
O
T
E
m
e
th
ods
a
r
e
a
s
f
ol
lo
w
s
[
23]
:
i)
S
M
O
T
E
:
it
is
one
of
th
e
m
e
th
ods
us
e
d
in
ML
a
nd
DM
to
de
a
l
w
it
h
th
e
c
h
a
ll
e
nge
of
c
la
s
s
im
ba
la
n
c
e
in
th
e
c
l
a
s
s
if
ic
a
ti
on
pr
obl
e
m
.
C
la
s
s
im
ba
l
a
nc
e
r
e
f
e
r
s
to
a
s
it
ua
ti
on w
he
r
e
by t
he
i
ns
ta
nc
e
s
be
lo
ng t
o one
c
la
s
s
a
r
e
ve
r
y f
e
w
t
ha
n t
hos
e
of
t
he
ot
he
r
c
la
s
s
. I
n or
de
r
t
o
s
ol
ve
th
e
pr
obl
e
m
,
S
M
O
T
E
c
r
e
a
te
ne
w
s
a
m
pl
e
s
ba
s
e
d
on
th
e
e
xi
s
ti
ng
in
s
ta
nc
e
s
of
th
e
m
in
or
it
ie
s
c
la
s
s
.
I
n
or
de
r
to
c
r
e
a
te
m
in
or
it
y
s
a
m
pl
e
s
S
M
O
T
E
,
a
ppl
ie
s
a
r
bi
tr
a
r
y
s
e
le
c
ti
on
of
a
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
a
nd
id
e
nt
if
yi
ng
a
gr
oup
of
it
s
k
-
ne
a
r
e
s
t
ne
ig
hbor
s
(
K
N
N
)
,
a
nd
th
e
n
c
r
e
a
ti
ng
ne
w
s
ynt
he
ti
c
in
s
ta
n
c
e
s
ba
s
e
d
on
th
e
s
e
r
e
la
ti
ons
hi
ps
[
24]
;
ii
)
B
or
de
r
li
ne
-
S
M
O
T
E
:
th
is
a
lg
o
r
it
hm
be
lo
ngs
to
th
e
e
xt
e
nde
d
S
M
O
T
E
a
lg
or
it
hm
f
a
m
il
y
w
hi
c
h i
s
a
m
e
th
od t
ha
t
or
ig
in
a
te
d f
r
om
t
he
ba
s
ic
S
M
O
T
E
a
lg
or
it
hm
but
w
it
h a
n a
im
of
ha
ndl
in
g
bor
de
r
li
ne
in
s
ta
nc
e
s
w
he
r
e
by
th
e
y
a
r
e
c
on
s
tr
uc
te
d
f
r
om
in
s
ta
nc
e
s
in
th
e
m
in
or
it
y
c
la
s
s
ne
a
r
by
th
e
de
c
is
io
n
bor
de
r
.
I
ns
te
a
d
of
us
in
g
a
ll
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
s
,
bor
de
r
li
ne
-
S
M
O
T
E
onl
y
s
e
le
c
ts
th
os
e
in
s
ta
nc
e
s
th
a
t
a
r
e
ne
a
r
th
e
de
c
is
io
n
bounda
r
y.
T
hi
s
a
ppr
oa
c
h a
im
s
to
ge
ne
r
a
te
s
ynt
he
ti
c
in
s
ta
nc
e
s
w
he
r
e
th
e
y
a
r
e
m
os
t
ne
e
de
d,
ne
a
r
th
e
bor
de
r
li
ne
be
twe
e
n
c
la
s
s
e
s
[
25]
;
ii
i)
a
da
pt
iv
e
s
ynt
he
ti
c
(
A
D
A
S
Y
N
)
s
a
m
pl
in
g
a
ppr
oa
c
h
:
i
t
is
a
n
a
d
a
pt
iv
e
e
xt
e
n
s
io
n
of
S
M
O
T
E
.
I
t
a
im
s
on
c
r
a
ti
ng
m
or
e
f
a
br
ic
a
te
d/
s
ynt
he
ti
c
s
a
m
pl
e
s
f
or
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
s
th
a
t
a
r
e
di
f
f
ic
ul
t
to
c
la
s
s
if
y,
s
uc
h
a
s
th
os
e
c
lo
s
e
to
th
e
de
c
is
io
n
bounda
r
y.
A
D
A
S
Y
N
us
e
s
a
de
n
s
it
y
di
s
tr
ib
ut
io
n
to
de
te
r
m
in
e
th
e
num
be
r
of
s
ynt
he
ti
c
s
a
m
pl
e
s
to
c
r
e
a
te
f
or
e
a
c
h
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
,
m
a
ki
ng
it
a
da
pt
iv
e
to
th
e
da
ta
di
s
tr
ib
ut
io
n
;
iv
)
S
M
O
T
E
w
it
h
e
di
te
d
ne
a
r
e
s
t
ne
ig
hbor
s
(
E
N
N
)
:
c
om
bi
ne
s
S
M
O
T
E
w
it
h
th
e
E
N
N
a
lg
or
it
hm
.
I
t
f
ir
s
t
ove
r
s
a
m
pl
e
s
th
e
m
in
or
i
ty
c
la
s
s
us
in
g
S
M
O
T
E
a
nd
th
e
n
us
e
s
E
N
N
to
c
le
a
n
th
e
s
ynt
he
ti
c
s
a
m
pl
e
s
by
r
e
m
ovi
ng
noi
s
y
one
s
a
nd
bor
de
r
li
ne
in
s
ta
nc
e
s
f
r
om
bot
h
th
e
m
in
or
it
y
a
nd
m
a
jo
r
it
y
c
la
s
s
e
s
[
26]
;
v)
S
M
O
T
E
-
T
om
e
k
li
nks
:
c
om
bi
na
ti
on
of
S
M
O
T
E
a
nd
th
e
T
om
e
k
li
nks
unde
r
s
a
m
pl
in
g
m
e
th
od.
I
t
us
e
s
S
M
O
T
E
to
ove
r
s
a
m
pl
e
th
e
m
in
or
it
y
c
la
s
s
a
nd
th
e
n
id
e
nt
if
ie
s
T
om
e
k
li
nks
(
pa
ir
s
of
i
ns
ta
nc
e
s
f
r
om
di
f
f
e
r
e
nt
c
la
s
s
e
s
th
a
t
a
r
e
c
lo
s
e
to
e
a
c
h
ot
he
r
a
nd
a
c
t
a
s
m
ut
ua
l
ne
a
r
e
s
t
ne
ig
hbor
s
)
.
T
h
e
T
om
e
k
li
nks
a
r
e
r
e
m
ove
d,
le
a
di
ng
to
a
c
le
a
ne
r
da
ta
s
e
t
[
27]
;
vi
)
s
a
f
e
-
le
ve
l
S
M
O
T
E
:
de
s
ig
ne
d
to
a
ddr
e
s
s
th
e
i
s
s
ue
of
ove
r
ge
ne
r
a
li
z
a
ti
on
w
he
n
ge
ne
r
a
ti
ng
s
ynt
he
ti
c
s
a
m
pl
e
s
.
I
t
c
a
lc
ul
a
te
s
a
“
s
a
f
e
le
v
e
l”
f
or
e
a
c
h
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
ba
s
e
d
on
th
e
de
ns
it
y
of
th
e
m
a
jo
r
it
y
c
la
s
s
in
s
ta
nc
e
s
in
it
s
n
e
ig
hbor
hood
.
S
ynt
he
ti
c
s
a
m
pl
e
s
a
r
e
ge
n
e
r
a
te
d
f
or
in
s
ta
nc
e
s
w
it
h
s
a
f
e
le
ve
l
s
a
bove
a
c
e
r
ta
in
th
r
e
s
hol
d
[
28]
;
vi
i)
e
vol
ut
io
na
r
y
m
ul
ti
-
di
s
tr
ib
ut
io
n
ove
r
-
s
a
m
pl
in
g
(
E
M
D
O
)
-
S
M
O
T
E
:
a
n
e
xt
e
ns
io
n
of
S
M
O
T
E
th
a
t
c
a
n
ha
ndl
e
da
ta
s
e
ts
w
it
h
m
ul
ti
pl
e
m
in
or
it
y
c
la
s
s
di
s
tr
ib
ut
io
ns
.
I
t
a
da
pt
s
th
e
ove
r
s
a
m
pl
in
g
s
tr
a
te
gy
ba
s
e
d
on
th
e
di
s
tr
ib
ut
io
n
of
th
e
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
s
,
a
ll
ow
in
g
it
to
ge
ne
r
a
te
s
ynt
he
ti
c
s
a
m
pl
e
s
m
or
e
e
f
f
e
c
ti
ve
ly
;
vi
ii
)
G
a
us
s
ia
n
S
M
O
T
E
:
it
is
a
n
a
dva
nc
e
d
va
r
ia
ti
on
of
th
e
or
ig
in
a
l
S
M
O
T
E
a
lg
or
it
hm
us
e
d
in
a
s
s
e
s
s
in
g
th
e
im
ba
la
n
c
e
d
c
la
s
s
pr
obl
e
m
in
ML
.
I
t
bui
ld
s
upon
th
e
c
onc
e
pt
of
S
M
O
T
E
by
in
tr
oduc
in
g
a
m
or
e
s
ophi
s
ti
c
a
te
d
a
ppr
oa
c
h
to
s
ynt
he
s
iz
in
g
a
ddi
ti
ona
l
s
a
m
pl
e
s
to
im
pr
ove
th
e
pe
r
s
oni
f
ic
a
t
io
n
of
th
e
m
in
or
it
y
c
la
s
s
la
be
ls
.
T
hi
s
a
ppr
oa
c
h
e
nha
nc
e
s
pr
oc
e
s
s
by
c
ons
id
e
r
in
g
th
e
di
s
tr
ib
ut
io
n
of
f
e
a
tu
r
e
va
lu
e
s
us
in
g
G
a
us
s
ia
n
di
s
tr
ib
ut
io
ns
.
