I
nd
o
ne
s
ia
n J
o
urna
l o
f
E
lect
rica
l En
g
ineering
a
nd
Co
m
pu
t
er
Science
Vo
l.
41
,
No
.
1
,
J
an
u
ar
y
20
26
,
p
p
.
2
5
8
~
2
6
9
I
SS
N:
2
502
-
4
7
52
,
DOI
: 1
0
.
1
1
5
9
1
/ijee
cs
.v
41.
i
1
.
p
p
258
-
2
6
9
258
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ee
cs
.
ia
esco
r
e.
co
m
A hybrid
div
isiv
e
K
-
mea
ns
framew
o
rk f
o
r big da
ta
–
driv
en
po
v
erty an
a
ly
sis
in
Cen
tral J
a
v
a
P
ro
v
ince
B
o
wo
Wina
rno
1
,2
,
B
ud
i Wa
r
s
it
o
1,
3
,
B
a
y
u Sura
rso
1,
3
1
D
o
c
t
o
r
a
l
P
r
o
g
r
a
m
o
f
I
n
f
o
r
ma
t
i
o
n
S
y
s
t
e
m,
S
c
h
o
o
l
o
f
P
o
s
t
g
r
a
d
u
a
t
e
S
t
u
d
i
e
s
,
D
i
p
o
n
e
g
o
r
o
U
n
i
v
e
r
si
t
y
,
S
e
mara
n
g
,
I
n
d
o
n
e
s
i
a
2
D
e
p
a
r
t
me
n
t
o
f
M
a
t
h
e
ma
t
i
c
s
,
S
e
b
e
l
a
s
M
a
r
e
t
U
n
i
v
e
r
s
i
t
y
,
S
u
r
a
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
3
F
a
c
u
l
t
y
o
f
S
c
i
e
n
c
e
a
n
d
M
a
t
h
e
m
a
t
i
c
s,
D
i
p
o
n
e
g
o
r
o
U
n
i
v
e
r
s
i
t
y
,
S
e
mar
a
n
g
,
I
n
d
o
n
e
s
i
a
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
No
v
4
,
2
0
2
5
R
ev
is
ed
Dec
5
,
2
0
2
5
Acc
ep
ted
Dec
13
,
2
0
2
5
Clu
ste
rin
g
is
e
ss
e
n
ti
a
l
in
b
i
g
d
a
ta
a
n
a
ly
ti
c
s,
e
sp
e
c
iall
y
f
o
r
p
a
rt
it
io
n
in
g
h
i
g
h
-
d
ime
n
sio
n
a
l
s
o
c
io
e
c
o
n
o
m
ic
d
a
ta
se
ts
to
su
p
p
o
rt
i
n
terp
re
tati
o
n
a
n
d
p
o
li
c
y
d
e
c
isio
n
s.
W
h
il
e
K
-
M
e
a
n
s
is
wid
e
ly
u
se
d
f
o
r
it
s
sim
p
li
c
it
y
a
n
d
sc
a
lab
il
it
y
,
it
s
stro
n
g
se
n
siti
v
it
y
t
o
in
it
ial
c
e
n
tro
id
se
lec
ti
o
n
o
ften
lea
d
s
t
o
u
n
sta
b
le
re
su
lt
s
a
n
d
slo
we
r
c
o
n
v
e
rg
e
n
c
e
.
P
re
v
io
u
s
h
y
b
rid
a
p
p
r
o
a
c
h
e
s,
su
c
h
a
s
Ag
g
lo
m
e
ra
ti
v
e
–
K
-
M
e
a
n
s,
a
tt
e
m
p
ted
to
a
d
d
re
ss
th
is
issu
e
b
y
u
si
n
g
h
iera
rc
h
ica
l
c
lu
ste
rin
g
f
o
r
c
e
n
tr
o
id
in
i
ti
a
li
z
a
ti
o
n
;
h
o
we
v
e
r,
th
e
s
e
m
e
th
o
d
s
re
ly
o
n
b
o
tt
o
m
-
u
p
m
e
rg
i
n
g
,
wh
ic
h
c
a
n
p
r
o
d
u
c
e
su
b
o
p
ti
m
a
l
i
n
it
ial
p
a
rti
ti
o
n
s
a
n
d
in
c
re
a
se
c
o
m
p
u
tati
o
n
a
l
o
v
e
rh
e
a
d
fo
r
larg
e
r
d
a
t
a
se
ts.
To
o
v
e
rc
o
m
e
th
e
se
li
m
it
a
ti
o
n
s,
th
is
stu
d
y
p
r
o
p
o
se
s
a
h
y
b
rid
d
i
v
isi
v
e
–
K
-
M
e
a
n
s
(DH
C)
m
o
d
e
l
th
a
t
e
m
p
l
o
y
s
to
p
-
d
o
wn
h
iera
rc
h
ica
l
sp
l
it
ti
n
g
to
g
e
n
e
ra
te
m
o
r
e
c
o
h
e
re
n
t
in
it
ial
c
e
n
tr
o
i
d
s
b
e
fo
re
re
fi
n
e
m
e
n
t
with
K
-
M
e
a
n
s.
Us
in
g
a
m
u
lt
i
d
i
m
e
n
sio
n
a
l
p
o
v
e
rt
y
d
a
tas
e
t
fro
m
Ce
n
tral
Ja
v
a
P
ro
v
in
c
e
p
r
o
v
i
d
e
d
b
y
t
h
e
I
n
d
o
n
e
sia
n
Ce
n
tral
Bu
re
a
u
o
f
S
tatisti
c
s
(BP
S
),
th
e
p
e
rfo
rm
a
n
c
e
o
f
DH
C
wa
s
e
v
a
lu
a
ted
a
g
a
in
st
sta
n
d
a
rd
K
-
M
e
a
n
s
a
n
d
Ag
g
lo
m
e
ra
ti
v
e
–
K
-
M
e
a
n
s.
Th
e
a
ss
e
ss
m
e
n
t
in
c
lu
d
e
d
e
x
e
c
u
t
io
n
ti
m
e
,
c
o
n
v
e
rg
e
n
c
e
it
e
ra
ti
o
n
s,
a
n
d
c
lu
ste
r
v
a
li
d
i
ty
i
n
d
ice
s
(S
il
h
o
u
e
tt
e
,
Da
v
ies
–
B
o
u
ld
i
n
,
a
n
d
Ca
li
n
sk
i
–
Ha
ra
b
a
sz
).
Ex
p
e
rime
n
t
a
l
re
su
lt
s
d
e
m
o
n
stra
te
th
a
t
DH
C
re
d
u
c
e
s
e
x
e
c
u
ti
o
n
ti
m
e
b
y
u
p
to
9
7
%
a
n
d
re
q
u
ires
4
0
%
fe
we
r
it
e
ra
ti
o
n
s th
a
n
sta
n
d
a
r
d
K
-
M
e
a
n
s,
wh
il
e
a
c
h
iev
i
n
g
c
o
m
p
a
ra
b
le
o
r
imp
ro
v
e
d
c
lu
ste
r
q
u
a
l
it
y
(e
.
g
.
,
CH
In
d
e
x
i
n
c
re
a
sin
g
fr
o
m
1
4
.
3
to
1
5
.
8
)
.
Th
e
se
fin
d
i
n
g
s
i
n
d
ica
te
th
a
t
th
e
DH
C
m
o
d
e
l
o
ffe
rs
a
m
o
re
e
fficie
n
t
a
n
d
sta
b
le
c
lu
ste
ri
n
g
so
l
u
ti
o
n
,
a
d
d
re
ss
in
g
th
e
sh
o
rtco
m
in
g
s
o
f
p
re
v
i
o
u
s
sta
n
d
a
rd
K
-
M
e
a
n
s
m
e
th
o
d
s
a
n
d
imp
ro
v
in
g
p
e
rfo
rm
a
n
c
e
f
o
r
l
a
rg
e
-
sc
a
l
e
so
c
io
e
c
o
n
o
m
ic d
a
ta an
a
ly
sis.
K
ey
w
o
r
d
s
:
B
ig
d
ata
C
lu
s
ter
in
g
Div
is
iv
e
h
ier
ar
ch
ical
Hy
b
r
id
m
o
d
el
K
-
Me
an
s
Po
v
er
ty
d
ata
a
n
aly
s
is
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
B
o
wo
W
in
ar
n
o
Do
cto
r
al
Pro
g
r
a
m
o
f
I
n
f
o
r
m
at
io
n
Sy
s
tem
,
Sch
o
o
l
o
f
Po
s
tg
r
a
d
u
ate
Stu
d
ies,
Dip
o
n
e
g
o
r
o
Un
iv
er
s
ity
Sem
ar
an
g
,
5
0
2
7
5
,
I
n
d
o
n
esia
E
m
ail: b
o
wo
win
ar
n
o
@
s
tu
d
en
ts
.
u
n
d
ip
.
ac
.
id
1.
I
NT
RO
D
UCT
I
O
N
C
lu
s
ter
in
g
is
a
wid
ely
u
s
ed
u
n
s
u
p
er
v
is
ed
lear
n
i
n
g
tech
n
iq
u
e
f
o
r
id
en
tify
i
n
g
h
id
d
en
s
tr
u
ctu
r
es
with
in
u
n
lab
eled
d
ata,
s
u
p
p
o
r
tin
g
ap
p
licatio
n
s
in
s
o
cio
-
ec
o
n
o
m
ic
an
aly
s
is
,
u
r
b
an
p
lan
n
in
g
,
en
v
ir
o
n
m
en
tal
m
o
n
ito
r
in
g
,
an
d
h
ea
lth
in
f
o
r
m
atics
[
1
]
.
Am
o
n
g
v
ar
io
u
s
clu
s
ter
in
g
alg
o
r
ith
m
s
,
K
-
Me
an
s
r
e
m
ain
s
p
o
p
u
lar
d
u
e
to
its
ef
f
icien
cy
an
d
s
ca
lab
ilit
y
;
h
o
wev
er
,
its
s
tr
o
n
g
s
en
s
itiv
ity
to
in
itial
ce
n
tr
o
id
s
elec
tio
n
o
f
ten
lead
s
to
in
co
n
s
is
ten
t
r
esu
lts
,
s
lo
w
co
n
v
er
g
e
n
ce
,
a
n
d
s
u
s
ce
p
tib
ilit
y
to
lo
ca
l
m
in
i
m
a
[
2
]
,
[
3
]
.
T
o
a
d
d
r
ess
th
ese
wea
k
n
ess
es,
s
ev
er
al
s
tu
d
ies h
a
v
e
p
r
o
p
o
s
ed
h
y
b
r
id
o
r
o
p
tim
iz
atio
n
b
ased
m
o
d
if
icatio
n
s
to
K
-
Me
an
s
,
in
clu
d
in
g
th
e
in
teg
r
atio
n
o
f
h
ier
ar
ch
ical
m
eth
o
d
s
,
g
e
n
etic
alg
o
r
ith
m
s
,
a
n
d
d
en
s
ity
-
b
ased
p
r
e
p
r
o
ce
s
s
in
g
[
4
]
–
[
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
h
yb
r
id
d
ivis
ive
K
-
mea
n
s
fr
a
mewo
r
k
fo
r
b
ig
d
a
ta
–
d
r
iven
p
o
ve
r
ty
a
n
a
lysi
s
in
C
en
tr
a
l
… (
B
o
w
o
Win
a
r
n
o
)
259
Alth
o
u
g
h
h
ier
a
r
ch
ical
clu
s
ter
in
g
p
r
o
v
id
es
d
eter
m
in
is
tic
p
ar
titi
o
n
in
g
an
d
av
o
id
s
r
an
d
o
m
in
itializatio
n
,
it
s
u
f
f
er
s
f
r
o
m
h
ig
h
c
o
m
p
u
tatio
n
al
co
m
p
lex
ity
wh
en
a
p
p
lied
to
lar
g
e
d
atas
ets
[
7
]
–
[
1
0
]
.
Prio
r
h
y
b
r
id
ap
p
r
o
ac
h
es,
s
u
c
h
as
h
ier
ar
ch
ical
–
K
-
Me
an
s
co
m
b
in
a
tio
n
s
,
h
av
e
attem
p
ted
to
m
er
g
e
th
e
s
tr
en
g
t
h
s
o
f
b
o
th
m
eth
o
d
s
.
Ho
wev
e
r
,
th
es
e
s
tu
d
ies
p
r
ed
o
m
in
a
n
tly
r
ely
o
n
ag
g
lo
m
er
ativ
e
(
b
o
tto
m
-
u
p
)
clu
s
ter
in
g
,
wh
ich
m
ay
r
esu
lt
in
s
u
b
o
p
tim
al
ce
n
tr
o
id
in
itializatio
n
d
u
e
to
its
m
er
g
in
g
-
b
ased
s
tr
u
ctu
r
e.
Mo
r
eo
v
er
,
ex
is
tin
g
h
y
b
r
id
s
r
ar
el
y
ev
alu
ate
wh
eth
er
th
e
h
ier
ar
ch
ica
l
s
tag
e
g
en
u
i
n
ely
im
p
r
o
v
es
in
itializatio
n
q
u
ality
o
r
s
ca
lab
ilit
y
wh
en
ap
p
lied
to
m
u
ltid
im
en
s
io
n
al
s
o
cio
-
ec
o
n
o
m
ic
d
atasets
[
1
1
]
–
[
1
6
]
.
T
h
ese
lim
itatio
n
s
h
ig
h
lig
h
t
a
clea
r
r
esear
ch
g
ap
: c
u
r
r
en
t
h
y
b
r
id
m
eth
o
d
s
h
av
e
n
o
t su
f
f
icien
tly
o
p
tim
ized
ce
n
tr
o
id
in
itializ
atio
n
wh
ile
m
ain
tain
in
g
co
m
p
u
tatio
n
al
e
f
f
icien
cy
,
esp
e
cially
f
o
r
co
m
p
lex
p
o
v
er
t
y
-
r
el
ated
in
d
icato
r
s
.
Giv
en
th
is
g
ap
,
th
e
u
n
d
er
l
y
in
g
p
r
o
b
lem
o
f
th
is
s
tu
d
y
em
er
g
es
n
atu
r
ally
:
d
esp
ite
th
e
ab
u
n
d
an
ce
o
f
h
y
b
r
id
clu
s
ter
in
g
ap
p
r
o
ac
h
es,
it r
em
ain
s
u
n
clea
r
h
o
w
ce
n
t
r
o
i
d
in
itializatio
n
ca
n
b
e
s
y
s
tem
a
tically
im
p
r
o
v
ed
to
en
h
an
ce
co
n
v
er
g
e
n
ce
s
p
ee
d
,
s
tab
ilit
y
,
an
d
clu
s
ter
in
g
q
u
ality
f
o
r
K
-
Me
an
s
wh
en
d
ea
lin
g
with
m
u
ltid
im
en
s
io
n
al
s
o
cio
-
ec
o
n
o
m
ic
d
ata.
E
x
is
tin
g
ev
id
en
ce
s
u
g
g
ests
th
at
a
m
o
r
e
g
lo
b
ally
in
f
o
r
m
ed
in
itializatio
n
s
tr
ateg
y
is
n
ee
d
ed
,
y
et
th
e
o
p
er
atio
n
al
ef
f
ec
t
iv
en
ess
o
f
s
u
ch
an
ap
p
r
o
ac
h
h
as
n
o
t
b
ee
n
f
u
lly
estab
lis
h
ed
in
p
r
io
r
w
o
r
k
.
Mo
tiv
ated
b
y
th
is
is
s
u
e,
th
e
p
r
esen
t
s
tu
d
y
a
d
v
an
ce
s
th
e
h
y
p
o
th
esis
th
at
a
d
iv
is
iv
e
h
ier
ar
ch
ical
p
r
o
ce
s
s
—
o
win
g
to
its
to
p
-
d
o
wn
,
r
ec
u
r
s
iv
el
y
s
p
litt
in
g
m
ec
h
an
is
m
—
ca
n
g
e
n
er
ate
m
o
r
e
c
o
h
er
e
n
t
an
d
r
ep
r
esen
tativ
e
in
itial
ce
n
t
r
o
id
s
.
T
h
is
,
in
tu
r
n
,
is
ex
p
ec
ted
to
r
ed
u
ce
ex
ec
u
tio
n
tim
e
an
d
co
n
v
er
g
en
ce
iter
atio
n
s
,
wh
ile
ac
h
iev
in
g
clu
s
ter
in
g
q
u
ality
co
m
p
a
r
ab
le
to
o
r
b
etter
th
an
co
n
v
en
tio
n
al
K
-
Me
an
s
an
d
ex
is
tin
g
Ag
g
lo
m
er
ativ
e
–
K
-
Me
an
s
h
y
b
r
id
s
[
1
1
]
,
[
1
7
]
,
[
1
8
]
.
