I
nte
rna
t
io
na
l J
o
urna
l o
f
E
lect
rica
l a
nd
Co
m
p
ute
r
E
ng
in
ee
ring
(
I
J
E
CE
)
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2020
,
p
p
.
447
~
454
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
1
0
i
1
.
pp
4
4
7
-
454
447
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ec
e.
ia
esco
r
e.
co
m/in
d
ex
.
p
h
p
/I
JE
C
E
An
e
nha
nce
d
k
e
r
nel
w
eig
hted
c
o
lla
bo
ra
tive
r
eco
mm
ended
s
y
ste
m
to
a
llev
ia
t
e
s
pa
rsity
S.
B
a
beet
ha
1
,
B
.
M
urug
a
na
n
t
ha
m
2
,
S.
G
a
nes
h K
u
m
a
r
3
,
A
.
M
urug
a
n
4
1
De
p
a
rtme
n
t
o
f
In
f
o
rm
a
ti
o
n
a
n
d
T
e
c
h
n
o
lo
g
y
,
S
RM
In
stit
u
te o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
In
d
ia
2
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter S
c
ien
c
e
a
n
d
E
n
g
in
e
e
rin
g
,
S
RM
In
stit
u
te
o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
In
d
ia
3
,4
S
RM
I
n
stit
u
te o
f
S
c
ien
c
e
a
n
d
T
e
c
h
n
o
l
o
g
y
,
In
d
ia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Feb
17
,
2
0
1
9
R
ev
i
s
ed
A
u
g
16
,
20
19
A
cc
ep
ted
A
u
g
2
8
,
2
0
19
Us
e
r
Re
v
ie
w
s
in
th
e
f
o
r
m
o
f
ra
ti
n
g
s
g
iv
in
g
a
n
o
p
p
o
rtu
n
it
y
to
ju
d
g
e
th
e
u
se
r
in
tere
st
o
n
t
h
e
a
v
a
il
a
b
le
p
r
o
d
u
c
ts
a
n
d
p
r
o
v
id
i
n
g
a
c
h
a
n
c
e
t
o
re
c
o
m
m
e
n
d
n
e
w
sim
il
a
r
it
e
m
s to
th
e
c
u
sto
m
e
rs.
P
e
rso
n
a
li
z
e
d
re
c
o
m
m
e
n
d
e
r
tec
h
n
iq
u
e
s p
lac
in
g
v
it
a
l
ro
le
in
th
is
g
ro
w
n
e
c
o
m
m
e
rc
e
c
e
n
tu
ry
to
p
re
d
ict
th
e
u
se
rs‟
in
tere
st.
Co
ll
a
b
o
ra
ti
v
e
F
il
terin
g
(CF
)
sy
ste
m
is
o
n
e
o
f
th
e
w
id
e
ly
u
se
d
d
e
m
o
c
ra
ti
c
re
c
o
m
m
e
n
d
e
r
s
y
ste
m
w
h
e
re
it
c
o
m
p
lete
l
y
r
e
l
y
o
n
u
se
r
ra
ti
n
g
s
t
o
p
ro
v
id
e
re
c
o
m
m
e
n
d
a
ti
o
n
s
f
o
r
th
e
u
se
rs.
In
t
h
is
p
a
p
e
r,
a
n
e
n
h
a
n
c
e
d
Co
l
lab
o
ra
ti
v
e
F
il
terin
g
sy
ste
m
is
p
ro
p
o
se
d
u
s
in
g
Ke
rn
e
l
Weig
h
ted
K
-
m
e
a
n
s
Clu
ste
rin
g
(KW
KC)
a
p
p
ro
a
c
h
u
sin
g
Ra
d
i
a
l
b
a
si
s
F
u
n
c
ti
o
n
s
(RBF
)
f
o
r
e
li
m
in
a
te
th
e
S
p
a
rsity
p
ro
b
lem
w
h
e
re
lac
k
o
f
ra
ti
n
g
is
th
e
c
h
a
ll
e
n
g
e
o
f
p
ro
v
id
in
g
th
e
a
c
c
u
ra
te
re
c
o
m
m
e
n
d
a
ti
o
n
t
o
th
e
u
se
r.
T
h
e
p
ro
p
o
se
d
sy
ste
m
h
a
v
in
g
t
w
o
p
h
a
se
s
o
f
sta
te
tr
a
n
siti
o
n
s:
C
o
n
n
e
c
ted
a
n
d
Disc
o
n
n
e
c
ted
.
Du
r
in
g
Co
n
n
e
c
ted
sta
te
th
e
f
o
rm
o
f
tran
siti
o
n
w
il
l
b
e
„Re
c
o
m
m
e
n
d
e
d
m
o
d
e
‟
w
h
e
re
th
e
a
c
ti
v
e
u
se
r
b
e
g
iv
e
n
w
it
h
th
e
P
re
d
icte
d
-
re
c
o
m
m
e
n
d
e
d
it
e
m
s.
In
Disc
o
n
n
e
c
ted
S
tate
th
e
f
o
r
m
o
f
tran
siti
o
n
w
il
l
b
e
„
L
e
a
rn
in
g
m
o
d
e
‟
w
h
e
r
e
th
e
h
y
b
r
id
lea
rn
in
g
a
p
p
ro
a
c
h
a
n
d
u
se
r
c
lu
ste
rs
w
il
l
b
e
u
se
d
to
d
e
f
in
e
th
e
si
m
il
a
r
u
se
r
m
o
d
e
ls.
Disc
o
n
n
e
c
ted
S
tate
a
c
ti
v
it
ies
w
il
l
b
e
p
e
rf
o
r
m
e
d
in
h
id
d
e
n
lay
e
r
o
f
RBF
a
n
d
Co
n
n
e
c
ted
S
a
te
a
c
ti
v
it
ies
w
il
l
b
e
p
e
rf
o
rm
e
d
in
o
u
t
p
u
t
L
a
y
e
r.
In
p
u
t
L
a
y
e
r
o
f
RBF
u
sin
g
o
rig
in
a
l
u
se
r
Ra
ti
n
g
s.
T
h
e
p
ro
p
o
se
d
KW
KC
u
se
d
to
s
m
o
o
th
e
n
th
e
sp
a
rse
o
rig
in
a
l
ra
ti
n
g
m
a
tr
ix
a
n
d
d
e
f
in
e
th
e
sim
il
a
r
u
se
r
c
lu
ste
rs.
A
b
e
n
c
h
m
a
rk
c
o
m
p
a
ra
ti
v
e
stu
d
y
a
lso
m
a
d
e
w
it
h
c
las
sic
a
l
lea
rn
in
g
a
n
d
p
re
d
ictio
n
tec
h
n
iq
u
e
s
in
term
s
o
f
a
c
c
u
ra
c
y
a
n
d
c
o
m
p
u
tatio
n
a
l
ti
m
e
.