T
hi
s
out
c
om
e
s
in
f
a
br
ic
a
te
d
s
a
m
pl
e
s
th
a
t
a
r
e
not
onl
y
lo
c
a
te
d
on
li
ne
s
e
gm
e
nt
s
b
e
twe
e
n
in
s
ta
nc
e
s
but
a
r
e
a
l
s
o
in
f
lu
e
nc
e
d
by
th
e
und
e
r
ly
in
g
di
s
tr
ib
ut
io
n
of
th
e
da
ta
[
29]
;
a
nd
ix
)
f
uz
z
y
ba
s
e
d
G
a
us
s
ia
n
(
F
G
)
-
S
M
O
T
E
:
it
is
ove
r
s
a
m
pl
in
g
te
c
hni
que
e
m
pl
oy
e
d
by
th
e
r
e
s
e
a
r
c
he
r
to
a
ddr
e
s
s
th
e
c
la
s
s
im
ba
la
nc
e
pr
obl
e
m
w
he
r
e
f
uz
z
y
G
a
us
s
ia
n
m
e
th
odol
ogy
w
a
s
us
e
d.
T
he
r
e
s
ul
ts
pr
ove
n
to
be
m
or
e
s
a
ti
s
f
yi
ng t
ha
n t
he
e
xi
s
ti
ng me
th
od
s
of
S
M
O
T
E
va
r
ia
nt
s
[
30]
.
E
a
c
h
of
th
e
s
e
va
r
ia
ti
ons
of
S
M
O
T
E
ha
s
it
s
s
tr
e
ngt
hs
a
nd
is
s
ui
te
d
to
di
f
f
e
r
e
nt
s
c
e
na
r
io
s
a
nd
da
ta
s
e
ts
.
T
he
pr
e
f
e
r
e
nc
e
of
w
hi
c
h
S
M
O
T
E
v
a
r
ia
nt
to
u
s
e
ba
s
e
d
on
th
e
pa
r
ti
c
ul
a
r
c
ha
r
a
c
te
r
is
ti
c
s
of
th
e
d
a
ta
s
e
t
a
nd
th
e
c
ha
ll
e
nge
s
po
s
e
d by c
la
s
s
i
m
ba
la
n
c
e
i
n t
he
M
L
ta
s
k.
3.
S
U
R
V
E
Y
O
N
C
L
A
S
S
I
M
B
A
L
A
N
C
E
P
R
O
B
L
E
M
S
A
ke
y
c
ha
ll
e
nge
in
th
e
im
ba
la
nc
e
d
c
la
s
s
pr
obl
e
m
is
th
e
r
e
li
a
bi
li
ty
of
a
c
c
ur
a
te
ne
s
s
a
nd
e
r
r
or
pr
opor
ti
on
a
s
m
e
tr
ic
s
f
or
e
v
a
lu
a
ti
ng
c
la
s
s
if
ie
r
pe
r
f
or
m
a
nc
e
.
F
i
gur
e
3
s
how
s
th
e
une
qua
l
di
s
tr
ib
ut
io
n
be
twe
e
n
th
e
two
c
la
s
s
e
s
,
w
he
r
e
gr
e
e
n
r
in
gs
r
e
pr
e
s
e
nt
th
e
m
a
jo
r
it
y
c
l
a
s
s
e
s
,
a
nd
th
e
bl
ue
c
ol
or
s
ta
r
r
e
pr
e
s
e
nt
s
th
e
m
in
or
it
y c
la
s
s
e
s
. F
ig
ur
e
4 s
how
s
t
he
s
a
m
pl
e
ge
ne
r
a
ti
on ove
r
i
m
ba
la
nc
e
d da
ta
s
e
ts
w
it
h r
e
s
pe
c
t
to
t
he
s
ynt
he
ti
c
s
a
m
pl
e
s
(
ye
ll
ow
di
a
m
ond)
.
C
ons
id
e
r
a
d
a
ta
s
e
t
w
he
r
e
th
e
99%
of
th
e
da
ta
w
it
h
m
a
jo
r
it
y
c
l
a
s
s
a
c
c
ount
s
a
nd
1%
of
th
e
m
in
or
it
y
c
la
s
s
of
th
e
da
ta
.
T
he
n
it
’
s
c
a
ll
e
d
a
d
a
ta
s
e
t
w
it
h
a
99:
1
im
ba
la
n
c
e
.
A
c
l
a
s
s
if
ie
r
li
ke
n
a
iv
e
,
w
hi
c
h
c
ons
i
s
te
nt
ly
pr
e
di
c
ts
th
e
m
a
jo
r
it
y
c
la
s
s
,
w
il
l
ha
ve
a
99%
a
c
c
ur
a
c
y
r
a
te
und
e
r
s
uc
h
c
ir
c
um
s
ta
nc
e
s
.
T
he
r
e
f
or
e
,
th
e
m
a
jo
r
it
y
c
la
s
s
if
ie
r
w
il
l
b
e
99.99%
a
c
c
ur
a
te
if
a
da
t
a
s
e
t
ha
s
a
b
a
la
nc
e
r
a
ti
o
of
9999:1.
C
on
s
id
e
r
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
ut
m
os
t
c
onve
nt
io
na
l
c
la
s
s
if
ie
r
s
w
h
e
n
us
e
d
in
th
e
im
ba
la
nc
e
d
c
la
s
s
pr
obl
e
m
to
il
lu
s
tr
a
te
on
e
e
f
f
e
c
t
of
th
is
c
ons
tr
a
in
t.
S
uc
h
c
a
s
e
s
c
l
a
s
s
if
ie
r
s
c
a
n
a
c
hi
e
ve
gr
e
a
t
a
c
c
ur
a
c
y,
b
ut
it
is
w
or
th
le
s
s
in
m
os
t
r
e
a
l
ti
m
e
a
ppl
ic
a
ti
ons
th
e
n
th
e
m
in
or
it
y
c
la
s
s
is
f
r
e
que
nt
ly
th
e
c
la
s
s
of
in
te
r
e
s
t
(
ot
he
r
w
is
e
,
f
or
th
is
a
c
la
s
s
if
ie
r
ne
e
d
not
be
r
e
qui
r
e
d,
a
s
th
e
c
l
a
s
s
of
in
te
r
e
s
t
oc
c
ur
s
pr
a
c
ti
c
a
ll
y
a
lwa
ys
)
[
31]
.
A
s
a
r
e
s
ul
t,
m
a
ny
s
tr
a
te
gi
e
s
ha
ve
r
e
m
a
in
e
d
c
r
e
a
t
e
d
to
a
ddr
e
s
s
th
e
is
s
ue
of
im
ba
la
nc
e
c
la
s
s
.
T
he
s
e
te
c
hni
que
s
c
a
n
be
di
vi
de
d
in
to
two
c
a
te
gor
ie
s
:
s
a
m
pl
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
1727
-
1742
1732
te
c
hni
que
s
a
nd
s
ke
w
-
in
s
e
ns
it
iv
e
c
l
a
s
s
if
ie
r
s
.
A
m
in
or
it
y
c
la
s
s
is
ty
pi
c
a
ll
y
ig
nor
e
d
or
m
is
s
e
d
by
a
s
ta
nda
r
d
c
la
s
s
if
ie
r
s
in
c
e
th
e
r
e
a
r
e
n’
t
m
a
ny
e
xa
m
pl
e
s
of
th
e
c
la
s
s
.
O
ne
o
ve
r
-
s
a
m
pl
in
g
m
e
th
od
th
a
t
a
ddr
e
s
s
e
s
th
is
is
s
ue
is
S
M
O
T
E
.
F
ig
ur
e
3. E
xa
m
pl
e
of
i
m
ba
la
nc
e
d da
ta
s
e
t
F
ig
ur
e
4. G
e
ne
r
a
ti
ng s
a
m
pl
e
ove
r
i
m
ba
la
nc
e
d da
ta
s
e
t
I
m
ba
la
nc
e
d
c
la
s
s
a
lg
or
it
hm
s
c
a
n
be
di
vi
de
d
in
to
th
r
e
e
ty
pe
s
s
uc
h
a
s
e
ns
e
m
bl
e
-
ba
s
e
d
m
ode
l,
da
ta
-
le
ve
l,
a
nd
a
lg
or
it
hm
le
ve
l
[
32]
.
A
m
ong
th
e
m
,
d
a
ta
-
le
ve
l
m
e
th
ods
w
hi
c
h
in
c
lu
de
th
e
th
r
e
e
s
ub
ty
pe
s
of
ove
r
s
a
m
pl
in
g,
unde
r
-
s
a
m
pl
in
g,
a
nd
hybr
id
s
c
he
m
e
s
a
r
e
w
id
e
l
y
us
e
d.
T
he
e
ns
e
m
bl
e
c
la
s
s
if
ie
r
s
c
a
n
im
pr
ove
th
e
a
c
c
ur
a
c
y
of
a
s
in
gl
e
c
la
s
s
if
ie
r
by
in
te
gr
a
ti
ng
m
ul
ti
pl
e
c
la
s
s
if
ie
r
s
to
ge
th
e
r
,
th
e
c
la
s
s
im
ba
l
a
nc
e
pr
obl
e
m
c
a
nnot
be
r
e
s
ol
ve
d
e
a
s
il
y
by
c
onve
nt
io
na
l
le
a
r
ni
ng
s
tr
a
te
gi
e
s
.
I
ns
te
a
d,
e
n
s
e
m
bl
e
l
e
a
r
ni
ng
a
lg
or
it
hm
s
m
us
t
be
s
pe
c
ia
ll
y
c
r
e
a
te
d
to
a
ddr
e
s
s
th
i
s
is
s
ue
.
B
y
a
tt
e
m
pt
in
g
to
r
e
s
to
r
e
th
e
e
qui
li
br
iu
m
be
twe
e
n
th
e
m
a
jo
r
it
y
a
nd
m
in
or
it
y
c
a
te
gor
ie
s
,
th
e
s
e
te
c
hni
que
s
a
lt
e
r
th
e
in
f
or
m
a
ti
on
it
s
e
lf
.
