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.
2
,
A
pr
il
2025
, pp.
1386
~
1400
I
S
S
N
:
2252
-
8938
,
D
O
I
:
10.11591/
ij
a
i.
v
14
.i
2
.pp
1386
-
1400
1386
Jou
r
n
al
h
om
e
page
:
ht
tp
:
//
ij
ai
.
ia
e
s
c
or
e
.c
om
A
n
ove
l
sc
al
ab
l
e
d
e
e
p
e
n
se
m
b
l
e
l
e
ar
n
i
n
g f
r
am
e
w
or
k
f
or
b
i
g d
at
a
c
l
ass
i
f
i
c
at
i
on
vi
a M
ap
R
e
d
u
c
e
i
n
t
e
gr
a
t
i
on
K
e
s
avan
M
e
t
t
u
r
V
ar
ad
h
ar
aj
an
1,
2
, Jos
e
p
h
in
e
P
r
e
m
K
u
m
a
r
3
,
N
an
d
a A
s
h
w
in
2
1
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng
,
V
i
s
ve
s
va
r
a
ya
T
e
c
hnol
ogi
c
a
l
U
ni
ve
r
s
i
t
y, B
e
l
a
ga
vi
, I
ndi
a
2
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng
, E
a
s
t
P
oi
nt
C
ol
l
e
ge
of
E
ngi
ne
e
r
i
ng a
nd T
e
c
hnol
ogy, B
e
nga
l
ur
u, I
ndi
a
3
D
e
pa
r
t
m
e
nt
of
C
om
put
e
r
S
c
i
e
nc
e
a
nd E
ngi
ne
e
r
i
ng
, C
a
m
br
i
dge
I
ns
t
i
t
ut
e
of
T
e
c
hnol
ogy, B
e
nga
l
ur
u, 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
M
a
r
14, 2024
R
e
vi
s
e
d
O
c
t
30, 2024
A
c
c
e
pt
e
d
N
ov 14, 2024
Big
d
ata
classification
involves
the
systematic
sorting
and
analy
sis
of
extensiv
e
datasets
that
are
aggregated
from
a
variety
of
sources.
These
datasets
may
include
but
are
not
limited
to,
electronic
records,
digital
imaging,
genetic
information
sequence
s,
transactio
nal
data,
resear
ch
o
utputs,
and
data
streams
from
wearable
technologies
and
connected
device
s.
This
paper introduces the
scalable
deep ensemble
learning
framework
for
b
ig data
classifi
cation
(SDELF
-
BDC),
a
novel
methodology
tailored
f
or
the
classifi
c
ation
of
large
-
scale
data.
At
its
core,
SDELF
-
BDC
lever
ages
a
Hadoop
-
based
map
-
reduce
framework
for
feature
s
election,
signif
icantly
reducing
feature
-
length
and
enhancing
computational
efficiency
.
The
methodology
is
further
augmented
by
a
deep
ensemble
mode
l
that
judiciously
applies
a
variety
of
deep
learning
classifier
s
based
o
n
data
characterist
ics,
thereby
ensuring
optimal
performance.
Each
clas
sifier'
s
output
undergoes
a
rigorous
optimization
-
based
ensemble
approa
ch
for
refinement,
utilizin
g
a
sophist
ic
ated
algorithm.
The
result
is
a
robust
classifi
cation
system
that
excels
in
predicti
ve
accur
acy
while
maint
aining
scalability
and
responsiveness
to
the
dynamic
requirements
of
bi
g
data
environm
ents.
Through
a
strategi
c
combinat
ion
of
classifi
ers
a
nd
an
innovative
reduction
phase,
SDELF
-
BDC
emerges
as
a
compreh
ensive
solution
for
big
data
classifi
cation
challenges
,
setting
new
benchma
rks
for
predictive analytics in diverse and da
ta
-
intensive doma
ins.
K
e
y
w
o
r
d
s
:
B
ig
da
ta
D
e
e
p
e
n
s
e
m
b
l
e
l
e
a
r
n
i
n
g
f
r
a
m
e
w
o
r
k
E
ns
e
m
bl
e
de
e
p ne
twor
k
L
a
r
ge
-
s
c
a
le
da
ta
M
a
pR
e
duc
e
i
nt
e
gr
a
ti
on
S
D
E
L
F
-
B
D
C
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
:
K
e
s
a
va
n
M
e
tt
ur
V
a
r
a
dha
r
a
ja
n
D
e
pa
r
tm
e
nt
of
C
om
put
e
r
S
c
ie
nc
e
a
nd E
ngi
ne
e
r
in
g, V
is
ve
s
v
a
r
a
ya
T
e
c
hnol
ogi
c
a
l
U
ni
ve
r
s
it
y
B
e
la
ga
vi
, I
ndi
a
E
m
a
il
:
ke
s
_m
v@
r
e
di
f
f
m
a
il
.c
om
1.
I
N
T
R
O
D
U
C
T
I
O
N
I
n
r
e
c
e
nt
ye
a
r
s
,
bi
g
da
ta
ha
s
be
c
om
e
th
e
m
a
in
te
c
h
ta
lk
be
twe
e
n
a
c
a
de
m
ia
a
nd
pr
a
c
ti
ti
one
r
s
in
th
e
di
gi
ta
l
c
om
pe
ti
ti
ve
pl
a
yya
r
d.
B
ig
da
ta
is
a
n
im
por
ta
nt
a
s
s
e
t
th
a
t
a
tt
r
a
c
ts
th
e
a
tt
e
nt
io
n
of
m
a
ny
c
hi
e
f
e
x
e
c
ut
iv
e
of
f
ic
e
r
s
(
C
E
O
s
)
in
di
f
f
e
r
e
nt
or
ga
ni
z
a
ti
ons
to
ga
in
f
a
s
te
r
in
s
ig
ht
s
a
nd
hi
gh
r
e
v
e
nue
[
1]
.
T
h
e
jo
ur
ne
y
of
bi
g
da
ta
s
ta
r
te
d w
he
n m
a
ny or
ga
ni
z
a
ti
ons
r
e
c
ogni
z
e
d t
ha
t
th
e
l
a
r
ge
vol
um
e
of
t
he
ir
da
ta
e
xc
e
e
de
d t
he
c
a
pa
bi
li
ti
e
s
of
th
e
ir
or
ga
ni
z
a
ti
ons
,
pr
oc
e
s
s
,
c
a
pa
c
it
y,
s
tr
uc
tu
r
e
,
t
e
c
hnol
ogy
i
nf
r
a
s
tr
uc
tu
r
e
,
a
nd
gove
r
na
nc
e
.
T
h
e
y s
tr
uggl
e
d
to
de
a
l
w
it
h t
he
r
e
qui
r
e
m
e
nt
s
f
or
a
na
ly
z
in
g t
he
hi
gh volum
e
of
va
r
io
us
da
ta
[
2]
.
A
c
c
or
di
ng
to
s
ta
ti
s
ti
c
a
l
r
e
por
ts
,
th
e
num
be
r
of
us
e
r
s
on
di
f
f
e
r
e
nt
s
oc
ia
l
m
e
di
a
pl
a
tf
or
m
s
ha
s
r
e
a
c
he
d
m
or
e
th
a
n
2
bi
ll
io
n.
W
ha
ts
A
pp,
f
or
e
xa
m
pl
e
,
ha
s
ov
e
r
600
m
il
li
on
us
e
r
s
,
m
or
e
th
a
n
ha
lf
a
bi
ll
io
n
phot
os
,
a
nd
one
hundr
e
d
m
il
li
on
vi
de
os
tr
a
ns
f
e
r
r
e
d
a
nd
s
ha
r
e
d
be
twe
e
n
us
e
r
s
da
il
y
[
3]
.
A
ls
o,
due
to
th
e
huge
a
dva
nc
e
s
in
s
m
a
r
tp
hone
te
c
hnol
ogy,
it
ha
s
be
c
om
e
e
a
s
ie
r
f
or
us
e
r
s
to
s
ha
r
e
im
a
ge
s
a
nd
w
r
it
e
pos
ts
on
s
u
c
h
s
oc
ia
l
m
e
di
a
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
A
nov
e
l
s
c
al
abl
e
de
e
p e
n
s
e
m
bl
e
l
e
a
r
ni
ng f
r
am
e
w
or
k
f
o
r
bi
g dat
a
…
(
K
e
s
av
an M
e
tt
ur
V
ar
adha
r
aj
an
)
1387
pl
a
tf
or
m
s
.
S
om
e
r
e
por
ts
s
how
th
a
t
th
e
num
be
r
of
po
s
ts
on
T
w
it
te
r
in
2007
w
a
s
5K
,
th
i
s
num
be
r
be
c
a
m
e
a
r
ound
500
m
il
li
on
a
f
te
r
a
bout
6
ye
a
r
s
,
w
hi
c
h
in
di
c
a
te
s
th
e
m
a
s
s
iv
e
a
m
ount
of
a
v
a
il
a
bl
e
da
t
a
on s
oc
ia
l
m
e
di
a
in
ge
ne
r
a
l.
T
hi
s
a
m
ount
of
da
ta
is
not
r
e
s
tr
ic
te
d
to
s
oc
ia
l
m
e
di
a
,
a
s
m
a
ny
ot
he
r
pl
a
tf
or
m
s
ge
ne
r
a
te
a
nd
s
to
r
e
huge
da
ta
vol
um
e
s
[
4]
.
T
hi
s
a
m
ount
of
da
ta
ne
e
ds
to
be
pr
oc
e
s
s
e
d
a
nd
a
na
ly
z
e
d
to
u
s
e
it
f
or
bui
ld
in
g
us
e
f
ul
knowle
dge
di
s
c
ove
r
y
a
nd
m
a
c
hi
ne
le
a
r
ni
ng
bi
g
da
ta
-
ba
s
e
d
a
ppl
ic
a
ti
ons
,
li
ke
f
a
c
ia
l
bi
g
da
ta
a
ppl
ic
a
ti
ons
,
s
ig
na
l
bi
g
da
ta
,
a
nd
va
r
io
us
in
dus
tr
y
bi
g
da
ta
-
ba
s
e
d
a
ppl
ic
a
ti
ons
[
5]
.
V
ol
um
e
(
bi
g)
,
va
r
ie
ty
,
a
nd
ve
lo
c
it
y
a
r
e
a
m
ong
th
e
m
os
t
di
s
ti
ngui
s
hi
ng
c
ha
r
a
c
te
r
is
ti
c
s
of
bi
g
da
ta
,
a
nd
a
s
a
r
e
s
ul
t,
it
is
a
tt
r
a
c
ti
ve
to
ha
ve
a
n
e
f
f
ic
ie
nt
c
la
s
s
if
ic
a
ti
on/
pr
e
di
c
ti
on
s
ys
t
e
m
to
le
a
r
n
f
r
om
s
uc
h
bi
g
da
ta
.
S
uc
h
a
ppl
ic
a
ti
ons
in
c
lu
de
,
but
a
r
e
not
li
m
it
e
d
to
,
m
e
di
c
a
l,
f
in
a
nc
ia
l,
s
e
c
ur
it
y, a
nd i
m
a
ge
-
ba
s
e
d a
ppl
ic
a
ti
on
s
[
6]
.
B
ig
da
ta
a
na
ly
s
is
of
f
e
r
s
s
e
r
vi
c
e
to
ol
s
,
s
uc
h
a
s
H
a
doop
di
s
tr
ib
ut
e
d
f
il
e
s
ys
te
m
(
H
D
F
S
)
w
hi
c
h
s
uppor
ts
m
a
na
gi
ng
a
nd
s
to
r
in
g
huge
a
m
ount
s
of
da
ta
,
f
a
s
t
a
ut
om
a
te
d
de
c
is
io
ns
a
nd
de
c
r
e
a
s
e
s
th
e
r
is
ks
of
hum
a
n
e
s
ti
m
a
ti
ons
.
T
h
e
H
D
F
S
i
s
a
c
c
e
pt
e
d
a
s
th
e
m
o
s
t
w
id
e
l
y
us
e
d
da
ta
s
e
t
to
ol
th
a
t
s
uppor
ts
r
e
dunda
nc
y,
r
e
li
a
bi
li
ty
,
s
c
a
la
bi
li
ty
,
pa
r
a
ll
e
l
pr
oc
e
s
s
in
g,
a
nd
di
s
tr
ib
ut
e
d
a
r
c
hi
te
c
tu
r
e
s
ys
t
e
m
s
[
7]
,
a
nd
is
de
s
ig
ne
d
to
ha
ndl
e
di
f
f
e
r
e
nt
bi
g
da
ta
ty
pe
s
;
s
tr
uc
tu
r
e
d,
s
e
m
i
-
s
tr
uc
tu
r
e
d,
a
nd
uns
t
r
uc
tu
r
e
d.