Alth
o
u
g
h
d
iv
is
iv
e
m
et
h
o
d
s
t
h
eo
r
etica
ll
y
p
r
o
v
id
e
a
b
r
o
a
d
er
s
tr
u
ctu
r
al
o
v
er
v
iew
th
an
a
g
g
l
o
m
er
ativ
e
ap
p
r
o
ac
h
es,
th
eir
p
o
ten
tial
b
en
ef
its
f
o
r
h
y
b
r
id
clu
s
ter
in
g
h
av
e
n
o
t
b
ee
n
co
m
p
r
eh
e
n
s
iv
ely
ev
alu
at
ed
in
p
r
e
v
io
u
s
s
tu
d
ies.
T
o
ev
alu
ate
th
is
h
y
p
o
th
esis
,
th
e
p
r
o
p
o
s
ed
DHC
m
o
d
el
is
ap
p
lied
to
a
m
u
ltid
im
en
s
io
n
al
p
o
v
er
ty
d
ataset
f
r
o
m
C
en
tr
al
J
av
a
Pr
o
v
in
ce
,
I
n
d
o
n
esia,
o
b
tain
e
d
f
r
o
m
th
e
C
en
tr
al
B
u
r
ea
u
o
f
Statis
tics
(
B
P
S).
T
h
e
d
ataset
co
n
s
is
ts
o
f
in
ter
r
elate
d
s
o
cio
-
e
co
n
o
m
ic
in
d
icato
r
s
,
in
clu
d
in
g
ed
u
ca
tio
n
,
in
c
o
m
e
,
em
p
lo
y
m
en
t,
a
n
d
liv
in
g
co
n
d
itio
n
s
,
wh
ich
ar
e
ch
allen
g
in
g
to
clu
s
ter
u
s
in
g
co
n
v
e
n
tio
n
al
m
et
h
o
d
s
.
Un
d
er
s
tan
d
in
g
p
o
v
er
t
y
d
is
tr
ib
u
tio
n
th
r
o
u
g
h
clu
s
ter
in
g
h
as
im
p
o
r
ta
n
t
im
p
licatio
n
s
f
o
r
p
o
licy
tar
g
etin
g
an
d
r
eg
io
n
al
d
ev
elo
p
m
en
t
p
lan
n
in
g
[
7
]
,
[
8
]
.
T
h
e
co
n
tr
i
b
u
tio
n
s
o
f
th
is
s
tu
d
y
ar
e
as f
o
llo
ws:
a)
Pro
p
o
s
in
g
a
h
y
b
r
id
d
iv
is
iv
e
–
K
-
m
ea
n
s
(
DHC)
alg
o
r
ith
m
t
o
i
m
p
r
o
v
e
ce
n
tr
o
i
d
in
itializatio
n
an
d
clu
s
ter
in
g
ef
f
icien
cy
f
o
r
m
u
ltid
im
en
s
io
n
al
s
o
cio
-
ec
o
n
o
m
ic
d
ata.
b)
C
o
m
p
ar
ativ
ely
ev
al
u
atin
g
K
-
Me
an
s
,
Ag
g
lo
m
er
ativ
e
–
K
-
Me
an
s
,
an
d
DHC
in
ter
m
s
o
f
e
x
ec
u
tio
n
tim
e,
co
n
v
er
g
en
ce
r
ate,
an
d
cl
u
s
ter
v
alid
ity
m
etr
ics.
c)
Dem
o
n
s
tr
atin
g
th
e
r
elev
an
ce
o
f
h
y
b
r
id
clu
s
ter
in
g
m
eth
o
d
s
f
o
r
r
eg
io
n
al
p
o
v
e
r
ty
an
aly
s
is
as
a
d
ec
is
io
n
-
s
u
p
p
o
r
t to
o
l f
o
r
s
o
cio
-
ec
o
n
o
m
ic
p
o
licy
f
o
r
m
u
latio
n
.
Ov
er
all,
th
is
r
esear
ch
e
x
ten
d
s
ex
is
tin
g
h
y
b
r
id
cl
u
s
ter
in
g
liter
atu
r
e
b
y
ad
d
r
ess
in
g
u
n
r
eso
lv
ed
lim
itatio
n
s
in
ce
n
tr
o
i
d
in
itiali
za
tio
n
an
d
d
e
m
o
n
s
tr
atin
g
th
at
co
m
b
i
n
in
g
d
eter
m
in
is
tic
h
ier
ar
ch
ical
s
tr
ateg
ies
with
p
ar
titi
o
n
in
g
tech
n
iq
u
es
ca
n
p
r
o
d
u
ce
m
o
r
e
s
tab
le
an
d
co
m
p
u
tatio
n
ally
ef
f
icien
t
clu
s
ter
in
g
r
esu
lts
[
1
1
]
,
[
1
3
]
,
[
1
5
]
,
[
1
9
]
.
2.
M
E
T
H
O
D
Fig
u
r
e
1
p
r
esen
ts
th
e
wo
r
k
f
lo
w
o
f
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
DH
C
f
r
am
ewo
r
k
f
o
r
b
ig
d
ata
–
d
r
iv
en
p
o
v
er
ty
an
aly
s
is
in
C
en
tr
al
J
av
a
Pro
v
in
ce
.
T
h
e
f
r
am
ewo
r
k
b
eg
i
n
s
with
th
e
ac
q
u
is
itio
n
o
f
in
p
u
t
d
a
ta
,
wh
ich
in
clu
d
es
lar
g
e
-
s
ca
le
p
o
v
e
r
ty
in
d
icat
o
r
s
s
u
ch
as
ed
u
ca
tio
n
attain
m
en
t
,
em
p
lo
y
m
en
t
s
tatu
s
,
an
d
h
o
u
s
eh
o
ld
ex
p
e
n
d
itu
r
e.
T
o
en
s
u
r
e
an
aly
tical
r
eliab
ilit
y
,
a
m
o
r
e
r
o
b
u
s
t
d
ata
p
r
ep
r
o
c
ess
in
g
p
ip
elin
e
is
em
p
lo
y
ed
.
T
h
is
s
tag
e
in
clu
d
es
s
y
s
tem
atic
h
an
d
lin
g
o
f
m
is
s
in
g
v
alu
es
th
r
o
u
g
h
m
u
ltiv
ar
iate
im
p
u
tatio
n
,
d
etec
tio
n
an
d
tr
e
atm
en
t
o
f
o
u
tlier
s
,
n
o
r
m
aliza
tio
n
o
f
h
ete
r
o
g
e
n
eo
u
s
n
u
m
er
ic
r
an
g
e
s
,
an
d
f
ea
tu
r
e
co
n
s
is
ten
cy
ch
ec
k
s
ac
r
o
s
s
d
is
tr
icts
.
Fo
llo
win
g
p
r
e
p
r
o
ce
s
s
in
g
,
a
d
iv
is
iv
e
h
ier
ar
ch
ical
clu
s
ter
in
g
p
r
o
ce
d
u
r
e
is
a
p
p
lied
u
s
in
g
a
to
p
-
d
o
w
n
s
tr
ateg
y
.
At
ea
ch
iter
atio
n
,
t
h
e
d
ataset
is
r
ec
u
r
s
iv
ely
s
p
lit
b
ased
o
n
m
ax
im
u
m
h
ete
r
o
g
e
n
eity
cr
iter
ia,
wit
h
ex
p
licit
alg
o
r
ith
m
ic
s
tep
s
d
ef
in
ed
f
o
r
s
elec
tin
g
s
p
litt
in
g
attr
ib
u
tes
an
d
ca
lcu
latin
g
s
u
b
g
r
o
u
p
ce
n
tr
o
i
d
s
.
T
h
ese
ce
n
tr
o
id
s
s
er
v
e
as st
r
u
ctu
r
e
d
,
d
ata
-
d
r
iv
e
n
in
itial seed
s
f
o
r
th
e
s
u
b
s
eq
u
en
t o
p
tim
izatio
n
s
tag
e.
T
h
e
n
ex
t
p
h
ase
p
e
r
f
o
r
m
s
K
-
Me
an
s
o
p
tim
izatio
n
u
s
in
g
a
p
r
e
d
ef
in
ed
n
u
m
b
er
o
f
clu
s
ter
s
(
k
=3
)
,
wh
er
e
th
e
d
iv
is
iv
e
-
g
en
er
ated
ce
n
tr
o
id
s
ar
e
r
ef
in
ed
th
r
o
u
g
h
iter
a
tiv
e
m
in
im
izatio
n
o
f
th
e
with
in
-
clu
s
ter
s
u
m
o
f
s
q
u
ar
es.
T
h
is
s
tep
en
h
an
ce
s
co
m
p
ac
tn
ess
an
d
r
ed
u
ce
s
s
en
s
itiv
ity
to
r
an
d
o
m
in
itializ
atio
n
,
ad
d
r
ess
in
g
a
co
m
m
o
n
lim
itatio
n
o
f
s
tan
d
ar
d
K
-
Me
an
s
.
C
o
n
v
er
g
en
ce
th
r
esh
o
ld
s
,
iter
atio
n
lim
its
,
an
d
d
is
tan
ce
m
etr
ics
ar
e
ex
p
licitly
d
ef
in
e
d
to
en
s
u
r
e
m
eth
o
d
o
lo
g
ical
tr
an
s
p
ar
en
c
y
.
T
o
ev
al
u
ate
clu
s
ter
in
g
r
o
b
u
s
tn
ess
,
th
e
r
esu
ltin
g
o
p
tim
iz
ed
clu
s
ter
s
ar
e
s
u
b
jecte
d
to
m
u
lti
p
le
b
en
ch
m
ar
k
in
g
tech
n
iq
u
es,
in
clu
d
in
g
c
o
m
p
ar
is
o
n
s
with
s
tan
d
ar
d
K
-
Me
an
s
an
d
alter
n
ativ
e
in
itializatio
n
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
1
,
J
an
u
ar
y
20
26
:
258
-
2
6
9
260
s
tr
ateg
ies.
C
lu
s
ter
q
u
ality
is
ass
es
s
ed
u
s
in
g
s
ev
er
al
v
ali
d
ity
in
d
ices
(
e.
g
.
,
Sil
h
o
u
ette
,
Dav
ies
–
B
o
u
ld
in
,
C
alin
s
k
i
–
Har
ab
asz)
,
en
ab
lin
g
b
r
o
ad
e
r
an
d
m
o
r
e
r
ig
o
r
o
u
s
p
er
f
o
r
m
an
ce
ev
alu
atio
n
.
Fin
ally
,
s
p
atial
an
d
s
o
cio
ec
o
n
o
m
ic
p
atter
n
s
ar
e
v
is
u
alize
d
at
th
e
d
is
tr
ict
lev
el
to
d
er
iv
e
p
o
licy
-
r
elev
an
t
in
s
ig
h
ts
.
W
h
ile
th
e
d
a
taset
u
s
ed
in
th
is
s
tu
d
y
is
m
o
d
est
in
s
ize,
lim
itin
g
th
e
d
em
o
n
s
tr
atio
n
o
f
f
u
ll
b
ig
-
d
ata
s
ca
lab
ilit
y
,
th
e
f
r
am
ewo
r
k
is
d
esig
n
ed
to
b
e
ex
ten
d
ab
le
to
lar
g
er
d
atasets
d
u
e
to
it
s
h
i
er
ar
ch
ical
r
ed
u
ctio
n
an
d
o
p
tim
ize
d
in
itializatio
n
s
tep
s
.
Fig
u
r
e
1
.
Hy
b
r
id
DHC
All
ex
p
er
im
en
ts
wer
e
ex
ec
u
te
d
in
th
e
G
o
o
g
le
C
o
lab
e
n
v
ir
o
n
m
en
t
u
s
in
g
Py
th
o
n
3
.
1
0
wit
h
s
tan
d
ar
d
h
ar
d
war
e
r
eso
u
r
ce
s
p
r
o
v
id
ed
b
y
th
e
p
latf
o
r
m
.
T
h
e
clu
s
ter
in
g
p
r
o
ce
d
u
r
es
wer
e
im
p
lem
en
ted
u
s
in
g
wid
el
y
ad
o
p
ted
s
cien
tific
lib
r
a
r
ies,
i
n
clu
d
in
g
s
cik
it
-
lear
n
,
Nu
m
P
y
,
p
a
n
d
as,
an
d
SciPy
.
T
h
ese
s
p
ec
if
icatio
n
s
ar
e
r
ep
o
r
ted
to
en
s
u
r
e
tr
an
s
p
ar
en
c
y
an
d
r
ep
r
o
d
u
cib
ilit
y
o
f
th
e
e
x
p
er
im
en
tal
wo
r
k
f
lo
w.
2
.
1
.
Da
t
a
s
et
T
h
is
s
tu
d
y
em
p
lo
y
s
th
e
p
o
v
er
ty
d
ataset
o
f
C
en
tr
al
J
av
a
Pro
v
in
ce
o
b
tain
ed
f
r
o
m
th
e
I
n
d
o
n
esian
C
en
tr
al
B
P
S
in
2
0
2
4
(
T
ab
le
1
)
[
8
]
.
T
h
e
d
ataset
co
n
s
is
ts
o
f
r
ec
o
r
d
s
f
r
o
m
3
5
d
is
tr
icts
an
d
m
u
n
icip
alities
with
in
th
e
p
r
o
v
in
ce
.
I
t c
o
n
tain
s
n
in
e
s
o
cio
-
ec
o
n
o
m
ic
in
d
icat
o
r
s
th
at
r
ep
r
esen
t m
u
ltid
im
en
s
io
n
al
asp
ec
ts
o
f
p
o
v
er
ty
.
T
h
is
d
ataset
was
s
elec
ted
b
ec
au
s
e
it
p
r
o
v
id
es
a
r
ea
l
-
wo
r
ld
ca
s
e
o
f
h
ig
h
-
d
im
en
s
io
n
al,
im
b
alan
ce
d
,
an
d
u
n
la
b
eled
s
o
cio
-
ec
o
n
o
m
i
c
d
ata
th
at
r
eq
u
ir
es
ac
cu
r
ate
clu
s
ter
in
g
to
s
u
p
p
o
r
t
r
eg
i
o
n
al
p
o
v
er
t
y
r
ed
u
ctio
n
p
o
licies
an
d
r
eso
u
r
ce
allo
ca
ti
o
n
[
7
]
,
[
8
]
,
[
2
0
]
.
Su
ch
m
u
lti
d
im
en
s
io
n
al
d
atasets
ar
e
o
f
ten
u
s
ed
in
b
ig
d
ata
clu
s
ter
in
g
r
esear
ch
to
ev
alu
ate
th
e
p
er
f
o
r
m
a
n
ce
an
d
s
ca
lab
ilit
y
o
f
clu
s
ter
in
g
alg
o
r
ith
m
s
in
r
ea
l
-
wo
r
ld
co
n
tex
ts
[
3
]
,
[
4
]
,
[
1
1
]
.
I
n
ad
d
itio
n
to
th
e
r
eg
i
o
n
al
d
ataset,
s
u
p
p
lem
en
tar
y
test
in
g
was
co
n
d
u
cted
u
s
in
g
b
en
c
h
m
ar
k
d
atasets
f
r
o
m
th
e
UC
I
m
ac
h
in
e
lear
n
in
g
r
e
p
o
s
ito
r
y
to
en
s
u
r
e
th
e
g
en
er
aliza
b
ilit
y
o
f
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
s
ac
r
o
s
s
d
if
f
er
en
t
d
o
m
ain
s
[
1
4
]
,
[
1
5
]
.
B
en
ch
m
a
r
k
d
atasets
ar
e
wid
ely
u
s
ed
f
o
r
ev
al
u
atin
g
cl
u
s
ter
in
g
alg
o
r
ith
m
s
u
n
d
er
s
tan
d
ar
d
ized
c
o
n
d
itio
n
s
to
v
alid
ate
co
n
s
is
ten
cy
,
ac
c
u
r
a
cy
,
an
d
ad
ap
tab
ilit
y
[
2
]
,
[
1
6
]
.
B
ef
o
r
e
p
er
f
o
r
m
in
g
clu
s
ter
in
g
,
d
ata
p
r
ep
r
o
ce
s
s
in
g
was
co
n
d
u
cted
to
h
an
d
le
m
is
s
in
g
o
r
in
co
m
p
lete
v
alu
es.
T
h
e
d
ataset
co
n
tain
e
d
s
ev
er
al
en
tr
ies
with
“NA
”
(
n
o
t
ap
p
licab
le
)
,
p
ar
ticu
lar
l
y
in
s
o
cio
ec
o
n
o
m
ic
in
d
icato
r
s
s
u
ch
as
em
p
lo
y
m
en
t
d
ata.