D
if
f
e
r
e
nt
ty
pe
s
of
c
la
s
s
im
ba
la
nc
e
d
a
lg
or
it
hm
s
ha
ve
de
ve
lo
pe
d
ba
s
e
d
on
th
e
da
ta
-
le
ve
l
a
ppr
oa
c
h
in
th
e
pa
s
t
de
c
a
de
s
a
s
s
how
n
in
F
ig
ur
e
5.
T
he
y
a
c
hi
e
ve
th
is
by
in
c
r
e
a
s
in
g
th
e
num
be
r
of
m
in
or
it
y
c
a
te
gor
y
s
a
m
pl
e
s
(
ove
r
s
a
m
pl
in
g)
or
by
de
c
r
e
a
s
in
g
th
e
num
be
r
of
m
a
jo
r
it
y
c
a
te
gor
y
s
a
m
pl
e
s
(
u
nde
r
-
s
a
m
pl
in
g)
[
33]
.
U
nde
r
-
s
a
m
pl
in
g
in
vol
ve
s
s
e
le
c
ti
ng
s
a
m
pl
e
s
a
t
r
a
ndom
or
e
m
pl
oyi
ng
he
ur
is
ti
c
m
e
th
ods
to
e
xc
lu
d
e
th
e
le
s
s
im
por
ta
nt
pa
tt
e
r
ns
.
H
ow
e
ve
r
,
unde
r
-
s
a
m
pl
in
g
is
s
om
e
ti
m
e
s
r
e
ga
r
de
d
a
s
da
nge
r
ous
due
to
th
e
in
c
r
e
a
s
e
d
li
ke
li
hood
of
lo
s
in
g
pot
e
nt
ia
ll
y
im
por
ta
nt
da
ta
.
O
ve
r
s
a
m
pl
in
g
is
th
e
ot
he
r
,
pos
s
ib
ly
m
or
e
e
f
f
e
c
ti
ve
c
hoi
c
e
.
C
r
e
a
ti
ng
ne
w
m
in
or
it
y
c
a
te
gor
y
pa
tt
e
r
ns
or
r
a
ndoml
y r
e
pr
oduc
in
g
m
in
or
it
y
c
a
te
gor
y pa
tt
e
r
ns
a
r
e
t
he
t
w
o
m
a
in
w
a
ys
t
ha
t
ove
r
s
a
m
pl
in
g i
s
pe
r
f
or
m
e
d.
T
he
c
r
uc
ia
l
be
ne
f
it
of
t
hi
s
da
ta
-
le
ve
l
a
ppr
oa
c
h i
s
t
ha
t
i
t
is
s
im
pl
e
t
o
ge
ne
r
a
te
ne
w
da
ta
t
o m
a
ke
up
f
or
t
he
m
in
or
it
y
c
a
te
gor
y’
s
da
ta
de
f
ic
ie
nc
y. A
s
a
r
e
s
ul
t,
ove
r
s
a
m
pl
in
g i
s
f
r
e
que
n
tl
y f
a
vor
e
d i
n da
ta
-
le
ve
l
a
ppr
oa
c
he
s
.
S
ynt
he
ti
c
s
a
m
pl
e
s
r
e
m
a
in
g
e
ne
r
a
te
d
f
or
th
e
m
in
or
it
y
c
la
s
s
in
ove
r
s
a
m
pl
in
g
te
c
hni
que
s
to
a
tt
a
in
th
e
ba
la
nc
e
d
da
ta
s
e
t.
A
w
e
ll
-
known
ove
r
s
a
m
pl
in
g
m
e
th
od
f
r
e
que
nt
ly
us
e
d
to
a
ddr
e
s
s
c
la
s
s
im
ba
la
nc
e
is
s
u
e
s
is
S
M
O
T
E
[
34]
.
I
t
is
qui
te
e
a
s
y
to
u
s
e
a
nd
ve
r
y
e
f
f
e
c
ti
ve
in
pr
a
c
ti
c
e
.
I
t
pr
oduc
e
s
s
ynt
h
e
ti
c
s
a
m
pl
e
s
pr
e
c
i
s
e
ly
w
he
r
e
th
e
m
in
or
it
y
c
la
s
s
jo
in
s
.
H
ow
e
ve
r
,
ove
r
s
a
m
pl
in
g
w
it
h
S
M
O
T
E
doe
s
not
r
e
a
ll
y
a
c
c
ount
f
or
th
e
di
s
tr
ib
ut
io
n
of
s
a
m
pl
e
s
f
r
om
th
e
m
a
jo
r
it
y
c
la
s
s
dur
in
g
th
e
d
e
ve
l
opm
e
nt
of
s
ynt
he
ti
c
s
a
m
pl
e
s
, w
hi
c
h
r
e
s
ul
ts
in
th
e
ge
ne
r
a
ti
on of
unne
c
e
s
s
a
r
y m
in
or
it
y s
a
m
pl
e
s
a
r
ound the
de
s
i
r
e
d
s
a
m
pl
e
s
[
35]
.
F
ig
ur
e
5. D
if
f
e
r
e
nt
t
ype
s
of
c
la
s
s
l
e
a
r
ne
r
c
ha
ll
e
nge
s
i
n
ML
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
v
ie
w
on c
la
s
s
i
m
bal
anc
e
t
e
c
hni
que
s
t
o s
tr
e
ngt
he
n m
ode
l
p
r
e
di
c
ti
on (
P
ut
ta
H
e
m
al
at
ha)
1733
3
.1
.
R
e
la
t
e
d
w
or
k
H
uge
a
m
ount
s
of
unpr
oc
e
s
s
e
d
d
a
ta
a
r
e
pr
oduc
e
d
in
th
e
ML
a
r
e
a
be
c
a
u
s
e
of
th
e
tr
e
m
e
ndou
s
pr
ogr
e
s
s
in
r
e
s
e
a
r
c
h
a
nd
te
c
hnol
ogy.
E
a
c
h
da
ta
s
e
t
ha
s
a
uni
que
di
s
tr
ib
ut
io
n
a
nd
in
f
or
m
a
ti
ona
l
s
tr
uc
tu
r
e
.
D
ue
to
th
e
s
ke
w
e
d
di
s
tr
ib
ut
io
n
of
c
la
s
s
e
s
in
s
om
e
da
ta
s
e
t
s
,
a
n
im
ba
la
nc
e
d
de
te
r
m
in
a
ti
on
is
pr
e
s
e
nt
e
d.
I
n
ot
he
r
w
or
ds
,
c
la
s
s
if
ie
r
s
s
houl
d
be
f
or
c
e
d
to
tr
e
a
t
m
in
or
c
a
s
e
s
a
s
bi
g
e
xa
m
pl
e
s
w
he
n
ha
ndl
in
g
im
ba
la
nc
e
d
in
f
or
m
a
ti
on
s
e
ts
.
D
M
is
th
e
pr
oc
e
s
s
of
di
s
c
ove
r
in
g
hi
dde
n
knowle
dge
f
r
om
th
e
da
ta
.
T
he
pr
im
a
r
y
m
e
th
ods
a
r
e
pa
tt
e
r
n
m
in
in
g,
a
s
s
oc
ia
ti
on
a
na
ly
s
is
,
gr
oupi
ng,
a
nd
c
la
s
s
if
ic
a
ti
on.
A
D
A
S
Y
N
is
a
ML
-
ba
s
e
d
ove
r
s
a
m
pl
in
g
te
c
hni
que
to
s
ol
ve
th
e
is
s
ue
of
im
ba
la
nc
e
d
da
ta
s
e
ts
[
36]
.
T
he
f
unda
m
e
nt
a
l
id
e
a
be
hi
nd
A
D
A
S
Y
N
is
to
e
m
pl
oy
a
w
e
ig
ht
e
d
di
s
tr
ib
ut
io
n
f
or
m
in
or
i
ty
c
la
s
s
in
s
ta
nc
e
s
de
p
e
ndi
ng
on
th
e
ir
le
a
r
ni
ng
e
xe
r
ti
on.
H
a
r
de
r
-
to
-
le
a
r
n
m
in
o
r
it
y
c
la
s
s
e
xa
m
pl
e
s
r
e
c
e
iv
e
m
or
e
s
ynt
he
ti
c
s
a
m
pl
e
s
a
s
s
oc
ia
te
d
to
th
o
s
e
th
a
t
a
r
e
s
im
pl
e
r
to
s
tu
dy.
A
ddi
ti
ona
ll
y,
A
D
A
S
Y
N
dyna
m
ic
a
ll
y
a
dj
us
ts
th
e
c
la
s
s
if
ie
r
'
s
de
c
is
io
n
bounda
r
y,
f
oc
us
in
g
m
or
e
on
c
ha
ll
e
ngi
ng
in
s
ta
nc
e
s
to
im
pr
ove
le
a
r
ni
ng
e
f
f
e
c
ti
ve
ne
s
s
.
T
hi
s
i
s
a
c
c
om
pl
is
he
d
th
r
ough
a
da
pt
iv
e
w
e
ig
ht
a
dj
u
s
tm
e
nt
s
a
nd
a
le
a
r
ni
ng
pr
oc
e
s
s
t
ha
t
c
on
s
id
e
r
s
t
he
da
ta
di
s
tr
ib
ut
io
n, e
ns
ur
in
g a
m
or
e
ba
la
nc
e
d a
nd e
f
f
e
c
ti
ve
c
la
s
s
if
ic
a
ti
on.
W
it
h
m
or
e
r
e
pr
e
s
e
nt
a
ti
ve
tr
a
in
in
g
da
ta
f
or
th
e
m
in
or
it
y
c
la
s
s
,
A
D
A
S
Y
N
e
nha
nc
e
s
th
e
qua
li
ty
of
ML
r
e
pr
e
s
e
nt
a
ti
ons
w
h
e
n
ha
ndl
in
g
w
it
h
im
b
a
la
nc
e
d
d
a
ta
s
e
ts
by
g
e
ne
r
a
ti
ng
f
a
br
ic
a
te
d
s
a
m
pl
e
s
f
or
th
e
m
in
or
it
y
c
la
s
s
,
e
ns
ur
in
g
a
m
or
e
ba
la
nc
e
d
r
e
pr
e
s
e
nt
a
ti
on
a
nd
im
pr
ovi
ng
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y.