Ex
p
e
rien
ti
a
l
se
tu
p
is
m
a
d
e
u
sin
g
M
o
v
ieL
e
n
s d
a
tas
e
t.
K
ey
w
o
r
d
s
:
C
las
s
ical
r
atin
g
p
r
ed
ictio
n
m
et
h
o
d
s
C
o
llab
o
r
ativ
e
f
i
lter
in
g
(
C
F)
Ker
n
el
C
F
K
-
W
eig
h
ted
cl
u
s
ter
i
n
g
R
ati
n
g
es
ti
m
atio
n
m
et
h
o
d
s
R
ec
o
m
m
e
n
d
er
s
y
s
te
m
(
R
S)
Sp
ar
s
it
y
Co
p
y
rig
h
t
©
2
0
2
0
In
stit
u
te o
f
A
d
v
a
n
c
e
d
E
n
g
i
n
e
e
rin
g
a
n
d
S
c
ien
c
e
.
Al
l
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
S.
B
ab
ee
th
a,
Dep
ar
t
m
en
t o
f
I
n
f
o
r
m
atio
n
an
d
T
ec
h
n
o
lo
g
y
,
SR
M
I
n
s
tit
u
te
o
f
Sc
ien
ce
a
n
d
T
ec
h
n
o
lo
g
y
,
C
h
e
n
n
ai,
I
n
d
ia.
E
m
ail:
b
ab
ee
s
1
2
0
7
@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
R
ec
o
m
m
e
n
d
atio
n
i
s
o
n
e
o
f
t
h
e
m
aj
o
r
ess
e
n
tialp
r
o
d
u
ctio
n
ac
ti
v
it
ies
to
att
r
ac
t
u
s
er
s
.
R
S
(
R
ec
o
m
m
e
n
d
atio
n
S
y
s
te
m
)
is
th
e
ap
p
r
o
ac
h
th
at
w
i
ll
len
d
a
h
an
d
in
R
ec
o
m
m
e
n
d
atio
n
s
y
s
te
m
s
ar
e
m
ec
h
a
n
i
s
m
t
h
at
le
n
d
a
h
a
n
d
i
n
m
ak
i
n
g
s
e
lectio
n
in
a
n
ar
ea
ev
e
n
w
it
h
lac
k
o
f
ac
q
u
ai
n
tan
c
e
in
it.
T
h
e
s
u
r
v
e
y
[
1
]
p
r
o
v
id
es
th
e
class
if
icatio
n
o
f
r
ec
o
m
m
en
d
er
s
y
s
te
m
s
.
B
ased
o
n
co
r
e
m
et
h
o
d
s
r
ec
o
m
m
en
d
er
s
y
s
te
m
i
s
b
r
o
ad
ly
class
i
f
ied
as
h
eu
r
i
s
tic
b
ased
/co
n
ten
t
b
ased
s
y
s
te
m
an
d
/u
s
er
p
r
ef
er
en
c
e
o
r
r
atin
g
lo
g
ic/co
llab
o
r
ativ
e
f
ilter
i
n
g
s
y
s
te
m
.
C
o
llab
o
r
ativ
e
Fil
ter
i
n
g
r
ec
o
m
m
en
d
er
s
y
s
te
m
s
ar
e
b
ased
o
n
th
e
r
ese
m
b
la
n
ce
b
et
w
ee
n
cu
s
to
m
er
f
av
o
r
ite
r
ati
n
g
s
f
o
r
w
o
r
k
o
u
t
r
ec
o
m
m
e
n
d
atio
n
s
.
Si
n
ce
t
h
is
m
eth
o
d
d
o
es
n
o
t
r
el
y
o
n
s
e
m
an
t
ics
o
f
t
h
e
p
r
o
d
u
cts,
it
is
co
m
p
li
m
en
tar
y
f
r
o
m
th
e
t
w
o
m
a
in
c
h
alle
n
g
e
s
o
f
t
h
e
co
n
ten
t
-
b
ased
ap
p
r
o
ac
h
:s
h
allo
w
-
a
n
al
y
s
i
s
p
r
o
b
lem
o
r
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
4
4
7
-
454
448
th
e
o
v
er
-
s
p
ec
ializa
tio
n
p
r
o
b
lem
[
2
]
f
o
r
w
h
ic
h
a
b
ad
c
h
o
ic
es
o
f
a
ttrib
u
te
is
th
e
r
o
o
t
ca
u
s
e.
Mo
s
t
f
r
eq
u
en
tl
y
u
s
ed
tec
h
n
iq
u
es
f
o
r
co
llab
o
r
ativ
e
r
ec
o
m
m
e
n
d
atio
n
ar
e;co
r
r
elatio
n
b
ased
m
e
th
o
d
s
[
3
,
4
]
,
L
aten
t
Se
m
a
n
tic
I
n
d
ex
i
n
g
(
L
SI)
[
1
,
2
]
,
B
ay
esi
an
L
ea
r
n
i
n
g
[
5
,
6
]
,
etc.
T
h
e
a
cc
u
r
ac
y
o
f
th
e
r
ec
o
m
m
e
n
d
ati
o
n
w
ill
b
e
ev
al
u
ated
o
n
th
e
q
u
an
tit
y
o
f
t
h
e
u
s
er
r
atin
g
p
r
o
v
id
ed
o
n
th
e
co
ll
ec
tio
n
o
f
p
r
o
d
u
c
ts
a
v
ailab
le.
T
h
e
ass
o
ciatio
n
to
th
e
clu
s
ter
s
w
ill
b
e
m
o
n
ito
r
ed
u
s
in
g
th
e
p
r
o
d
u
ct
co
v
er
ag
e
w
it
h
s
i
g
n
i
f
ican
t
o
v
er
lap
s
i
n
r
atin
g
s
.
B
u
t
in
r
ea
lit
y
th
is
w
ill
n
o
t
b
e
t
h
e
ca
s
e,
w
h
er
e
th
er
e
w
i
ll
b
e
lack
o
f
lar
g
e
c
u
s
to
m
er
p
o
o
l
an
d
tr
en
d
o
f
in
tr
o
d
u
cin
g
n
e
w
p
r
o
d
u
cts,
th
u
s
ca
u
s
in
g
Sp
ar
s
it
y
P
r
o
b
lem
.
I
f
t
h
e
o
v
er
lap
u
s
er
co
m
b
in
at
io
n
i
s
v
er
y
lo
w
,
t
h
e
n
t
h
e
s
it
u
atio
n
w
ill
l
ea
d
in
to
p
o
o
r
r
ec
o
m
m
en
d
atio
n
[
7
].
a.