R
ath
er
th
an
d
eletin
g
r
e
co
r
d
s
with
m
is
s
in
g
v
alu
es
w
h
ich
m
ay
lead
t
o
th
e
lo
s
s
o
f
m
ea
n
i
n
g
f
u
l
in
f
o
r
m
atio
n
an
d
d
is
to
r
tio
n
o
f
d
ata
d
is
tr
ib
u
tio
n
th
is
s
tu
d
y
a
p
p
lied
im
p
u
t
atio
n
tech
n
i
q
u
es
to
r
ep
lace
m
is
s
in
g
en
tr
ies
with
esti
m
ated
v
alu
es
d
er
iv
e
d
f
r
o
m
ex
is
tin
g
d
ata
p
atter
n
s
[
9
]
,
[
1
0
]
,
[
1
2
]
,
[
1
8
]
.
Sp
ec
if
ically
,
f
o
r
t
h
e
d
ata
p
r
esen
ted
in
T
a
b
le
1
,
m
is
s
in
g
v
alu
es
wer
e
im
p
u
te
d
u
s
in
g
th
e
m
ea
n
o
f
t
h
e
co
r
r
esp
o
n
d
in
g
v
ar
iab
le
wh
ic
h
is
o
n
e
o
f
t
h
e
m
o
s
t
co
m
m
o
n
ly
u
s
ed
s
tatis
tical
im
p
u
tatio
n
m
eth
o
d
s
in
u
n
s
u
p
er
v
is
ed
lear
n
in
g
task
s
[
1
0
]
,
[
1
2
]
.
T
h
is
ap
p
r
o
ac
h
e
n
s
u
r
ed
th
at
th
e
d
ataset
r
em
ai
n
ed
co
n
s
is
ten
t
an
d
co
m
p
lete
f
o
r
clu
s
ter
in
g
an
al
y
s
is
.
Data
im
p
u
tatio
n
is
g
en
er
ally
m
o
r
e
ef
f
ec
tiv
e
th
an
d
eletio
n
b
ec
au
s
e
it
p
r
eser
v
es
d
ataset
in
te
g
r
ity
,
r
o
b
u
s
tn
ess
,
an
d
co
m
p
leten
ess
,
e
s
p
ec
ially
wh
en
d
ea
lin
g
with
m
u
ltid
im
en
s
io
n
al
o
r
b
ig
d
ata
s
ce
n
ar
io
s
wh
e
r
e
ea
ch
r
ec
o
r
d
c
o
n
tr
ib
u
tes to
m
o
d
el
p
e
r
f
o
r
m
a
n
ce
[
9
]
,
[
1
0
]
,
[
1
2
]
,
[
2
1
]
.
2
.
2
.
Div
is
iv
e
h
iera
rc
hica
l + K
-
M
ea
ns
(
H
y
brid DH
C
-
K
M
ea
ns
)
T
h
e
d
iv
is
iv
e
h
ier
ar
ch
ical
+
K
-
Me
an
s
(
DHC
–
KM
ea
n
s
)
m
eth
o
d
is
an
o
t
h
er
h
y
b
r
i
d
clu
s
ter
in
g
ap
p
r
o
ac
h
th
at
co
m
b
in
es
d
i
v
is
iv
e
h
ier
ar
c
h
ical
clu
s
ter
in
g
with
K
-
Me
an
s
.
Un
lik
e
th
e
a
g
g
lo
m
e
r
ativ
e
m
eth
o
d
,
wh
ic
h
s
tar
ts
f
r
o
m
i
n
d
iv
id
u
al
p
o
in
ts
an
d
m
e
r
g
es
th
em
s
tep
b
y
s
tep
,
th
e
d
iv
is
iv
e
ap
p
r
o
ac
h
wo
r
k
s
in
th
e
o
p
p
o
s
ite
d
ir
ec
t
io
n
.
I
t
b
eg
in
s
with
t
h
e
en
tire
d
ataset
as
a
s
in
g
le
lar
g
e
clu
s
ter
an
d
t
h
en
r
ec
u
r
s
iv
ely
s
p
lits
it
in
to
s
m
aller
s
u
b
-
clu
s
ter
s
u
n
til th
e
d
esire
d
n
u
m
b
e
r
o
f
clu
s
ter
s
(
k
)
is
r
ea
ch
ed
[
3
]
,
[
2
2
]
,
[
2
3
]
.
2
.
2
.
1
.
Div
is
iv
e
hiera
rc
hica
l st
a
g
e
T
o
b
e
g
in
th
e
p
r
o
ce
s
s
o
f
d
iv
is
iv
e
h
ier
a
r
ch
ical
cl
u
s
ter
in
g
,
th
e
alg
o
r
ith
m
a
d
o
p
ts
a
to
p
-
d
o
wn
ap
p
r
o
ac
h
th
at
s
y
s
tem
atica
lly
p
ar
titi
o
n
s
th
e
d
ataset
in
to
p
r
o
g
r
ess
iv
ely
s
m
aller
an
d
m
o
r
e
h
o
m
o
g
en
e
o
u
s
g
r
o
u
p
s
.
I
n
th
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
h
yb
r
id
d
ivis
ive
K
-
mea
n
s
fr
a
mewo
r
k
fo
r
b
ig
d
a
ta
–
d
r
iven
p
o
ve
r
ty
a
n
a
lysi
s
in
C
en
tr
a
l
… (
B
o
w
o
Win
a
r
n
o
)
261
m
eth
o
d
,
all
d
ata
p
o
in
ts
ar
e
in
itially
g
r
o
u
p
e
d
in
to
a
s
in
g
le,
co
m
p
r
eh
e
n
s
iv
e
clu
s
ter
,
r
ep
r
esen
tin
g
th
e
en
tire
d
ataset
as
o
n
e
u
n
it.
T
h
e
alg
o
r
ith
m
th
en
an
aly
ze
s
th
e
in
ter
n
al
d
is
s
im
ilar
itie
s
am
o
n
g
th
e
d
ata
p
o
in
ts
to
id
en
tify
th
e
m
o
s
t
d
is
tin
ct
s
ep
ar
atio
n
.
B
ased
o
n
t
h
ese
d
is
s
im
ilar
iti
es,
th
e
clu
s
ter
is
d
i
v
id
ed
i
n
t
o
two
s
u
b
clu
s
ter
s
,
en
s
u
r
in
g
th
at
o
b
jects
with
in
ea
ch
g
r
o
u
p
ar
e
as
s
im
ilar
a
s
p
o
s
s
ib
le
wh
ile
m
ain
tain
in
g
clea
r
s
ep
ar
atio
n
f
r
o
m
th
e
o
th
er
g
r
o
u
p
[
2
4
]
,
[
2
5
]
.
T
h
is
r
ec
u
r
s
iv
e
s
p
litt
in
g
p
r
o
c
ess
co
n
tin
u
es
iter
ativ
ely
,
wit
h
ea
ch
r
esu
ltin
g
clu
s
ter
b
ei
n
g
f
u
r
t
h
er
di
v
id
ed
ac
co
r
d
in
g
to
th
e
s
am
e
d
is
s
im
ilar
ity
cr
iter
ia.
T
h
e
p
r
o
ce
d
u
r
e
p
r
o
ce
e
d
s
u
n
til
th
e
d
esire
d
n
u
m
b
er
o
f
clu
s
ter
s
(
k
)
is
r
ea
ch
ed
,
r
esu
lti
n
g
in
a
s
tr
u
ctu
r
ed
h
ier
a
r
ch
y
t
h
at
r
ef
lects
th
e
n
atu
r
al
d
iv
is
io
n
s
with
in
th
e
d
ataset
[
2
2
]
.
T
h
is
to
p
-
d
o
wn
s
tr
ateg
y
a
llo
ws
t
h
e
alg
o
r
ith
m
to
u
n
c
o
v
e
r
m
ea
n
in
g
f
u
l
clu
s
ter
b
o
u
n
d
ar
i
es
ef
f
icien
tly
wh
ile
p
r
eser
v
in
g
t
h
e
g
lo
b
al
s
tr
u
ctu
r
e
o
f
th
e
d
ata.
T
ab
le
1
.
E
m
p
lo
y
s
th
e
p
o
v
e
r
ty
d
ataset
o
f
C
en
tr
al
J
av
a
Pro
v
in
ce
R
e
g
e
n
c
y
P
e
r
c
e
n
t
a
g
e
o
f
p
o
o
r
p
o
p
u
l
a
t
i
o
n
(
%)
D
i
d
N
o
t
/
H
a
v
e
N
o
t
C
o
m
p
l
e
t
e
d
P
r
i
ma
r
y
S
c
h
o
o
l
(
>
1
5
Y
e
a
r
s
O
l
d
)
Li
t
e
r
a
c
y
R
a
t
e
(
1
5
–
5
5
Y
e
a
r
s
O
l
d
)
S
c
o
l
P
a
r
t
i
c
i
p
a
t
i
o
n
R
a
t
e
(
1
3
–
1
5
Y
e
a
r
s
O
l
d
)
N
o
t
Emp
l
o
y
e
d
(
>
1
5
Y
e
a
r
s
O
l
d
)
Emp
l
o
y
e
d
i
n
t
h
e
A
g
r
i
c
u
l
t
u
r
a
l
S
e
c
t
o
r
(
>
1
5
Y
e
a
r
s
O
l
d
)
Emp
l
o
y
e
d
i
n
t
h
e
I
n
f
o
r
mal
S
e
c
t
o
r
(
>
1
5
Y
e
a
r
s
O
l
d
)
P
e
r
C
a
p
i
t
a
M
o
n
t
h
l
y
Ex
p
e
n
d
i
t
u
r
e
o
n
F
o
o
d
C
o
mm
o
d
i
t
i
e
s
U
si
n
g
P
r
i
v
a
t
e
/
S
h
a
r
e
d
To
i
l
e
t
C
i
l
a
c
a
p
1
0
.
6
8
1
7
.
7
6
9
5
.
2
8
9
5
.
3
7
3
9
.
5
4
3
2
.
2
5
3
8
.
9
6
6
5
.
6
6
9
3
.
8
8
B
a
n
y
u
ma
s
1
1
.
9
5
1
5
.
2
5
9
5
.
9
1
9
9
.
5
9
3
8
.
0
5
1
4
.
9
3
37
6
4
.
0
3
9
1
.
5
6
P
u
r
b
a
l
i
n
g
g
a
1
4
.
1
8
1
8
.
4
5
9
3
.
8
7
9
4
.
7
7
3
2
.
0
5
1
9
.
0
4
3
9
.
8
9
6
3
.
9
8
9
4
.
1
6
B
a
n
j
a
r
n
e
g
a
r
a
1
4
.
7
1
1
7
.
1
6
9
2
.
9
2
9
4
.
6
4
3
1
.
9
7
3
0
.
7
3
4
6
.
3
4
6
0
.
8
4
9
3
.
5
8
K
e
b
u
me
n
1
5
.
7
1
1
5
.
6
6
9
4
.
8
7
9
9
.
4
7
3
0
.
7
7
2
7
.
6
4
8
.
9
1
6
0
.
0
4
9
8
.
3
5
P
u
r
w
o
r
e
j
o
1
0
.
8
7
1
2
.
1
3
9
5
.
8
1
9
9
.
1
2
9
.
4
3
2
7
.
7
1
4
6
.
7
7
6
1
.
9
2
9
9
.
1
2
W
o
n
o
s
o
b
o
1
5
.
2
8
1
8
.
7
1
9
3
.
0
3
9
0
.
0
8
2
9
.
4
2
3
6
.
4
5
4
7
.
9
1
6
2
.
5
2
9
3
.
8
M
a
g
e
l
a
n
g
1
0
.
8
3
1
4
.
3
7
9
3
.
7
2
9
3
.
9
9
2
8
.
7
9
3
0
.
9
1
4
6
.
6
4
5
8
.
0
5
9
2
.
6
1
B
o
y
o
l
a
l
i
9
.
6
3
1
4
.
6
6
9
2
.
7
7
9
6
.
2
6
2
7
.
3
1
2
6
.
8
4
4
5
.
5
5
6
2
.
1
9
2
.
5
3
K
l
a
t
e
n
1
2
.
0
4
1
2
.
4
8
9
4
.
2
7
9
9
.
9
9
3
3
.
4
3
1
5
.
5
2
3
4
.
1
9
5
9
.
1
9
8
.
1
3
S
u
k
o
h
a
r
j
o
7
.
4
7
8
.
9
8
9
5
.
5
1
9
9
.
9
8
3
3
.
4
8
9
.
3
7
2
8
.
0
3
6
1
.
1
9
8
.
3
3
W
o
n
o
g
i
r
i
1
0
.
7
1
1
7
.
7
4
9
2
.
4
3
9
7
.
0
4
3
1
.
6
7
3
2
.
6
6
4
5
.
7
5
6
2
.
0
8
1
0
0
K
a
r
a
n
g
a
n
y
a
r
9
.
5
9
1
0
.
2
1
9
4
.
0
7
9
8
.
7
3
3
3
.
0
8
1
5
.
8
2
3
1
.
9
5
6
0
.
3
7
9
7
.
5
7
S
r
a
g
e
n
1
2
.
4
1
1
6
.
4
6
8
9
.
7
9
9
4
.
2
1
2
9
.
3
2
8
.
2
9
4
3
.
7
5
5
8
.
4
1
0
0
G
r
o
b
o
g
a
n
1
1
.
4
3
1
1
.
9
4
9
4
.
1
6
9
7
.
0
2
2
9
.
0
1
3
2
.
7
2
5
0
.
2
8
6
5
.
6
8
9
3
.
7
B
l
o
r
a
1
1
.
4
2
1
9
.
8
9
8
8
.
5
6
9
7
.
0
4
2
7
.
1
5
3
8
.
9
4
5
4
.
7
4
6
2
.
3
9
9
4
.
1
4
R
e
m
b
a
n
g
1
4
.
0
2
1
4
.
4
3
9
4
.
9
7
9
9
.
1
7
3
2
.
1
2
6
.
6
6
3
9
.
2
6
6
2
.
2
1
9
0
.
7
4
P
a
t
i
9
.
1
7
1
4
.
7
9
9
4
.
0
1
9
7
.
6
3
5
.
0
0
2
4
.
0
2
4
1
.
4
5
6
1
.
8
3
9
8
.
5
2
K
u
d
u
s
7
.
2
3
8
.
5
6
9
5
.
9
1
9
9
.
9
9
3
0
.
5
7
6
.
9
7
2
5
.
6
6
5
9
.
4
3
9
8
.
6
3
Jep
a
r
a
6
.
0
9
1
1
.
2
2
9
6
.
2
6
9
8
.
9
1
3
4
.
2
6
1
0
.
3
5
3
0
.
5
3
6
3
.
5
9
9
4
.
1
4
D
e
mak
1
1
.
8
9
1
1
.
2
6
9
5
.
9
2
9
8
.
2
5
3
1
.
1
9
1
7
.
7
5
3
3
.
4
2
6
0
.
8
2
9
3
.
7
4
S
e
mara
n
g
6
.
9
6
1
3
.
7
9
9
5
.
6
7
9
6
.
3
1
2
7
.
9
5
1
7
.
5
7
3
6
.
8
1
5
9
.
2
8
9
6
.
3
3
Te
ma
n
g
g
u
n
g
8
.
6
7
1
6
.
7
7
9
6
.
0
7
9
5
.
7
8
2
5
.
9
4
1
.
5
1
5
2
.
6
8
6
3
.
3
9
5
.
0
2
K
e
n
d
a
l
9
.
3
5
1
6
.
3
8
9
5
.
4
7
9
7
.
3
8
3
4
.
3
9
1
9
.
5
7
3
4
.
9
6
2
.
0
6
9
3
.
3
5
B
a
t
a
n
g
8
.
7
3
1
8
.
0
8
9
3
.
5
9
9
3
.
1
3
2
9
.
7
2
0
.
2
5
3
9
.
1
9
6
3
.
9
5
9
4
.
7
3
P
e
k
a
l
o
n
g
a
n
8
.
9
5
1
5
.
9
9
3
.
4
7
9
6
.
3
3
0
.
6
6
1
1
.
2
2
3
3
.
7
5
6
2
.
2
9
9
2
.
2
2
P
e
mal
a
n
g
1
4
.
9
2
2
0
.
9
9
8
9
.
2
7
8
8
.
8
5
3
6
.
4
5
2
0
.
8
3
4
1
.
0
4
6
3
.
1
4
9
1
.
2
6
Te
g
a
l
6
.
8
1
1
9
.
2
1
9
3
.
6
2
9
9
.
9
3
4
0
.
8
5
1
2
.
5
2
30
6
3
.
6
3
9
5
.
1
9
B
r
e
b
e
s
1
5
.
6
2
4
.
7
2
9
2
.
3
7
9
7
.
1
8
3
4
.