I
t
s
houl
d
be
u
s
e
d
w
is
e
ly
,
th
ough,
pa
r
ti
c
ul
a
r
ly
in
c
onj
unc
ti
on
w
it
h
ot
h
e
r
m
e
th
o
ds
f
or
de
a
li
ng
w
it
h
c
la
s
s
im
ba
l
a
nc
e
,
s
uc
h
a
s
us
in
g
va
r
io
us
a
s
s
e
s
s
m
e
nt
m
e
tr
ic
s
or
ut
il
iz
in
g
ot
he
r
s
a
m
pl
in
g
te
c
hni
que
s
li
ke
u
nde
r
-
s
a
m
pl
in
g
or
c
om
bi
ni
ng
ove
r
s
a
m
pl
in
g
a
nd
und
e
r
s
a
m
pl
in
g.
I
n
th
e
c
a
s
e
of
m
ul
ti
-
c
la
s
s
d
a
ta
s
e
t
s
,
it
i
s
m
uc
h
m
or
e
di
f
f
ic
ul
t
to
de
f
in
e
th
e
m
a
jo
r
it
y
a
nd
m
in
or
it
y
c
la
s
s
e
s
pr
opos
e
d
a
nd
a
ddr
e
s
s
e
d
im
ba
la
nc
e
d
da
ta
c
om
pr
is
in
g
m
a
ny
c
l
a
s
s
e
s
b
a
s
e
d
on
c
lu
s
te
r
in
g
a
nd
S
M
O
T
E
.
T
he
s
tr
a
te
gy
put
in
to
pr
a
c
ti
c
e
,
a
nd
th
e
r
e
s
ul
ts
of
th
e
e
xpe
r
im
e
nt
s
de
m
on
s
tr
a
te
th
a
t
it
is
a
u
s
e
f
ul
to
ol
f
or
ha
ndl
in
g
m
ul
ti
c
la
s
s
im
ba
la
nc
e
d
da
ta
s
e
ts
.
I
t
c
a
n
a
ls
o
e
nha
nc
e
th
e
m
in
or
it
y
c
la
s
s
’
s
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
bot
h
w
it
hi
n
a
nd
th
r
oughout
th
e
e
nt
ir
e
da
ta
s
e
t.
U
s
in
g s
e
ve
r
a
l
ba
la
n
c
in
g
te
c
hni
que
s
c
r
e
a
te
d
a
S
M
O
T
E
f
r
a
m
e
w
or
k
ba
s
e
d
on
a
ge
ne
ti
c
a
lg
or
it
hm
f
or
le
a
r
ni
ng
he
a
lt
hc
a
r
e
da
ta
th
a
t
ha
s
a
n
im
ba
la
nc
e
d
di
s
tr
ib
ut
io
n.
I
n
he
a
lt
hc
a
r
e
da
t
a
a
na
ly
ti
c
s
,
f
in
di
ng
in
f
r
e
que
nt
but
im
por
ta
nt
he
a
lt
hc
a
r
e
e
ve
nt
s
in
e
nor
m
ous
uns
tr
uc
tu
r
e
d
da
ta
s
e
t
s
ha
s
be
c
om
e
a
ty
pi
c
a
l
pr
ob
le
m
.
H
ow
e
ve
r
,
be
c
a
us
e
m
os
t
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
s
im
pl
ic
it
ly
a
s
s
um
e
a
n
e
qua
l
oc
c
ur
r
e
nc
e
of
c
la
s
s
e
s
a
n
d
a
r
e
c
r
e
a
te
d
to
im
p
r
ove
ove
r
a
ll
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y,
une
v
e
n
c
la
s
s
di
s
tr
ib
ut
io
n
in
m
a
ny
pr
a
c
ti
c
a
l
d
a
ta
s
e
ts
s
ig
ni
f
ic
a
nt
ly
hi
nde
r
s
th
e
de
te
c
ti
on
of
unc
om
m
on oc
c
ur
r
e
nc
e
s
.
F
uz
z
y
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
F
S
V
M
)
is
w
id
e
ly
a
c
knowle
dge
d
a
s
a
not
e
w
or
th
y
e
nha
nc
e
m
e
nt
to
s
of
t
m
a
r
gi
n
S
V
M
s
,
s
u
c
h
a
s
C
-
S
V
M
,
a
s
th
e
la
tt
e
r
pr
oduc
e
s
s
u
bopt
im
a
l
out
c
om
e
s
w
he
n
out
li
e
r
s
a
r
e
pr
e
s
e
nt
.
F
uz
z
y
-
S
M
O
T
E
br
oa
de
ns
th
e
m
in
or
it
y
de
c
i
s
io
n
bounda
r
y
by
pr
oduc
in
g
m
or
e
m
in
or
it
y
c
la
s
s
tr
a
in
in
g
in
s
ta
nc
e
s
,
w
hi
c
h
e
nha
n
c
e
s
c
la
s
s
if
ie
r
pe
r
f
or
m
a
nc
e
s
on
im
ba
la
nc
e
d
da
ta
s
e
ts
.
T
he
f
uz
z
y
-
S
M
O
T
E
a
lg
or
it
hm
a
s
s
ig
ns
a
m
e
m
be
r
s
hi
p
de
gr
e
e
to
e
a
c
h
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
ba
s
e
d
on
th
e
or
ig
in
a
l
c
la
s
s
di
s
tr
ib
ut
io
n
in
th
e
tr
a
in
in
g
da
ta
s
e
t.
I
t
th
e
n
ge
ne
r
a
te
s
a
hi
ghe
r
num
be
r
of
s
ynt
he
ti
c
s
a
m
pl
e
s
f
or
in
s
ta
nc
e
s
w
it
h
lo
w
e
r
m
e
m
be
r
s
hi
p
de
gr
e
e
s
,
e
ns
ur
in
g
a
m
or
e
ba
la
nc
e
d
r
e
pr
e
s
e
nt
a
ti
on
of
th
e
m
in
or
it
y
c
la
s
s
.
I
n
or
de
r
to
pr
oduc
e
ne
w
tr
a
in
in
g
da
ta
s
e
t,
t
he
ol
d t
r
a
in
in
g da
ta
s
e
t
is
c
om
bi
ne
d w
it
h f
r
e
s
h s
ynt
h
e
ti
c
i
ns
ta
nc
e
s
f
r
om
t
he
m
in
or
it
y c
la
s
s
[
37]
.
A
M
L
m
e
th
od
-
ba
s
e
d c
la
s
s
i
m
ba
la
nc
e
t
e
c
hni
que
i
s
pr
opos
e
d w
h
ic
h us
e
s
a
c
os
t
-
s
e
ns
it
iv
e
m
odi
f
ic
a
ti
on
of
th
e
le
a
s
t
m
e
a
n
s
qua
r
e
(
L
M
S
)
a
lg
or
it
hm
to
pe
na
li
z
e
e
r
r
or
s
of
va
r
io
us
s
a
m
pl
e
s
w
it
h
va
r
io
us
w
e
ig
ht
s
a
nd
s
om
e
ge
ne
r
a
l
gui
de
li
ne
s
to
c
hoos
e
th
os
e
w
e
ig
ht
s
.
F
ol
lo
w
in
g
th
e
ba
la
nc
in
g
s
ta
ge
,
e
m
pl
oys
va
r
io
us
c
la
s
s
if
ic
a
ti
on a
lg
or
it
hm
s
. W
he
n
c
om
pa
r
e
d t
o t
he
ot
he
r
c
la
s
s
e
s
, t
he
s
e
a
ppl
ic
a
ti
on
s
f
r
e
que
nt
ly
ha
ve
one
or
m
or
e
m
in
or
it
y
c
la
s
s
e
s
w
it
h
ve
r
y
f
e
w
s
a
m
pl
e
s
.
N
um
e
r
ou
s
s
tu
di
e
s
ha
v
e
de
m
ons
tr
a
te
d
th
a
t
DM
te
c
hni
que
s
f
a
ll
s
hor
t
of
de
s
ir
e
d
pe
r
f
or
m
a
nc
e
w
he
n
th
e
tr
a
in
in
g
da
ta
is
im
ba
la
nc
e
d.
T
r
a
di
ti
ona
l
M
L
a
ppr
oa
c
he
s
r
e
ly
on
ba
la
nc
e
d
c
a
te
gor
y
di
s
tr
ib
ut
io
ns
f
or
tr
a
in
in
g
da
ta
a
nd,
in
a
ddi
ti
on,
a
r
e
li
ke
ly
to
a
c
c
ur
a
te
ly
di
s
ti
ngui
s
h
th
e
m
a
jo
r
it
y
c
a
te
gor
y
w
hi
le
m
is
c
la
s
s
if
yi
ng
th
e
m
in
or
it
y
c
a
te
gor
y.
T
hi
s
is
pr
im
a
r
il
y
due
to
th
e
di
s
pa
r
it
y
in
c
la
s
s
.
T
he
da
ta
-
le
ve
l
a
ppr
oa
c
h
is
of
te
n
us
e
d
in
ha
ndl
in
g
c
la
s
s
im
ba
la
n
c
e
du
e
to
it
s
e
f
f
e
c
ti
ve
pe
r
f
or
m
a
nc
e
a
nd
c
la
s
s
if
ie
r
in
de
pe
nde
nc
e
.
B
y
a
tt
e
m
pt
in
g
to
r
e
s
to
r
e
th
e
e
qui
li
br
iu
m
be
twe
e
n
th
e
m
a
jo
r
it
y
a
nd
m
in
or
it
y
c
a
te
gor
ie
s
,
th
e
s
e
te
c
hni
que
s
a
lt
e
r
th
e
in
f
or
m
a
ti
on
it
s
e
lf
.