C
o
n
te
n
t
-
b
ased
S
y
s
te
m
: P
r
ed
ictio
n
s
b
ased
o
n
t
h
e
it
e
m
s
w
h
ic
h
ar
e
s
i
m
ilar
to
th
eir
co
n
te
n
t o
r
s
e
m
an
tics
.
b.
C
o
llab
o
r
ativ
e
S
y
s
te
m
: Ca
lcu
la
tio
n
s
ar
e
b
ased
o
n
th
e
s
i
m
i
lar
r
atin
g
o
p
ted
p
r
ev
io
u
s
l
y
b
e
th
e
u
s
er
s
.
c.
H
y
b
r
id
S
y
s
te
m
:
T
h
e
co
m
b
in
atio
n
o
f
co
n
te
n
t
b
ased
a
n
d
co
llab
o
r
ativ
e
f
il
ter
in
g
m
et
h
o
d
s
f
o
r
m
H
y
b
r
i
d
r
ec
o
m
m
e
n
d
er
s
y
s
te
m
s
.
T
h
e
h
u
b
o
f
t
h
is
p
ap
er
is
p
r
i
m
ar
il
y
o
n
ite
m
b
ased
r
ati
n
g
p
r
o
p
h
ec
y
w
h
ic
h
ar
e
C
o
llab
o
r
ativ
e
Fil
ter
i
n
g
r
ec
o
m
m
e
n
d
atio
n
s
;
w
h
er
e
it
p
r
o
v
id
es
th
e
s
c
h
e
m
in
g
lo
o
m
to
r
ec
o
m
m
e
n
d
er
s
y
s
te
m
s
.
T
h
e
aim
o
f
C
o
llab
o
r
ativ
e
Fil
ter
in
g
(
C
F)
s
y
s
te
m
o
n
p
r
ed
ictin
g
a
u
s
er
‟
s
m
i
n
d
o
n
co
llect
iv
e
ite
m
s
ex
i
s
ts
in
th
e
ar
ea
,
u
s
i
n
g
u
s
er
s
‟
p
r
ev
io
u
s
av
ailab
le
o
p
in
io
n
s
o
r
r
atin
g
o
n
ite
m
s
.
C
o
llab
o
r
ativ
e
f
ilter
i
n
g
i
s
th
e
w
id
el
y
u
s
ed
p
r
ed
ictio
n
ap
p
r
o
ac
h
[1
].
T
h
e
C
lass
ic
C
F
ap
p
r
o
ac
h
w
o
r
k
s
u
s
i
n
g
th
eo
r
y
o
f
p
r
ed
ictin
g
t
h
e
u
s
er
r
atin
g
u
s
in
g
t
h
e
u
s
er
-
r
atin
g
m
atr
i
x
[
8
]
.
I
n
t
h
is
p
ap
er
,
Ker
n
el
W
ei
g
h
t
ed
K
-
m
ea
n
s
C
l
u
s
ter
i
n
g
(
KW
KC
)
ap
p
r
o
ac
h
is
u
s
ed
i
n
R
a
d
ial
b
asis
Net
w
o
r
k
(
R
B
N)
L
a
y
er
to
alle
v
iate
t
h
e
s
p
ar
s
it
y
p
r
o
b
le
m
.
T
h
e
p
r
o
p
o
s
ed
t
w
o
p
h
ases
w
ill
m
ak
e
t
h
e
p
r
ed
ictio
n
p
r
o
ce
s
s
as
a
n
i
n
cr
e
m
e
n
tal
ap
p
r
o
ac
h
,
w
h
er
e
ea
c
h
s
tate
w
il
l
i
n
cr
ea
s
e
t
h
e
ac
c
u
r
ac
y
o
f
t
h
e
p
r
ed
icatio
n
.
T
h
e
f
ir
s
t
s
tate
„
Di
s
co
n
n
ec
ted
‟
w
ill
s
m
o
o
th
en
th
e
Sp
ar
s
e
d
ata
m
atr
ix
u
s
i
n
g
KW
K
C
a
n
d
t
h
e
s
ec
o
n
d
s
tate
„
C
o
n
n
ec
ted
‟
w
il
l
p
r
o
v
id
e
th
e
r
ec
o
m
m
en
d
at
io
n
to
th
e
cu
r
r
en
t
u
s
er
.
E
x
p
er
im
en
tal
s
et
u
p
is
m
ad
e
u
s
i
n
g
M
o
v
ieL
e
n
s
Data
s
e
t
an
d
th
e
r
es
u
lts
ar
e
co
m
p
ar
ed
w
it
h
c
lass
ic
s
y
s
te
m
s
li
k
e
Si
n
g
le
Val
u
e
d
ec
o
m
p
o
s
i
tio
n
(
SV
D)
,
Su
p
p
o
r
t
Vec
to
r
m
ac
h
in
e
s
(
SVM)
a
n
d
KF
C
M
(
Ker
n
el
F
u
zz
y
C
-
Me
a
n
s
)
.
T
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
s
p
r
o
v
ed
w
it
h
co
m
p
ar
ativ
el
y
h
ig
h
ac
c
u
r
ac
y
a
n
d
q
u
alit
y
.
T
h
e
r
est
o
f
th
is
p
ap
er
is
o
r
g
a
n
ized
as:
d
is
cu
s
s
io
n
s
o
n
ex
is
ti
n
g
s
i
m
ilar
i
t
y
f
i
n
d
in
g
tec
h
n
iq
u
e
s
o
f
C
F
i
n
Sectio
n
2
,
p
r
o
p
o
s
ed
s
y
s
te
m
an
d
ar
ch
itect
u
r
e
f
lo
w
A
l
g
o
r
ith
m
s
u
s
ed
at
e
v
er
y
s
tag
e
o
f
p
r
o
p
o
s
ed
s
y
s
te
m
i
n
Sectio
n
3
,
ex
p
er
i
m
e
n
tal
r
es
u
lt
s
in
Sect
io
n
4
an
d
Sectio
n
5
c
o
n
clu
d
es t
h
e
s
y
s
te
m
alo
n
g
w
i
t
h
F
u
tu
r
e
e
x
ten
s
io
n
s
.
2.
RE
L
AT
E
D
WO
RK
As
p
er
[
9
]
,
th
er
e
ar
e
t
w
o
t
y
p
es
in
C
o
llab
o
r
ativ
e
Fi
lter
in
g
s
y
s
te
m
:
m
o
d
el(
User
/ite
m
)
an
d
m
e
m
o
r
y
b
ased
(
o
r
h
eu
r
is
tic
-
b
ased
)
.