M
or
e
ove
r
,
th
e
H
a
doop
M
a
pR
e
duc
e
jo
b
-
s
c
he
dul
in
g
a
lg
or
it
hm
[
6]
s
uppor
ts
c
lu
s
te
r
in
g
bi
g
da
ta
in
a
s
pr
e
a
d
ne
twor
k
e
nvi
r
onm
e
nt
.
I
n
a
ddi
ti
on,
bi
g
da
ta
a
na
ly
s
is
pr
ovi
de
s
s
ig
ni
f
ic
a
nt
oppor
tu
ni
ti
e
s
f
or
s
ol
vi
ng
di
f
f
e
r
e
nt
in
f
or
m
a
ti
on
s
e
c
ur
it
y
pr
obl
e
m
s
by
u
s
in
g
H
a
doop
te
c
hnol
ogi
e
s
a
nd
H
D
F
S
to
ol
s
.
T
h
e
da
ta
va
lu
e
th
a
t
is
ge
ne
r
a
te
d
f
r
om
bi
g
da
ta
th
r
ough
th
e
a
na
ly
s
is
pha
s
e
is
of
e
xt
r
e
m
e
im
por
ta
nc
e
[
8
]
.
D
ur
in
g
th
e
pr
e
vi
ous
f
e
w
de
c
a
de
s
,
c
la
s
s
if
ie
r
s
ha
ve
be
e
n
in
te
ns
iv
e
ly
s
tu
di
e
d
a
nd
a
na
ly
z
e
d.
C
la
s
s
if
ie
r
s
a
r
e
w
id
e
ly
u
s
e
d
in
m
a
ny
m
o
de
r
n
a
ppl
ic
a
ti
ons
a
s
ke
y
c
om
put
e
r
te
c
hnol
ogy.
M
a
ny c
la
s
s
if
ie
r
s
e
xi
s
t,
s
uc
h a
s
k
-
ne
a
r
e
s
t
ne
ig
hbor
(
K
N
N
)
,
s
uppor
t
ve
c
to
r
m
a
c
hi
ne
(
S
V
M
)
,
n
a
iv
e
B
a
ye
s
(
N
B
)
,
r
a
ndom
f
or
e
s
t
(
R
F
)
,
de
c
is
io
n
t
r
e
e
(
D
T
)
,
a
nd
m
a
ny
m
or
e
.
I
n
te
r
m
s
of
a
c
c
ur
a
c
y a
nd
ti
m
e
c
on
s
um
in
g
bui
ld
in
g
a
tr
a
in
e
d
m
ode
l,
e
a
c
h
c
la
s
s
if
ie
r
ha
s
a
dv
a
nt
a
ge
s
a
nd
li
m
it
a
ti
ons
;
s
om
e
a
r
e
m
or
e
e
f
f
e
c
ti
ve
w
it
h
s
pe
c
if
ic
da
ta
s
e
ts
th
a
n ot
he
r
s
, a
nd he
nc
e
t
he
r
e
i
s
no opti
m
a
l
c
la
s
s
if
ie
r
t
ha
t
c
a
n pe
r
f
e
c
tl
y c
la
s
s
if
y a
ll
t
ype
s
of
da
ta
[
8]
.
D
e
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
s
uc
h
a
s
c
onvolut
io
na
l
ne
ur
a
l
ne
twor
ks
(
C
N
N
s
)
a
nd
r
e
c
ur
r
e
nt
ne
ur
a
l
ne
twor
ks
(
R
N
N
s
)
,
in
c
lu
di
ng
lo
ng
s
hor
t
-
te
r
m
m
e
m
or
y
ne
twor
k
s
(
L
S
T
M
s
)
,
of
f
e
r
ta
il
or
e
d
a
dva
nt
a
ge
s
f
or
bi
g
da
ta
c
la
s
s
if
ic
a
ti
on.
C
N
N
s
e
xc
e
l
in
s
pa
ti
a
l
hi
e
r
a
r
c
hy
le
a
r
ni
ng
f
r
om
im
a
ge
s
,
e
na
bl
in
g
r
obus
t
f
e
a
tu
r
e
e
xt
r
a
c
ti
on
w
it
hout
m
a
nua
l
in
te
r
ve
nt
io
n,
m
a
ki
ng
th
e
m
id
e
a
l
f
or
im
a
ge
-
ba
s
e
d
c
la
s
s
if
ic
a
ti
on
ta
s
ks
.
T
he
ir
s
tr
uc
tu
r
e
,
c
om
pos
e
d
of
c
onvolut
io
na
l,
pool
in
g,
a
nd
f
ul
ly
c
onne
c
te
d
la
y
e
r
s
,
e
ns
ur
e
s
e
f
f
ic
ie
nc
y
a
nd
s
c
a
la
bi
li
ty
,
e
ve
n
w
it
h
hi
gh
-
di
m
e
ns
io
na
l
da
ta
.
O
n
th
e
ot
he
r
ha
nd,
R
N
N
s
a
nd
L
S
T
M
s
s
hi
ne
in
s
e
que
n
c
e
da
ta
a
na
ly
s
is
,
s
uc
h
a
s
ti
m
e
s
e
r
ie
s
,
s
p
e
e
c
h,
or
te
xt
,
by
e
f
f
e
c
ti
ve
ly
c
a
pt
ur
in
g
te
m
por
a
l
de
pe
nde
nc
ie
s
a
nd
h
a
ndl
in
g
va
r
ia
bl
e
-
le
ngt
h
in
put
s
.
L
S
T
M
s
f
ur
th
e
r
m
it
ig
a
te
t
he
va
ni
s
hi
ng gr
a
di
e
nt
p
r
obl
e
m
, a
ll
ow
i
ng f
or
t
he
l
e
a
r
ni
ng
of
l
ong
-
te
r
m
de
pe
nde
nc
ie
s
.
T
he
s
e
a
r
c
hi
te
c
tu
r
e
s
c
ol
le
c
ti
ve
ly
e
nh
a
nc
e
bi
g
da
ta
c
la
s
s
if
ic
a
ti
o
n
by
of
f
e
r
in
g
s
c
a
la
bl
e
,
e
f
f
ic
ie
nt
,
a
nd
a
c
c
ur
a
te
m
ode
li
ng
c
a
pa
bi
li
ti
e
s
,
c
a
pa
bl
e
of
e
xt
r
a
c
ti
ng
de
e
p
in
s
ig
ht
s
f
r
om
c
om
pl
e
x
da
ta
s
e
ts
,
th
e
r
e
by
dr
iv
in
g
in
nova
ti
on
a
c
r
os
s
di
ve
r
s
e
doma
in
s
[
9]
.
T
he
in
te
gr
a
ti
on
of
de
e
p
le
a
r
ni
ng
w
it
h
M
a
pR
e
du
c
e
,
a
pr
ogr
a
m
m
in
g
m
ode
l
f
or
pr
oc
e
s
s
in
g
a
nd
ge
ne
r
a
ti
ng
la
r
ge
da
ta
s
e
ts
w
it
h
a
pa
r
a
ll
e
l,
di
s
tr
ib
ut
e
d
a
lg
or
i
th
m
on
a
c
lu
s
te
r
,
of
f
e
r
s
a
pow
e
r
f
ul
a
pp
r
oa
c
h
f
or
e
nha
nc
e
d
c
la
s
s
if
ic
a
ti
on
ta
s
ks
on
bi
g
da
ta
.
T
hi
s
in
te
gr
a
ti
on
le
ve
r
a
ge
s
th
e
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y
o
f
M
a
pR
e
duc
e
to
ha
ndl
e
th
e
m
a
s
s
iv
e
s
c
a
la
bi
li
ty
r
e
qui
r
e
m
e
nt
s
o
f
de
e
p
le
a
r
ni
ng
a
lg
or
it
hm
s
,
a
ll
ow
in
g
f
or
th
e
di
s
tr
ib
ut
e
d
pr
oc
e
s
s
in
g
of
da
ta
a
c
r
os
s
m
ul
ti
pl
e
node
s
,
w
hi
c
h
s
i
gni
f
ic
a
nt
ly
s
pe
e
ds
up
th
e
tr
a
in
in
g
of
c
om
pl
e
x
m
ode
ls
on l
a
r
ge
da
ta
s
e
t
s
. F
or
i
ns
ta
nc
e
, de
e
p l
e
a
r
ni
ng mode
ls
, s
uc
h a
s
C
N
N
s
f
or
i
m
a
ge
c
la
s
s
if
ic
a
ti
on or
R
N
N
s
f
or
s
e
que
nc
e
da
ta
,
c
a
n
be
tr
a
in
e
d
m
or
e
e
f
f
ic
ie
nt
ly
u
s
in
g
th
is
in
te
gr
a
te
d
a
ppr
oa
c
h,
e
na
bl
in
g
m
or
e
s
ophi
s
ti
c
a
te
d
a
nd
a
c
c
ur
a
te
c
la
s
s
if
ic
a
ti
on
c
a
pa
bi
li
ti
e
s
.
H
ow
e
ve
r
,
th
i
s
in
te
gr
a
ti
on
is
not
w
it
hout
it
s
li
m
it
a
ti
ons
.
T
he
c
om
pl
e
xi
ty
of
c
onf
ig
ur
in
g
a
nd
m
a
na
gi
ng
a
di
s
tr
ib
ut
e
d
c
om
put
in
g
e
nvi
r
onm
e
nt
c
a
n
in
tr
oduc
e
ove
r
he
a
d
a
nd
pot
e
nt
ia
l
bot
tl
e
ne
c
ks
,
e
s
p
e
c
ia
ll
y
in
te
r
m
s
of
ne
twor
k
c
om
m
uni
c
a
ti
on
a
nd
da
ta
tr
a
ns
f
e
r
s
pe
e
d
s
.
A
ddi
ti
ona
ll
y,
th
e
in
he
r
e
nt
c
ha
ll
e
nge
s
of
pa
r
a
ll
e
li
z
in
g
de
e
p
le
a
r
ni
ng
a
lg
or
it
hm
s
,
s
uc
h
a
s
s
ync
hr
oni
z
a
ti
on
of
m
ode
l
upda
te
s
a
nd
th
e
non
-
uni
f
or
m
di
s
tr
ib
ut
io
n
of
da
ta
,
c
a
n
a
f
f
e
c
t
th
e
e
f
f
ic
i
e
nc
y
a
nd
s
c
a
la
bi
li
ty
of
th
e
s
ol
ut
io
n.
D
e
s
pi
te
th
e
s
e
c
ha
ll
e
nge
s
,
th
e
c
om
bi
na
ti
on
of
de
e
p
le
a
r
ni
ng
a
nd
M
a
pR
e
duc
e
hol
ds
s
ig
ni
f
ic
a
nt
pr
om
is
e
f
or
a
dva
nc
in
g
bi
g da
ta
c
la
s
s
if
ic
a
ti
on by ha
r
ne
s
s
in
g t
he
s
tr
e
ngt
hs
of
bot
h t
e
c
hn
ol
ogi
e
s
[
10]
–
[
12]
.
T
he
e
xpone
nt
ia
l
gr
ow
th
of
bi
g
da
ta
a
c
r
os
s
va
r
io
us
s
e
c
to
r
s
,
c
oupl
e
d
w
it
h
it
s
p
r
of
ound
i
m
pa
c
t
on
de
c
is
io
n
-
m
a
ki
ng,
in
nova
ti
on,
a
nd
c
om
pe
ti
ti
ve
a
dva
nt
a
ge
,
ha
s
s
pa
r
ke
d
s
ig
ni
f
ic
a
nt
in
te
r
e
s
t
a
m
ong
r
e
s
e
a
r
c
he
r
s
a
nd
pr
a
c
ti
ti
one
r
s
a
li
ke
.
T
hi
s
bur
ge
oni
ng
in
te
r
e
s
t
i
s
r
oot
e
d
in
th
e
r
e
a
li
z
a
ti
on
th
a
t
bi
g
da
t
a
,
c
ha
r
a
c
te
r
iz
e
d
by
it
s
vol
um
e
,
ve
lo
c
it
y,
a
nd
va
r
ie
ty
,
hol
ds
th
e
ke
y
to
unl
oc
ki
ng
nov
e
l
in
s
ig
ht
s
a
nd
f
os
te
r
in
g
a
dva
nc
e
m
e
nt
s
a
c
r
os
s
va
r
io
us
f
ie
ld
s
in
c
lu
di
ng
he
a
lt
hc
a
r
e
,
f
in
a
nc
e
,
s
e
c
ur
it
y,
a
nd
be
yo
nd.