6
1
2
7
.
0
2
4
3
.
5
6
6
2
.
2
7
9
1
.
1
5
M
a
g
e
l
a
n
g
C
i
t
y
5
.
9
4
2
.
8
8
9
9
.
4
3
9
9
.
8
1
4
1
.
7
2
0
.
5
2
4
.
1
8
5
9
.
2
2
7
6
.
8
6
S
u
r
a
k
a
r
t
a
C
i
t
y
8
.
3
1
4
.
2
7
9
8
.
3
3
9
9
.
9
6
3
8
.
3
8
NA
2
3
.
5
2
5
6
.
5
8
7
8
.
0
6
S
a
l
a
t
i
g
a
C
i
t
y
4
.
5
7
4
.
4
1
9
8
.
6
3
9
9
.
0
2
3
2
.
0
5
2
.
7
5
2
5
.
4
5
5
.
2
4
1
0
0
.
0
0
S
e
mara
n
g
C
i
t
y
4
.
0
3
5
.
8
6
9
7
.
8
8
9
9
.
9
8
3
4
.
2
9
0
.
7
1
2
1
.
0
6
5
5
.
3
9
9
4
.
9
9
P
e
k
a
l
o
n
g
a
n
C
i
t
y
6
.
7
1
8
.
4
7
9
8
.
6
9
9
6
.
5
2
3
1
.
3
7
1
.
6
7
2
5
.
6
6
2
.
7
8
3
.
9
7
Te
g
a
l
C
i
t
y
7
.
6
4
1
3
.
5
4
9
7
.
9
4
9
8
.
5
6
3
6
.
7
4
.
4
7
2
3
.
9
4
6
0
.
6
3
9
4
.
6
1
2
.
2
.
2
.
Cent
ro
id
ini
t
ia
liza
t
io
n
On
ce
th
e
d
iv
is
iv
e
h
ier
ar
ch
ica
l
p
r
o
ce
s
s
h
as
s
u
cc
ess
f
u
lly
p
a
r
titi
o
n
ed
th
e
d
ataset
i
n
to
k
cl
u
s
ter
s
,
th
e
n
ex
t
s
tep
f
o
cu
s
es
o
n
i
n
teg
r
a
tin
g
th
ese
r
esu
lts
in
to
th
e
h
y
b
r
id
Div
is
iv
e
–
K
-
Me
a
n
s
f
r
a
m
ewo
r
k
f
o
r
f
u
r
th
e
r
r
ef
in
em
en
t.
I
n
th
is
s
tag
e,
th
e
m
ea
n
(
ce
n
tr
o
i
d
)
o
f
ea
c
h
c
lu
s
ter
p
r
o
d
u
ce
d
b
y
th
e
h
ier
a
r
ch
ical
d
iv
is
io
n
is
co
m
p
u
ted
to
r
ep
r
esen
t
th
e
ce
n
tr
al
ten
d
en
c
y
o
f
t
h
e
d
ata
p
o
i
n
ts
with
in
th
at
clu
s
ter
.
T
h
ese
c
alcu
lated
ce
n
tr
o
id
s
s
er
v
e
as
s
tr
ateg
ic
an
d
r
ep
r
esen
tat
iv
e
in
itial
s
ee
d
s
f
o
r
t
h
e
s
u
b
s
eq
u
en
t
K
-
Me
an
s
alg
o
r
ith
m
,
ef
f
ec
tiv
ely
elim
in
atin
g
th
e
r
a
n
d
o
m
n
ess
ty
p
ically
ass
o
ciate
d
with
ce
n
tr
o
id
in
itializatio
n
[
1
9
]
,
[
2
1
]
,
[
2
6
]
.
B
y
u
s
in
g
ce
n
tr
o
id
s
d
e
r
iv
ed
f
r
o
m
th
e
h
ie
r
ar
ch
ical
s
tag
e,
th
e
h
y
b
r
id
m
o
d
el
en
s
u
r
es
th
at
K
-
Me
an
s
b
eg
in
s
its
o
p
tim
izatio
n
f
r
o
m
m
o
r
e
ac
c
u
r
ate
an
d
d
ata
-
i
n
f
o
r
m
e
d
s
tar
tin
g
p
o
in
ts
,
th
er
e
b
y
en
h
an
cin
g
th
e
p
r
ec
is
io
n
an
d
co
n
s
is
ten
cy
o
f
th
e
f
in
al
clu
s
ter
in
g
r
esu
lts
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
1
,
J
an
u
ar
y
20
26
:
258
-
2
6
9
262
2
.
2
.
2
.
K
-
M
ea
ns
re
f
inem
ent
s
t
a
g
e
Fo
llo
win
g
th
e
in
itializatio
n
p
h
ase
u
s
in
g
ce
n
tr
o
id
s
o
b
tain
ed
f
r
o
m
th
e
d
iv
is
iv
e
h
ier
ar
ch
ic
al
p
r
o
ce
s
s
,
th
e
K
-
Me
an
s
alg
o
r
ith
m
is
em
p
lo
y
ed
to
f
u
r
th
e
r
r
ef
in
e
th
e
cl
u
s
ter
ass
ig
n
m
en
ts
an
d
en
h
an
c
e
th
e
o
v
er
all
q
u
ality
o
f
clu
s
ter
in
g
.
At
th
is
s
tag
e,
K
-
Me
an
s
u
tili
ze
s
th
e
ce
n
tr
o
id
s
d
er
iv
ed
f
r
o
m
th
e
d
i
v
is
iv
e
clu
s
ter
in
g
r
esu
lts
as
it
s
in
itial
r
ef
er
en
ce
p
o
i
n
ts
f
o
r
o
p
tim
izatio
n
.
B
ec
au
s
e
th
ese
c
en
tr
o
id
s
alr
ea
d
y
r
ef
lect
th
e
n
atu
r
al
d
iv
is
io
n
s
a
n
d
in
h
er
en
t
s
tr
u
ct
u
r
e
o
f
th
e
d
ata
s
et,
th
e
alg
o
r
ith
m
b
e
g
in
s
with
a
m
o
r
e
in
f
o
r
m
e
d
s
tar
tin
g
c
o
n
f
ig
u
r
atio
n
.
As
a
r
esu
lt,
K
-
Me
an
s
co
n
v
er
g
es
m
o
r
e
r
ap
id
l
y
t
o
war
d
s
tab
le
clu
s
ter
b
o
u
n
d
a
r
ies
an
d
ty
p
ic
ally
r
eq
u
ir
es
f
ewe
r
iter
atio
n
s
to
ac
h
iev
e
o
p
tim
al
r
esu
lts
[
2
]
,
[
2
7
]
,
[
2
8
]
.
T
h
is
in
teg
r
atio
n
en
s
u
r
es
m
o
r
e
ac
cu
r
ate
an
d
r
eliab
le
clu
s
ter
in
g
o
u
tco
m
es c
o
m
p
ar
e
d
to
co
n
v
en
tio
n
al
r
a
n
d
o
m
in
itia
lizatio
n
.
2
.
2
.
3
.
Adv
a
nta
g
es o
f
t
he
hy
b
rid m
et
ho
d
T
h
e
in
teg
r
atio
n
o
f
th
e
Div
is
iv
e
Hier
ar
ch
ical
C
lu
s
ter
in
g
ap
p
r
o
ac
h
with
K
-
Me
an
s
o
f
f
er
s
s
ev
er
al
s
ig
n
if
ican
t
ad
v
a
n
tag
es
th
at
en
h
an
ce
b
o
th
th
e
s
tab
ilit
y
a
n
d
ef
f
icien
cy
o
f
th
e
clu
s
ter
in
g
p
r
o
c
ess
.
First,
th
e
u
s
e
o
f
r
ep
r
esen
tativ
e
in
itial
ce
n
tr
o
id
s
o
b
tain
ed
f
r
o
m
t
h
e
d
i
v
is
iv
e
s
tag
e
p
r
o
v
id
es
a
to
p
-
d
o
wn
an
a
ly
tical
p
er
s
p
ec
tiv
e
,
en
s
u
r
in
g
th
at
th
e
in
itial
p
o
in
t
s
s
elec
ted
f
o
r
K
-
Me
an
s
a
r
e
al
r
ea
d
y
p
o
s
itio
n
ed
clo
s
e
to
th
e
tr
u
e
clu
s
ter
ce
n
ter
s
[
3
]
,
[
2
9
]
.
T
h
is
s
tr
ateg
ic
in
itializatio
n
m
in
im
izes
th
e
r
an
d
o
m
n
e
s
s
th
at
ty
p
ically
af
f
ec
ts
th
e
tr
ad
itio
n
al
K
-
Me
an
s
m
eth
o
d
.
C
o
n
s
eq
u
e
n
tly
,
th
e
clu
s
ter
in
g
r
esu
lts
ex
h
ib
it
g
r
ea
ter
s
tab
ilit
y
an
d
co
n
s
is
ten
cy
,
as
th
e
alg
o
r
ith
m
p
r
o
d
u
ce
s
s
im
ilar
o
u
tco
m
es
ac
r
o
s
s
m
u
ltip
le
r
u
n
s
r
ath
e
r
t
h
an
f
lu
ctu
atin
g
d
u
e
to
r
an
d
o
m
ce
n
t
r
o
id
s
elec
ti
o
n
[
1
3
]
,
[
3
0
]
.
Mo
r
eo
v
er
,
b
ec
au
s
e
th
e
ce
n
tr
o
id
s
ar
e
i
n
itialized
b
ased
o
n
m
ea
n
in
g
f
u
l
s
tr
u
ct
u
r
al
d
iv
is
io
n
s
with
in
th
e
d
ataset,
K
-
Me
an
s
co
n
v
er
g
es
f
aster
an
d
r
eq
u
ir
es
f
ewe
r
iter
at
io
n
s
to
r
ea
ch
an
o
p
tim
al
s
o
lu
ti
o
n
[
1
5
]
,
[
3
1
]
,
[
3
2
]
.
T
h
is
im
p
r
o
v
em
en
t
n
o
t
o
n
ly
r
ed
u
ce
s
co
m
p
u
tatio
n
al
tim
e
b
u
t
also
en
h
an
ce
s
th
e
o
v
er
all
ac
cu
r
ac
y
an
d
in
ter
p
r
etab
ilit
y
o
f
th
e
r
esu
lti
n
g
clu
s
ter
s
,
m
a
k
in
g
th
e
h
y
b
r
id
Div
is
iv
e
–
K
-
Me
an
s
m
et
h
o
d
a
m
o
r
e
r
o
b
u
s
t
alter
n
ativ
e
f
o
r
c
o
m
p
le
x
d
ata
a
n
aly
s
is
.
2
.
2
.
4
.
L
im
it
a
t
io
ns
Desp
ite
its
ad
v
an
tag
es
in
p
r
o
d
u
cin
g
m
o
r
e
ac
cu
r
ate
an
d
well
-
s
tr
u
ctu
r
ed
clu
s
ter
s
,
th
e
Div
is
iv
e
–
K
-
Me
an
s
h
y
b
r
id
m
eth
o
d
also
p
r
esen
ts
s
ev
er
al
co
m
p
u
tatio
n
al
l
im
itatio
n
s
th
at
m
u
s
t
b
e
c
o
n
s
id
er
ed
wh
e
n
a
p
p
lied
to
lar
g
e
-
s
ca
le
d
ata
an
aly
s
is
.
On
e
o
f
th
e
p
r
im
ar
y
d
r
awb
ac
k
s
is
th
at
d
iv
i
s
iv
e
clu
s
ter
in
g
is
co
m
p
u
tatio
n
ally
d
em
an
d
in
g
,
as
it
in
v
o
lv
es
a
to
p
-
d
o
wn
s
p
litt
in
g
p
r
o
ce
s
s
th
at
r
eq
u
ir
es
ev
alu
atin
g
n
u
m
er
o
u
s
p
o
s
s
ib
le
p
ar
titi
o
n
in
g
s
tr
ateg
ies
b
ef
o
r
e
d
eter
m
in
in
g
th
e
o
p
tim
al
d
iv
is
i
o
n
o
f
d
ata
[
6
]
,
[
7
]
,
[
2
3
]
.
T
h
is
ev
alu
atio
n
p
r
o
ce
s
s
ca
n
s
ig
n
if
ican
tly
in
cr
ea
s
e
c
o
m
p
u
tatio
n
al
l
o
ad
,
esp
ec
ially
wh
en
d
ea
lin
g
with
h
ig
h
-
d
im
e
n
s
io
n
al
o
r
co
m
p
lex
d
atasets
.
Ad
d
itio
n
ally
,
s
im
ilar
to
t
h
e
a
g
g
lo
m
er
ativ
e
ap
p
r
o
ac
h
,
d
iv
is
iv
e
clu
s
ter
in
g
is
n
o
t
id
ea
l
f
o
r
v
er
y
lar
g
e
d
atasets
,
as
th
e
r
ec
u
r
s
iv
e
s
p
lit
tin
g
an
d
d
is
tan
ce
ca
lcu
latio
n
s
d
em
an
d
s
u
b
s
tan
tial
co
m
p
u
tati
o
n
al
r
eso
u
r
ce
s
an
d
m
em
o
r
y
ca
p
ac
ity
[
2
1
]
,
[
3
3
]
.
C
o
n
s
eq
u
en
tly
,
wh
ile
th
e
m
eth
o
d
o
f
f
er
s
im
p
r
o
v
ed
ac
cu
r
ac
y
an
d
s
tab
ilit
y
in
clu
s
ter
in
g
r
esu
lts
,
it
m
ay
b
e
co
m
e
im
p
r
ac
tical
f
o
r
lar
g
e
-
s
ca
le
o
r
r
ea
l
-
tim
e
ap
p
licatio
n
s
with
o
u
t
f
u
r
th
er
o
p
tim
izat
io
n
o
r
th
e
u
s
e
o
f
p
a
r
allel
p
r
o
ce
s
s
in
g
tech
n
iq
u
es.
T
h
e
co
m
b
i
n
atio
n
o
f
h
ier
a
r
ch
i
ca
l
an
d
p
a
r
titi
o
n
in
g
tech
n
iq
u
es
in
b
o
th
h
y
b
r
id
m
eth
o
d
s
is
d
esig
n
ed
to
ad
d
r
ess
th
e
wea
k
n
ess
es
o
f
K
-
Me
an
s
wh
ile
m
ain
tain
in
g
co
m
p
u
tatio
n
al
e
f
f
icien
cy
[
1
3
]
,
[
2
9
]
,
[
3
4
]
.
2
.
3
.
E
v
a
lua
t
i
o
n
m
et
rics
T
o
co
m
p
r
eh
en
s
iv
ely
ass
ess
cl
u
s
ter
in
g
p
er
f
o
r
m
a
n
ce
,
th
is
s
tu
d
y
ap
p
lies
th
r
ee
ca
teg
o
r
ies
o
f
ev
alu
atio
n
m
etr
ics:
ex
ec
u
tio
n
tim
e,
co
n
v
er
g
en
ce
iter
atio
n
s
,
an
d
clu
s
ter
v
alid
ity
in
d
ices
.
T
h
ese
m
etr
i
cs
ca
p
tu
r
e
n
o
t
o
n
ly
co
m
p
u
tatio
n
al
ef
f
icien
cy
b
u
t
also
th
e
s
tab
ili
ty
an
d
q
u
ality
o
f
clu
s
ter
in
g
r
esu
lts
,
wh
ich
ar
e
ess
en
tial
f
o
r
ev
alu
atin
g
clu
s
ter
in
g
alg
o
r
ith
m
s
in
b
ig
d
ata
e
n
v
ir
o
n
m
en
ts
[
1
]
,
[
2
]
,
[
1
5
]
,
[
2
0
]
.
2
.
3
.
1
.
E
x
ec
utio
n
t
im
e
E
x
ec
u
tio
n
tim
e
r
e
f
er
s
to
th
e
to
tal
am
o
u
n
t
o
f
tim
e
ta
k
en
b
y
e
ac
h
clu
s
ter
in
g
alg
o
r
ith
m
to
co
m
p
lete
th
e
clu
s
ter
in
g
p
r
o
ce
s
s
,
m
ea
s
u
r
ed
in
s
ec
o
n
d
s
.
I
n
th
is
s
tu
d
y
,
ex
e
cu
tio
n
tim
e
was
r
ec
o
r
d
e
d
u
s
i
n
g
Py
th
o
n
’
s
b
u
ilt
-
in
tim
e
f
u
n
ctio
n
,
wh
ich
ca
p
tu
r
es
th
e
d
u
r
atio
n
f
r
o
m
th
e
in
itializatio
n
o
f
th
e
alg
o
r
ith
m
to
its
co
n
v
er
g
en
ce
.