Me
m
o
r
y
b
ased
p
r
o
ce
d
u
r
e
[
9
-
11
]
p
r
im
ar
il
y
ar
e
s
e
m
a
n
tict
h
e
p
r
ed
ictio
n
r
elies
o
n
th
e
co
m
p
letese
t
o
f
ite
m
s
r
ated
b
y
c
u
s
to
m
er
s
b
ef
o
r
eh
a
n
d
.
T
h
e
m
e
m
o
r
y
-
b
ased
C
F
alg
o
r
it
h
m
u
s
es
t
h
e
f
o
llo
w
i
n
g
s
tep
s
:
Step
1
: Fin
d
in
g
Si
m
ilar
i
t
y
u
s
i
n
g
s
i
m
ilar
it
y
f
i
n
d
i
n
g
tec
h
n
iq
u
es lik
e
P
ea
r
s
o
n
co
r
r
elatio
n
o
r
C
o
s
i
n
e
s
i
m
ilar
it
y
.
Step
2
: P
r
e
d
ict
R
atin
g
f
o
r
th
e
cu
r
r
en
t a
cti
v
e
u
s
er
o
n
T
o
p
N
it
e
m
s
f
o
u
n
d
.
Mo
s
t
o
f
th
e
t
i
m
e,
t
h
e
q
u
a
n
t
if
ied
a
m
o
u
n
t
o
f
r
ati
n
g
w
i
ll
n
o
t
b
e
av
ailab
le
to
o
b
tain
t
h
e
asp
ir
ed
r
ec
o
m
m
e
n
d
atio
n
s
.
T
h
is
u
n
s
ati
s
f
ac
to
r
y
n
u
m
b
er
o
f
r
atin
g
s
ca
u
s
e
s
th
e
s
p
ar
s
i
t
y
p
r
o
b
le
m
[
1
2
,
13
]
.
I
n
[
1
,
2
],
a
d
im
e
n
s
io
n
a
lit
y
r
ed
u
ctio
n
m
eth
o
d
o
lo
g
y
f
o
r
tack
li
n
g
w
ith
in
ad
eq
u
ate
r
atin
g
m
atr
i
x
w
as
d
is
cu
s
s
ed
,
Si
n
g
u
lar
Valu
e
Dec
o
m
p
o
s
itio
n
[
1
4
]
is
a
s
o
u
n
d
m
et
h
o
d
o
n
m
a
tr
ix
f
a
cto
r
izatio
n
th
at
g
i
v
es
lo
w
er
r
an
k
e
s
ti
m
atio
n
s
f
o
r
o
r
ig
in
al
s
p
ar
s
e
m
atr
i
x
[1
].
T
h
e
t
w
o
p
o
s
s
ib
le
s
o
lu
tio
n
s
av
a
ilab
le
f
o
r
ad
d
r
ess
in
g
s
p
a
r
s
it
y
p
r
o
b
le
m
ar
e:
t
he
f
ir
s
t
o
n
e
u
s
es
s
m
o
o
th
e
n
[
1
5
,
16
]
th
e
s
p
ar
s
e
m
atr
i
x
o
r
s
h
r
i
n
k
i
n
g
th
e
d
i
m
e
n
s
io
n
f
o
r
m
i
n
i
m
izi
n
g
t
h
e
s
p
ar
s
i
t
y
o
f
r
ati
n
g
m
atr
i
x
.
T
h
e
Seco
n
d
r
eso
lu
tio
n
p
r
o
v
id
es
th
e
i
m
p
r
o
v
ed
m
et
h
o
d
s
to
in
cr
ea
s
e
th
e
ef
f
icie
n
c
y
w
it
h
o
u
t
ch
a
n
g
i
n
g
th
e
s
p
ar
s
it
y
o
f
th
e
m
atr
ix
[
1
7
].
Fo
r
S
m
o
o
th
e
n
i
n
g
[
1
5
,
1
8
]
w
h
ic
h
w
i
ll
r
ed
u
ce
th
e
s
p
ar
s
it
y
,
R
ad
ial
B
asis
Fu
n
ctio
n
Ne
t
w
o
r
k
is
u
s
ed
b
y
p
r
o
v
id
in
g
t
h
e
ap
p
r
o
x
i
m
ate
r
at
in
g
f
o
r
m
ea
g
er
Ker
n
el
Fu
zz
y
C
m
ea
n
s
cl
u
s
ter
ed
m
atr
i
x
.
T
o
p
r
ed
ict
th
e
m
is
s
i
n
g
v
al
u
e
i
n
le
s
s
d
en
s
e
r
ati
n
g
m
atr
ix
,
a
t
w
o
s
tep
s
o
l
u
tio
n
is
i
n
tr
o
d
u
ce
d
in
[
8
,
19
]
u
s
i
n
g
C
o
-
cl
u
s
ter
i
n
g
an
d
R
ad
ia
l B
asis
Fu
n
ctio
n
.
I
n
th
i
s
s
itu
a
ti
o
n
,
if
a
n
e
w
u
s
er
is
n
o
t r
ated
s
u
f
f
icien
t
n
u
m
b
er
o
f
ite
m
s
p
r
o
p
er
ly
t
h
en
t
h
e
p
r
o
ce
s
s
o
f
p
r
ed
ictin
g
t
h
e
r
ec
o
m
m
e
n
d
atio
n
s
to
th
at
u
s
er
is
d
i
f
f
ic
u
lt
j
o
b
to
th
e
s
y
s
te
m
.
T
h
e
co
ld
s
tar
t
p
r
o
b
lem
(
n
e
w
u
s
er
p
r
o
b
le
m
)
[
1
2
,
2
0
]
w
ill
b
e
r
aised
,
if
th
er
e
i
s
n
o
m
atc
h
b
et
w
ee
n
th
e
n
e
w
u
s
er
‟
s
r
ati
n
g
s
w
it
h
th
e
p
r
ev
io
u
s
l
y
p
r
ese
n
t u
s
er
s
.
T
h
is
p
r
o
b
lem
b
e
p
r
ese
n
t
i
n
b
u
n
ch
o
f
i
n
d
u
s
tr
ial
d
o
m
a
in
s
.
C
o
l
d
s
tar
t
is
d
e
f
i
n
ed
as,
t
h
e
s
itu
a
t
io
n
w
h
er
e
eith
er
th
er
e
ar
e
n
o
t
ad
eq
u
ate
d
ata
to
ex
a
m
i
n
e
o
r
th
e
u
s
er
ar
e
m
ea
g
er
.