T
he
m
ot
iv
a
ti
on
be
hi
nd
th
is
r
e
s
e
a
r
c
h
s
te
m
s
f
r
om
th
e
c
ha
ll
e
nge
s
a
nd
oppor
tu
ni
ti
e
s
pr
e
s
e
nt
e
d
by
th
e
v
a
s
t
a
m
ount
s
of
da
t
a
ge
ne
r
a
te
d
by
s
oc
ia
l
m
e
di
a
pl
a
tf
or
m
s
,
I
oT
de
vi
c
e
s
,
a
nd
di
gi
ta
l
tr
a
ns
a
c
ti
ons
,
w
hi
c
h
e
xc
e
e
d
th
e
pr
oc
e
s
s
in
g
c
a
pa
bi
li
ti
e
s
of
tr
a
di
ti
ona
l
da
ta
a
na
ly
s
is
to
ol
s
.
A
s
or
ga
ni
z
a
ti
ons
s
tr
iv
e
to
ha
r
ne
s
s
th
e
f
ul
l
pot
e
nt
ia
l
of
bi
g
da
ta
f
or
knowle
dge
di
s
c
ove
r
y,
m
a
c
hi
ne
le
a
r
ni
ng,
a
nd
pr
e
di
c
ti
ve
a
na
ly
ti
c
s
,
th
e
r
e
a
r
is
e
s
a
c
r
it
ic
a
l
ne
e
d
f
or
e
f
f
ic
ie
nt
c
la
s
s
if
ic
a
ti
on
a
nd
pr
e
di
c
ti
on
s
ys
te
m
s
c
a
p
a
bl
e
of
m
a
na
gi
ng
th
is
c
om
pl
e
xi
ty
.
T
he
in
te
gr
a
ti
on
of
de
e
p
le
a
r
ni
ng
a
r
c
hi
te
c
tu
r
e
s
a
nd
M
a
pR
e
duc
e
f
r
a
m
e
w
or
ks
pr
e
s
e
nt
s
a
pr
om
is
in
g
a
ve
nue
f
or
a
d
dr
e
s
s
in
g
th
e
s
e
c
ha
ll
e
nge
s
,
of
f
e
r
in
g
s
c
a
la
bl
e
,
e
f
f
ic
ie
nt
s
ol
ut
io
ns
f
or
bi
g
da
ta
c
la
s
s
if
ic
a
ti
on.
T
hi
s
r
e
s
e
a
r
c
h
is
dr
iv
e
n
by
th
e
a
m
bi
ti
on
to
c
ont
r
ib
ut
e
to
th
e
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.
2
,
A
pr
il
2025
:
1386
-
1400
1388
de
ve
lo
pm
e
nt
of
a
dva
nc
e
d
bi
g
da
ta
a
na
ly
s
i
s
to
ol
s
th
a
t
c
a
n
not
onl
y
m
a
na
ge
th
e
s
he
e
r
s
c
a
le
of
da
t
a
but
a
ls
o
pr
ovi
de
a
c
ti
ona
bl
e
in
s
ig
ht
s
,
th
e
r
e
by
e
n
a
bl
in
g
or
ga
ni
z
a
ti
ons
to
m
a
ke
in
f
or
m
e
d
de
c
is
io
ns
,
e
nha
nc
e
ope
r
a
ti
ona
l
e
f
f
ic
ie
nc
y, a
nd a
c
hi
e
ve
s
tr
a
te
gi
c
goa
ls
.
A
dv
a
n
c
e
d
d
e
e
p
e
n
s
e
m
b
le
l
e
a
r
ni
n
g
a
r
c
hi
te
c
t
ur
e
:
d
e
vi
s
e
d
a
s
oph
is
t
ic
a
t
e
d
f
r
a
m
e
w
o
r
k
th
a
t
in
t
e
g
r
a
t
e
s
a
s
u
it
e
o
f
de
e
p
l
e
a
r
ni
ng
c
la
s
s
if
ie
r
s
,
or
c
he
s
tr
a
te
d
w
it
hi
n
a
H
a
doo
p
-
b
a
s
e
d
in
f
r
a
s
tr
u
c
t
ur
e
.
T
hi
s
a
ppr
o
a
c
h
i
s
e
ng
in
e
e
r
e
d
to
r
e
f
i
ne
c
l
a
s
s
if
i
c
a
ti
o
n
by
c
o
ndu
c
t
in
g
pr
e
c
i
s
e
f
e
a
tu
r
e
s
e
le
c
t
io
n
a
n
d
e
m
pl
oyi
ng
a
hy
br
i
d
of
a
lg
or
it
hm
i
c
s
tr
a
te
gi
e
s
.
T
he
r
e
s
ul
t
i
s
a
s
u
bs
ta
nt
i
a
l
im
pr
o
ve
m
e
nt
i
n
a
c
c
ur
a
c
y
a
nd
e
f
f
ic
ie
nc
y,
p
a
r
ti
c
u
la
r
ly
s
ui
t
e
d
t
o
th
e
c
om
p
le
xi
t
ie
s
of
bi
g
d
a
t
a
c
la
s
s
if
ic
a
t
io
n
.
I
nno
va
ti
v
e
hy
br
id
he
u
r
i
s
ti
c
opt
im
i
z
a
ti
o
n
t
e
c
hn
iq
ue
:
in
tr
o
du
c
e
d
a
c
ut
ti
n
g
-
e
dg
e
op
ti
m
i
z
a
ti
on
m
e
th
od
t
ha
t
is
i
m
pl
e
m
e
n
te
d w
it
hi
n
a
m
a
p
-
r
e
du
c
e
f
r
a
m
e
w
or
k,
t
hi
s
t
e
c
hni
qu
e
e
n
ha
n
c
e
s
th
e
s
e
le
c
ti
on
a
nd
c
l
a
s
s
if
i
c
a
ti
on
of
f
e
a
tu
r
e
s
,
t
h
e
r
e
b
y
e
l
e
v
a
ti
ng
th
e
e
f
f
e
c
t
iv
e
n
e
s
s
a
nd
pr
e
c
i
s
io
n o
f
b
ig
da
ta
s
y
s
t
e
m
s
.
T
he
r
e
s
e
a
r
c
h
is
or
ga
ni
z
e
d
in
th
is
pa
pe
r
in
to
f
our
s
e
c
ti
ons
.
T
h
e
f
ir
s
t
s
e
c
ti
on
gi
ve
s
a
br
ie
f
ove
r
vi
e
w
of
th
e
bi
g
da
ta
c
la
s
s
if
ic
a
ti
on
pr
oc
e
s
s
,
a
nd
th
e
s
e
c
ond
s
e
c
ti
on
di
s
c
us
s
e
s
th
e
e
xi
s
ti
ng
li
te
r
a
tu
r
e
a
nd
s
hor
tc
om
in
gs
a
s
s
oc
ia
t
e
d
w
it
h
e
a
c
h.
I
n
th
e
th
ir
d
s
e
c
ti
on,
a
m
ode
l
is
pr
opos
e
d
to
ove
r
c
om
e
th
is
a
nd
in
th
e
f
our
th
s
e
c
ti
on
a
c
om
pa
r
is
on
is
c
a
r
r
ie
d
out
w
it
h
th
e
e
xi
s
ti
ng
a
nd
th
e
pr
opos
e
d
t
e
c
hni
que
s
,
a
nd
th
e
r
e
s
ul
ts
a
r
e
ta
bul
a
te
d
in
th
e
f
or
m
of
gr
a
phs
.
2.
R
E
L
A
T
E
D
WORK
I
n
th
e
r
e
a
lm
of
bi
g
da
ta
c
la
s
s
if
ic
a
ti
on,
r
e
c
e
nt
s
tu
di
e
s
ha
v
e
in
nova
te
d
m
e
th
ods
to
e
nh
a
nc
e
f
e
a
tu
r
e
s
e
le
c
ti
on
a
nd
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y
in
la
r
ge
da
ta
s
e
ts
.
F
or
e
x
a
m
pl
e
,
a
nove
l
a
ppr
oa
c
h
in
te
gr
a
t
e
s
M
a
pR
e
du
c
e
w
it
h
f
e
a
tu
r
e
s
ubs
e
t
s
e
le
c
ti
on
a
nd
hype
r
pa
r
a
m
e
te
r
-
tu
ne
d
de
e
p
be
li
e
f
ne
twor
ks
(
D
B
N
)
,
a
im
in
g
to
a
ddr
e
s
s
th
e
c
om
pl
e
xi
ti
e
s
of
bi
g
da
ta
pr
oc
e
s
s
in
g
a
nd
im
pr
ove
c
la
s
s
if
ic
a
ti
o
n
pe
r
f
or
m
a
nc
e
.
A
c
la
s
s
if
ic
a
ti
on
m
ode
l
na
m
e
d
r
a
ndom
f
or
e
s
t
-
ba
s
e
d
f
e
a
tu
r
e
s
e
le
c
ti
on
a
nd
e
xt
r
a
c
ti
on
(
R
F
S
E
)
-
ga
te
d
r
e
c
ur
r
e
nt
uni
t
(
G
R
U
)
[
13]
w
a
s
de
ve
lo
pe
d,
in
te
gr
a
ti
ng
G
R
U
w
it
h
a
s
tr
a
te
gi
c
a
ppr
oa
c
h
f
or
f
e
a
tu
r
e
s
e
le
c
ti
on
a
nd
da
ta
ba
la
nc
e
.
T
hi
s
m
ode
l
le
ve
r
a
ge
s
th
e
RF
a
lg
or
it
hm
to
id
e
nt
if
y
th
e
m
os
t
im
pa
c
tf
ul
f
e
a
tu
r
e
s
f
or
c
la
s
s
if
ic
a
ti
on.
T
o
m
it
ig
a
te
th
e
c
ha
ll
e
nge
s
pos
e
d
by
da
ta
im
ba
la
nc
e
,
it
e
m
pl
oys
a
c
om
bi
na
ti
on
of
th
e
s
ynt
he
ti
c
m
i
nor
it
y
ove
r
s
a
m
pl
in
g
te
c
hni
que
(
S
M
O
T
E
)
f
or
ove
r
s
a
m
pl
in
g
a
nd
th
e
e
di
te
d
ne
a
r
e
s
t
ne
ig
hbor
(
E
N
N
)
te
c
hni
que
f
or
unde
r
-
s
a
m
pl
in
g,
e
nha
nc
in
g
th
e
m
ode
l
'
s
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y.
T
e
c
hnol
ogi
e
s
s
uc
h
a
s
te
xt
a
nd
da
ta
m
in
in
g,
onl
in
e
a
nd
m
obi
le
m
in
in
g,
pr
oc
e
s
s
m
in
in
g,
s
ta
ti
s
ti
c
a
l
a
na
ly
s
is
,
ne
twor
k
a
na
ly
ti
c
s
,
s
oc
ia
l
m
e
di
a
a
na
ly
ti
c
s
,
a
udi
o
a
nd
vi
d
e
o
a
na
ly
ti
c
s
,
a
nd
w
e
b
a
na
ly
ti
c
s
a
r
e
a
ll
in
c
lu
de
d
in
bi
g
da
ta
a
na
ly
ti
c
s
.
T
o
e
nha
nc
e
he
a
lt
hc
a
r
e
da
ta
s
e
t
s
,
a
r
a
nge
of
da
ta
m
in
in
g
te
c
hni
que
s
m
a
y
be
u
s
e
d,
s
u
c
h
a
s
s
um
m
a
r
iz
a
ti
on,
a
s
s
oc
ia
ti
on
r
ul
e
s
,
c
lu
s
te
r
in
g,
c
la
s
s
if
ic
a
ti
on,
a
nom
a
ly
de
te
c
ti
on,
a
nd
la
r
ge
-
s
c
a
le
da
ta
vi
s
ua
li
z
a
ti
on
[
14]
.
M
ode
r
n
da
ta
a
na
ly
ti
c
s
a
lg
or
it
hm
s
e
m
pl
oy
c
e
r
ta
in
da
ta
pr
ope
r
ti
e
s
to
e
va
lu
a
te
s
e
ns
or
a
nd
hi
gh
-
s
pe
e
d
da
t
a
s
tr
e
a
m
s
.