T
h
is
m
etr
ic
d
ir
ec
tly
ev
alu
ates
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
wh
ich
i
s
p
ar
ticu
lar
ly
cr
itical
in
th
e
c
o
n
tex
t
o
f
b
ig
d
ata
an
aly
s
is
,
wh
er
e
clu
s
ter
in
g
m
e
th
o
d
s
m
u
s
t
b
e
b
o
t
h
ac
cu
r
ate
an
d
s
ca
lab
le
[
3
]
,
[
1
1
]
,
[
2
0
]
.
Hy
b
r
id
ap
p
r
o
ac
h
es,
s
u
ch
as
Ag
g
lo
m
er
ati
v
e
K
-
Me
an
s
an
d
Div
is
iv
e
K
-
Me
an
s
,
a
r
e
ex
p
ec
ted
t
o
r
e
d
u
ce
o
v
er
all
c
o
m
p
u
tatio
n
tim
e
b
y
im
p
r
o
v
in
g
ce
n
tr
o
id
in
itializatio
n
,
wh
ich
lead
s
to
f
aster
co
n
v
er
g
en
ce
d
esp
ite
th
e
ad
d
itio
n
al
h
ier
ar
c
h
ical
o
v
er
h
ea
d
[
1
]
,
[
3
1
]
,
[
3
5
]
.
Pr
ev
io
u
s
s
tu
d
ies
h
av
e
d
em
o
n
s
t
r
ated
th
at
s
u
ch
h
y
b
r
id
h
ier
a
r
ch
ical
–
p
ar
titi
o
n
in
g
m
eth
o
d
s
ca
n
ac
h
iev
e
a
b
alan
c
e
b
etwe
en
ac
cu
r
ac
y
an
d
ef
f
ici
en
cy
,
o
u
tp
e
r
f
o
r
m
in
g
tr
ad
itio
n
a
l K
-
Me
an
s
in
lar
g
e
-
s
ca
le
d
atasets
[
1
3
]
,
[
3
2
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
h
yb
r
id
d
ivis
ive
K
-
mea
n
s
fr
a
mewo
r
k
fo
r
b
ig
d
a
ta
–
d
r
iven
p
o
ve
r
ty
a
n
a
lysi
s
in
C
en
tr
a
l
… (
B
o
w
o
Win
a
r
n
o
)
263
2
.
3
.
2
.
Co
nv
er
g
ence
it
er
a
t
io
ns
C
o
n
v
er
g
en
ce
iter
atio
n
s
r
ep
r
esen
t
th
e
n
u
m
b
er
o
f
r
ef
in
e
m
en
t
s
tep
s
K
-
Me
an
s
r
eq
u
ir
es
to
s
ta
b
ilize
af
ter
ce
n
tr
o
id
in
itializatio
n
.
A
lo
we
r
n
u
m
b
e
r
o
f
iter
atio
n
s
in
d
icate
s
th
at
ce
n
tr
o
id
s
wer
e
in
itialize
d
clo
s
er
to
o
p
tim
al
p
o
s
itio
n
s
,
lead
in
g
to
f
aster
c
o
n
v
er
g
en
ce
a
n
d
r
e
d
u
ce
d
co
m
p
u
tatio
n
al
lo
ad
[
3
1
]
,
[
1
9
]
,
[
2
6
]
.
I
n
s
tan
d
ar
d
K
-
Me
an
s
,
p
o
o
r
ce
n
tr
o
i
d
in
itializatio
n
ca
n
lead
to
m
u
ltip
le
r
ed
u
n
d
an
t
iter
atio
n
s
,
in
cr
ea
s
in
g
b
o
th
ex
ec
u
tio
n
tim
e
an
d
th
e
r
is
k
o
f
s
u
b
o
p
tim
al
clu
s
ter
in
g
[
2
]
,
[
1
5
]
,
[
3
6
]
.
I
n
c
o
n
tr
ast,
h
y
b
r
id
m
eth
o
d
s
s
u
ch
as
h
ier
ar
ch
ical
-
b
ased
o
r
m
etah
eu
r
is
tic
-
ass
is
ted
ce
n
tr
o
id
in
itializatio
n
im
p
r
o
v
e
co
n
v
er
g
en
ce
b
y
p
r
o
v
id
in
g
b
etter
s
tar
tin
g
ce
n
tr
o
id
s
,
th
er
eb
y
ac
ce
ler
atin
g
th
e
s
tab
il
izatio
n
p
r
o
ce
s
s
[
1
]
,
[
2
6
]
,
[
2
7
]
,
[
3
7
]
.
T
h
is
m
etr
ic
is
ess
en
tial
f
o
r
co
m
p
a
r
in
g
th
e
ef
f
icie
n
cy
an
d
s
tab
ilit
y
b
etwe
en
co
n
v
en
tio
n
al
an
d
h
y
b
r
id
c
lu
s
ter
in
g
alg
o
r
ith
m
s
,
as
it
h
ig
h
lig
h
ts
th
e
r
o
le
o
f
in
itializatio
n
in
th
e
o
p
tim
izati
o
n
o
f
clu
s
ter
in
g
p
e
r
f
o
r
m
an
ce
[
1
9
]
,
[
2
1
]
,
[
3
8
]
.
2
.
4
.
Clus
t
er
v
a
lid
it
y
ind
ices
T
o
ev
alu
ate
th
e
q
u
ality
o
f
th
e
r
esu
ltin
g
clu
s
ter
s
,
th
r
ee
in
ter
n
al
v
alid
atio
n
in
d
ices
ar
e
em
p
lo
y
ed
:
Sil
h
o
u
ette
C
o
ef
f
icien
t
,
Dav
i
es
–
B
o
u
ld
in
I
n
d
ex
(
DB
I
)
,
an
d
C
alin
s
k
i
–
Har
ab
asz
I
n
d
e
x
(
C
H
I
n
d
ex
)
.
T
h
ese
in
d
ices m
ea
s
u
r
e
clu
s
ter
co
h
esio
n
,
s
ep
ar
atio
n
,
a
n
d
v
ar
ia
n
ce
s
tr
u
ctu
r
e
to
p
r
o
v
id
e
a
c
o
m
p
r
e
h
e
n
s
iv
e
ev
alu
atio
n
o
f
clu
s
ter
in
g
q
u
ality
[
1
0
]
,
[
1
4
]
,
[
2
8
]
.
T
h
e
Sil
h
o
u
ette
C
o
ef
f
icien
t
:
is
o
n
e
o
f
th
e
m
o
s
t
wid
ely
u
s
ed
in
ter
n
al
v
alid
atio
n
i
n
d
ices
f
o
r
clu
s
ter
in
g
ev
alu
atio
n
.
I
t
p
r
o
v
id
es
a
q
u
a
n
titativ
e
m
ea
s
u
r
e
o
f
h
o
w
wel
l
ea
ch
d
a
ta
p
o
in
t
f
its
with
in
its
ass
ig
n
ed
clu
s
ter
co
m
p
ar
ed
to
o
t
h
er
clu
s
ter
s
.
T
h
e
in
d
ex
co
m
b
in
es
two
k
ey
a
s
p
ec
ts
o
f
clu
s
ter
in
g
q
u
ality
:
c
o
h
esio
n
(
th
e
d
eg
r
ee
o
f
s
im
ilar
ity
b
etwe
en
a
d
ata
p
o
in
t
an
d
o
t
h
er
p
o
i
n
ts
in
th
e
s
am
e
clu
s
ter
)
an
d
s
ep
ar
at
io
n
(
th
e
d
eg
r
ee
o
f
d
is
s
im
ilar
ity
b
etwe
en
a
d
ata
p
o
in
t a
n
d
p
o
in
ts
in
th
e
n
ea
r
est n
eig
h
b
o
r
i
n
g
clu
s
ter
)
.
Fo
r
ea
ch
d
ata
p
o
in
t
i
,
t
h
e
Si
lh
o
u
ette
v
al
u
e
s
(
i)
is
d
ef
i
n
ed
as:
(
)
=
(
)
−
(
)
m
ax
{
(
)
,
(
)
}
th
e
Sil
h
o
u
ette
C
o
ef
f
icien
t
f
o
r
m
u
latio
n
,
two
m
ain
co
m
p
o
n
e
n
ts
ar
e
u
s
ed
t
o
ev
alu
ate
th
e
clu
s
ter
in
g
q
u
ality
o
f
ea
ch
d
ata
p
o
in
t.
T
h
e
f
ir
s
t c
o
m
p
o
n
en
t,
a(
i)
,
r
ep
r
esen
ts
th
e
av
er
ag
e
d
is
tan
ce
b
e
twee
n
a
s
p
ec
if
ic
d
ata
p
o
in
t
(
i)
an
d
all
o
th
er
p
o
in
ts
with
in
th
e
s
am
e
clu
s
ter
.
T
h
is
v
alu
e
m
ea
s
u
r
es
co
h
esio
n
,
o
r
h
o
w
clo
s
ely
r
elate
d
th
e
p
o
i
n
t
is
to
th
e
m
em
b
er
s
o
f
its
o
wn
clu
s
ter
.
T
h
e
s
ec
o
n
d
c
o
m
p
o
n
en
t,
b
(
i)
,
d
en
o
tes
th
e
m
in
im
u
m
av
e
r
ag
e
d
is
tan
ce
b
et
wee
n
th
e
s
am
e
p
o
in
t
(
i)
an
d
all
p
o
in
ts
b
elo
n
g
in
g
to
o
th
er
clu
s
ter
s
,
wh
ich
r
ef
lects
s
ep
ar
atio
n
,
o
r
h
o
w
d
is
t
in
ct
th
e
p
o
in
t
is
f
r
o
m
o
th
er
clu
s
ter
s
.
B
y
co
m
p
ar
in
g
th
es
e
two
v
alu
es,
th
e
Sil
h
o
u
ette
C
o
ef
f
icien
t
ass
es
s
es
wh
eth
er
a
d
ata
p
o
in
t
is
ap
p
r
o
p
r
iately
ass
ig
n
ed
to
its
cl
u
s
ter
,
b
alan
cin
g
b
o
th
in
ter
n
al
s
im
ilar
ity
an
d
ex
ter
n
al
d
is
s
im
ilar
ity
.
T
h
e
Sil
h
o
u
ette
C
o
ef
f
icie
nt
is
a
wid
ely
u
s
ed
m
etr
ic
f
o
r
e
v
alu
atin
g
clu
s
ter
in
g
q
u
ality
,
w
ith
v
alu
es
r
an
g
in
g
f
r
o
m
–
1
to
+1
.
A
co
e
f
f
icien
t
v
alu
e
clo
s
e
to
+1
in
d
i
ca
tes
th
at
a
d
ata
p
o
in
t
is
well
-
m
atch
ed
to
its
o
wn
clu
s
ter
an
d
d
is
tin
ctly
s
ep
ar
ated
f
r
o
m
n
eig
h
b
o
r
in
g
cl
u
s
ter
s
,
r
ef
lectin
g
a
well
-
d
ef
in
ed
a
n
d
c
o
h
esiv
e
clu
s
ter
in
g
s
tr
u
ctu
r
e.
C
o
n
v
er
s
ely
,
a
v
al
u
e
n
ea
r
0
s
u
g
g
ests
th
at
th
e
d
ata
p
o
in
t
lies
o
n
th
e
b
o
u
n
d
a
r
y
b
et
wee
n
two
o
r
m
o
r
e
clu
s
ter
s
,
in
d
icatin
g
p
o
ten
tial o
v
er
lap
o
r
am
b
i
g
u
ity
in
clu
s
ter
m
em
b
er
s
h
ip
.
Me
an
wh
ile,
a
v
a
lu
e
ap
p
r
o
ac
h
in
g
–
1
i
m
p
lies
th
at
th
e
d
ata
p
o
in
t
m
a
y
h
av
e
b
ee
n
in
co
r
r
ec
tly
ass
ig
n
ed
to
its
cu
r
r
en
t
clu
s
ter
,
as
it
is
m
o
r
e
s
im
ilar
to
p
o
in
ts
in
an
o
th
er
clu
s
ter
[
9
]
.
T
h
is
r
an
g
e
allo
ws
r
esear
ch
er
s
to
ass
e
s
s
b
o
th
th
e
o
v
er
all
clu
s
ter
in
g
p
er
f
o
r
m
an
ce
an
d
th
e
a
p
p
r
o
p
r
iaten
ess
o
f
in
d
iv
id
u
al
d
ata
p
o
in
t a
s
s
ig
n
m
en
ts
with
in
th
e
m
o
d
el.
Hig
h
Sil
h
o
u
ette
s
co
r
es
s
u
g
g
est
co
m
p
ac
t
an
d
well
-
s
ep
ar
a
ted
clu
s
ter
s
,
wh
ile
lo
w
s
co
r
es
in
d
icate
o
v
er
lap
o
r
wea
k
s
tr
u
ctu
r
e
[
1
0
]
,
[
1
4
]
.
T
h
is
m
etr
ic
is
f
r
eq
u
en
tly
u
s
ed
in
c
o
m
p
ar
ativ
e
s
tu
d
ies
o
f
clu
s
ter
in
g
alg
o
r
ith
m
s
to
ev
al
u
ate
th
e
ef
f
e
ctiv
en
ess
o
f
in
itializatio
n
an
d
t
h
e
o
p
tim
al
n
u
m
b
er
o
f
cl
u
s
ter
s
[
1
]
,
[
1
5
]
,
[
2
8
]
.
T
h
e
Dav
ies
–
B
o
u
ld
in
I
n
d
ex
(
DB
I
)
:
is
an
in
ter
n
al
clu
s
ter
v
alid
ity
m
etr
ic
th
at
ev
alu
ates
t
h
e
av
er
a
g
e
s
im
ilar
ity
b
etwe
en
clu
s
ter
s
b
y
c
o
n
s
id
er
in
g
b
o
th
th
e
co
m
p
ac
tn
ess
with
in
clu
s
ter
s
an
d
th
e
s
ep
ar
atio
n
b
etwe
en
clu
s
ter
s
.
I
t
was
f
ir
s
t
in
tr
o
d
u
ce
d
b
y
Dav
ies
an
d
B
o
u
ld
i
n
(
1
9
7
9
)
an
d
h
as
s
in
ce
b
ee
n
wid
el
y
u
s
ed
f
o
r
ass
ess
in
g
clu
s
ter
in
g
q
u
ality
in
u
n
s
u
p
er
v
is
ed
lear
n
in
g
.
Fo
r
ea
ch
cl
u
s
ter
i
,
th
e
DB
I
is
ca
lcu
lated
as
th
e
av
er
ag
e
o
f
t
h
e
m
ax
im
u
m
s
im
ilar
ity
v
alu
es
b
e
twee
n
clu
s
ter
i
an
d
all
o
th
er
clu
s
ter
s
j
.
T
h
e
s
im
ilar
ity
m
ea
s
u
r
e
is
d
ef
in
ed
as
th
e
r
atio
b
etwe
en
th
e
with
in
-
clu
s
ter
s
ca
tter
(
h
o
w
co
m
p
ac
t
t
h
e
clu
s
ter
is
)
an
d
th
e
d
is
tan
ce
b
etwe
en
clu
s
ter
ce
n
tr
o
id
s
(
h
o
w
f
ar
a
p
ar
t two
c
lu
s
ter
s
ar
e)
.
Ma
th
em
atica
lly
,
t
h
e
DB
I
is
ex
p
r
ess
ed
as:
=
1
∑
ma
x
≠
(
+
)
=
1
I
n
th
e
f
o
r
m
u
latio
n
o
f
th
e
DB
I
,
s
ev
er
al
p
a
r
am
eter
s
ar
e
u
s
ed
t
o
ev
alu
ate
clu
s
ter
in
g
p
er
f
o
r
m
an
ce
b
ased
o
n
in
tr
a
-
clu
s
ter
s
im
ilar
ity
a
n
d
in
ter
-
clu
s
ter
s
ep
ar
atio
n
.
Her
e,
k
r
ep
r
esen
ts
th
e
to
tal
n
u
m
b
er
o
f
clu
s
ter
s
f
o
r
m
ed
b
y
th
e
alg
o
r
ith
m
.
T
h
e
ter
m
S
ᵢ
d
en
o
tes
th
e
a
v
er
ag
e
d
is
tan
ce
o
f
all
d
ata
p
o
in
ts
with
in
clu
s
ter
i
to
its
ce
n
tr
o
id
,
wh
ich
m
ea
s
u
r
es
th
e
in
tr
a
-
clu
s
ter
d
is
tan
ce
o
r
h
o
w
co
m
p
ac
t
ea
ch
clu
s
ter
is
.