E
.
g
.
a
n
e
w
ac
t
iv
e
u
s
er
in
a
n
o
n
l
in
e
co
m
m
u
n
it
y
w
eb
p
ag
e,
t
h
e
u
s
e
r
d
o
esn
‟
t
h
a
v
e
e
v
e
n
a
s
i
n
g
le
co
m
p
an
io
n
o
r
li
k
el
y
ite
m
,
a
n
d
it
‟
s
n
o
t
ea
s
y
to
p
r
o
v
id
e
r
ec
o
m
m
en
d
atio
n
s
to
s
u
ch
u
s
er
.
T
h
er
e
co
m
es t
h
e
m
aj
o
r
n
atu
r
e
o
f
co
ld
s
tar
t p
r
o
b
le
m
s
in
C
F,
Ne
w
User
o
r
Ne
w
I
te
m
.
H
y
b
r
id
C
F
s
y
s
te
m
s
,
lik
e
co
n
t
en
t
en
h
an
ce
d
C
F
tec
h
n
iq
u
e
[
2
1
]
,
u
s
ed
to
ad
d
r
ess
th
e
s
p
ar
s
ity
p
r
o
b
le
m
;
w
h
er
e
e
x
ter
io
r
co
n
te
n
t
ca
n
b
e
u
s
ed
to
g
e
n
er
at
e
m
i
s
s
i
n
g
r
ati
n
g
s
f
o
r
n
e
w
u
s
er
/ite
m
.
Du
r
i
n
g
t
h
e
R
B
F
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
n
en
h
a
n
ce
d
ke
r
n
el
w
eig
h
ted
co
lla
b
o
r
a
tive
r
ec
o
mme
n
d
ed
s
ystem
to
a
llevia
te
s
p
a
r
s
ity
(
S
.
B
a
b
ee
th
a
)
449
r
ec
o
m
m
e
n
d
atio
n
ap
p
r
o
ac
h
[
8
,
18
]
,
if
th
e
ac
ti
v
eu
s
er
is
n
e
w
to
th
i
s
s
y
s
te
m
o
r
ap
p
licatio
n
w
i
t
h
o
u
t
h
a
v
i
n
g
r
atin
g
(
Ne
w
u
s
er
co
ld
s
tar
t p
r
o
b
lem
)
;
th
e
to
p
r
ated
item
s
o
n
v
er
y
cl
u
s
ter
is
s
u
g
g
ested
f
o
r
th
at
ac
ti
v
e
u
s
er
.
T
h
e
Mo
d
el
-
b
ased
[3
,
4
,
2
2
]
C
F
ap
p
r
o
ac
h
ad
d
r
ess
es
t
h
e
s
e
p
r
o
b
lem
s
u
s
in
g
th
e
s
et
o
f
r
a
tin
g
s
to
b
e
tr
ain
ed
a
m
o
d
el
an
d
it
w
ill
b
e
u
tili
ze
d
to
p
r
ed
ict
r
atin
g
.
T
h
e
li
k
elih
o
o
d
th
at
u
s
er
p
r
ate
th
e
p
ar
ticu
lar
ite
m
i
g
iv
e
n
th
at
p
r
ev
io
u
s
r
ati
n
g
o
f
u
s
er
o
n
t
h
e
r
ated
ite
m
s
,
is
u
s
u
all
y
d
eter
m
in
ed
b
y
t
w
o
p
r
o
b
ab
ilis
tic
m
o
d
els:
clu
s
ter
m
o
d
els
a
n
d
B
a
y
esia
n
n
et
w
o
r
k
s
[
17
]
.
T
h
e
b
asic
B
ay
esia
n
co
llab
o
r
ativ
e
F
ilter
i
n
g
alg
o
r
ith
m
[1
7
]
u
s
e
s
a
NB
(
n
aiv
e
B
a
y
es)
s
tr
atag
e
m
o
n
p
r
ed
ictio
n
s
.
W
ith
t
h
e
g
i
v
e
n
is
o
lated
class
e
s
,
th
e
cla
s
s
h
a
v
i
n
g
p
ee
k
p
r
o
b
ab
ilit
y
w
il
l
b
e
ca
te
g
o
r
ized
as
t
h
e
p
r
ed
icted
class
[
1
]
.
P
lace
th
e
b
ase
B
a
y
esia
n
al
g
o
r
ith
m
to
d
iv
er
s
ed
ata
f
o
r
C
F
o
b
j
ec
tiv
es
[
1
7
]
,
p
r
o
d
u
ce
s
p
r
ed
ictiv
ea
cc
u
r
ac
y
i
n
w
o
r
s
e
m
a
n
n
er
w
ith
ad
v
is
ab
le
s
ca
lab
ilit
y
th
an
t
h
e
P
ea
r
s
o
n
co
r
r
elatio
n
.
C
lu
s
ter
i
n
g
C
o
llab
o
r
atin
g
Fil
t
er
in
g
m
e
m
o
r
y
b
ased
A
p
p
r
o
ac
h
:
AC
l
u
s
ter
i
s
d
ef
i
n
ed
as
a
s
et
o
f
d
ata
ite
m
s
w
h
ich
ar
e
co
r
r
elate
d
to
w
ar
d
s
ea
c
h
o
th
er
w
i
th
in
th
e
cu
r
r
en
t
cl
u
s
ter
an
d
ar
e
u
n
co
n
n
ec
ted
to
t
h
e
ite
m
s
a
m
o
n
g
r
est
o
f
t
h
ec
l
u
s
ter
s
[
23
]
.
L
eg
ac
y
co
llab
o
r
ativ
e
filt
er
in
g
h
a
v
in
g
les
s
s
ca
lab
ilit
y
t
h
a
n
C
lu
s
ter
i
n
g
m
o
d
el
s
,
b
ec
au
s
e
th
e
y
m
a
k
e
p
r
ed
ictio
n
s
w
it
h
i
n
co
m
p
ar
ati
v
el
y
s
m
a
ll
clu
s
ter
s
w
h
ich
i
s
k
n
o
w
n
u
s
r
ed
u
ce
d
d
im
e
n
s
io
n
d
ata
s
et
[
7
,
2
1
,
2
4
].
A
s
t
h
e
u
s
er
b
ase
g
r
o
w
s
i
n
v
o
lu
m
e
s
,
th
e
n
User
–
u
s
er
co
llab
o
r
ativ
e
Fil
t
er
,
w
h
ile
e
ff
ec
ti
v
e,
su
ff
er
f
r
o
m
s
ca
lab
ilit
y
i
s
s
u
e
s
.
T
o
p
-
N
ite
m
b
ased
r
ec
o
m
m
en
d
atio
n
m
eth
o
d
s
w
i
ll
b
e
u
s
ed
to
alle
v
iate
th
e
s
ca
lab
ilit
y
p
r
o
b
lem
o
f
to
p
-
N
u
s
er
b
ased
r
ec
o
m
m
en
d
at
io
n
[
25
].