A
s
a
r
e
s
ul
t,
ove
r
s
a
m
pl
in
g
is
f
r
e
que
nt
ly
f
a
vor
e
d
in
da
ta
-
le
ve
l
a
ppr
oa
c
he
s
.
S
ynt
he
ti
c
s
a
m
pl
e
s
a
r
e
c
r
e
a
te
d
f
or
th
e
m
in
or
it
y
gr
oup
in
ove
r
s
a
m
pl
in
g
pr
oc
e
dur
e
s
to
ba
la
nc
e
a
n
im
ba
la
nc
e
d da
ta
ba
s
e
.
C
la
s
s
im
ba
la
n
c
e
le
a
r
ni
ng
us
in
g
in
tu
it
io
ni
s
ti
c
f
uz
z
y
twi
n
s
u
ppor
t
ve
c
to
r
m
a
c
hi
ne
(
C
I
L
-
F
A
R
T
-
I
F
T
S
V
M
)
a
nd
f
uz
z
y
a
da
pt
iv
e
r
e
s
ona
nc
e
th
e
or
y
(
A
R
T
)
i
s
pr
opos
e
d
to
ove
r
c
om
e
th
e
c
h
a
ll
e
nge
s
pos
e
d
by
noi
s
e
im
ba
la
nc
e
.
T
hi
s
a
ppr
oa
c
h
e
f
f
e
c
ti
ve
ly
a
ddr
e
s
s
e
s
c
la
s
s
im
ba
la
nc
e
in
da
ta
s
e
ts
c
ont
a
in
in
g
noi
s
e
out
li
e
r
s
a
nd
la
r
ge
-
s
c
a
le
da
ta
,
e
nha
nc
in
g
m
ode
l
r
obus
tn
e
s
s
a
nd
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
.
R
e
s
a
m
pl
in
g
te
c
hni
que
s
a
r
e
f
r
e
que
nt
ly
us
e
d
to
c
r
e
a
te
s
ynt
he
ti
c
da
ta
.
G
e
ne
r
a
ti
ve
a
dve
r
s
a
r
ia
l
ne
twor
ks
(
G
A
N
s
)
c
a
n
a
ls
o
be
us
e
d
a
s
a
n
a
lt
e
r
na
ti
ve
to
in
f
or
m
a
ti
on
ga
in
a
ddr
e
s
s
th
e
is
s
ue
of
c
la
s
s
im
ba
la
nc
e
by
pr
oduc
in
g
ta
bul
a
r
da
ta
,
de
s
pi
te
th
e
ir
pr
im
a
r
y
pur
pos
e
be
in
g
th
e
ge
ne
r
a
ti
on
of
im
a
ge
da
ta
.
I
n
th
is
w
or
k,
r
e
s
a
m
pl
in
g
te
c
hni
que
s
a
nd
G
A
N
-
ba
s
e
d
te
c
hni
que
s
a
r
e
c
om
pa
r
e
d.
W
he
n
r
e
s
a
m
pl
in
g
m
e
th
ods
a
r
e
us
e
d
t
o
ba
la
nc
e
th
e
d
a
ta
,
M
L
m
e
th
od
s
pe
r
f
or
m
27%
be
tt
e
r
.
O
ne
popula
r
a
lg
or
it
hm
f
o
r
ba
la
nc
in
g
tr
a
in
da
ta
th
a
t
a
d
ds
s
ynt
he
ti
c
da
ta
on
m
in
or
c
la
s
s
da
t
a
is
c
a
ll
e
d
S
M
O
T
E
.
O
ne
of
th
e
ke
y
s
te
ps
in
th
e
S
M
O
T
E
pr
oc
e
s
s
is
id
e
nt
if
yi
ng
th
e
K
N
N
us
in
g
E
uc
li
de
a
n
di
s
ta
nc
e
to
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
1727
-
1742
1734
ge
ne
r
a
te
s
ynt
he
ti
c
d
a
ta
.
H
ow
e
v
e
r
,
r
e
ly
in
g
s
ol
e
ly
on
E
uc
li
de
a
n
di
s
ta
nc
e
w
it
hout
c
ons
id
e
r
in
g
a
tt
r
ib
ut
e
c
or
r
e
la
ti
ons
m
a
y l
e
a
d t
o t
he
s
e
le
c
ti
on of
unr
e
pr
e
s
e
nt
a
ti
ve
ne
ig
h
bor
s
, e
s
pe
c
ia
ll
y w
he
n
c
e
r
ta
in
a
tt
r
ib
ut
e
s
e
xhi
bi
t
hi
ghe
r
c
or
r
e
la
ti
on
va
lu
e
s
th
a
n
ot
he
r
s
.
T
o
a
ddr
e
s
s
th
is
li
m
it
a
ti
on,
th
is
s
tu
dy
de
ve
lo
pe
d
a
tt
r
ib
ut
e
w
e
ig
ht
e
d
a
nd
K
N
N
hub
on
S
M
O
T
E
(
A
W
H
-
S
M
O
T
E
)
,
it
in
c
or
po
r
a
te
s
a
tt
r
ib
ut
e
w
e
ig
ht
in
g
a
nd
KNN
hub
s
e
le
c
ti
on
to
e
nha
nc
e
th
e
qua
li
ty
of
s
ynt
he
ti
c
da
ta
ge
ne
r
a
ti
on.,
w
hi
c
h
im
pr
ove
s
S
M
O
T
E
by
us
in
g
o
c
c
ur
r
e
nc
e
d
a
ta
in
th
e
KNN
hub
to
im
pr
ove
th
e
s
e
l
e
c
ti
ve
s
a
m
pl
in
g
m
e
th
od
a
nd
by
u
s
in
g
a
tt
r
ib
ut
e
w
e
ig
ht
in
g
to
im
pr
ove
ne
ig
hbor
’
s
a
nd
noi
s
e
id
e
nt
if
ic
a
ti
on.
T
he
m
e
th
ods
of
in
f
or
m
a
ti
on
g
a
in
a
r
e
a
ppl
ie
d
to
a
tt
r
ib
ut
e
w
e
ig
ht
in
g,
bor
de
r
li
ne
-
S
M
O
T
E
1
a
nd
bor
de
r
li
ne
-
S
M
O
T
E
2
a
r
e
va
r
ia
nt
s
of
th
e
S
M
O
T
E
m
e
th
od
th
a
t
s
pe
c
if
ic
a
ll
y
ove
r
s
a
m
pl
e
m
in
or
it
y
c
la
s
s
s
a
m
pl
e
s
lo
c
a
te
d
ne
a
r
th
e
de
c
is
io
n
bounda
r
y.
E
xpe
r
im
e
nt
a
l
r
e
s
ul
ts
in
di
c
a
te
th
a
t
th
e
s
e
a
ppr
oa
c
h
e
s
out
pe
r
f
or
m
bot
h
S
M
O
T
E
a
nd
r
a
ndom
ove
r
s
a
m
pl
in
g
in
te
r
m
s
of
tr
ue
pos
it
iv
e
(
T
P
)
r
a
te
a
nd
F
-
va
lu
e
f
or
th
e
m
in
or
it
y
c
la
s
s
.
S
in
c
e
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
s
ne
a
r
th
e
bound
a
r
y
a
r
e
m
or
e
s
us
c
e
pt
ib
le
to
m
is
c
la
s
s
if
ic
a
ti
on,
th
e
s
e
te
c
hni
que
s
e
nha
nc
e
th
e
ir
r
e
pr
e
s
e
nt
a
ti
on
a
nd
im
pr
ove
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
.
T
hus
,
w
hi
le
S
M
O
T
E
a
nd
r
a
ndom
ove
r
-
s
a
m
pl
in
g
b
r
in
g
a
ll
e
xa
m
pl
e
s
f
r
om
th
e
m
in
or
i
ty
c
la
s
s
or
a
r
a
ndom
s
ubs
e
t
of
th
e
s
a
m
pl
e
s
of
th
e
m
in
or
it
y c
la
s
s
, t
he
pr
opos
e
d a
ppr
oa
c
he
s
onl
y ove
r
s
a
m
pl
e
t
h
e
s
a
m
pl
e
s
of
t
he
m
in
or
it
y
c
la
s
s
t
ha
t
is
c
lo
s
e
t
o
th
e
de
c
is
io
n
s
ur
f
a
c
e
.
T
he
r
e
f
or
e
,
th
is
pa
pe
r
de
ve
lo
ps
a
f
uz
z
y
-
ba
s
e
d
S
M
O
T
E
a
lg
or
it
hm
w
it
h
th
e
pur
pos
e
of
a
ddi
ng
th
e
f
uz
z
y
s
e
t
th
e
or
y
to
th
e
S
M
O
T
E
.
T
he
f
uz
z
y
-
S
M
O
T
E
pr
oduc
e
s
a
ddi
ti
ona
l
f
a
br
ic
a
te
s
a
m
pl
e
s
f
or
m
in
or
it
y c
la
s
s
i
ns
ta
nc
e
s
, pa
r
ti
c
ul
a
r
ly
i
n t
he
f
uz
z
y r
e
gi
on, whe
r
e
m
in
or
it
y e
xa
m
pl
e
s
e
xhi
bi
t
lo
w
e
r
m
e
m
be
r
s
hi
p
de
gr
e
e
s
t
o t
he
m
in
or
it
y c
la
s
s
a
nd a
r
e
m
or
e
pr
one
to
m
is
c
la
s
s
if
ic
a
ti
on. As
t
he
s
a
m
pl
e
s
iz
e
of
t
he
m
a
jo
r
it
y c
la
s
s
in
c
r
e
a
s
e
s
r
e
la
ti
ve
to
th
e
m
in
or
it
y
c
la
s
s
,
f
uz
z
y
-
S
M
O
T
E
e
n
s
ur
e
s
a
m
or
e
ba
la
n
c
e
d
di
s
tr
ib
ut
io
n
by
c
r
e
a
ti
ng
s
ynt
he
ti
c
in
s
ta
nc
e
s
th
a
t
e
xpa
nd
th
e
de
c
is
io
n
bounda
r
y
of
c
la
s
s
i
f
ie
r
s
.