B
ig
da
ta
ha
s
s
e
v
e
r
a
l
a
ppl
ic
a
ti
on
s
,
in
c
lu
di
ng
im
pr
ovi
ng
di
a
gnos
is
,
a
ve
r
ti
ng
il
ln
e
s
s
,
m
oni
to
r
in
g
pa
ti
e
nt
s
f
r
om
a
di
s
ta
nc
e
,
r
e
duc
in
g
hos
pi
ta
l
s
ta
ys
,
in
te
gr
a
ti
ng
m
e
di
c
a
l
im
a
gi
ng,
r
e
duc
in
g
f
r
a
ud,
s
tr
e
ngt
he
ni
ng
da
t
a
s
e
c
ur
it
y,
a
nd
m
or
e
.
I
nt
r
oduc
in
g
a
br
a
nd
-
ne
w
de
e
p l
e
a
r
ni
ng
-
ba
s
e
d m
obi
le
t
r
a
f
f
ic
da
ta
c
a
te
gor
iz
a
ti
on
s
ol
ut
io
n.
T
he
pr
opos
e
d
R
F
S
E
-
G
R
U
m
ode
l
is
a
c
la
s
s
if
ic
a
ti
on
m
ode
l
th
a
t
in
c
or
por
a
te
s
f
e
a
tu
r
e
s
e
le
c
ti
on,
da
ta
ba
la
nc
in
g,
a
nd
th
e
G
R
U
a
lg
or
it
hm
[
15]
.
T
he
RF
m
e
th
od
s
e
le
c
ts
f
e
a
tu
r
e
s
a
c
c
or
di
ng
to
th
e
ir
s
ig
ni
f
ic
a
nc
e
f
or
th
e
c
la
s
s
if
ic
a
ti
on
pr
oc
e
s
s
.
A
ddi
ti
ona
ll
y,
th
e
c
om
bi
na
ti
on
of
th
e
S
M
O
T
E
ove
r
s
a
m
pl
in
g
s
tr
a
te
gy
w
it
h
th
e
E
N
N
unde
r
-
s
a
m
pl
in
g
m
e
th
od
r
e
duc
e
d
th
e
ne
ga
ti
ve
im
pa
c
t
of
da
ta
i
m
ba
la
nc
e
on
c
la
s
s
if
ic
a
ti
on
pe
r
f
or
m
a
nc
e
.
T
hi
s
pa
pe
r
pr
e
s
e
nt
s
a
r
e
de
s
ig
ne
d
K
N
N
te
c
hni
que
a
nd
c
om
pa
r
e
s
it
to
th
e
tr
a
di
ti
ona
l
K
N
N
a
lg
or
it
hm
.
W
it
hi
n
th
e
vi
c
in
it
y
of
th
e
que
r
y
in
s
ta
nc
e
,
th
e
tr
a
di
ti
ona
l
K
N
N
c
la
s
s
if
ie
r
i
s
e
m
pl
oye
d
to
do
th
e
c
la
s
s
if
ic
a
ti
on,
a
s
s
ig
ni
ng
w
e
ig
ht
s
to
e
a
c
h
c
l
a
s
s
.
T
he
te
c
hni
que
c
on
s
id
e
r
s
th
e
c
la
s
s
di
s
tr
ib
ut
io
n
s
ur
r
ounding
th
e
que
r
y
in
s
ta
nc
e
to
pr
e
ve
nt
th
e
w
e
ig
ht
a
s
s
ig
nm
e
nt
f
r
om
a
dv
e
r
s
e
ly
a
f
f
e
c
ti
ng
th
e
out
li
e
r
s
[
16]
.
T
he
c
ur
r
e
nt
s
tu
dy
a
ddr
e
s
s
e
s
th
e
s
hor
tc
om
in
gs
of
t
he
t
r
a
di
ti
ona
l
K
N
N
t
e
c
hni
que
w
it
h
l
a
r
ge
da
ta
s
e
ts
by i
nt
r
oduc
in
g a
n
i
m
pr
ove
d K
N
N
s
tr
a
te
gy
th
a
t
c
om
bi
ne
s
de
n
s
it
y
c
r
oppi
ng
a
nd
c
lu
s
te
r
de
noi
s
in
g.
T
hi
s
a
ppr
oa
c
h
us
e
s
c
lu
s
te
r
in
g
to
im
pr
ove
de
noi
s
in
g
pr
oc
e
s
s
in
g
a
nd
boos
t
th
e
c
la
s
s
if
ic
a
ti
on
e
f
f
ic
ie
nc
y
of
th
e
K
N
N
a
lg
or
it
hm
by
a
c
c
e
le
r
a
ti
ng
th
e
K
NN
s
e
a
r
c
h
w
it
hout
c
om
pr
om
is
in
g
th
e
a
lg
or
it
hm
'
s
c
la
s
s
if
ic
a
ti
on
a
c
c
ur
a
c
y.
J
ia
ng
a
nd
L
i
[
17]
pr
opos
a
l'
s
e
dge
pr
oc
e
s
s
in
g
uni
t
is
c
om
pos
e
d
of
two
p
r
im
a
r
y
c
om
pone
nt
s
:
a
da
ta
t
r
a
ns
m
is
s
io
n
uni
t
th
a
t
us
e
s
a
ppr
opr
ia
te
c
om
m
uni
c
a
ti
o
n
m
e
th
ods
to
s
e
nd
d
a
ta
to
r
a
il
w
a
y
c
ont
r
ol
c
e
nt
e
r
s
b
a
s
e
d
on
th
e
t
ype
of
da
ta
r
e
c
e
iv
e
d,
a
nd
a
da
ta
c
la
s
s
if
ic
a
ti
on
m
ode
l
th
a
t
s
e
pa
r
a
te
s
in
te
r
ne
t
of
th
in
gs
(
I
oT
)
da
ta
in
to
two
c
a
t
e
gor
ie
s
:
m
a
in
te
na
nc
e
-
c
r
it
ic
a
l
da
ta
(
M
C
D
)
a
nd
m
a
in
te
na
nc
e
-
non
-
c
r
it
ic
a
l
da
ta
(
M
N
C
D
)
.
F
or
m
ul
ti
c
la
s
s
c
la
s
s
if
i
c
a
ti
on
w
it
h
la
r
ge
da
ta
s
e
ts
,
w
e
m
a
y
r
e
duc
e
th
e
te
m
por
a
l
c
om
pl
e
xi
ty
a
nd
in
c
r
e
a
s
e
th
e
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y
of
e
ne
r
gy
ba
la
nc
e
-
r
e
la
te
d
be
ha
vi
or
s
(
E
B
R
B
s
)
by
us
in
g
a
dom
a
in
di
vi
s
io
n
-
ba
s
e
d
r
ul
e
r
e
duc
ti
on
te
c
hni
que
,
a
m
or
e
s
tr
a
ig
ht
f
or
w
a
r
d
e
vi
de
nt
ia
l
r
e
a
s
oni
ng
a
lg
or
it
hm
,
a
nd
a
m
e
th
od
to
do
a
w
a
y
w
it
h
r
ul
e
w
e
ig
ht
c
a
l
c
u
la
ti
on.
T
hi
s
is
a
M
ic
r
o
-
E
B
R
B
S
,
w
hi
c
h
i
s
a
n
E
B
R
B
S
th
a
t
ha
s
be
e
n
s
hr
unk
dow
n.
F
ur
th
e
r
m
or
e
,
A
pa
c
he
S
pa
r
k,
a
w
e
ll
-
li
ke
d
c
lu
s
te
r
c
om
put
in
g
to
ol
,
is
us
e
d
in
th
e
de
ve
lo
pm
e
nt
of
m
ic
r
o
-
E
B
R
B
S
'
s
pa
r
a
ll
e
l
r
ul
e
ge
ne
r
a
ti
on
a
nd
in
f
e
r
e
nc
e
m
e
th
odol
ogi
e
s
f
or
bi
g
da
ta
m
ul
ti
c
la
s
s
c
la
s
s
if
ic
a
ti
on
is
s
ue
s
[
18]
.
O
f
f
e
r
in
g
a
nove
l
a
ppr
oa
c
h
to
f
a
ul
t
li
ne
s
e
le
c
ti
on
th
a
t
us
e
s
bi
g
da
ta
a
nd
f
e
a
tu
r
e
c
la
s
s
if
ic
a
ti
on
to
ove
r
c
om
e
th
e
s
hor
tc
om
in
gs
of
e
xi
s
ti
ng
te
c
hni
que
s
.
T
hi
s
m
e
th
od
a
ddr
e
s
s
e
s
th
e
f
a
ul
t
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
A
nov
e
l
s
c
al
abl
e
de
e
p e
n
s
e
m
bl
e
l
e
a
r
ni
ng f
r
am
e
w
or
k
f
o
r
bi
g dat
a
…
(
K
e
s
av
an M
e
tt
ur
V
ar
adha
r
aj
an
)
1389
li
ne
s
e
le
c
ti
on
i
s
s
ue
a
s
a
c
la
s
s
if
ic
a
ti
on
ta
s
k
by
f
e
e
di
ng
la
r
ge
da
t
a
s
e
ts
a
s
s
oc
ia
te
d
w
it
h
f
a
ul
ts
in
to
th
e
c
la
s
s
if
ie
r
s
dur
in
g
tr
a
in
in
g.
T
he
f
our
m
a
in
s
te
ps
in
th
e
p
r
oc
e
dur
e
a
r
e
da
ta
ga
th
e
r
in
g,
tr
a
in
in
g,
c
la
s
s
if
ic
a
ti
on,
a
nd
a
s
s
e
s
s
m
e
nt
.
D
a
ta
pr
e
pa
r
a
ti
on
is
in
c
ha
r
ge
of
pr
e
pr
oc
e
s
s
in
g
a
nd
da
ta
c
ol
le
c
ti
on,
w
he
r
e
a
s
tr
a
in
in
g
e
m
pl
oys
pr
oc
e
s
s
e
d
da
ta
to
tr
a
in
c
la
s
s
if
ie
r
s
[
19]
,
[
20]
p
r
e
s
e
nt
s
a
m
or
e
a
dva
nc
e
d
K
N
N
te
c
hni
que
a
nd
c
om
pa
r
e
s
it
w
it
h
th
e
tr
a
di
ti
ona
l
K
N
N
a
lg
or
it
hm
.
W
it
hi
n
th
e
vi
c
in
it
y
of
th
e
que
r
y
in
s
ta
nc
e
,
th
e
tr
a
di
ti
ona
l
K
N
N
c
la
s
s
if
ie
r
is
e
m
pl
oye
d
to
do
th
e
c
la
s
s
if
ic
a
ti
on,
a
s
s
ig
ni
ng
w
e
ig
ht
s
to
e
a
c
h
c
la
s
s
.
T
he
te
c
hni
que
c
ons
id
e
r
s
th
e
c
la
s
s
di
s
tr
ib
ut
io
n
s
ur
r
ounding
th
e
que
r
y
in
s
ta
nc
e
to
pr
e
ve
nt
th
e
w
e
ig
ht
a
s
s
ig
nm
e
nt
f
r
om
a
dve
r
s
e
ly
a
f
f
e
c
ti
ng
th
e
out
li
e
r
s
.
B
y
in
c
lu
di
ng
c
lu
s
te
r
de
noi
s
in
g
a
nd
de
n
s
it
y
c
r
oppi
ng,
th
is
s
tu
dy
pr
e
s
e
nt
s
a
n
im
pr
ove
d
K
N
N
m
e
th
od
th
a
t
ove
r
c
om
e
s
th
e
s
hor
tc
om
in
gs
of
th
e
tr
a
di
ti
ona
l
K
N
N
s
tr
a
te
gy
w
he
n
w
or
ki
ng
w
it
h
la
r
ge
da
ta
s
e
ts
.
T
hi
s
a
ppr
oa
c
h e
nha
nc
e
s
de
noi
s
in
g c
a
pa
bi
li
ti
e
s
a
nd i
nc
r
e
a
s
e
s
K
N
N
a
l
gor
it
hm
e
f
f
ic
ie
nc
y by e
m
pl
oyi
ng c
lu
s
te
r
in
g t
o
s
pe
e
d up the
s
e
a
r
c
h f
or
ne
a
r
e
s
t
ne
ig
hbor
s
w
it
hout
l
os
in
g c
l
a
s
s
if
i
c
a
ti
on a
c
c
ur
a
c
y
[
21]
–
[
24]
.
3.