I
n
s
tead
o
f
f
i
n
d
in
g
u
s
er
-
u
s
er
s
i
m
ilar
it
y
,
it p
r
o
p
o
s
es to
ca
lcu
late
s
i
m
i
lar
ities
o
f
t
w
o
s
et
s
an
d
t
h
en
s
o
r
t th
e
s
i
m
ilar
it
y
v
a
lu
e
s
in
d
ec
r
ea
s
in
g
o
r
d
er
[
26
]
.
3.
P
RO
P
O
SE
D
SYS
T
E
M
USI
NG
K
WK
C
T
h
e
p
r
o
p
o
s
ed
s
y
s
te
m
is
s
h
o
w
n
i
n
Fi
g
u
r
e
1
.
As
s
h
o
w
n
,
th
e
p
h
a
s
e
„
Dis
co
n
n
ec
ted
‟
w
i
ll
p
er
f
o
r
m
th
e
lear
n
i
n
g
ac
tiv
it
y
f
r
o
m
t
h
e
ex
is
t
in
g
r
atin
g
m
atr
i
x
w
h
er
e
t
h
r
ee
m
aj
o
r
s
tep
s
w
il
l b
e
ca
r
r
ied
o
u
t:
Step
1
:
UC
C
(
U
n
s
u
p
er
v
is
ed
C
o
r
r
elatio
n
C
lu
s
ter
in
g
)
Step
2
: S
m
o
o
th
e
n
i
n
g
u
s
i
n
g
E
u
clid
ea
n
No
r
m
Step
3
:
KW
KNN
clu
s
ter
i
n
g
(
Ker
n
el
W
eig
h
ted
K
-
m
ea
n
s
Nea
r
est Ne
ig
h
b
o
u
r
)
T
h
e
p
h
ase
„
C
o
n
n
ec
ted
‟
w
ill p
e
r
f
o
r
m
th
e
o
n
li
n
e
r
ec
o
m
m
e
n
d
at
io
n
to
th
e
cu
r
r
en
t a
ctiv
e
u
s
er
.
Step
4
: P
r
ed
ict
th
e
r
ec
o
m
m
e
n
d
atio
n
s
f
o
r
cu
r
r
en
t u
s
er
Fig
u
r
e
1
.
KW
KC
ap
p
r
o
ac
h
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
4
4
7
-
454
450
3
.
1
.
Dis
co
nn
ec
t
ed
s
t
a
t
e
3
.
1
.
1
.
Alg
o
r
it
hm
1
:
UCC
(
un
s
up
er
v
is
ed
co
rr
el
a
t
io
n
clus
t
er
ing
)
p
ha
s
e
I
np
ut
:
is
th
e
S
p
a
r
s
e
C
u
s
to
m
er
-
I
tem
r
a
tin
g
M
at
r
ix
.
A
v
er
ag
e
R
atin
g
o
f
u
s
e
r
s
o
n
item
is
̅
O
utpu
t
:
-
n
u
m
b
er
o
f
u
s
e
r
g
r
o
u
p
s
-
C
al
le
d
C
lu
s
te
r
s
Ste
ps
:
1.
Fo
r
x
=
1
t
o
N
//
Sim
ila
r
ity
is
g
o
in
g
to
b
e
c
alcu
l
ate
d
am
o
n
g
N
u
s
er
s
//
1
.
1
.
Sim
ilar
ity
_
I
n
d
ex
(
x
)
=0
//
I
n
it
ia
lize
b
y
Z
er
o
w
h
ich
w
ill d
ete
r
m
in
e
th
e
s
im
ilar
u
s
e
r
s
1
.
2
.
Fo
r
y
=1
t
o
N
//F
in
d
Sim
ilar
ity
b
e
tw
ee
n
u
s
er
s
u
s
in
g
P
e
ar
s
o
n
co
r
r
el
ati
o
n
f
u
n
cti
o
n
//
∑
̅
̅
√
∑
̅
√
∑
̅
(
1
)
w
h
er
e
p
an
d
q
ar
e
u
s
er
s
,
U
is
th
e
ite
m
co
llectio
n
r
ated
b
y
u
s
er
s
p
&
q
,
̅
is
av
er
ag
e
r
ati
n
g
o
f
u
s
er
(
p
o
r
q
)
,
u
s
er
u
„
s
r
ati
n
g
o
n
ite
m
.
1
.
3
.
I
f
=1
th
en
Si
m
ilar
it
y
_
I
n
d
ex
(
x
)
=
Si
m
ilar
it
y
_
I
n
d
ex
(
x
)
+1
2.
Fetch
k
u
s
er
s
‟
w
h
o
s
e
Si
m
ilar
it
y
_
I
n
d
e
x
v
al
u
e
i
s
t
h
e
h
ig
h
es
t
a
m
o
n
g
N.
Ass
ig
n
t
h
o
s
e
k
u
s
er
s
as
cl
u
s
ter
C
en
ter
s
.
3.
C
lu
b
th
e
r
est
o
f
th
e
u
s
er
s
(
N
-
k
)
to
co
r
r
esp
o
n
d
in
g
clu
s
ter
,
wh
er
e
th
e
=1
.
(
if
p
–
is
clu
s
ter
ce
n
ter
th
en
)
3
.
1
.
2
.
Alg
o
r
it
hm
2
:
Sm
o
o
t
h
enin
g
u
s
ing
eu
clid
ea
n
n
o
rm
I
np
ut:
Sp
ar
s
e
C
u
s
to
m
er
-
I
te
m
r
atin
g
Ma
tr
i
x
.
;N
-
T
o
tal
n
u
m
b
e
r
o
f
u
s
er
s
,
-
T
o
tal
Nu
m
b
er
o
f
I
t
e
m
s
.
an
d
ar
e
Min
im
u
m
a
n
d
Ma
x
i
m
u
m
Val
u
es
o
f
th
e
ac
ti
v
ati
o
n
o
r
Ob
j
ec
tiv
e
f
u
n
ctio
n
o
f
a
p
ar
ticu
lar
p
ar
am
e
ter
i (
w
h
er
e
i c
an
b
e
eith
er
ite
m
o
r
u
s
er
)
.
O
utput
:
S
m
o
o
th
e
n
ed
r
atin
g
Ma
tr
ix
Ste
ps
:
1.
C
o
llectio
n
=(
R
ati
n
g
_
Ma
x
-
R
ati
n
g
-
m
i
n
)
+1
2.
Der
iv
e
k
cl
u
s
ter
s
s
u
ch
t
h
at
⌊
⌋
3.