B
y
in
c
or
por
a
ti
ng
f
uz
z
y
-
ba
s
e
d
s
a
m
pl
in
g,
f
uz
z
y
-
S
M
O
T
E
r
e
duc
e
s
c
la
s
s
if
ie
r
bi
a
s
to
w
a
r
d
th
e
m
a
jo
r
it
y
c
la
s
s
,
a
ll
ow
in
g
m
ode
ls
to
c
on
s
tr
uc
t
br
oa
de
r
de
c
is
io
n
r
e
gi
ons
w
it
h
e
nh
a
nc
e
d
r
e
pr
e
s
e
nt
a
ti
on
of
m
in
or
it
y
c
a
s
e
s
.
C
om
pa
r
a
ti
ve
a
na
ly
s
is
de
m
on
s
tr
a
te
s
th
a
t
f
uz
z
y
-
S
M
O
T
E
out
pe
r
f
or
m
s
tr
a
di
ti
ona
l
S
M
O
T
E
a
nd
bor
de
r
li
ne
-
S
M
O
T
E
in
te
r
m
s
of
a
c
c
ur
a
c
y,
e
f
f
e
c
ti
ve
ly
im
pr
ovi
ng
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
f
or
bot
h
th
e
m
in
or
it
y
c
la
s
s
a
nd
th
e
ove
r
a
ll
da
ta
s
e
t.
C
I
L
-
F
A
R
T
-
I
F
T
S
V
M
a
nd
f
uz
z
y
A
R
T
is
pr
opos
e
d
a
s
a
n
e
f
f
e
c
ti
ve
a
ppr
oa
c
h f
or
a
ddr
e
s
s
in
g
c
la
s
s
im
ba
la
nc
e
,
pa
r
ti
c
ul
a
r
ly
in
th
e
pr
e
s
e
nc
e
of
noi
s
e
,
out
li
e
r
s
,
a
nd
l
a
r
ge
-
s
c
a
le
da
ta
s
e
ts
.
T
o
t
a
c
kl
e
th
e
im
ba
la
n
c
e
i
s
s
ue
,
f
uz
z
y
A
R
T
is
ut
il
iz
e
d
a
s
a
c
lu
s
t
e
r
in
g
te
c
hni
que
to
m
odi
f
y
th
e
da
ta
s
e
t
di
s
tr
ib
ut
io
n,
e
ns
ur
in
g
a
m
or
e
ba
la
nc
e
d
r
e
pr
e
s
e
nt
a
ti
on
of
m
in
or
it
y
a
nd
m
a
jo
r
it
y
c
la
s
s
e
s
.
F
ol
lo
w
in
g
th
is
,
I
F
T
S
V
M
is
a
ppl
ie
d
to
c
ons
tr
uc
t
opt
im
a
l
non
-
pa
r
a
ll
e
l
hype
r
pl
a
ne
s
on
th
e
n
e
w
ly
tr
a
ns
f
or
m
e
d
d
a
ta
s
e
t
a
f
te
r
da
ta
r
e
duc
ti
on.
F
in
a
ll
y,
to
e
nha
nc
e
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y,
a
c
oor
di
na
te
de
s
c
e
nt
s
ys
te
m
w
it
h
s
hr
in
ki
ng
vi
a
a
n
a
c
ti
ve
s
e
t
is
e
m
pl
oye
d,
s
ig
ni
f
ic
a
nt
ly
r
e
duc
in
g
th
e
c
om
put
a
ti
ona
l
c
om
pl
e
xi
ty
of
th
e
le
a
r
ni
ng
pr
oc
e
s
s
.
T
hi
s
a
ppr
oa
c
h
e
ns
ur
e
s
im
pr
ove
d
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
i
n c
ha
ll
e
ngi
ng i
m
ba
la
nc
e
d da
ta
s
c
e
na
r
io
s
.
3
.2
.
S
M
O
T
E
S
M
O
T
E
is
a
popula
r
pr
e
-
pr
oc
e
s
s
in
g
m
e
th
od
f
or
c
or
r
e
c
ti
ng
c
la
s
s
im
ba
la
nc
e
in
M
L
da
ta
s
e
t
s
,
e
s
pe
c
ia
ll
y
in
bi
na
r
y
c
l
a
s
s
if
ic
a
ti
on
pr
obl
e
m
s
.
S
M
O
T
E
c
r
e
a
t
e
s
s
ynt
he
ti
c
s
a
m
pl
e
s
f
or
th
e
m
in
or
it
y
c
la
s
s
to
ba
la
nc
e
th
e
di
s
tr
ib
ut
io
n
of
th
e
c
la
s
s
e
s
.
T
hi
s
he
lp
s
a
voi
d
m
ode
l
bi
a
s
in
f
a
vor
of
th
e
dom
in
a
nt
c
la
s
s
a
nd
c
a
n
e
nha
nc
e
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
.
T
h
e
S
M
O
T
E
pr
oc
e
s
s
’
s
goa
l
w
a
s
to
in
te
r
pol
a
te
da
ta
f
r
om
di
ve
r
s
e
m
in
or
it
y
c
la
s
s
s
a
m
pl
e
s
.
G
e
ne
r
a
ti
ng
nove
l
m
in
or
it
y
c
la
s
s
s
a
m
pl
e
s
in
th
e
ne
a
r
by
lo
c
a
ti
on,
w
il
l
in
c
r
e
a
s
e
th
e
num
be
r
of
m
in
or
it
y
c
la
s
s
oc
c
ur
r
e
nc
e
s
a
nd
he
lp
c
la
s
s
if
ie
r
s
b
e
c
om
e
m
or
e
ge
n
e
r
ic
.
T
he
S
M
O
T
E
m
ode
l
is
a
not
a
bl
e
pi
one
e
r
in
im
ba
la
nc
e
d
c
la
s
s
if
ic
a
ti
on. T
he
S
M
O
T
E
m
e
t
hodology
in
th
e
d
a
ta
pr
e
-
pr
oc
e
s
s
in
g
ha
s
gr
ow
n
in
pr
om
in
e
nc
e
a
m
ong
a
c
a
de
m
ic
s
.
T
o
im
pr
ove
th
e
pe
r
f
or
m
a
nc
e
of
S
M
O
T
E
a
lg
or
it
hm
s
in
va
r
io
us
s
it
ua
ti
ons
,
num
e
r
ous
a
lt
e
r
na
ti
ve
s
a
nd
t
e
c
hnol
ogi
c
a
l
a
dva
nc
e
m
e
nt
s
w
e
r
e
pr
ovi
de
d.
S
M
O
T
E
w
a
s
w
id
e
ly
us
e
d
a
nd
a
c
c
e
pt
e
d,
m
a
ki
ng
it
on
e
of
th
e
pot
e
nt
ia
l
m
e
c
ha
ni
s
m
s
f
or
s
a
m
pl
in
g
a
nd
da
ta
pr
e
-
pr
oc
e
s
s
in
g
in
DM
a
nd
M
L
a
ppl
ic
a
ti
ons
.
T
o
ba
la
nc
e
th
e
pr
e
li
m
in
a
r
y
tr
a
in
in
g
s
e
t,
th
e
S
M
O
T
E
m
e
th
od
a
ppl
yi
ng
a
n
ove
r
s
a
m
pl
in
g
s
tr
a
te
gy
by
ge
ne
r
a
ti
ng
s
ynt
he
ti
c
s
a
m
pl
e
s
in
s
te
a
d
of
m
e
r
e
ly
dupl
ic
a
ti
ng
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
s
.
T
he
c
or
e
id
e
a
be
hi
nd
S
M
O
T
E
is
to
c
r
e
a
t
e
ne
w
-
f
a
ngl
e
d
da
ta
poi
nt
s
th
r
ough
e
xc
la
m
a
ti
on
be
twe
e
n
e
xi
s
ti
ng
m
in
or
it
y
c
la
s
s
s
a
m
pl
e
s
w
it
hi
n
a
de
f
in
e
d
ne
ig
hbor
hood.
U
nl
ik
e
tr
a
di
ti
ona
l
m
e
th
od
s
t
ha
t
f
oc
us
on
in
di
vi
dua
l
da
t
a
poi
nt
s
,
S
M
O
T
E
ope
r
a
te
s
in
th
e
f
e
a
tu
r
e
s
pa
c
e
,
le
ve
r
a
gi
ng
th
e
va
lu
e
s
a
nd
r
e
la
ti
o
ns
hi
ps
of
f
e
a
tu
r
e
s
to
s
ynt
he
s
iz
e
ne
w
s
a
m
pl
e
s
.
T
hi
s
di
s
ti
nc
ti
on
hi
ghl
ig
ht
s
th
e
ne
e
d
f
or
a
c
om
pl
e
te
a
s
s
e
s
s
m
e
nt
of
th
e
th
e
or
e
ti
c
a
l
a
s
s
oc
i
a
ti
on
be
twe
e
n
r
e
a
l
a
nd
s
ynt
he
ti
c
i
ns
ta
nc
e
s
, c
on
s
id
e
r
in
g t
he
i
m
pa
c
t
of
da
ta
di
m
e
ns
io
na
li
ty
on t
he
e
f
f
ic
ie
nc
y of
t
he
ge
ne
r
a
te
d s
a
m
pl
e
s
.
T
he
li
nk
be
twe
e
n
th
e
di
s
tr
ib
ut
io
ns
of
tr
a
in
in
g
a
nd
te
s
t
in
s
ta
nc
e
s
,
a
s
w
e
ll
a
s
ot
he
r
f
e
a
tu
r
e
s
li
ke
va
r
ia
nc
e
a
nd
c
or
r
e
la
ti
on
in
th
e
da
ta
a
nd
f
e
a
tu
r
e
s
pa
c
e
,
m
us
t
be
c
ons
id
e
r
e
d
a
s
il
lu
s
tr
a
te
d
in
F
ig
ur
e
6.