P
R
O
P
O
S
E
D
M
E
T
H
O
D
O
L
O
G
Y
T
he
pr
opos
e
d
s
c
a
la
bl
e
de
e
p
e
n
s
e
m
bl
e
le
a
r
ni
ng
f
r
a
m
e
w
or
k
f
or
bi
g
da
ta
c
la
s
s
if
ic
a
ti
on
(
S
D
E
L
F
-
B
D
C
)
m
e
th
odol
ogy
be
gi
ns
w
it
h
th
e
f
ir
s
t
s
te
p
w
hi
c
h
i
s
f
e
a
tu
r
e
s
e
le
c
ti
o
n,
w
he
r
e
r
e
le
va
nt
da
t
a
a
tt
r
ib
ut
e
s
a
r
e
id
e
nt
if
ie
d
us
in
g
a
H
a
doop
-
ba
s
e
d
m
a
p
-
r
e
duc
e
f
r
a
m
e
w
or
k
to
m
in
im
iz
e
f
e
a
tu
r
e
le
ngt
h.
T
he
c
hos
e
n
f
e
a
tu
r
e
s
a
r
e
th
e
n
pr
oc
e
s
s
e
d
th
r
ough
a
de
e
p
e
n
s
e
m
bl
e
m
ode
l
th
a
t
ut
il
iz
e
s
a
va
r
i
e
ty
of
de
e
p
le
a
r
ni
ng
c
la
s
s
if
ie
r
s
s
uc
h
a
s
C
N
N
,
D
B
N
,
L
S
T
M
,
e
xt
r
e
m
e
le
a
r
ni
ng
m
a
c
hi
ne
(
ELM
)
,
a
nd
de
e
p
ne
ur
a
l
ne
twor
k
(
DNN
)
.
T
he
s
e
c
la
s
s
if
ie
r
s
a
r
e
a
ppl
ie
d
c
ondi
ti
ona
ll
y,
ba
s
e
d
on
th
e
ir
s
ui
ta
bi
li
ty
to
th
e
da
ta
c
ha
r
a
c
te
r
is
ti
c
s
.
I
n
th
e
c
la
s
s
if
ic
a
ti
on
a
nd
r
e
duc
ti
on
pha
s
e
,
th
e
r
e
s
ul
ts
f
r
om
e
a
c
h
c
la
s
s
if
ie
r
a
r
e
c
om
bi
ne
d,
a
nd
a
n
opt
im
iz
a
ti
on
-
ba
s
e
d
e
ns
e
m
bl
e
a
ppr
oa
c
h
is
e
m
pl
oye
d
to
r
e
f
in
e
th
e
r
e
s
ul
ts
f
ur
th
e
r
.
T
hi
s
c
om
bi
na
ti
on
a
im
s
to
ge
ne
r
a
te
a
de
c
is
iv
e
s
tr
a
te
gy
th
a
t
m
a
xi
m
iz
e
s
th
e
pr
e
di
c
ti
on
m
e
tr
ic
s
w
it
hi
n
th
e
m
a
p
-
r
e
duc
e
f
r
a
m
e
w
or
k.
T
h
e
f
in
a
l
out
put
is
a
r
obus
t
pr
e
di
c
ti
on
m
ode
l.
F
ig
ur
e
1 s
how
s
t
he
bl
oc
k di
a
gr
a
m
.
3.1.
F
e
at
u
r
e
s
e
le
c
t
io
n
T
he
pr
opos
e
d H
a
doop
-
ba
s
e
d m
a
p r
e
duc
ti
on mode
l
c
hoos
e
s
t
he
r
e
le
va
nt
f
e
a
tu
r
e
s
f
r
om
t
he
m
a
p pha
s
e
w
hi
c
h
is
ge
ne
r
a
te
d
by
th
e
pr
opos
e
d
m
ode
l
to
m
in
im
iz
e
e
a
c
h
f
e
a
tu
r
e
'
s
le
ngt
h
w
hi
c
h
ge
ts
m
a
ny
e
f
f
e
c
ti
ve
f
e
a
tu
r
e
s
.
T
he
r
e
le
va
nt
f
e
a
tu
r
e
s
a
r
e
c
hos
e
n
us
in
g
th
e
pr
opos
e
d
a
lg
or
it
hm
th
a
t
m
in
im
iz
e
s
th
e
c
om
put
a
ti
on
ti
m
e
a
nd
ove
r
f
it
ti
ng
is
s
ue
.
T
hi
s
is
e
s
s
e
nt
ia
l
to
e
nha
nc
e
th
e
a
c
c
ur
a
c
y
,
w
it
h
a
dde
d
r
e
le
va
nt
f
e
a
tu
r
e
s
us
e
d
to
r
e
duc
e
th
e
num
be
r
of
in
put
va
r
ia
bl
e
s
th
a
t
e
li
m
in
a
te
non
-
r
e
le
va
nt
f
e
a
tu
r
e
s
.
T
he
s
e
f
e
a
tu
r
e
s
a
r
e
s
e
le
c
te
d
f
r
om
two
di
f
f
e
r
e
nt
da
ta
s
e
ts
.
T
hu
s
th
e
opt
im
a
l
f
e
a
tu
r
e
s
f
or
th
is
m
ode
l
a
r
e
e
xpr
e
s
s
e
d
a
s
h
u
f
in
e
−
tu
n
e
,
w
he
r
e
in
u
=
1
,
2
,
…
…
…
…
.
.
,
U
us
in
g t
he
pr
opos
e
d a
lg
or
it
hm
.
3.2.
P
r
op
os
e
d
al
gor
it
h
m
T
he
pr
opos
e
d
he
ur
is
ti
c
opt
im
iz
a
ti
on
m
e
th
od,
w
hi
c
h
is
ge
n
e
r
a
te
d
by
th
e
m
a
p
r
e
duc
ti
on
f
r
a
m
e
w
or
k
-
ba
s
e
d c
la
s
s
if
ic
a
ti
on s
y
s
te
m
, a
dj
us
t
s
s
e
ve
r
a
l
p
a
r
a
m
e
te
r
s
t
o ge
t
th
e
be
s
t
-
pr
oj
e
c
te
d r
e
s
ul
ts
.
T
he
r
e
a
s
on t
hi
s
m
ode
l
c
hoos
e
s
a
nove
l
opt
im
iz
a
ti
on a
lg
or
it
hm
is
th
a
t
it
c
a
n
gua
r
a
nt
e
e
qui
c
k
c
onve
r
ge
nc
e
a
nd
pr
e
ve
nt
pr
obl
e
m
s
w
it
h
lo
c
a
l
opt
im
a
.
H
ow
e
ve
r
,
it
s
in
a
bi
li
ty
to
ge
t
th
e
be
s
t
out
c
om
e
s
w
or
ld
w
id
e
r
e
nde
r
s
it
in
e
f
f
e
c
ti
ve
f
or
a
va
r
ie
ty
o
f
opt
im
iz
a
ti
on
ta
s
ks
.
I
t
ha
s
be
e
n
c
oupl
e
d
w
it
h
a
n
opt
im
iz
a
ti
on
a
ppr
oa
c
h
to
ove
r
c
om
e
th
e
s
hor
tc
om
in
gs
of
th
e
pr
e
s
e
nt
opt
im
iz
a
ti
on
a
ppr
oa
c
h
be
c
a
us
e
of
it
s
c
a
pa
c
it
y
to
boos
t
pe
r
f
or
m
a
nc
e
w
he
n
a
ddr
e
s
s
in
g
e
ngi
ne
e
r
in
g
opt
im
iz
a
ti
on
c
ha
ll
e
nge
s
a
nd
to
in
c
r
e
a
s
e
e
f
f
ic
ie
nc
y
in
gl
ob
a
l
s
e
a
r
c
h.
T
h
e
pr
opos
e
d
a
lg
or
it
hm
in
c
r
e
a
s
e
s
th
e
c
la
s
s
if
ic
a
ti
on
u
s
in
g
e
f
f
ic
ie
nc
y
by
ut
il
iz
in
g
th
e
m
a
p
-
r
e
duc
e
a
r
c
hi
te
c
tu
r
e
.
T
he
pa
r
a
m
e
te
r
s
R
u1
a
nd
R
u2
in
th
e
pr
opos
e
d
a
lg
or
it
hm
a
r
e
c
om
put
e
d
us
in
g
th
e
de
vi
a
ti
on
of
th
e
op
ti
m
iz
a
ti
on
a
ppr
oa
c
h
a
nd
a
not
he
r
opt
im
iz
a
ti
on
a
ppr
oa
c
h.
T
he
out
c
om
e
of
th
e
de
vi
a
ti
on
is
s
how
n
in
(
1)
.
T
he
SD
f
or
R
u1
a
n
d
R
u2
a
r
e
e
va
lu
a
te
d
to
be
SD
(
R
u1
,
R
u2
)
, f
in
a
ll
y t
he
upda
te
d va
lu
e
i
s
de
not
e
d by
D
a
s
s
how
n i
n (
2)
.
=
(
1
,
2
)
+
(
1
,
2
)
(
1)
=
+
(
2)
H
e
r
e
F
in
a
l
o
ut
de
not
e
s
th
e
out
c
om
e
of
th
e
s
ol
ut
io
n.
T
hi
s
m
e
th
od
ut
il
iz
e
s
pa
r
a
m
e
t
e
r
s
to
d
e
f
in
e
a
s
e
t
of
c
oyot
e
pa
c
ks
,
or
Re
il
.
E
a
c
h
pa
c
k
c
on
s
is
ts
of
uni
que
c
oyot
e
s
,
EA
am
,
s
how
c
a
s
in
g
s
pe
c
if
ic
s
o
c
ia
l
be
h
a
vi
or
s
w
it
hi
n
th
e
ir
gr
oup,
il
,
ove
r
a
s
pe
c
if
ic
ti
m
e
f
r
a
m
e
,
Vm
.
A
n
a
lp
ha
c
o
yot
e
,
known
f
or
i
ts
s
upe
r
io
r
a
da
pt
a
bi
li
ty
a
nd
be
ha
vi
or
th
a
t
s
ui
ts
c
ha
ngi
ng
e
nvi
r
onm
e
nt
a
l
c
ondi
ti
ons
,
a
s
s
um
e
s
le
a
de
r
s
hi
p
in
e
ve
r
y
pa
c
k.
In
(
3)
s
um
m
a
r
iz
e
s
th
e
le
a
de
r
s
hi
p c
ha
r
a
c
te
r
is
ti
c
s
of
th
e
a
lp
ha
c
oyot
e
.
T
he
in
f
or
m
a
ti
on
r
e
le
va
nt
to
th
e
c
oyot
e
s
g
a
in
e
d
f
r
om
th
e
gr
oups
f
o
r
pe
r
f
or
m
in
g
c
ul
tu
r
a
l
te
nde
nc
y
is
s
how
n
in
(
4)
.
H
e
r
e
S
e
il
,
Vm
de
pi
c
ts
th
e
r
a
nke
d
s
oc
ia
l
s
ta
tu
s
of
th
e
c
oyot
e
in
vol
ve
d
th
r
ough
th
e
s
e
a
r
c
h
di
m
e
n
s
io
n
km
w
it
hi
n
th
e
pa
c
k
il
in
th
e
pe
r
io
d
Vm
.
T
he
bi
r
th
r
a
te
i
s
e
va
lu
a
te
d a
nd t
e
r
m
e
d a
s
a
l
if
e
e
ve
nt
f
or
t
he
ne
w
c
oyot
e
d
e
not
e
d by the
(
5)
.
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.
2
,
A
pr
il
2025
:
1386
-
1400
1390
α
EA
il
,
Vm
=
Bs
am
il
,
Vm
f
or
m
in
Hs
am
il
,
Vm
(
3)
t
km
il
,
Vm
=
{
S
e
EA
am
+
1
2
il
,
Vm
,
k
EA
am
is
odd
Se
EA
am
+
1
2
il
,
Vm
,
km
o
r
e
l
s
e
}
(
4)
t
km
il
,
Vm
=
{
B
t
tp
1
j
l
tt
jl
<
D
uw
or
km
=
k
m
1
B
t
t
p
2
j
l
tt
jl
<
D
uw
or
km
=
k
m
2
il
,
Vm
tf
jl
e
l
s
e
il
,
Vm
}
(
5)
T
he
de
s
ig
n
di
m
e
ns
io
n
de
not
e
d
by
k
m
1
or
k
m
2
.
T
he
te
r
m
D
uv
is
e
va
lu
a
te
d
to
s
c
a
tt
e
r
pr
oba
bi
li
ti
e
s
a
nd
D
cr
is
in
di
c
a
te
d
a
s
th
e
a
s
s
oc
ia
ti
on
of
pr
oba
bi
li
ti
e
s
.