Fo
r
i=1
to
3
.
1
.
Sep
ar
a
te
th
e
u
s
er
s
in
to
k
clu
s
t
er
s
u
s
i
n
g
alg
o
r
it
h
m
1
.
3
.
2
.
C
alcu
late
E
u
clid
ea
n
ce
n
ter
s
∑
(
2
)
w
h
er
e
is
r
ep
r
esen
ti
n
g
n
u
m
b
er
o
f
u
s
er
s
w
it
h
in
a
cl
u
s
ter
.
3
.
3
.
Fin
d
t
h
e
E
u
clid
ea
n
d
is
tan
ce
m
atr
ix
‖
‖
w
h
er
e
3
.
4
.
C
o
m
p
u
te
th
e
o
b
j
ec
tiv
e
f
u
n
c
tio
n
u
s
in
g
Ga
u
s
s
ian
p
o
s
iti
v
e
d
ef
in
ite
f
u
n
ctio
n
w
h
er
e
(
3
)
3
.
5
.
Dete
r
m
i
n
e
th
e
w
e
ig
h
ts
u
s
in
g
p
s
eu
d
o
r
an
d
o
m
w
ei
g
h
t f
u
n
c
tio
n
(
)
⁄
∑
(
)
⁄
(
4
)
3
.
6
.
C
alcu
late
̃
(
)
w
h
er
e
(
)
∑
(
‖
‖
)
(
5
)
is
th
e
ac
ti
v
atio
n
f
u
n
ctio
n
a
n
d
,
is
th
e
cl
u
s
t
er
ce
n
ter
.
3
.
7
.
Der
iv
e
th
e
S
m
o
o
t
h
en
ed
r
atin
g
Ma
tr
ix
{
̃
(
6
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
n
en
h
a
n
ce
d
ke
r
n
el
w
eig
h
ted
co
lla
b
o
r
a
tive
r
ec
o
mme
n
d
ed
s
ystem
to
a
llevia
te
s
p
a
r
s
ity
(
S
.
B
a
b
ee
th
a
)
451
3
.
1
.
3
.
Alg
o
r
it
hm
3
:
K
WK
N
N
clus
t
er
in
g
(
k
er
n
el
wei
g
h
t
ed
k
-
m
e
a
ns
n
e
a
re
s
t
nei
g
hb
o
u
r)
I
np
ut:
S
m
o
o
th
e
n
ed
r
atin
g
Ma
tr
ix
O
utput
:
Ker
n
el
w
ei
g
h
ted
C
l
u
s
ter
ce
n
ter
s
,
m
e
m
b
er
s
h
ip
d
eg
r
ee
f
o
r
u
s
er
c
i
n
cl
u
s
ter
cl.
T
h
e
m
e
m
b
er
s
h
ip
r
an
g
e
w
ill b
e
0
-
1
w
h
ic
h
in
d
icate
s
th
e
i
n
ter
est o
f
u
s
er
o
n
t
h
e
clu
s
ter
.
Ste
ps
:
1.
I
n
itialize
∑
⁄
s
et
q
=1
,
w
h
ic
h
d
eter
m
i
n
e
t
h
e
n
u
m
b
er
o
f
er
r
o
r
cy
cle
s
.
2.
C
alcu
late
Ker
n
el
w
ei
g
h
ted
C
l
u
s
ter
ce
n
ter
s
∑
∑
(
7
)
3.
w
h
er
e
4.
C
o
m
p
u
te
m
em
b
er
s
h
ip
v
alu
e
∑
(
8
)
5.
C
alcu
la
te
E
r
r
o
r
in
d
ex
=m
ax
(
o
l
d
,
n
ew
6.
I
f
E
r
r
o
r
in
d
ex
>0
.
5
,
th
en
q
=
q
+
1
;
g
o
t
o
s
te
p
2
3
.
2
.
Co
nn
ec
t
ed
s
t
a
t
e
Alg
o
rit
h
m
4
:
P
re
dict
t
he
re
co
mm
enda
t
io
ns
f
o
r
curr
ent
u
s
er
(
o
nli
ne
a
ct
iv
it
y
)
I
np
ut:
C
u
r
r
en
t
ac
ti
v
e
u
s
er
R
ati
n
g
ar
r
a
y
an
d
d
en
o
tes
t
h
e
p
r
ed
icted
r
ec
o
m
m
en
d
ed
it
e
m
s
,
Ker
n
el
w
ei
g
h
ted
C
l
u
s
ter
ce
n
ter
s
an
d
th
e
Gau
s
s
ia
n
ac
ti
v
atio
n
f
u
n
c
tio
n
.
O
utput
:
R
ec
o
m
m
e
n
d
ed
I
tem
s
f
o
r
th
e
cu
r
r
en
t
u
s
er
Ste
ps
:
1.
I
f
t
h
e
c
u
r
r
en
t
u
s
er
is
a
n
e
w
u
s
er
an
d
t
h
e
R
atin
g
ar
r
a
y
i
s
NU
L
L
(
Ne
w
u
s
er
co
ld
s
tar
t)
,
t
h
e
n
p
r
o
v
id
e
T
o
p
N
r
ated
ite
m
s
f
r
o
m
ea
ch
cl
u
s
te
r
to
th
e
r
ec
o
m
m
e
n
d
er
en
g
i
n
e.
2.
I
f
th
e
cu
r
r
en
t
u
s
er
is
a
n
e
w
u
s
er
an
d
h
av
e
r
ated
f
e
w
ite
m
s
(
an
d
also
R
atin
g
ar
r
a
y
is
n
o
t
NUL
L
)
th
e
n
p
er
f
o
r
m
b
elo
w
s
tep
s
f
o
llo
w
ed
b
y
s
h
ar
i
n
g
o
f
T
o
p
N
r
ated
ite
m
s
f
r
o
m
ea
c
h
clu
s
ter
to
th
e
r
ec
o
m
m
e
n
d
er
en
g
in
e.
2
.
1
.
C
alcu
la
te
d
eg
r
ee
o
f
m
e
m
b
er
s
h
ip
w
ith
all
c
lu
s
te
r
s
2
.
2
.
P
r
ed
ict
th
e
u
n
k
n
o
w
n
r
atin
g
o
f
th
e
cu
r
r
en
t a
ctiv
e
u
s
er
∑
∑
(
9
)
3.
I
f
th
e
c
u
r
r
en
t
u
s
er
is
a
n
e
x
is
tin
g
u
s
er
an
d
w
it
h
n
o
ad
d
ed
r
atin
g
t
h
e
n
p
r
o
v
id
e
T
o
p
N
tr
ain
ed
p
h
ase
p
r
ed
icted
r
atin
g
to
th
e
r
ec
o
m
m
en
d
er
en
g
i
n
e.