T
o
pr
oduc
e
ne
w
s
ynt
he
ti
c
da
ta
poi
nt
s
,
m
in
or
it
y
c
la
s
s
in
s
ta
nc
e
xi
is
c
hos
e
n
a
s
th
e
f
ounda
ti
on.
S
e
ve
r
a
l
ne
a
r
e
s
t
ne
ig
hbor
s
of
th
e
s
a
m
e
c
la
s
s
xi
1
-
xi
4
a
r
e
de
s
ig
na
te
d
f
r
om
th
e
tr
a
in
in
g
s
e
t
ba
s
e
d
on
a
di
s
ta
nc
e
m
e
a
s
ur
e
.
T
o
obt
a
in
f
r
e
s
h i
ns
ta
nc
e
s
r
1
–
r
4, a
r
a
ndomi
z
e
d i
nt
e
r
pol
a
ti
on i
s
t
he
n
pe
r
f
or
m
e
d.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
nt
J
A
r
ti
f
I
nt
e
ll
I
S
S
N
:
2252
-
8938
R
e
v
ie
w
on c
la
s
s
i
m
bal
anc
e
t
e
c
hni
que
s
t
o s
tr
e
ngt
he
n m
ode
l
p
r
e
di
c
ti
on (
P
ut
ta
H
e
m
al
at
ha)
1735
F
ig
ur
e
6. A
s
im
pl
e
i
ll
us
tr
a
ti
on of
S
M
O
T
E
a
lg
or
it
hm
3
.3. Ge
n
e
r
al
w
or
k
f
lo
w
o
f
S
M
O
T
E
S
M
O
T
E
is
a
w
id
e
ly
us
e
d
ove
r
s
a
m
pl
in
g
te
c
hni
que
th
a
t
ge
ne
r
a
te
s
s
ynt
he
ti
c
s
a
m
pl
e
s
r
a
th
e
r
th
a
n
s
im
pl
y
dupl
ic
a
ti
ng
m
in
or
it
y
c
la
s
s
in
s
ta
n
c
e
s
,
e
f
f
e
c
ti
ve
ly
m
it
ig
a
ti
ng
c
la
s
s
im
ba
la
n
c
e
.
A
de
t
a
il
e
d
w
or
kf
lo
w
in
s
how
n
in
F
ig
ur
e
7.
I
t
is
im
por
ta
nt
to
not
e
th
a
t
w
hi
le
S
M
O
T
E
is
a
us
e
f
ul
te
c
hni
que
f
or
a
ddr
e
s
s
in
g
c
l
a
s
s
im
ba
la
nc
e
,
it
m
a
y
not
be
s
ui
ta
bl
e
f
or
a
ll
ty
p
e
s
of
d
a
ta
s
e
t
s
or
im
ba
la
nc
e
s
c
e
n
a
r
io
s
.
A
ddi
ti
ona
ll
y,
th
e
r
e
a
r
e
va
r
ia
ti
ons
a
nd
e
xt
e
ns
io
n
s
of
S
M
O
T
E
,
s
u
c
h
a
s
bor
de
r
li
ne
-
S
M
O
T
E
,
A
D
A
S
Y
N
,
f
uz
z
y
-
ba
s
e
d,
a
nd
L
or
e
nt
z
ia
n
ba
s
e
d S
M
O
T
E
w
hi
c
h a
ddr
e
s
s
s
pe
c
if
ic
c
ha
ll
e
nge
s
i
n h
a
ndl
in
g i
m
ba
la
nc
e
d da
ta
i
s
di
s
c
us
s
e
d i
n
T
a
bl
e
1.
F
ig
ur
e
7. G
e
ne
r
a
l
W
or
kf
lo
w
of
S
M
O
T
E
M
ode
l
Evaluation Warning : The document was created with Spire.PDF for Python.
I
S
S
N
:
2252
-
8938
I
nt
J
A
r
ti
f
I
nt
e
ll
, V
ol
.
14
, N
o.
3
,
J
une
20
25
:
1727
-
1742
1736
T
a
bl
e
1. S
um
m
a
r
y of
S
M
O
T
E
ba
s
e
d t
e
c
hni
que
s
R
e
l
a
t
e
d w
or
k
A
ppr
oa
c
he
s
M
e
t
r
i
c
s
us
e
d
R
e
m
a
r
ks
D
a
t
a
s
e
t
s
u
s
e
d
[
3
8
]
A
-
S
M
O
T
E
F
-
m
e
a
s
ur
e
,
a
c
c
u
r
a
c
y
A
c
hi
e
ve
d
gr
e
a
t
e
r
r
e
s
ul
t
i
n
bi
na
r
y
c
l
a
s
s
i
f
i
c
a
t
i
on pr
obl
e
m
s
.
44 hi
ghl
y i
m
ba
l
a
nc
e
d
[
39]
A
M
D
O
AUC
O
ut
pe
r
f
or
m
e
d
i
n
ha
ndl
i
ng
m
ul
t
i
c
l
a
s
s
i
m
ba
l
a
nc
e
d da
t
a
10 s
e
t
s
[
14]
S
M
O
T
E
G
e
ne
r
a
l
m
e
t
r
i
c
s
E
xhi
bi
t
e
d
be
t
t
e
r
pe
r
f
or
m
a
nc
e
i
n
ha
ndl
i
ng
m
ul
t
i
c
l
a
s
s
pr
obl
e
m
s
t
ha
n
c
onve
nt
i
ona
l
m
e
t
hod.
15
U
C
I
r
e
pos
i
t
or
y
da
t
a
s
e
t
s
[
18]
S
M
O
T
E
a
nd
k
-
m
e
a
ns
A
c
c
ur
a
c
y
m
e
t
r
i
c
I
t
pr
e
ve
nt
s
t
he
i
nt
r
oduc
t
i
on
of
noi
s
e
a
nd
e
f
f
e
c
t
i
ve
l
y
a
ddr
e
s
s
e
s
bot
h
i
nt
e
r
-
c
l
a
s
s
a
nd i
nt
r
a
-
c
l
a
s
s
i
m
ba
l
a
nc
e
s
.
L
U
C
A
S
hi
ghl
y
i
m
ba
l
a
nc
e
d da
t
a
s
e
t
[
8]
D
-
S
M
O
T
E
R
e
c
a
l
l
,
f
-
va
l
ue
,
pr
e
c
i
s
i
o
n
E
vol
ve
d
ba
s
e
d
on
D
B
-
S
C
A
N
a
nd
S
M
O
T
E
f
or
m
ul
t
i
-
c
l
a
s
s
i
m
ba
l
a
nc
e
pr
obl
e
m
.
1
6
H
i
g
h
d
i
m
e
n
s
i
o
n
a
l
d
a
t
a
s
e
t
s
[
40]
D
yna
m
i
c
S
M
O
T
E
M
i
ni
m
um
r
e
c
a
l
l
,
M
c
N
e
m
a
r
’
s
T
e
s
t
,
G
-
M
e
a
n
O
ut
pe
r
f
or
m
e
d
t
he
bor
de
r
l
i
ne
S
M
O
T
E
a
nd
t
r
a
di
t
i
ona
l
S
M
O
T
E
be
s
t
a
l
gor
i
t
hm
f
or
m
ul
t
i
-
c
l
a
s
s
i
m
ba
l
a
nc
e
pr
obl
e
m
29
da
t
a
s
e
t
s
f
r
om
di
f
f
e
r
e
nt
s
of
t
w
a
r
e
r
e
pos
i
t
or
i
e
s
[
41]
C
l
us
t
e
r
o
r
i
e
n
t
e
d
S
M
O
T
E
G
-
m
e
a
n
,
f
-
me
a
s
ur
e
,
a
c
c
u
r
a
c
y
C
l
us
t
e
r
B
a
s
e
d
S
ynt
he
t
i
c
O
ve
r
s
a
m
pl
i
ng
obt
a
i
ne
d gr
e
a
t
e
r
r
e
s
ul
t
on ot
he
r
m
e
t
hods
20 r
e
a
l
w
or
l
d da
t
a
s
e
t
s
[
12]
O
ve
r
s
a
m
pl
i
n
g
s
c
he
m
e
ba
s
e
d
on
w
e
i
g
ht
s
e
t
t
i
ng
a
n
d
c
l
a
s
s
i
f
i
c
a
t
i
o
n
r
a
nki
ng
A
c
c
u
r
a
c
y
A
c
hi
e
ve
d
90%
a
c
c
ur
a
c
y
f
or
ha
ndl
i
ng
m
ul
t
i
c
l
a
s
s
s
a
m
pl
e
s
.
12
da
t
a
s
e
t
s
f
r
om
U
C
I
r
e
pos
i
t
or
y a
r
e
us
e
d
[
31]
S
yn
t
h
e
t
i
c
o
v
e
r
-
s
a
m
pl
i
ng s
c
he
m
e
G
e
ne
r
a
l
m
ode
l
R
e
l
i
a
bl
e
s
pe
e
d
f
or
bi
na
r
y c
l
a
s
s
da
t
a
26
be
nc
h
m
a
r
k
da
t
a
s
e
t
s
a
r
e
u
s
e
d
[
4]
G
-
S
M
O
T
E
p
-
va
l
ue
,
G
-
m
e
t
r
i
c
,
A
c
c
u
r
a
c
y
N
ove
l
da
t
a
ge
ne
r
a
t
i
on
pr
oc
e
dur
e
out
pe
r
f
o
r
m
e
d
in
c
l
a
s
s
i
f
i
c
a
t
i
on
13 U
C
I
da
t
a
s
e
t
s
[
28]
S
a
f
e
-
l
e
ve
l
-
S
M
O
T
E
G
-
m
e
a
n,
A
U
C
,
f
-
m
e
a
s
u
r
e
A
c
hi
e
ve
d
gr
e
a
t
e
r
r
e
s
ul
t
i
n
s
oc
i
a
l
m
e
di
a
da
t
a
a
na
l
ys
i
s
R
a
w
da
t
a
s
e
t
s
a
r
e
us
e
d
[
34]
R
e
ve
r
s
e
-
S
M
O
T
E
G
-
m
e
a
n,
A
U
C
,
f
-
m
e
a
s
u
r
e
E
f
f
i
c
i
e
nt
f
o
r
c
l
a
s
s
i
f
i
c
a
t
i
on pr
obl
e
m
3
i
m
ba
l
a
nc
e
d
da
t
a
s
e
t
s
f
r
om
K
E
E
L
r
e
pos
i
t
or
y
a
r
e
us
e
d.