Me
an
wh
ile,
M
ᵢⱼ
r
ef
er
s
to
th
e
d
is
tan
ce
b
etwe
en
th
e
ce
n
tr
o
id
s
o
f
cl
u
s
ter
s
i
a
n
d
j
,
ca
p
tu
r
in
g
th
e
in
ter
-
cl
u
s
ter
d
is
tan
c
e
o
r
t
h
e
d
e
g
r
ee
o
f
s
ep
ar
atio
n
b
etwe
en
clu
s
ter
s
.
B
y
an
aly
zin
g
th
e
b
alan
ce
b
et
wee
n
th
ese
two
d
is
tan
ce
s
,
th
e
D
B
I
p
r
o
v
id
es
an
o
v
er
all
m
ea
s
u
r
e
o
f
cl
u
s
ter
in
g
ef
f
ec
tiv
en
ess
,
wh
er
e
l
o
wer
v
alu
es
in
d
icate
b
etter
-
d
e
f
in
ed
an
d
m
o
r
e
d
is
tin
ct
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
1
,
J
an
u
ar
y
20
26
:
258
-
2
6
9
264
clu
s
ter
s
.
L
o
wer
DB
I
v
alu
es
in
d
icate
b
etter
clu
s
ter
in
g
co
m
p
ac
t
clu
s
ter
s
an
d
h
ig
h
s
ep
ar
at
io
n
[
1
4
]
,
[
2
8
]
.
T
h
e
DB
I
is
p
ar
ticu
lar
ly
ef
f
ec
tiv
e
f
o
r
im
b
alan
ce
d
o
r
v
a
r
iab
le
-
d
en
s
ity
d
atasets
,
m
ak
in
g
it
a
s
u
itab
le
in
d
ex
f
o
r
ev
alu
atin
g
h
y
b
r
id
clu
s
ter
in
g
al
g
o
r
ith
m
s
in
b
ig
d
ata
a
n
aly
s
is
[
1
]
,
[
1
0
]
,
[
2
0
]
.
T
h
e
C
al
in
s
k
i
–
Har
ab
asz
I
n
d
ex
(
C
H
I
n
d
ex
)
:
also
r
ef
er
r
e
d
to
a
s
th
e
Var
ian
ce
R
atio
C
r
iter
io
n
(
VR
C
)
,
is
a
n
in
ter
n
al
clu
s
ter
in
g
v
alid
atio
n
m
etr
ic
th
at
ev
alu
ates th
e
q
u
ality
o
f
a
clu
s
ter
in
g
s
tr
u
ctu
r
e
b
ased
o
n
th
e
r
atio
o
f
b
etwe
en
-
clu
s
ter
d
is
p
er
s
io
n
t
o
with
in
-
clu
s
ter
d
is
p
er
s
io
n
.
I
t
was
f
ir
s
t
p
r
o
p
o
s
ed
b
y
C
aliń
s
k
i
an
d
Har
a
b
asz
(
1
9
7
4
)
an
d
h
as
s
in
ce
b
ee
n
w
id
ely
ad
o
p
te
d
as
a
r
eliab
le
m
ea
s
u
r
e
f
o
r
d
eter
m
in
in
g
th
e
o
p
tim
al
n
u
m
b
er
o
f
clu
s
ter
s
in
u
n
s
u
p
er
v
is
ed
lear
n
i
n
g
.
Ma
th
em
atica
lly
,
th
e
C
H
I
n
d
ex
is
d
ef
in
ed
as:
(
)
=
(
)
(
)
−
−
1
I
n
th
e
co
m
p
u
tatio
n
o
f
th
e
C
alin
s
k
i
–
Har
ab
asz
I
n
d
e
x
(
C
H
I
n
d
ex
)
,
s
ev
er
al
k
ey
p
ar
am
ete
r
s
ar
e
em
p
l
o
y
ed
to
m
ea
s
u
r
e
th
e
b
alan
ce
b
etwe
en
clu
s
ter
s
ep
ar
atio
n
an
d
co
m
p
ac
tn
ess
.
T
h
e
v
ar
iab
le
N
d
en
o
tes
th
e
to
tal
n
u
m
b
er
o
f
d
ata
p
o
in
ts
i
n
th
e
d
ataset,
wh
il
e
k
r
ep
r
esen
ts
th
e
n
u
m
b
er
o
f
c
lu
s
ter
s
f
o
r
m
ed
b
y
th
e
al
g
o
r
it
h
m
.
T
h
e
ter
m
T
r
(
B
ₖ)
r
ef
er
s
to
th
e
tr
ac
e
o
f
th
e
b
et
wee
n
-
clu
s
ter
d
is
p
er
s
io
n
m
atr
i
x
,
wh
ich
q
u
an
tifie
s
th
e
v
ar
ia
n
ce
o
f
th
e
clu
s
ter
ce
n
tr
o
id
s
r
elativ
e
to
t
h
e
o
v
e
r
all
m
ea
n
a
m
ea
s
u
r
e
o
f
h
o
w
well
clu
s
ter
s
ar
e
s
ep
ar
ate
d
f
r
o
m
ea
c
h
o
th
er
.
C
o
n
v
er
s
ely
,
T
r
(
W
ₖ)
in
d
icate
s
th
e
tr
ac
e
o
f
th
e
with
in
-
clu
s
ter
d
is
p
er
s
io
n
m
atr
ix
,
r
ef
lectin
g
th
e
v
ar
ian
ce
o
f
d
ata
p
o
in
ts
with
in
ea
ch
clu
s
ter
,
o
r
h
o
w
tig
h
tly
g
r
o
u
p
e
d
th
e
m
em
b
er
s
o
f
a
clu
s
ter
ar
e.
A
h
ig
h
e
r
C
alin
s
k
i
–
Har
ab
asz
I
n
d
ex
v
alu
e
s
u
g
g
ests
th
at
t
h
e
clu
s
ter
in
g
s
tr
u
ctu
r
e
ex
h
i
b
its
b
o
th
s
tr
o
n
g
in
ter
-
clu
s
ter
s
ep
ar
a
tio
n
an
d
lo
w
in
tr
a
-
clu
s
ter
v
ar
ian
ce
,
in
d
icatin
g
b
et
ter
clu
s
ter
in
g
q
u
ality
.
A
h
ig
h
er
C
H
I
n
d
ex
v
alu
e
in
d
icate
s
b
etter
clu
s
ter
in
g
q
u
ality
,
as
it
r
ef
lects
clu
s
ter
s
th
at
ar
e
well
-
s
ep
ar
ated
f
r
o
m
ea
ch
o
th
er
(
h
ig
h
b
etwe
en
-
clu
s
ter
v
ar
ian
c
e)
an
d
in
ter
n
ally
c
o
m
p
ac
t
(
l
o
w
with
in
-
clu
s
ter
v
ar
ian
ce
)
.
Un
lik
e
th
e
DB
I
,
w
h
er
e
lo
wer
v
al
u
es
ar
e
p
r
e
f
er
r
e
d
,
th
e
C
H
I
n
d
e
x
f
av
o
r
s
h
ig
h
e
r
v
alu
es
as
a
s
ig
n
o
f
o
p
tim
al
p
ar
titi
o
n
in
g
[
1
]
,
[
1
4
]
.
Hig
h
er
C
H
v
alu
es
in
d
icate
b
etter
clu
s
ter
in
g
,
s
ig
n
if
y
in
g
h
ig
h
i
n
ter
-
clu
s
ter
v
ar
ian
ce
an
d
lo
w
in
tr
a
-
clu
s
te
r
v
ar
ian
ce
[
1
0
]
,
[
2
8
]
.
T
h
e
C
H
I
n
d
ex
is
wid
ely
ap
p
lied
f
o
r
m
o
d
el
s
ele
ctio
n
to
d
eter
m
in
e
t
h
e
o
p
tim
al
n
u
m
b
e
r
o
f
clu
s
ter
s
,
co
m
p
lem
en
tin
g
th
e
Sil
h
o
u
ette
a
n
d
DB
I
m
etr
i
cs
[
1
0
]
,
[
1
5
]
,
[
2
8
]
.
Prio
r
s
tu
d
ies
h
av
e
co
n
f
ir
m
e
d
its
ef
f
ec
tiv
en
ess
in
h
y
b
r
id
h
ier
ar
ch
ical
–
p
ar
titi
o
n
i
n
g
clu
s
ter
in
g
,
esp
ec
ially
f
o
r
m
ed
iu
m
to
lar
g
e
-
s
ca
le
d
atasets
,
d
u
e
to
its
s
en
s
itiv
ity
to
b
o
th
co
m
p
ac
tn
ess
an
d
s
ep
ar
atio
n
[
1
]
,
[
1
3
]
,
[
3
0
]
.
Alo
n
g
s
id
e
th
e
Sil
h
o
u
ette
C
o
ef
f
icien
t
an
d
DB
I
,
th
e
CH
I
n
d
e
x
o
f
f
e
r
s
a
co
m
p
lem
e
n
tar
y
p
er
s
p
ec
tiv
e
in
ev
alu
atin
g
th
e
o
v
e
r
all
p
er
f
o
r
m
an
ce
an
d
q
u
ality
o
f
cl
u
s
ter
in
g
r
esu
lts
.
T
o
g
eth
er
,
t
h
es
e
m
etr
ics
p
r
o
v
i
d
e
a
b
alan
ce
d
ass
ess
m
en
t a
cr
o
s
s
d
if
f
er
en
t a
s
p
ec
ts
o
f
cl
u
s
ter
in
g
ef
f
ec
tiv
en
ess
.
T
h
e
ex
ec
u
tio
n
tim
e
m
etr
ic
m
e
asu
r
es
co
m
p
u
tatio
n
al
ef
f
icien
cy
,
in
d
icatin
g
h
o
w
q
u
ick
ly
a
n
alg
o
r
ith
m
ca
n
p
r
o
d
u
ce
r
esu
lts
with
o
u
t
c
o
m
p
r
o
m
is
in
g
ac
cu
r
ac
y
[
1
]
,
[
3
]
,
[
1
1
]
,
[
3
2
]
.
Me
a
n
wh
ile,
C
o
n
v
er
g
en
ce
I
ter
atio
n
s
r
ef
lect
b
o
th
th
e
ef
f
ec
tiv
en
es
s
o
f
ce
n
tr
o
id
i
n
itializatio
n
an
d
th
e
s
tab
ilit
y
o
f
t
h
e
alg
o
r
ith
m
,
wh
er
e
f
ewe
r
iter
at
io
n
s
g
en
er
ally
s
ig
n
if
y
a
m
o
r
e
o
p
tim
ized
an
d
co
n
s
is
t
en
t
p
r
o
ce
s
s
[
1
9
]
,
[
2
6
]
,
[
2
7
]
,
[
3
1
]
.
Fin
ally
,
th
e
co
m
b
in
atio
n
o
f
C
lu
s
ter
Valid
ity
I
n
d
ices
in
clu
d
in
g
Sil
h
o
u
e
tte
,
DB
I
,
an
d
CH
s
er
v
es
to
ev
alu
ate
clu
s
ter
in
g
ac
cu
r
ac
y
a
n
d
s
tr
u
ct
u
r
al
q
u
ality
,
p
r
o
v
id
in
g
in
s
ig
h
ts
in
t
o
h
o
w
well
clu
s
ter
s
ar
e
f
o
r
m
ed
a
n
d
h
o
w
d
is
tin
ct
th
ey
ar
e
f
r
o
m
o
n
e
a
n
o
th
er
[
1
0
]
,
[
1
4
]
,
[
1
5
]
,
[
2
8
]
.
T
o
g
eth
er
,
t
h
ese
ev
alu
atio
n
m
etr
ics
f
o
r
m
a
co
m
p
r
eh
en
s
iv
e
f
r
am
ewo
r
k
f
o
r
ass
ess
in
g
clu
s
ter
in
g
p
er
f
o
r
m
an
ce
f
r
o
m
m
u
ltip
le
d
im
en
s
io
n
s
:
ac
cu
r
a
cy
,
s
tab
ilit
y
,
an
d
c
o
m
p
u
tatio
n
al
e
f
f
icien
cy
.
T
h
is
in
teg
r
ated
e
v
alu
atio
n
f
r
am
ew
o
r
k
en
s
u
r
es
a
b
alan
ce
d
an
d
o
b
jectiv
e
co
m
p
ar
is
o
n
b
etwe
en
s
tan
d
ar
d
K
-
Me
an
s
an
d
h
y
b
r
id
ap
p
r
o
ac
h
es,
r
ev
ea
lin
g
th
e
tr
ad
e
-
o
f
f
s
b
etwe
en
ef
f
icien
cy
,
s
tab
ilit
y
,
an
d
clu
s
ter
q
u
ality
[
1
]
,
[
2
]
,
[
1
5
]
,
[
2
0
]
,
[
2
8
]
.
3.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
T
o
illu
s
tr
ate
th
e
o
p
er
atio
n
al
wo
r
k
f
lo
w
o
f
th
e
p
r
o
p
o
s
ed
Hy
b
r
id
Div
is
iv
e
–
K
-
Me
an
s
m
o
d
e
l,
th
e
co
d
e
s
eg
m
en
t
in
Alg
o
r
ith
m
1
p
r
esen
ts
th
e
alg
o
r
ith
m
ic
s
tep
s
u
s
ed
in
th
e
e
x
p
er
im
e
n
t.
T
h
is
im
p
lem
en
tatio
n
d
em
o
n
s
tr
ates
h
o
w
th
e
d
iv
is
iv
e
s
p
lit
tin
g
p
r
o
ce
s
s
is
ex
ec
u
ted
to
o
b
tain
in
itial
clu
s
ter
p
ar
titi
o
n
s
,
h
o
w
ce
n
tr
o
id
s
ar
e
co
m
p
u
ted
f
r
o
m
th
e
h
ier
a
r
c
h
ical
r
esu
lts
,
an
d
h
o
w
th
ese
ce
n
tr
o
id
s
ar
e
s
u
b
s
eq
u
en
tly
r
ef
i
n
ed
u
s
in
g
K
-
Me
an
s
.
T
h
e
co
d
e
r
ef
lects
th
e
ex
ac
t
p
r
o
ce
d
u
r
e
a
p
p
lied
in
th
e
s
tu
d
y
to
en
s
u
r
e
m
e
th
o
d
o
l
o
g
ical
clar
ity
a
n
d
r
ep
r
o
d
u
cib
ilit
y
.
Alg
o
r
ith
m
1
.
A
lg
o
r
ith
m
ic
s
tep
s
u
s
ed
in
th
e
ex
p
er
im
en
t
def hybrid_divisive_kmeans(X, n_clusters=3):
cluster_labels = np.zeros(len(X), dtype=int)
clusters = [np.arange(len(X))]
#Step 1: Divisive → split until the number of clusters reaches
n_clusters
.
while len(clusters) < n_clusters:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
h
yb
r
id
d
ivis
ive
K
-
mea
n
s
fr
a
mewo
r
k
fo
r
b
ig
d
a
ta
–
d
r
iven
p
o
ve
r
ty
a
n
a
lysi
s
in
C
en
tr
a
l
… (
B
o
w
o
Win
a
r
n
o
)
265
sizes = [len(c) for c in clusters]
idx_split = np.argmax(sizes)
indices = clusters.pop(idx_split)
if len(in
dices) <= 1:
clusters.append(indices)
continue
km = KMeans(n_clusters=2, random_state=42, n_init=10)
split_labels = km.fit_predict(X[indices])
clusters.append(indices[split_labels == 0])
clusters.append(indices[split_labels == 1])
# mapping divisive cluster results to cluster_labels
for cid, idx in enumerate(clusters):
cluster_labels[idx] = cid
# Compute the centroid from the divisive results
centroids = np.array([X
[cluster_labels == i].mean(axis=0) for i in range(n_clusters)])
# Step 2: K
-
Means with initialization from the divisive centroids
kmeans = KMeans(n_clusters=n_clusters, init=centroids, n_init=1, random_state=42)
return kmeans.fit_predict(X)
3.
1
.
E
x
ec
utio
n
t
i
m
e
co
m
pa
ri
s
o
n
Fig
u
r
e
2
p
r
esen
ts
t
h
e
f
ir
s
t
3
5
d
ata
p
o
in
ts
o
f
th
e
clu
s
ter
in
g
r
e
s
u
lts
s
h
o
w
th
e
as
s
ig
n
ed
clu
s
te
r
lab
els
f
o
r
ea
ch
m
eth
o
d
:
K
-
Me
an
s
p
r
e
d
o
m
in
an
tly
ass
ig
n
s
m
o
s
t
p
o
in
ts
t
o
clu
s
ter
2
,
with
s
o
m
e
p
o
in
ts
in
clu
s
ter
s
0
an
d
1
;
Ag
g
lo
m
er
ativ
e
K
-
Me
an
s
s
h
o
ws
m
o
r
e
v
ar
iatio
n
ac
r
o
s
s
clu
s
ter
s
0
,
1
,
an
d
2
;
wh
ile
Div
is
iv
e
K
-
Me
an
s
p
r
o
d
u
ce
s
a
p
atter
n
v
er
y
s
im
ilar
to
Ag
g
lo
m
er
ativ
e
K
-
Me
an
s
,
in
d
icatin
g
co
m
p
ar
a
b
le
clu
s
ter
ass
ig
n
m
en
ts
b
etwe
en
th
e
two
h
y
b
r
i
d
ap
p
r
o
ac
h
es.