T
he
r
a
ndom
va
r
ia
bl
e
s
de
not
e
d
by
tt
jl
a
nd
tf
jl
in
th
e
r
a
nge
[
0,
1]
.
T
he
f
in
a
l
s
ta
tu
s
is
de
te
r
m
in
e
d
by
c
or
r
e
la
ti
ng
th
e
pr
io
r
a
nd
upgr
a
de
d
s
ta
tu
s
a
s
s
how
n
in
(
6)
.
T
he
opt
im
a
l
s
ol
ut
io
n
is
f
in
a
li
z
e
d
th
r
ough
th
e
c
ondi
ti
on
w
hi
c
h
is
in
c
or
por
a
te
d
by
th
is
s
e
a
r
c
hi
ng
di
m
e
ns
io
n.
A
c
c
or
di
ng
to
th
e
hunt
in
g
m
e
c
ha
ni
s
m
th
e
pr
e
y
e
n
c
lo
s
e
d
by
th
e
a
lp
ha
pa
r
a
m
e
te
r
s
by
d
e
te
r
m
in
in
g
th
e
ir
pos
it
io
n
w
it
h t
he
a
s
s
is
t
a
nc
e
of
a
s
p
e
c
if
ic
a
ge
nt
, t
he
e
n
c
ir
c
li
ng s
tr
a
te
gy i
s
gi
ve
n by (
7)
.
Bt
am
il
,
Vm
+
1
=
{
Bt
am
il
,
Vm
+
1
p
y
H
t
am
il
,
Vm
+
1
<
H
t
il
,
Vm
Bt
am
il
,
Vm
e
l
s
e
}
(
6)
ms
tr
(
ws
+
1
)
=
μ
1
∑
[
τ
q
(
ws
)
−
ms
tr
(
ws
)
]
ps
ps
0
=
1
−
ms
∗
(
ws
)
(
7)
T
he
a
r
bi
tr
a
r
y
num
be
r
is
de
not
e
d
by
μ
1
in
th
e
r
a
nge
[
-
2
,
2]
th
e
be
s
t
s
ol
ut
io
n
obs
e
r
ve
d
f
r
om
th
e
pr
e
vi
ous
s
te
p
i
s
de
not
e
d
by
ms
∗
(
ws
)
w
he
r
e
a
s
th
e
pr
e
s
e
nt
s
ol
ut
io
n
is
ms
tr
a
nd
th
e
popula
ti
on
is
d
e
not
e
d
by
W
s
.
T
he
s
ubs
e
t
of
th
e
s
ol
ut
io
n
is
de
not
e
d
by
τ
q
(
ws
)
,
th
e
ne
w
pos
it
io
n
w
it
hi
n
th
e
s
ol
ut
io
n
is
de
not
e
d
by
ms
tr
(
ws
+
1
)
w
he
r
e
a
s
th
e
r
a
ndom
va
r
ia
bl
e
is
de
not
e
d
by
ps
,
th
e
y
c
ha
s
e
th
e
pr
e
y
by
tr
a
c
ki
ng
th
e
pos
it
io
n
th
e
y
a
r
e
w
it
hi
n,
th
is
f
e
a
tu
r
e
is
d
e
not
e
d
in
(
8)
.
T
he
s
e
a
r
c
h
a
ge
nt
is
r
e
f
e
r
r
e
d
to
a
s
th
e
ms
g
(
ws
)
,
ms
tr
(
ws
+
1
)
de
pi
c
ts
th
e
m
ove
m
e
nt
a
nd
th
e
a
r
bi
tr
a
r
y
num
be
r
de
not
e
d
by
μ
1
in
th
e
r
a
nge
[
-
1
,
1]
.
T
hi
s
c
la
s
s
if
ic
a
ti
on
is
obs
e
r
ve
d a
s
i
n (
9)
w
it
h t
he
s
ur
vi
va
l
r
a
te
t
he
c
om
put
a
ti
on i
s
pe
r
f
or
m
e
d i
n (
10)
.
ms
tr
(
ws
+
1
)
=
ms
∗
(
ws
)
+
μ
1
∗
zs
μ
2
∗
(
ms
g
(
ws
)
−
ms
tr
(
ws
)
)
(
8)
ms
tr
(
ws
+
1
)
=
0
.
5
[
as
μ
2
∗
ms
g
(
ws
)
−
(
−
1
)
γ
∗
ms
tr
(
ws
)
]
(
9)
ut
(
tc
)
=
f
m
a
x
−
f
(
t
)
f
m
a
x
−
f
m
i
n
(
10)
T
he
f
(
t
)
de
s
ig
na
te
s
th
e
c
ons
i
s
te
nt
va
lu
e
of
t
vj
s
e
a
r
c
h
a
ge
nt
.
Ms
∗
(
ws
)
i
s
de
pi
c
te
d
by
th
e
be
s
t
s
ol
ut
io
n
th
r
ough
th
e
pr
e
vi
ous
it
e
r
a
ti
on.
T
he
va
r
ia
bl
e
s
f
ma
x
a
nd
f
min
w
hi
c
h
r
e
pr
e
s
e
nt
s
th
e
w
or
s
t
-
c
a
s
e
opt
im
a
l
f
it
ne
s
s
va
lu
e
,
w
h
e
r
e
a
s
th
e
opt
im
a
l
f
it
ne
s
s
va
lu
e
s
a
nd
th
e
p
a
r
a
m
e
te
r
γ
a
r
e
r
e
pr
e
s
e
nt
e
d
by
th
e
bi
na
r
y
num
be
r
.
T
he
a
lg
or
it
hm
i
s
gi
ve
n a
s
s
how
n i
n
A
lg
or
it
hm
1
. F
ig
ur
e
2 s
how
s
t
he
pr
opos
e
d a
lg
or
it
hm
f
lo
w
c
ha
r
t.
A
lg
or
it
hm
1. P
r
opos
e
d a
lg
or
it
hm
Step 1
I
n
i
t
i
a
l
i
z
a
t
i
o
n
o
f
p
o
p
u
l
a
t
i
o
n
Step 2
C
o
m
p
u
t
e
f
i
t
n
e
s
s
f
o
r
a
l
l
s
o
l
u
t
i
o
n
s
Step 3
I
f
t
e
r
m
i
n
a
t
i
o
n
i
s
n
o
t
m
e
t
,
t
h
e
n
:
-
F
o
r
e
v
e
r
y
s
o
l
u
t
i
o
n
i
n
t
h
e
p
o
p
u
l
a
t
i
o
n
:
•
U
p
d
a
t
e
t
h
e
p
o
s
i
t
i
o
n
f
o
l
l
o
w
i
n
g
t
h
e
p
r
o
c
e
d
u
r
e
-
E
v
a
l
u
a
t
e
t
h
e
R
u
1
•
U
p
d
a
t
e
t
h
e
p
o
s
i
t
i
o
n
a
c
c
o
r
d
i
n
g
t
o
t
h
e
p
r
o
c
e
d
u
r
e
-
E
v
a
l
u
a
t
e
t
h
e
R
u
2
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
A
nov
e
l
s
c
al
abl
e
de
e
p e
n
s
e
m
bl
e
l
e
a
r
ni
ng f
r
am
e
w
or
k
f
o
r
bi
g dat
a
…
(
K
e
s
av
an M
e
tt
ur
V
ar
adha
r
aj
an
)
1391
-
C
a
l
c
u
l
a
t
e
t
h
e
f
i
n
a
l
d
e
v
i
a
t
i
o
n
u
s
i
n
g
(
1
)
-
C
o
n
t
i
n
u
e
e
x
e
c
u
t
i
o
n
Else
-
U
p
g
r
a
d
e
f
i
n
a
l
p
o
s
i
t
i
o
n
b
y
(
2
)
-
I
m
p
r
o
v
i
s
a
t
i
o
n
R
e
t
u
r
n
t
o
s
t
e
p
3
t
o
c
h
e
c
k
f
o
r
t
e
r
m
i
n
a
t
i
o
n
a
g
a
i
n
Step 4
Else:
-
O
b
t
a
i
n
t
h
e
b
e
s
t
o
p
t
i
m
a
l
s
o
l
u
t
i
o
n
a
n
d
t
e
r
m
i
n
a
t
e
t
h
e
a
l
g
o
r
i
t
h
m
F
ig
ur
e
1. B
lo
c
k di
a
gr
a
m
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.
2
,
A
pr
il
2025
:
1386
-
1400
1392
F
ig
ur
e
2. P
r
opos
e
d a
lg
or
it
hm
f
lo
w
c
ha
r
t
3.3.
D
e
e
p
e
n
s
e
m
b
le
m
od
e
l
T
he
de
e
p
e
ns
e
m
bl
e
m
o
de
l
f
or
th
e
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
ha
r
ne
s
s
e
s
th
e
pow
e
r
of
m
ul
ti
pl
e
de
e
p
le
a
r
ni
ng
c
la
s
s
if
ie
r
s
to
pr
e
di
c
t
phys
ic
a
l
a
c
ti
vi
ti
e
s
ba
s
e
d
on
s
e
ns
or
da
ta
,
s
uc
h
a
s
a
c
c
e
le
r
om
e
te
r
r
e
a
di
ngs
.
T
hi
s
a
ppr
oa
c
h
is
d
e
s
ig
ne
d
to
im
pr
ove
pr
e
di
c
ti
on
a
c
c
ur
a
c
y
by
le
ve
r
a
gi
ng
th
e
c
ol
le
c
ti
ve
s
tr
e
ngt
hs
a
nd
c
om
pe
ns
a
ti
ng
f
or
th
e
w
e
a
kne
s
s
e
s
of
in
di
vi
dua
l
m
ode
ls
.
T
he
pr
oc
e
s
s
be
gi
ns
w
it
h
th
e
pr
e
-
pr
oc
e
s
s
in
g
of
in
put
da
ta
,
e
n
s
ur
in
g
it
is
nor
m
a
li
z
e
d
a
nd
pa
r
ti
ti
one
d
in
to
tr
a
in
in
g,
va
li
da
ti
on,
a
nd
te
s
t
s
e
ts
to
f
a
c
il
it
a
te
e
f
f
e
c
ti
ve
m
ode
l
tr
a
in
in
g a
nd e
va
lu
a
ti
on.
T
he
c
or
e
of
th
e
a
lg
or
it
hm
li
e
s
in
it
s
c
ondi
ti
ona
l
a
ppl
ic
a
ti
on
o
f
di
f
f
e
r
e
nt
de
e
p
le
a
r
ni
ng
te
c
hni
qu
e
s
,
e
a
c
h
s
e
le
c
t
e
d
ba
s
e
d
on
th
e
s
pe
c
if
ic
c
ha
r
a
c
te
r
is
ti
c
s
of
th
e
da
ta
a
t
ha
nd.
F
or
da
ta
s
e
ts
w
he
r
e
s
pa
ti
a
l
pa
tt
e
r
n
r
e
c
ogni
ti
on
is
pa
r
a
m
ount
,
th
e
C
N
N
is
e
m
pl
oye
d.
C
N
N
s
a
r
e
a
de
pt
a
t
e
xt
r
a
c
ti
ng
hi
e
r
a
r
c
hi
c
a
l
s
pa
ti
a
l
f
e
a
tu
r
e
s
f
r
om
da
ta
,
m
a
ki
ng
th
e
m
id
e
a
l
f
or
a
na
ly
z
in
g
im
a
ge
s
or
s
e
ns
or
da
ta
th
a
t
ha
ve
a
s
pa
ti
a
l
s
tr
uc
tu
r
e
.
O
n
th
e
ot
he
r
ha
nd,
w
he
n
th
e
d
a
ta
in
vol
ve
s
te
m
por
a
l
de
pe
nde
n
c
ie
s
or
s
e
que
nc
e
s
w
he
r
e
th
e
or
de
r
of
da
ta
poi
nt
s
is
c
r
uc
ia
l,
th
e
L
S
T
M
ne
twor
k
is
ut
il
iz
e
d.