4.
R
O
L
E
O
F
C
L
A
S
S
I
F
I
E
R
O
N
I
M
B
A
L
A
N
C
E
D
A
T
A
T
hi
s
s
e
c
ti
on
e
xpl
or
e
s
th
e
v
a
r
io
us
c
la
s
s
if
ie
r
s
us
e
d
in
th
is
s
tu
dy
to
a
ddr
e
s
s
th
e
c
ha
ll
e
nge
of
c
la
s
s
im
ba
la
nc
e
.
T
h
e
s
e
c
la
s
s
if
ie
r
s
a
r
e
tr
a
in
e
d
on
da
ta
s
e
ts
s
pe
c
if
ic
a
ll
y
de
s
ig
ne
d
to
h
a
ndl
e
s
k
e
w
e
d
da
ta
di
s
tr
ib
ut
io
ns
.
R
e
s
e
a
r
c
h
ha
s
c
ons
is
te
nt
ly
hi
ghl
ig
ht
e
d
th
a
t
im
ba
l
a
nc
e
d
or
uns
t
r
uc
tu
r
e
d
da
ta
pos
e
s
a
s
ig
ni
f
ic
a
nt
c
ha
ll
e
nge
in
DM
,
pa
r
ti
c
ul
a
r
ly
w
he
n
th
e
c
la
s
s
of
in
te
r
e
s
t
ha
s
f
a
r
f
e
w
e
r
s
a
m
pl
e
s
th
a
n
th
e
ot
he
r
s
.
T
hi
s
im
ba
la
nc
e
c
a
n
le
a
d
to
bi
a
s
e
d
pr
e
di
c
ti
ons
a
nd
r
e
duc
e
d
m
ode
l
pe
r
f
or
m
a
nc
e
,
m
a
ki
ng
it
c
r
uc
ia
l
to
im
pl
e
m
e
nt
s
tr
a
te
gi
e
s
th
a
t
e
f
f
e
c
ti
ve
ly
m
it
ig
a
te
th
e
s
e
is
s
ue
s
.
D
if
f
e
r
e
nt
c
la
s
s
if
ie
r
s
a
r
e
us
e
d
to
h
a
ndl
e
t
he
c
la
s
s
im
ba
la
nc
e
d
a
ta
s
e
t.
H
ow
e
ve
r
,
f
or
th
e
s
tu
dy pur
pos
e
w
e
w
e
r
e
ut
il
iz
e
d
s
om
e
di
f
f
e
r
e
nt
c
la
s
s
if
ie
r
s
a
s
de
s
c
r
ib
e
d a
s
f
ol
lo
w
s
:
4
.1
.
K
-
n
e
ar
e
s
t
n
e
ig
h
b
or
s
A
w
e
ll
-
li
ke
d
a
nd
a
da
pt
a
bl
e
c
la
s
s
if
ic
a
ti
on
a
ppr
oa
c
h
f
or
i
m
ba
la
nc
e
d
da
ta
i
s
K
N
N
[
42]
.
T
o
s
uc
c
e
s
s
f
ul
ly
m
a
na
ge
th
e
d
a
ta
s
e
t’
s
im
ba
la
nc
e
,
n
e
ve
r
th
e
le
s
s
,
a
f
e
w
f
a
c
to
r
s
a
nd
s
tr
a
te
gi
e
s
m
us
t
b
e
us
e
d.
H
e
r
e
a
r
e
a
f
e
w
m
e
th
ods
f
o
r
a
ppl
yi
ng
K
N
N
to
da
ta
th
a
t
is
im
ba
la
nc
e
d
:
i)
c
la
s
s
ba
la
nc
in
g
te
c
hni
que
s
:
unde
r
s
a
m
pl
in
g:
e
xt
r
a
c
t
in
s
ta
nc
e
s
b
e
lo
ngi
ng
to
th
e
m
a
jo
r
it
y
c
la
s
s
to
de
c
r
e
a
s
e
th
e
num
be
r
of
in
s
ta
nc
e
s
a
nd
in
c
r
e
a
s
e
th
e
numbe
r
of
i
ns
ta
nc
e
s
of
t
he
m
in
or
i
ty
c
la
s
s
. H
ow
e
ve
r
, do not
th
r
ow
a
w
a
y t
he
ba
by w
it
h
ba
th
w
a
te
r
t
o
a
voi
d
im
por
ta
nt
da
ta
di
s
a
ppe
a
r
in
g
s
a
m
e
w
a
y.
O
v
e
r
s
a
m
pl
in
g:
m
or
e
in
s
ta
nc
e
s
f
r
om
th
e
m
in
or
it
y
c
la
s
s
s
houl
d
b
e
s
ynt
he
s
iz
e
d
by
m
e
th
ods
li
ke
S
M
O
T
E
a
nd
A
D
A
S
Y
N
.
S
a
m
pl
in
g
te
c
hni
que
s
:
t
he
r
e
a
r
e
a
num
be
r
of
te
c
hni
que
s
s
uc
h
a
s
gi
vi
ng
m
or
e
w
e
ig
ht
s
to
th
e
c
la
s
s
e
s
w
hi
c
h
a
r
e
le
s
s
r
e
pr
e
s
e
nt
e
d,
th
e
n
m
a
ki
ng
th
e
m
ode
l
to
f
oc
u
s
m
or
e
on
th
e
m
in
or
it
y
c
la
s
s
;
ii
)
di
s
ta
nc
e
m
e
tr
ic
s
e
le
c
ti
on:
c
hoos
e
a
p
pr
opr
ia
te
di
s
ta
nc
e
m
e
tr
ic
s
th
a
t
a
r
e
s
e
ns
it
iv
e
to
th
e
na
tu
r
e
of
th
e
da
ta
.
E
xpe
r
im
e
nt
w
it
h
di
f
f
e
r
e
nt
di
s
ta
nc
e
m
e
tr
i
c
s
(
e
.g.,
E
uc
li
de
a
n
a
nd
M
a
nha
tt
a
n
)
to
f
in
d
th
e
m
os
t
s
ui
ta
bl
e
one
f
or
your
da
ta
s
e
t
;
ii
i
)
tu
ne
K
va
lu
e
:
e
xpe
r
im
e
nt
w
it
h
di
f
f
e
r
e
nt
va
lu
e
s
of
K
(
th
e
num
be
r
o
f
ne
a
r
e
s
t
ne
ig
hbor
s
)
t
o f
in
d t
he
opt
im
a
l
va
lu
e
f
or
your
da
ta
s
e
t.
T
oo s
m
a
ll
a
K
m
a
y
le
a
d t
o ove
r
f
it
ti
ng, while
t
o
o
la
r
ge
a
K
m
a
y
c
a
us
e
unde
r
f
it
ti
ng
;
iv
)
pr
oba
bi
li
s
ti
c
out
put
:
obt
a
in
pr
oba
bi
li
ti
e
s
f
or
e
a
c
h
c
la
s
s
pr
e
di
c
ti
on
in
s
te
a
d
of
di
r
e
c
t
c
la
s
s
la
be
ls
.
T
hi
s
a
ll
ow
s
f
or
a
dj
us
ti
ng
th
e
c
la
s
s
if
ic
a
ti
on
th
r
e
s
hol
d
to
im
pr
ove
s
e
ns
it
iv
it
y
or
s
pe
c
if
ic
it
y
ba
s
e
d
on
th
e
pr
obl
e
m
’
s
r
e
qui
r
e
m
e
nt
s
;
v)
a
lg
or
it
hm
tu
ni
ng:
e
xpe
r
im
e
nt
w
it
h
di
f
f
e
r
e
nt
K
N
N
va
r
ia
nt
s
or
e
xt
e
ns
io
ns
t
ha
t
a
r
e
s
ui
ta
bl
e
f
or
i
m
ba
la
nc
e
d da
ta
s
e
ts
, s
uc
h a
s
E
N
N
or
c
onde
ns
e
d ne
a
r
e
s
t
ne
ig
hbor
s
;
vi
)
e
ns
e
m
bl
e
a
ppr
oa
c
he
s
:
c
om
bi
ne
K
N
N
w
it
h
ot
he
r
c
la
s
s
if
ie
r
s
or
e
ns
e
m
bl
e
m
e
th
ods
li
ke
ba
ggi
ng
or
boo
s
ti
ng
to
e
nha
nc
e
pe
r
f
or
m
a
nc
e
,
e
s
pe
c
ia
ll
y
f
or
im
ba
la
nc
e
d
da
ta
s
e
t
s
;
vi
i)
c
r
os
s
-
va
li
da
ti
on
a
nd
e
va
lu
a
ti
on
m
e
tr
ic
s
:
in
or
de
r
to
r
e
duc
e
th
e
r
is
k
of
a
hi
gh
va
r
ia
nc
e
th
e
f
ol
lo
w
in
g
m
e
th
ods
s
houl
d
be
e
m
pl
oye
d,
f
or
in
s
ta
nc
e
,
s
tr
a
ti
f
ie
d
k
-
f
ol
d
c
r
os
s
va
li
da
ti
on
e
ns
ur
e
s
th
a
t
c
la
s
s
e
s
in
th
e
da
ta
di
vi
s
io
n
a
r
e
pr
opor
ti
ona
te
ly
s
pl
it
.
I
n
e
va
lu
a
ti
ng
th
e
Evaluation Warning : The document was created with Spire.PDF for Python.