T
ab
le
2
p
r
esen
ts
th
e
e
x
ec
u
tio
n
tim
e
o
f
th
e
two
clu
s
ter
in
g
m
eth
o
d
s
:
s
tan
d
ar
d
K
-
Me
an
s
a
s
s
h
o
wn
in
F
i
g
u
r
e
3
,
an
d
Div
is
iv
e
K
-
Me
an
s
as
s
h
o
wn
in
F
ig
u
r
e
4
.
T
h
e
r
esu
lts
ar
e
r
ep
o
r
ted
in
s
ec
o
n
d
s
an
d
r
ep
r
esen
t
th
e
av
er
ag
e
o
f
m
u
ltip
le
ex
p
er
im
e
n
tal
r
u
n
s
to
r
ed
u
ce
th
e
ef
f
ec
t
o
f
r
an
d
o
m
v
ar
iatio
n
s
.
Fig
u
r
es
5
an
d
6
p
r
esen
t
th
e
r
u
n
tim
e
s
ca
lab
ilit
y
an
aly
s
is
o
f
th
e
p
r
o
p
o
s
ed
Hy
b
r
id
Div
is
iv
e
–
K
-
Me
an
s
f
r
am
ewo
r
k
,
illu
s
tr
atin
g
h
o
w
co
m
p
u
tatio
n
al
p
er
f
o
r
m
an
ce
ch
an
g
es with
in
cr
ea
s
in
g
d
ataset
s
ize
(
n
)
an
d
f
ea
tu
r
e
d
im
en
s
io
n
ality
(
d
)
.
Fig
u
r
e
2.
First 3
5
d
ata
p
o
i
n
ts
o
f
clu
s
ter
in
g
r
esu
lts
T
ab
le
2
.
E
x
ec
u
tio
n
tim
e
c
o
m
p
ar
is
o
n
o
f
clu
s
ter
in
g
m
eth
o
d
s
M
e
t
h
o
d
Ex
e
c
u
t
i
o
n
T
i
me
(
ms)
I
t
e
r
a
t
i
o
n
s
t
o
c
o
n
v
e
r
g
e
n
c
e
S
i
l
h
o
u
e
t
t
e
sc
o
r
e
D
B
I
C
H
I
n
d
e
x
K
M
e
a
n
s
5
4
.
9
8
5
0
.
1
9
6
1
.
4
5
4
1
4
.
3
D
i
v
i
si
v
e
K
M
e
a
n
s (D
H
C
)
1
.
4
5
3
0
.
1
9
5
1
4
.
0
5
1
5
.
8
Fig
u
r
e
3
.
T
h
e
K
-
m
ea
n
s
clu
s
ter
in
g
Fig
u
r
e
4
.
T
h
e
h
y
b
r
id
d
iv
is
iv
e
K
-
m
ea
n
s
clu
s
ter
in
g
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
5
0
2
-
4
7
52
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
,
Vo
l.
41
,
No
.
1
,
J
an
u
ar
y
20
26
:
258
-
2
6
9
266
Fig
u
r
e
5
.
R
u
n
tim
e
s
ca
lin
g
wit
h
d
ataset
s
ize(
n
)
Fig
u
r
e
6
.
R
u
n
tim
e
s
ca
lin
g
wit
h
d
im
en
s
io
n
ality
(
d
)
3
.
2
.
I
nte
rpre
t
a
t
io
n
o
f
clus
t
er
v
a
lid
it
y
ind
ices
T
h
e
clu
s
ter
v
alid
ity
in
d
ices
p
r
o
v
id
e
d
ee
p
er
i
n
s
ig
h
t
in
to
t
h
e
q
u
ality
o
f
th
e
clu
s
ter
in
g
r
esu
lts
.
T
h
e
Sil
h
o
u
ette
Sco
r
e
s
h
o
ws
th
at
s
t
an
d
ar
d
K
-
Me
an
s
a
n
d
Div
is
iv
e
K
-
Me
an
s
p
r
o
d
u
ce
n
ea
r
ly
id
e
n
tical
v
alu
es
(
0
.
1
9
6
v
s
.
0
.
1
9
5
)
,
s
u
g
g
esti
n
g
s
im
ila
r
lev
els
o
f
s
ep
a
r
atio
n
a
n
d
c
o
h
esio
n
am
o
n
g
clu
s
ter
s
.
Alth
o
u
g
h
th
e
s
co
r
e
is
r
elativ
ely
m
o
d
est,
it
alig
n
s
with
th
e
n
atu
r
e
o
f
co
m
p
lex
s
o
cio
ec
o
n
o
m
ic
d
atasets
,
wh
ich
o
f
ten
ex
h
ib
it
o
v
er
lap
p
i
n
g
g
r
o
u
p
c
h
ar
ac
ter
is
tics
.
Me
an
wh
ile,
th
e
DB
I
in
d
icat
es
th
at
K
-
Me
an
s
ac
h
iev
es
b
etter
s
ep
ar
atio
n
b
etwe
en
clu
s
ter
s
with
a
lo
wer
v
alu
e
(
1
.
4
5
4
)
co
m
p
ar
ed
to
Di
v
is
iv
e
K
-
Me
an
s
(
1
4
.
0
5
)
.
T
h
e
u
n
u
s
u
ally
h
ig
h
DB
I
f
o
r
Div
is
iv
e
K
-
Me
an
s
m
a
y
r
ef
lect
co
m
p
ac
t
y
et
clo
s
ely
p
o
s
itio
n
ed
clu
s
ter
s
,
u
n
d
er
s
co
r
in
g
th
e
n
ee
d
f
o
r
ad
d
itio
n
al
p
ar
am
eter
tu
n
in
g
o
r
ex
p
an
d
ed
b
en
ch
m
a
r
k
in
g
to
f
u
lly
u
n
d
er
s
tan
d
th
is
b
eh
av
i
o
r
.
I
n
co
n
tr
ast,
th
e
CH
I
n
d
ex
f
av
o
r
s
th
e
Div
is
iv
e
K
-
Me
an
s
m
eth
o
d
,
w
h
ich
attain
s
a
h
ig
h
er
s
co
r
e
(
1
5
.
8
)
th
an
s
tan
d
ar
d
K
-
Me
an
s
(
1
4
.
3
)
,
d
em
o
n
s
tr
atin
g
s
u
p
e
r
io
r
o
v
er
all
d
is
p
er
s
io
n
an
d
c
o
m
p
ac
tn
ess
—
an
o
u
tco
m
e
co
n
s
is
ten
t
with
its
m
o
r
e
s
tr
u
ctu
r
ed
in
itializatio
n
ap
p
r
o
ac
h
.
T
o
g
eth
er
,
t
h
ese
in
d
ices
s
u
g
g
est
th
at
wh
ile
th
e
p
r
o
p
o
s
e
d
h
y
b
r
id
im
p
r
o
v
es
co
m
p
u
tatio
n
al
ef
f
icien
c
y
an
d
ce
n
tr
o
id
s
tab
ilit
y
,
th
e
o
v
er
all
clu
s
ter
in
g
q
u
ality
v
ar
ies
d
ep
e
n
d
in
g
o
n
th
e
m
etr
ic
u
s
ed
,
war
r
an
tin
g
b
r
o
ad
er
c
o
m
p
ar
ativ
e
v
alid
atio
n
.
3
.
3
.
Dis
cus
s
io
n a
nd
lim
it
a
t
io
ns
O
v
e
r
a
l
l
,
t
h
e
h
y
b
r
i
d
m
o
d
e
l
d
e
m
o
n
s
t
r
a
t
es
e
n
h
a
n
c
e
d
c
o
n
v
e
r
g
e
n
c
e
s
p
e
e
d
a
n
d
s
t
a
b
i
l
it
y
,
c
o
n
f
i
r
m
i
n
g
t
h
e
a
d
v
a
n
t
a
g
e
o
f
d
i
v
i
s
i
v
e
i
n
i
t
i
al
i
za
t
i
o
n
f
o
r
l
a
r
g
e
a
n
d
h
i
g
h
-
d
i
m
e
n
s
i
o
n
a
l
d
a
t
as
e
ts
.
H
o
w
e
v
e
r
,
th
e
f
i
n
d
i
n
g
s
r
e
m
a
i
n
l
a
r
g
e
l
y
d
e
s
c
r
i
p
ti
v
e
a
n
d
a
r
e
b
a
s
e
d
o
n
a
m
o
d
e
s
t
-
s
i
z
e
d
d
a
t
a
s
e
t
,
l
im
i
t
i
n
g
t
h
e
a
b
i
l
it
y
t
o
f
u
l
l
y
v
a
l
i
d
a
t
e
t
h
e
f
r
a
m
e
w
o
r
k
’
s
s
c
a
l
a
b
i
l
it
y
c
l
ai
m
s
.
A
d
d
it
i
o
n
a
l
e
x
p
e
r
i
m
e
n
t
s
w
it
h
l
a
r
g
e
r
d
a
t
ase
t
s
p
o
t
e
n
t
i
al
l
y
e
x
ce
e
d
i
n
g
m
il
l
io
n
s
o
f
o
b
s
e
r
v
at
i
o
n
s
w
o
u
l
d
b
e
n
e
c
e
s
s
a
r
y
t
o
e
m
p
i
r
i
ca
l
l
y
c
o
n
f
i
r
m
t
h
e
m
o
d
el
’
s
s
u
i
ta
b
i
l
i
t
y
f
o
r
b
i
g
d
at
a
a
p
p
l
ic
a
t
i
o
n
s
.
M
o
r
e
o
v
e
r
,
w
h
i
l
e
t
h
e
c
l
u
s
t
e
r
i
n
g
r
e
s
u
l
ts
r
e
v
e
a
l
m
e
an
i
n
g
f
u
l
s
t
r
u
c
t
u
r
a
l
p
a
tt
e
r
n
s
,
t
h
e
i
r
d
i
r
e
c
t
r
e
le
v
a
n
c
e
t
o
p
o
v
e
r
ty
p
o
l
i
c
y
i
s
n
o
t
y
e
t
s
t
r
o
n
g
l
y
e
s
t
a
b
l
is
h
e
d
.
F
u
t
u
r
e
wo
r
k
s
h
o
u
l
d
i
n
t
e
g
r
a
t
e
s
p
at
i
a
l
a
n
a
l
y
s
is
,
d
is
t
r
i
ct
-
l
e
v
e
l
s
o
c
i
o
e
c
o
n
o
m
i
c
p
r
o
f
i
l
i
n
g
,
a
n
d
d
o
m
a
i
n
e
x
p
e
r
t
v
a
l
i
d
a
t
i
o
n
t
o
t
r
an
s
l
a
t
e
t
h
e
c
l
u
s
t
e
r
i
n
g
o
u
t
c
o
m
es
i
n
t
o
a
c
t
i
o
n
a
b
l
e
p
o
li
c
y
i
n
s
i
g
h
t
s
.
4.
CO
NCLU
SI
O
N
T
h
e
Hy
b
r
i
d
Div
is
iv
e
–
K
-
Me
an
s
m
o
d
el
d
em
o
n
s
t
r
ates
clea
r
im
p
r
o
v
e
m
en
ts
o
v
e
r
th
e
s
tan
d
a
r
d
K
-
Me
an
s
ap
p
r
o
ac
h
,
p
a
r
ticu
lar
ly
t
h
r
o
u
g
h
f
aster
co
n
v
er
g
e
n
ce
an
d
m
o
r
e
s
tab
le
ce
n
tr
o
id
in
itializatio
n
.
T
h
e
ex
ec
u
ti
o
n
tim
e
d
r
o
p
p
ed
f
r
o
m
5
4
.
9
8
m
s
to
1
.
4
5
m
s
,
th
o
u
g
h
th
is
g
ain
s
h
o
u
ld
b
e
v
iewe
d
in
r
elatio
n
to
t
h
e
r
elativ
e
ly
s
m
all
d
ataset
u
s
ed
.
Ad
d
itio
n
ally
,
th
e
d
ec
r
ea
s
e
in
iter
atio
n
s
f
r
o
m
f
iv
e
to
th
r
ee
s
h
o
ws
th
at
th
e
d
iv
is
iv
e
in
itializatio
n
s
tep
ef
f
ec
tiv
ely
r
ed
u
ce
s
ce
n
tr
o
id
r
an
d
o
m
n
ess
,
lead
in
g
t
o
a
m
o
r
e
co
n
s
is
ten
t a
n
d
r
eliab
le
clu
s
ter
in
g
p
r
o
ce
s
s
.
T
h
e
clu
s
ter
v
alid
ity
in
d
ic
es
also
d
em
o
n
s
tr
ate
th
at
th
e
h
y
b
r
id
ap
p
r
o
ac
h
m
ain
tain
s
o
r
s
lig
h
tly
en
h
an
ce
s
clu
s
ter
in
g
q
u
ality
.
A
h
ig
h
er
C
alin
s
k
i
–
Har
ab
asz
I
n
d
ex
(
1
5
.
8
)
an
d
a
co
m
p
ar
ab
le
Sil
h
o
u
ette
Sco
r
e
(
0
.
1
9
5
)
s
u
g
g
est
th
at
th
e
r
esu
ltin
g
cl
u
s
ter
s
ar
e
r
ea
s
o
n
ab
ly
co
m
p
a
ct
an
d
well
-
s
ep
ar
ated
,
ev
en
t
h
o
u
g
h
th
e
Da
v
ies
–
B
o
u
ld
in
I
n
d
ex
in
cr
ea
s
ed
.
I
n
a
s
o
cio
-
ec
o
n
o
m
ic
c
o
n
tex
t,
a
h
ig
h
er
Sil
h
o
u
ette
Sco
r
e
in
d
icate
s
th
at
d
is
tr
icts
with
s
im
ilar
p
o
v
er
ty
c
h
ar
ac
ter
is
tics
s
u
ch
as
u
n
em
p
l
o
y
m
en
t
r
ate,
ed
u
ca
tio
n
lev
el,
o
r
ac
ce
s
s
to
s
er
v
ices
ar
e
g
r
o
u
p
ed
m
o
r
e
c
o
n
s
is
t
en
tly
,
im
p
ly
i
n
g
th
at
th
e
m
o
d
el
ca
p
tu
r
es
m
ea
n
in
g
f
u
l
p
atter
n
s
o
f
d
ep
r
i
v
atio
n
.
Similar
ly
,
h
ig
h
er
C
alin
s
k
i
–
Har
ab
asz
v
alu
es
im
p
ly
th
at
th
e
d
is
tin
ctio
n
s
b
etwe
e
n
lo
w
-
,
m
o
d
er
ate
-
,
an
d
h
ig
h
-
p
o
v
er
ty
cl
u
s
ter
s
ar
e
m
o
r
e
s
tr
u
ctu
r
ally
d
e
f
in
ed
,
w
h
ich
ca
n
s
u
p
p
o
r
t
p
o
licy
m
a
k
er
s
in
id
en
tify
in
g
p
r
io
r
ity
r
eg
io
n
s
.
C
o
n
v
er
s
ely
,
an
in
cr
ea
s
e
in
Dav
ies
–
B
o
u
ld
i
n
s
co
r
es
s
u
g
g
ests
th
at
s
o
m
e
clu
s
ter
s
m
ay
s
till
o
v
er
lap
s
o
cio
-
ec
o
n
o
m
ically
,
in
d
icatin
g
th
at
ce
r
tain
d
is
tr
icts
s
h
ar
e
m
ix
ed
ch
ar
ac
ter
is
tics
th
at
co
m
p
licate
clea
r
p
o
licy
ca
teg
o
r
izatio
n
.
T
h
is
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
d
o
n
esian
J
E
lec
E
n
g
&
C
o
m
p
Sci
I
SS
N:
2
5
0
2
-
4
7
52
A
h
yb
r
id
d
ivis
ive
K
-
mea
n
s
fr
a
mewo
r
k
fo
r
b
ig
d
a
ta
–
d
r
iven
p
o
ve
r
ty
a
n
a
lysi
s
in
C
en
tr
a
l
… (
B
o
w
o
Win
a
r
n
o
)
267
m
ix
ed
p
er
f
o
r
m
a
n
ce
ac
r
o
s
s
v
alid
ity
m
etr
ics
h
ig
h
lig
h
ts
th
e
n
ee
d
f
o
r
b
r
o
ad
er
b
en
ch
m
ar
k
in
g
an
d
ex
ter
n
al
v
alid
atio
n
b
ef
o
r
e
g
en
er
alizin
g
th
e
m
o
d
el’
s
ef
f
ec
tiv
en
ess
to
lar
g
er
o
r
m
o
r
e
h
eter
o
g
en
eo
u
s
d
a
tasets
.