L
S
T
M
s
e
xc
e
l
in
r
e
m
e
m
be
r
in
g
in
f
or
m
a
ti
on
ove
r
e
xt
e
nde
d
pe
r
io
ds
,
m
a
ki
ng
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
A
nov
e
l
s
c
al
abl
e
de
e
p e
n
s
e
m
bl
e
l
e
a
r
ni
ng f
r
am
e
w
or
k
f
o
r
bi
g dat
a
…
(
K
e
s
av
an M
e
tt
ur
V
ar
adha
r
aj
an
)
1393
th
e
m
s
ui
te
d
f
or
ti
m
e
-
s
e
r
ie
s
da
ta
or
s
e
qu
e
nc
e
s
w
he
r
e
pa
s
t
in
f
or
m
a
ti
on
in
f
lu
e
nc
e
s
f
ut
ur
e
pr
e
di
c
ti
ons
.
F
or
s
c
e
na
r
io
s
de
m
a
ndi
ng
r
a
pi
d
tr
a
in
in
g
a
nd
s
tr
a
ig
ht
f
or
w
a
r
d
pr
e
di
c
ti
ons
w
it
hout
th
e
ne
e
d
f
or
c
a
pt
ur
in
g
c
om
pl
e
x
r
e
la
ti
ons
hi
ps
or
ti
m
e
d
e
pe
nde
nc
ie
s
,
th
e
E
L
M
of
f
e
r
s
a
f
a
s
t
a
nd
e
f
f
ic
ie
nt
s
ol
ut
io
n.
E
L
M
s
a
r
e
s
in
gl
e
-
la
ye
r
f
e
e
df
or
w
a
r
d
ne
twor
ks
th
a
t
c
a
n
a
c
hi
e
ve
hi
gh
-
s
pe
e
d
tr
a
in
in
g
a
nd
ge
ne
r
a
li
z
a
ti
on
pe
r
f
or
m
a
nc
e
,
m
a
ki
ng t
he
m
s
ui
ta
bl
e
f
or
s
im
pl
e
r
t
a
s
ks
or
w
he
n c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y i
s
a
pr
io
r
it
y.
T
he
e
ns
e
m
bl
e
m
ode
l'
s
s
tr
a
te
gy
in
vol
ve
s
e
va
lu
a
ti
ng
th
e
pe
r
f
or
m
a
nc
e
of
e
a
c
h
c
la
s
s
if
ie
r
on
th
e
va
li
da
ti
on
s
e
t
a
nd
th
e
n
c
om
bi
ni
ng
th
e
s
e
m
ode
ls
in
a
w
a
y
th
a
t
le
ve
r
a
ge
s
th
e
ir
s
tr
e
ngt
hs
.
I
f
th
e
r
e
'
s
c
ons
e
ns
us
a
m
ong
th
e
m
ode
l
s
,
th
e
a
lg
or
it
hm
m
ig
ht
s
im
pl
y
t
a
ke
th
e
m
a
jo
r
it
y
vot
e
a
s
th
e
f
in
a
l
pr
e
di
c
ti
on.
H
ow
e
ve
r
,
in
c
a
s
e
s
of
di
s
a
gr
e
e
m
e
nt
,
th
e
a
lg
or
it
hm
opt
s
f
or
a
m
or
e
nua
nc
e
d
a
ppr
oa
c
h
by
w
e
ig
ht
in
g
th
e
out
pu
ts
of
e
a
c
h
m
ode
l
ba
s
e
d
on
th
e
ir
va
li
da
ti
on
a
c
c
ur
a
c
y
a
nd
c
om
put
in
g
a
w
e
i
ght
e
d
a
ve
r
a
ge
pr
e
di
c
ti
on. T
hi
s
m
e
th
od
e
ns
ur
e
s
th
a
t
m
or
e
a
c
c
ur
a
te
m
ode
l
s
ha
ve
a
gr
e
a
te
r
in
f
lu
e
nc
e
on
th
e
f
in
a
l
pr
e
di
c
ti
on,
e
nha
nc
in
g
th
e
ov
e
r
a
ll
pr
e
di
c
ti
on
pe
r
f
or
m
a
nc
e
of
th
e
e
n
s
e
m
bl
e
.
A
f
te
r
s
e
le
c
ti
ng
th
e
c
om
bi
na
ti
on
s
tr
a
te
gy
ba
s
e
d
on
m
ode
l
a
gr
e
e
m
e
nt
a
nd
pe
r
f
or
m
a
nc
e
w
e
ig
ht
in
g,
th
e
e
n
s
e
m
bl
e
m
ode
l
is
tr
a
in
e
d
on
th
e
e
nt
ir
e
tr
a
in
in
g
s
e
t.
T
hi
s
in
te
gr
a
te
d
tr
a
in
in
g
a
ppr
oa
c
h
a
ll
ow
s
th
e
e
ns
e
m
bl
e
to
l
e
a
r
n
f
r
om
th
e
c
om
pr
e
he
ns
iv
e
da
ta
a
va
il
a
bl
e
,
e
ns
ur
in
g
it
is
w
e
ll
-
a
dj
us
te
d
to
m
a
ke
a
c
c
ur
a
te
pr
e
di
c
ti
ons
.
T
he
f
in
a
l
s
te
p
in
vol
ve
s
e
va
lu
a
ti
ng
t
he
e
ns
e
m
bl
e
m
ode
l'
s
pe
r
f
or
m
a
nc
e
on
th
e
te
s
t
s
e
t,
pr
ovi
di
ng
a
n
a
s
s
e
s
s
m
e
nt
of
it
s
a
bi
li
ty
to
ge
ne
r
a
li
z
e
to
uns
e
e
n
da
ta
.
F
or
ne
w
in
put
s
,
th
e
m
ode
l
pr
e
di
c
ts
da
ta
c
la
s
s
e
s
by
pr
e
pr
oc
e
s
s
in
g
th
e
da
ta
to
m
a
tc
h
th
e
tr
a
in
in
g
f
or
m
a
t
a
nd
th
e
n
a
ppl
yi
ng
th
e
t
r
a
in
e
d
e
ns
e
m
bl
e
m
ode
l
to
ge
ne
r
a
te
pr
e
di
c
ti
ons
.
T
hi
s
a
lg
or
it
hm
r
e
pr
e
s
e
nt
s
a
n
e
f
f
ic
ie
nt
a
ppr
oa
c
h
to
pr
e
di
c
ti
ve
m
ode
li
ng
in
th
e
c
ont
e
xt
of
da
ta
c
la
s
s
if
ic
a
ti
on,
l
e
ve
r
a
gi
ng
th
e
uni
que
c
a
pa
bi
li
ti
e
s
of
di
f
f
e
r
e
nt
de
e
p
le
a
r
ni
ng
m
ode
ls
to
c
r
e
a
te
a
r
obu
s
t,
a
c
c
ur
a
te
,
a
nd
e
f
f
ic
ie
nt
to
ol
f
or
da
ta
c
la
s
s
pr
e
di
c
ti
on.
B
y
c
om
bi
ni
ng
th
e
pr
e
di
c
ti
ons
f
r
om
va
r
io
us
m
ode
ls
,
th
e
e
ns
e
m
bl
e
m
e
th
od
a
c
hi
e
ve
s
a
ba
la
nc
e
be
twe
e
n
th
e
de
pt
h
of
le
a
r
ni
ng
a
nd
c
om
put
a
ti
ona
l
e
f
f
ic
ie
nc
y,
ul
ti
m
a
te
ly
e
nha
nc
in
g
th
e
r
e
li
a
bi
li
ty
a
nd
a
c
c
ur
a
c
y
of
da
ta
c
la
s
s
pr
e
di
c
ti
ons
.
A
lg
or
it
hm
2
s
how
s
th
e
de
e
p
e
ns
e
m
bl
e
a
lg
or
it
hm
. F
ig
ur
e
3
de
e
p e
ns
e
m
bl
e
f
lo
w
c
ha
r
t
f
or
pr
e
di
c
ti
on.
A
lg
or
it
hm
2. D
e
e
p e
ns
e
m
bl
e
a
lg
or
it
hm
Input
D
a
t
a
s
e
t
Step 1
Pre
-
p
r
o
c
e
s
s
i
n
g
:
-
N
o
r
m
a
l
i
z
e
t
h
e
d
a
t
a
t
o
e
n
s
u
r
e
u
n
i
f
o
r
m
i
t
y
a
n
d
b
e
t
t
e
r
m
o
d
e
l
t
r
a
i
n
i
n
g
Step 2
I
n
i
t
i
a
l
i
z
a
t
i
o
n
:
-
CN
N
_
m
od
e
l
=
T
r
a
i
n
a
c
o
n
v
o
l
u
t
i
o
n
a
l
n
e
u
r
a
l
n
e
t
w
o
r
k
o
n
t
h
e
d
a
t
a
.
-
D
B
N
_
m
od
e
l
=
T
r
a
i
n
a
d
e
e
p
b
e
l
i
e
f
n
e
t
w
o
r
k
o
n
t
h
e
d
a
t
a
.
-
L
S
TM
_
m
od
e
l
=
T
r
a
i
n
a
l
o
n
g
s
h
o
r
t
-
t
e
r
m
m
e
m
o
r
y
n
e
t
w
o
r
k
o
n
t
h
e
d
a
t
a
.
-
E
L
M
_
m
od
e
l
=
T
r
a
i
n
a
n
e
x
t
r
e
m
e
l
e
a
r
n
i
n
g
m
a
c
h
i
n
e
o
n
t
h
e
d
a
t
a
.
-
D
N
N
_
m
od
e
l
=
T
r
a
i
n
a
d
e
e
p
n
e
u
r
a
l
n
e
t
w
o
r
k
o
n
t
h
e
d
a
t
a
.
Step 3
D
e
f
i
n
e
t
h
e
e
n
s
e
m
b
l
e
p
r
e
d
i
c
t
i
o
n
t
e
c
h
n
i
q
u
e
:
-
C
o
l
l
e
c
t
p
r
e
d
i
c
t
i
o
n
s
f
r
o
m
e
a
c
h
m
o
d
e
l
.
-
F
o
r
e
a
c
h
i
n
s
t
a
n
c
e
i
n
t
h
e
t
e
s
t
d
a
t
a
,
d
o
t
h
e
f
o
l
l
o
w
i
n
g
:
•
if
CN
N
_
m
od
e
l
confiden
ce >
thres
hold,
use
CN
N
_
m
od
e
l
predicti
on.
•
e
l
s
e
i
f
D
B
N
_
m
od
e
l
confiden
ce >
thres
hold
, use
D
B
N
_
m
od
e
l
predicti
on.
•
e
l
s
e
i
f
L
S
TM
_
m
od
e
l
confiden
ce >
thres
hold
, use
L
S
TM
_
m
od
e
l
predicti
on.
•
e
l
s
e
i
f
E
L
M
_
m
od
e
l
confiden
ce >
thres
hold
, use
E
L
M
_
m
od
e
l
predicti
on.
-
e
l
s
e
,
u
s
e
D
N
N
_
m
od
e
l
p
r
e
d
i
c
t
i
o
n
a
s
a
d
e
f
a
u
l
t
.
Step 4
E
v
a
l
u
a
t
e
t
h
e
p
e
r
f
o
r
m
a
n
c
e
o
f
e
a
c
h
m
o
d
e
l
o
n
t
h
e
v
a
l
i
d
a
t
i
o
n
s
e
t
.
Step 5
C
o
m
b
i
n
e
p
r
e
d
i
c
t
i
o
n
s
f
o
r
t
h
e
f
i
n
a
l
d
e
c
i
s
i
o
n
:
-
i
f
t
h
e
m
o
d
e
l
s
a
g
r
e
e
o
n
t
h
e
p
r
e
d
i
c
t
i
o
n
:
•
A
c
c
e
p
t
t
h
e
m
a
j
o
r
i
t
y
v
o
t
e
a
s
t
h
e
f
i
n
a
l
p
r
e
d
i
c
t
i
o
n
.
-
else
•
W
e
i
g
h
t
t
h
e
o
u
t
p
u
t
s
b
a
s
e
d
o
n
t
h
e
i
r
v
a
l
i
d
a
t
i
o
n
a
c
c
u
r
a
c
y
,
a
n
d
c
o
m
p
u
t
e
a
w
e
i
g
h
t
e
d
a
v
e
r
a
g
e
p
r
e
d
i
c
t
i
o
n
.
Step 6
T
r
a
i
n
t
h
e
e
n
s
e
m
b
l
e
m
o
d
e
l
o
n
t
h
e
e
n
t
i
r
e
t
r
a
i
n
i
n
g
s
e
t
u
s
i
n
g
t
h
e
s
e
l
e
c
t
e
d
s
t
r
a
t
e
g
y
f
r
o
m
s
t
e
p
5
.
Step 7
E
v
a
l
u
a
t
e
t
h
e
e
n
s
e
m
b
l
e
m
o
d
e
l
o
n
t
h
e
t
e
s
t
s
e
t
t
o
a
s
s
e
s
s
i
t
s
p
r
e
d
i
c
t
i
o
n
p
e
r
f
o
r
m
a
n
c
e
.
Step 8
f
o
r
n
e
w
s
e
n
s
o
r
d
a
t
a
i
n
p
u
t
s
:
-
Pre
-
p
r
o
c
e
s
s
t
h
e
d
a
t
a
a
s
i
n
s
t
e
p
1
.
•
P
r
e
d
i
c
t
t
h
e
c
l
a
s
s
e
s
u
s
i
n
g
t
h
e
t
r
a
i
n
e
d
e
n
s
e
m
b
l
e
m
o
d
e
l
.
output
T
h
e
p
r
e
d
i
c
t
e
d
c
l
a
s
s
.
3.4.
C
la
s
s
if
i
c
at
io
n
an
d
r
e
d
u
c
t
io
n
p
h
as
e
T
h
e
pr
o
po
s
e
d
f
r
a
m
e
w
or
k
f
or
m
o
ni
t
or
in
g
d
e
v
e
lo
ps
a
n
o
pt
im
iz
a
t
io
n
-
ba
s
e
d
e
n
s
e
m
bl
e
a
p
pr
o
a
c
h
f
or
th
e
pr
e
di
c
ti
o
n
of
ph
y
s
i
c
a
l
a
c
ti
vi
ti
e
s
th
r
oug
h
th
e
op
ti
m
a
l
f
e
a
t
ur
e
s
kno
w
n
a
s
h
u
f
in
e
−
tu
n
e
f
r
om
t
he
pr
op
os
e
d
a
l
gor
i
th
m
.
T
h
e
pr
o
po
s
e
d
a
l
gor
it
h
m
i
nc
or
p
or
a
te
s
a
ll
th
i
s
in
t
he
c
om
b
in
i
ng
ph
a
s
e
.
T
he
d
e
v
e
l
op
e
d
f
r
a
m
e
w
or
k
i
s
u
s
e
d
to
m
a
x
im
i
z
e
t
h
e
m
e
tr
ic
s
in
th
e
m
a
p
-
r
e
d
uc
e
f
r
a
m
e
w
or
k
f
or
pr
e
di
c
ti
o
n.
T
h
e
pr
op
o
s
e
d m
a
p
-
r
e
d
uc
e
s
ys
te
m
a
ggr
e
g
a
t
e
s
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.
2
,
A
pr
il
2025
:
1386
-
1400
1394
r
e
s
ul
ts
f
r
o
m
t
he
c
om
bi
n
in
g
s
t
a
g
e
t
o
b
e
pr
o
c
e
s
s
e
d
i
n
t
he
r
e
du
c
ti
on
s
t
a
g
e
,
w
he
r
e
a
ll
out
c
o
m
e
s
a
r
e
c
o
m
bi
ne
d
to
ge
n
e
r
a
t
e
a
de
c
i
s
i
ve
s
tr
a
t
e
gy
.
T
h
i
s
f
r
a
m
e
w
or
k
i
s
c
r
u
c
i
a
l
f
or
bi
g d
a
t
a
c
l
a
s
s
if
i
c
a
ti
o
n
w
it
h d
if
f
e
r
e
nt
a
p
pl
i
c
a
ti
on
.
F
ig
ur
e
3. D
e
e
p e
ns
e
m
bl
e
f
lo
w
c
ha
r
t
f
or
pr
e
di
c
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
A
nov
e
l
s
c
al
abl
e
de
e
p e
n
s
e
m
bl
e
l
e
a
r
ni
ng f
r
am
e
w
or
k
f
o
r
bi
g dat
a
…
(
K
e
s
av
an M
e
tt
ur
V
ar
adha
r
aj
an
)
1395
4.
P
E
R
F
O
R
M
A
N
C
E
E
V
A
L
U
A
T
I
O
N
T
he
pe
r
f
or
m
a
nc
e
e
va
lu
a
ti
on
in
vol
ve
s
a
s
s
e
s
s
in
g
th
e
a
c
c
ur
a
c
y
of
di
f
f
e
r
e
nt
m
e
th
odol
ogi
e
s
,
in
c
lu
di
ng
th
e
pr
opos
e
d
m
ode
l,
a
c
r
os
s
f
our
da
ta
s
e
ts
:
s
upe
r
s
ym
m
e
tr
y
(
S
U
S
Y
)
,
hi
ggs
,
m
odi
f
ie
d
na
ti
ona
l
in
s
ti
tu
te
o
f
s
ta
nda
r
ds
a
nd
te
c
hnol
ogy
(
M
ni
s
t
)
,
a
nd
U
ni
te
d
S
ta
te
s
pos
ta
l
s
e
r
vi
c
e
(
U
S
P
S
)
.
A
ddi
ti
ona
ll
y,
th
e
e
va
lu
a
ti
on
in
c
lu
de
s
c
om
pa
r
in
g
th
e
ti
m
e
ta
ke
n
a
nd
s
pe
e
d
-
up
a
c
hi
e
ve
d
s
pe
c
if
ic
a
ll
y
f
or
th
e
S
U
S
Y
a
nd
H
ig
gs
da
ta
s
e
ts
.
T
he
r
e
s
ul
ts
a
r
e
pr
e
s
e
nt
e
d
c
om
pr
e
he
ns
iv
e
ly
th
r
ough
ta
bl
e
s
a
nd
gr
a
phs
to
pr
ovi
de
a
c
le
a
r
c
om
pa
r
is
on
of
th
e
m
e
th
odol
ogi
e
s
'
e
f
f
e
c
ti
ve
ne
s
s
a
c
r
os
s
t
he
va
r
io
us
da
ta
s
e
ts
a
nd pe
r
f
or
m
a
nc
e
m
e
tr
ic
s
.
4.1.
A
c
c
u
r
ac
y
A
na
ly
z
in
g
th
e
pe
r
f
or
m
a
nc
e
of
va
r
io
us
m
e
th
ods
on
th
e
S
U
S
Y
da
ta
s
e
t,
w
e
s
e
e
a
r
a
nge
of
a
c
c
ur
a
c
y
s
c
or
e
s
f
r
om
th
e
s
e
c
la
s
s
if
ic
a
ti
on
a
lg
or
it
hm
s
.
M
ul
ti
va
r
ia
te
de
c
i
s
io
n
tr
e
e
2
(
M
D
T
2
)
de
m
ons
tr
a
te
s
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
a
m
ong
th
e
m
a
jo
r
it
y
of
th
e
t
r
a
di
ti
ona
l
a
lg
or
it
h
m
s
,
w
it
h
a
s
c
or
e
of
0.749,
s
ugge
s
ti
ng
th
a
t
it
s
m
e
th
od
of
c
la
s
s
if
ic
a
ti
on
is
pa
r
ti
c
ul
a
r
ly
w
e
ll
-
s
ui
te
d
to
th
is
da
ta
s
e
t.
C
ons
ta
nt
-
ti
m
e
e
n
s
e
m
bl
e
le
a
r
ni
ng
c
l
a
s
s
if
ie
r
(
C
T
E
L
C
)
[
(
e
xi
s
ti
ng
s
ys
te
m
)
ES
]
,
a
n
e
ns
e
m
bl
e
s
tr
a
te
gy,
s
li
ght
ly
out
pe
r
f
or
m
s
M
D
T
2
w
it
h
a
s
c
or
e
of
0.758
,
in
di
c
a
ti
ng
th
e
e
f
f
e
c
ti
ve
ne
s
s
of
c
om
bi
ni
ng
m
ul
ti
pl
e
m
ode
l
s
to
e
nha
nc
e
pr
e
di
c
ti
ve
pe
r
f
or
m
a
nc
e
.
N
ot
a
bl
y,
th
e
pr
opos
e
d
s
y
s
te
m
(
P
S
)
m
e
th
od
s
ta
nd
s
out
w
it
h
th
e
hi
ghe
s
t
a
c
c
ur
a
c
y
of
0.8063,
w
hi
c
h
c
oul
d
im
pl
y
a
n
a
dva
nc
e
d
f
e
a
tu
r
e
s
e
le
c
ti
on
or
opt
im
iz
a
ti
on
pr
oc
e
s
s
th
a
t
s
ig
ni
f
ic
a
nt
ly
be
ne
f
it
s
th
e
m
ode
l'
s
pe
r
f
or
m
a
nc
e
on
th
is
da
ta
s
e
t.
O
n
th
e
ot
he
r
e
nd
of
th
e
s
pe
c
tr
um
,
n
a
ti
ona
l
be
nc
hm
a
r
k
te
s
ts
(
N
B
T
)
s
c
or
e
s
th
e
lo
w
e
s
t
a
t
0.594,
w
hi
c
h
m
a
y
in
di
c
a
te
th
a
t
it
s
pr
oba
bi
li
s
ti
c
a
ppr
oa
c
h
is
le
s
s
e
f
f
e
c
ti
ve
f
or
th
e
pa
tt
e
r
ns
pr
e
s
e
nt
in
th
e
S
U
S
Y
da
ta
.
T
he
id
e
nt
ic
a
l
s
c
or
e
s
of
m
ovi
ng
r
a
nge
k
-
ne
a
r
e
s
t
ne
ig
hbor
(
MR
-
K
N
N
)
a
nd
K
N
N
-
I
S
(
K
N
N
de
s
ig
n
ba
s
e
d
on
s
pa
r
k)
s
ugge
s
t
s
im
il
a
r
c
a
pa
bi
li
ti
e
s
in
ha
ndl
in
g
th
e
da
ta
,
pot
e
nt
ia
ll
y
d
ue
to
s
ha
r
e
d
r
e
li
a
nc
e
on
th
e
pr
oxi
m
it
y
of
da
ta
poi
nt
s
.
F
uz
z
y
c
la
s
s
if
ie
r
s
s
ho
w
r
obus
tn
e
s
s
w
it
h
a
s
c
or
e
of
0.735,
w
hi
c
h
m
ig
ht
be
due
to
th
e
ir
a
bi
li
ty
to
ha
ndl
e
unc
e
r
ta
in
or
im
pr
e
c
is
e
in
f
or
m
a
ti
on.
L
a
s
tl
y,
f
ur
th
e
s
t
-
pa
ir
-
ba
s
e
d
bi
na
r
y
s
e
a
r
c
h
tr
e
e
(
F
P
B
S
T
)
a
nd
m
in
im
um
/m
a
xi
m
um
nor
m
s
-
ba
s
e
d
bi
na
r
y
tr
e
e
(
M
N
B
T
)
a
r
e
ti
e
d
a
t
0.71,
w
hi
c
h
c
oul
d
s
ugge
s
t
a
pa
r
it
y
in
th
e
ir
a
bi
li
ty
to
ge
ne
r
a
li
z
e
or
a
n
unde
r
ly
in
g
s
im
il
a
r
it
y
in
th
e
ir
a
ppr
oa
c
h
to
th
e
S
U
S
Y
da
ta
s
e
t.
T
a
bl
e
1
a
nd
F
ig
ur
e
4
s
how
t
he
a
c
c
ur
a
c
y c
om
pa
r
is
on f
or
t
he
S
U
S
Y
da
ta
s
e
t.
T
a
bl
e
1
.
A
c
c
ur
a
c
y c
om
pa
r
is
on
M
e
t
hod
S
U
S
Y
M
D
T
1
[
25]
0.729
M
D
T
2
[
25]
0.749
F
P
B
S
T
[
26]
0.71
MR
-
KNN
[
27]
0.694
KNN
-
IS
[
28]
0.694
F
uz
z
y
[
29]
0.735
N
B
T
[
30]
0.594
M
N
B
T
[
30]
0.71
C
T
E
L
C
[
E
S
]
[
31]
0.758
PS
0.8063
F
ig
ur
e
4
.
A
c
c
ur
a
c
y c
om
pa
r
is
on f
or
t
he
S
U
S
Y
da
ta
s
e
t
0
0
.1
0
.2
0
.3
0
.4
0
.5
0
.6
0
.7
0
.8
0
.9
M
D
T
1
M
D
T
2
F
P
B
S
T
M
R
-
K
N
N
K
N
N
-
IS
F
u
z
z
y
N
B
T
M
N
B
T
CT
E
L
C
[
E
S
]
PS
A
c
c
u
r
a
c
y
v
a
l
u
e
M
et
h
o
d
o
l
o
gy
S
U
S
Y
Evaluation Warning : The document was created with Spire.PDF for Python.