B
ey
o
n
d
tech
n
ical
p
er
f
o
r
m
an
ce
,
th
e
clu
s
ter
in
g
r
esu
lts
h
av
e
m
ea
n
in
g
f
u
l
im
p
licatio
n
s
f
o
r
r
eg
io
n
al
s
o
cio
-
ec
o
n
o
m
ic
p
o
licy
.
B
y
g
r
o
u
p
in
g
d
is
tr
icts
b
ased
o
n
m
u
ltid
im
en
s
io
n
al
p
o
v
er
t
y
in
d
ic
ato
r
s
,
p
o
licy
m
ak
er
s
ca
n
id
e
n
tify
p
r
io
r
ity
r
e
g
io
n
s
,
allo
ca
te
r
eso
u
r
ce
s
m
o
r
e
ef
f
e
ctiv
ely
,
an
d
d
esig
n
ta
r
g
eted
p
o
v
er
ty
allev
iatio
n
p
r
o
g
r
a
m
s
.
T
h
e
DHC f
r
am
ewo
r
k
th
u
s
p
r
o
v
id
es a
n
an
aly
tical
f
o
u
n
d
atio
n
th
at
s
u
p
p
o
r
ts
ev
id
e
n
ce
-
b
ased
p
lan
n
in
g
f
o
r
r
e
g
io
n
al
d
e
v
elo
p
m
e
n
t.
Ho
wev
er
,
th
is
s
tu
d
y
also
ac
k
n
o
wled
g
es
k
e
y
lim
itatio
n
s
.
T
h
e
d
ataset
co
n
tain
s
a
lim
ited
n
u
m
b
er
o
f
v
ar
iab
les,
wh
ich
m
ay
n
o
t
f
u
ll
y
ca
p
tu
r
e
th
e
co
m
p
lex
ity
o
f
p
o
v
er
ty
d
y
n
am
ics.
I
n
ad
d
itio
n
,
th
e
r
elativ
ely
s
m
all
d
ataset
r
estricts
th
e
ev
alu
atio
n
o
f
s
ca
lab
ilit
y
claim
s
,
an
d
e
x
ter
n
al
v
alid
atio
n
u
s
in
g
in
d
e
p
en
d
en
t
d
atasets
was
n
o
t
co
n
d
u
cted
.
Fu
tu
r
e
r
esear
ch
s
h
o
u
ld
th
er
ef
o
r
e
ex
p
lo
r
e
th
e
a
p
p
licatio
n
o
f
th
e
m
o
d
el
to
lar
g
er
d
atasets
,
in
co
r
p
o
r
ate
ad
d
itio
n
al
s
o
cio
-
ec
o
n
o
m
ic
i
n
d
icato
r
s
,
an
d
a
s
s
es
s
s
ca
lab
ilit
y
u
s
in
g
p
ar
allel
o
r
d
is
tr
ib
u
ted
co
m
p
u
tin
g
f
r
a
m
ewo
r
k
s
.
E
x
p
a
n
d
in
g
th
e
a
n
aly
s
is
to
o
th
e
r
p
r
o
v
in
ce
s
o
r
s
ec
to
r
s
wo
u
l
d
f
u
r
t
h
er
s
tr
en
g
t
h
en
th
e
r
o
b
u
s
tn
ess
an
d
p
o
licy
r
elev
a
n
ce
o
f
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
clu
s
ter
in
g
ap
p
r
o
ac
h
.
ACK
NO
WL
E
DG
M
E
N
T
T
h
e
au
th
o
r
s
wo
u
ld
also
lik
e
t
o
ex
p
r
ess
th
eir
s
in
ce
r
e
g
r
atitu
d
e
to
Un
iv
e
r
s
ity
Dip
o
n
eg
o
r
o
,
Sem
ar
an
g
,
f
o
r
its
s
u
p
p
o
r
t a
n
d
c
o
llab
o
r
ati
o
n
.
F
UNDING
I
NF
O
R
M
A
T
I
O
N
T
h
is
r
esear
ch
was
f
u
n
d
e
d
b
y
th
e
2
0
2
5
R
KAT
o
f
Un
iv
e
r
s
itas
Seb
elas
Ma
r
et
th
r
o
u
g
h
th
e
Do
cto
r
al
Dis
s
er
tatio
n
R
e
s
ea
r
ch
Sch
em
e
(
PDD
-
UNS)
u
n
d
er
R
esear
ch
Ass
ig
n
m
en
t
Ag
r
ee
m
en
t
No
.
3
6
9
/UN2
7
.
2
2
/PT.
0
1
.
0
3
/2
0
2
5
.
CO
NF
L
I
C
T
O
F
I
N
T
E
R
E
S
T
ST
A
T
E
M
E
NT
T
h
e
au
th
o
r
s
s
tate
n
o
co
n
f
lict o
f
in
ter
est.
DATA AV
AI
L
AB
I
L
I
T
Y
Data
av
ailab
ilit
y
d
o
es n
o
t a
p
p
l
y
to
th
is
p
ap
e
r
as n
o
n
ew
d
ata
wer
e
cr
ea
ted
o
r
an
aly
ze
d
in
th
is
s
tu
d
y
.
RE
F
E
R
E
NC
E
S
[
1
]
A
.
E.
Ez
u
g
w
u
,
S
.
E
l
si
s
i
,
A
.
G
.
H
u
s
sa
i
n
,
a
n
d
O
.
S
.
O
l
a
y
e
mi
,
“
H
y
b
r
i
d
f
i
r
e
f
l
y
a
l
g
o
r
i
t
h
ms
f
o
r
c
l
u
st
e
r
i
n
g
,
”
I
EEE
A
c
c
e
ss
,
v
o
l
.
8
,
p
p
.
1
2
1
0
8
9
–
1
2
1
1
1
8
,
J
u
l
.
2
0
2
0
,
d
o
i
:
1
0
.
1
1
0
9
/
A
C
C
ESS
.
2
0
2
0
.
3
0
0
6
0
3
0
.
[
2
]
M
.
A
h
me
d
,
R
.
S
e
r
a
j
,
a
n
d
S
.
M
.
S
.
I
sl
a
m,
“
Th
e
k
-
me
a
n
s
a
l
g
o
r
i
t
h
m
:
A
c
o
m
p
r
e
h
e
n
si
v
e
s
u
r
v
e
y
a
n
d
p
e
r
f
o
r
m
a
n
c
e
e
v
a
l
u
a
t
i
o
n
,
”
El
e
c
t
r
o
n
i
c
s
,
v
o
l
.
9
,
n
o
.
8
,
p
.
1
2
9
5
,
A
u
g
.
2
0
2
0
,
d
o
i
:
1
0
.
3
3
9
0
/
e
l
e
c
t
r
o
n
i
c
s
9
0
8
1
2
9
5
.
[
3
]
A
.
E.
E
z
u
g
w
u
,
I
.
A
.
A
k
i
n
o
l
a
,
M
.
H
.
El
si
s
i
,
a
n
d
O
.
A
.
O
y
e
b
a
d
e
,
“
A
c
o
mp
r
e
h
e
n
s
i
v
e
s
u
r
v
e
y
o
f
c
l
u
st
e
r
i
n
g
a
l
g
o
r
i
t
h
ms,”
En
g
i
n
e
e
r
i
n
g
Ap
p
l
i
c
a
t
i
o
n
s
o
f
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
1
1
0
,
p
.
1
0
4
7
4
3
,
M
a
r
.
2
0
2
2
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
a
p
p
a
i
.
2
0
2
2
.
1
0
4
7
4
3
.
[
4
]
L.
B
a
i
,
J.
L
i
a
n
g
,
a
n
d
F
.
C
a
o
,
“
A
mu
l
t
i
p
l
e
k
-
m
e
a
n
s
c
l
u
s
t
e
r
i
n
g
e
n
s
e
m
b
l
e
a
l
g
o
r
i
t
h
m
t
o
f
i
n
d
n
o
n
l
i
n
e
a
r
l
y
se
p
a
r
a
b
l
e
c
l
u
s
t
e
r
s,
”
I
n
f
o
rm
a
t
i
o
n
F
u
si
o
n
,
v
o
l
.
6
1
,
p
p
.
3
6
–
4
7
,
J
a
n
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
f
f
u
s.
2
0
2
0
.
0
3
.
0
0
6
.
[
5
]
R
.
J.
G
.
B
.
C
a
m
p
e
l
l
o
,
D
.
M
o
u
l
a
v
i
,
a
n
d
J.
S
.
S
a
n
d
e
r
,
“
H
i
e
r
a
r
c
h
i
c
a
l
d
e
n
s
i
t
y
e
st
i
ma
t
e
s
f
o
r
d
a
t
a
c
l
u
s
t
e
r
i
n
g
,
”
A
C
M
T
r
a
n
sa
c
t
i
o
n
s
o
n
K
n
o
w
l
e
d
g
e
D
i
sc
o
v
e
ry
f
r
o
m
D
a
t
a
,
v
o
l
.
1
0
,
n
o
.
1
,
p
p
.
1
–
5
1
,
J
u
l
.
2
0
1
5
,
d
o
i
:
1
0
.
1
1
4
5
/
2
7
3
3
3
8
1
.
[
6
]
S
.
C
h
a
k
r
a
b
o
r
t
y
,
M
.
D
a
s
,
a
n
d
S
.
B
a
n
d
y
o
p
a
d
h
y
a
y
,
“
H
i
e
r
a
r
c
h
i
c
a
l
c
l
u
st
e
r
i
n
g
w
i
t
h
o
p
t
i
mal
t
r
a
n
s
p
o
r
t
,
”
S
t
a
t
i
s
t
i
c
s
and
Pr
o
b
a
b
i
l
i
t
y
L
e
t
t
e
rs
,
v
o
l
.
1
6
3
,
p
.
1
0
8
7
8
1
,
A
p
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
sp
l
.
2
0
2
0
.
1
0
8
7
8
1
.
[
7
]
A
.
B
e
l
h
a
d
i
,
S
.
T.
T.
N
g
u
y
e
n
,
A
.
G
.
B
o
u
d
h
i
r
,
a
n
d
A
.
H
a
me
u
r
l
a
i
n
,
“
S
p
a
c
e
–
t
i
me
ser
i
e
s
c
l
u
s
t
e
r
i
n
g
:
A
l
g
o
r
i
t
h
ms,
t
a
x
o
n
o
m
y
,
a
n
d
c
a
s
e
st
u
d
y
o
n
u
r
b
a
n
smar
t
c
i
t
i
e
s,”
E
n
g
i
n
e
e
ri
n
g
Ap
p
l
i
c
a
t
i
o
n
s
o
f
Ar
t
i
f
i
c
i
a
l
I
n
t
e
l
l
i
g
e
n
c
e
,
v
o
l
.
9
5
,
p
.
1
0
3
8
5
7
,
M
a
r
.
2
0
2
0
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
e
n
g
a
p
p
a
i
.
2
0
2
0
.
1
0
3
8
5
7
.
[
8
]
B
a
d
a
n
P
u
sat
S
t
a
t
i
st
i
k
(
B
P
S
)
,
P
o
v
e
r
t
y
D
a
t
a
a
n
d
I
n
f
o
r
m
a
t
i
o
n
f
o
r
D
i
s
t
r
i
c
t
s/
C
i
t
i
e
s
i
n
2
0
2
4
(
i
n
I
n
d
o
n
e
si
a
n
:
D
a
t
a
d
a
n
I
n
f
o
rm
a
si
K
e
m
i
s
k
i
n
a
n
K
a
b
u
p
a
t
e
n
/
K
o
t
a
T
a
h
u
n
2
0
2
4
)
,
Ja
k
a
r
t
a
,
I
n
d
o
n
e
s
i
a
:
B
P
S
,
N
o
v
.
2
0
2
4
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
w
w
w
.
b
p
s.
g
o
.
i
d
/
i
d
/
p
u
b
l
i
c
a
t
i
o
n
/
2
0
2
4
/
1
1
/
2
9
/
d
2
8
4
8
c
3
9
9
0
f
0
8
1
1
8
2
1
2
5
a
4
1
6
/
d
a
t
a
-
d
a
n
-
i
n
f
o
r
m
a
si
-
k
e
m
i
sk
i
n
a
n
-
k
a
b
u
p
a
t
e
n
--
k
o
t
a
-
t
a
h
u
n
-
2
0
2
4
.
h
t
ml
.
[
9
]
T.
E
mm
a
n
u
e
l
,
“
A
s
u
r
v
e
y
o
n
mi
s
si
n
g
d
a
t
a
i
n
mac
h
i
n
e
l
e
a
r
n
i
n
g
,
”
J
o
u
r
n
a
l
o
f
B
i
g
D
a
t
a
,
v
o
l
.
8
,
n
o
.
1
,
p
.
1
4
0
,
Ja
n
.
2
0
2
1
,
d
o
i
:
1
0
.
1
1
8
6
/
s
4
0
5
3
7
-
0
2
1
-
0
0
5
1
6
-
9.
[
1
0
]
Y
.
Zh
o
u
,
S
.
A
r
y
a
l
,
a
n
d
M
.
R
.
B
o
u
a
d
j
e
n
e
k
,
“
A
c
o
m
p
r
e
h
e
n
si
v
e
r
e
v
i
e
w
o
f
h
a
n
d
l
i
n
g
m
i
ssi
n
g
d
a
t
a
,
”
a
rXi
v
p
re
p
ri
n
t
a
r
X
i
v
:
2
4
0
4
.
0
4
9
0
5
,
A
p
r
.
2
0
2
4
.
[
O
n
l
i
n
e
]
.
A
v
a
i
l
a
b
l
e
:
h
t
t
p
s
:
/
/
a
r
x
i
v
.
o
r
g
/
a
b
s/
2
4
0
4
.
0
4
9
0
5
[
1
1
]
Y
.
C
h
e
n
,
L.
W
a
n
g
,
a
n
d
J
.
L
i
,
“
A
c
l
u
s
t
e
r
i
n
g
-
b
a
se
d
a
p
p
r
o
a
c
h
u
s
i
n
g
K
-
m
e
a
n
s
a
n
d
h
i
e
r
a
r
c
h
i
c
a
l
me
t
h
o
d
s
f
o
r
l
a
r
g
e
-
s
c
a
l
e
d
a
t
a
,
”
I
n
f
o
rm
a
t
i
o
n
,
v
o
l
.
1
6
,
n
o
.
6
,
p
.
4
4
1
,
J
u
n
.
2
0
2
5
,
d
o
i
:
1
0
.
3
3
9
0
/
i
n
f
o
1
6
0
6
0
4
4
1
.
[
1
2
]
Y
.
Z
h
a
n
g
,
“
A
c
o
mp
r
e
h
e
n
s
i
v
e
s
u
r
v
e
y
o
n
t
r
a
f
f
i
c
m
i
ssi
n
g
d
a
t
a
i
m
p
u
t
a
t
i
o
n
,
”
I
EE
E
T
r
a
n
sa
c
t
i
o
n
s
o
n
I
n
t
e
l
l
i
g
e
n
t
T
ra
n
s
p
o
rt
a
t
i
o
n
S
y
s
t
e
m
s
,
v
o
l
.
2
5
,
n
o
.
2
,
p
p
.
1
2
3
4
–
1
2
4
5
,
F
e
b
.
2
0
2
4
,
d
o
i
:
1
0
.
1
1
0
9
/
TI
TS.2
0
2
4
.
3
4
7
8
8
1
6
.
[
1
3
]
A
.
S
o
l
a
n
o
a
n
d
F
.
J.
B
e
r
l
a
n
g
a
,
“
A
h
y
b
r
i
d
c
l
u
s
t
e
r
i
n
g
me
t
h
o
d
b
a
se
d
o
n
s
e
v
e
r
a
l
d
i
v
e
r
se
c
l
u
st
e
r
i
n
g
a
p
p
r
o
a
c
h
e
s
,
”
C
o
m
m
u
n
i
c
a
t
i
o
n
s
i
n
S
t
a
t
i
st
i
c
s
–
S
i
m
u
l
a
t
i
o
n
a
n
d
C
o
m
p
u
t
a
t
i
o
n
,
v
o
l
.
5
3
,
n
o
.
3
,
p
p
.
7
0
7
–
7
2
3
,
2
0
2
4
,
d
o
i
:
1
0
.
1
0
8
0
/
0
3
6
1
0
9
1
8
.
2
0
2
2
.
2
1
0
4
7
6
1
.
Evaluation Warning : The document was created with Spire.PDF for Python.