4.
I
f
th
e
c
u
r
r
en
t u
s
er
is
a
n
ex
i
s
ti
n
g
u
s
er
an
d
w
it
h
ad
d
ed
n
e
w
r
ati
n
g
s
th
e
g
o
to
s
tep
2
.
4.
E
XP
E
R
I
M
E
NT
A
L
SE
T
UP
AND
RE
SUL
T
S
4
.
1
.
T
est
s
et
T
h
e
ex
p
er
im
e
n
tal
s
et
u
p
w
a
s
m
ad
e
u
s
i
n
g
Mo
v
ie
L
en
s
d
ata
s
et
w
it
h
,
1
0
0
,
0
0
0
u
s
er
r
atin
g
s
f
r
o
m
1
0
0
0
o
d
d
u
s
er
s
f
o
r
1
7
8
3
m
o
v
ie
s
.
F
o
r
co
m
p
ar
ativ
e
p
u
r
p
o
s
e,
th
e
d
ata
is
s
p
lit
w
it
h
d
i
f
f
er
e
n
t
le
v
el
o
f
s
p
ar
s
it
y
(
2
0
%
to
9
0
%).
T
o
an
aly
ze
t
h
e
r
esu
lts
t
h
e
d
ata
s
et
d
iv
id
ed
in
to
t
w
o
s
e
ts
: tr
ain
i
n
g
d
ataset
a
n
d
test
i
n
g
d
ataset.
4
.
2
.
P
er
f
o
r
m
a
nce
m
ea
s
ure
s
4
.
2
.
1
.
M
e
a
n
a
bs
o
lute
er
ro
r
(
M
AE
)
MA
E
is
t
h
e
tr
ad
itio
n
al
cla
s
s
if
icatio
n
ac
cu
r
ac
y
to
m
ea
s
u
r
e
h
o
w
clo
s
e
t
h
e
r
ec
o
m
m
e
n
d
er
s
y
s
t
e
m
s
p
r
ed
icted
r
atin
g
s
ar
e
to
th
e
tr
u
e
u
s
er
r
atin
g
s
(
f
o
r
all
t
h
e
r
atin
g
s
in
th
e
te
s
t
s
e
t)
.
I
n
o
u
r
ex
p
er
i
m
e
n
ts
,
w
e
ca
lc
u
late
th
e
ac
c
u
r
ac
y
f
o
r
ex
is
ti
n
g
C
F
alg
o
r
it
h
m
s
an
d
KW
KC
u
s
i
n
g
Me
a
n
A
b
s
o
lu
te
E
r
r
o
r
(
MA
E
)
as
s
h
o
w
n
in
F
ig
u
r
e
2
u
s
i
n
g
(
1
0
)
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
4
4
7
-
454
452
N
i
M
j
i
i
M
N
r
p
M
A
E
1
1
*
|
|
(
1
0
)
w
h
er
e,
N
is
t
h
e
n
u
m
b
er
o
f
I
te
m
s
,
M
is
t
h
e
n
u
m
b
er
o
f
User
s
,
is
th
e
P
r
ed
icted
r
atin
g
an
d
is
th
e
ac
tu
al
o
r
o
r
ig
in
al
r
ati
n
g
g
i
v
en
b
y
th
e
u
s
er
in
itiall
y
.
4
.
2
.
2
.
P
re
cisi
o
n a
nd
r
ec
a
ll
T
h
e
s
ec
o
n
d
m
ea
s
u
r
e
is
t
h
e
r
elev
an
t
r
ec
o
m
m
e
n
d
atio
n
s
m
ea
s
u
r
e
u
s
i
n
g
P
r
ec
is
io
n
an
d
R
ec
all
.
P
r
ec
is
io
n
an
d
r
ec
all
ar
e
th
e
m
o
s
t
p
o
p
u
la
r
m
etr
ics
f
o
r
ev
al
u
ati
n
g
in
f
o
r
m
atio
n
r
etr
iev
al
s
y
s
te
m
s
.
P
r
ec
is
io
n
i
s
t
h
e
r
atio
o
f
r
elev
an
t
i
te
m
s
s
elec
ted
b
y
t
h
e
r
ec
o
m
m
e
n
d
er
to
th
e
n
u
m
b
er
o
f
ite
m
s
s
e
lecte
d
u
s
i
n
g
(
1
1
)
.
R
ec
all
is
th
e
r
atio
o
f
r
elev
an
t i
te
m
s
s
elec
ted
to
th
e
n
u
m
b
er
o
f
r
elev
a
n
t
u
s
i
n
g
(
1
2
)
.
N
N
N
N
e
c
i
s
i
o
n
N
i
is
s
s
1
Pr
(
1
1
)
N
N
N
N
c
a
l
l
N
i
rn
s
s
1
Re
(
1
2
)
w
h
er
e
th
e
v
al
u
es a
r
e
tak
e
n
as
s
h
o
w
n
T
ab
le1
.
T
ab
le
1
.
I
tem
c
ateg
o
r
ies
S
e
l
e
c
t
e
d
N
o
t
S
e
l
e
c
t
e
d
T
o
t
a
l
R
e
l
e
v
a
n
t
rs
N
rn
N
r
N
I
r
r
e
l
e
v
a
n
t
is
N
in
N
i
N
T
o
t
a
l
s
N
n
N
N
5.
RE
SU
L
T
S AN
D
D
I
SCU
SS
I
O
N
5
.
1
.
Chec
k
po
int
f
o
r
co
m
pu
t
a
t
io
na
l
er
ro
r
T
o
in
v
esti
g
ate
t
h
e
Me
an
Ab
s
o
lu
te
E
r
r
o
r
,
th
e
p
r
im
ar
y
tech
n
iq
u
es
li
k
e
SVD
(
Si
n
g
u
lar
Valu
e
Dec
o
m
p
o
s
itio
n
)
,
SVM
(
Si
n
g
u
lar
Vec
to
r
Ma
ch
in
e)
,
R
B
FN
(
R
ad
ial
B
asis
Fu
n
ctio
n
al
Net
wo
r
k
)
w
it
h
P
ea
r
s
o
n
co
r
r
elatio
n
an
d
Ker
n
el
F
u
zz
y
C
-
Me
an
s
(
KF
C
M)
o
f
C
F
ar
e
co
m
p
ar
ed
w
ith
th
e
p
r
o
p
o
s
ed
alg
o
r
it
h
m
KW
KC
.
T
h
e
er
r
o
r
c
o
m
p
u
tatio
n
m
ad
e
f
o
r
b
o
th
in
Dis
co
n
n
ec
ted
p
h
ase
(
Mo
d
elin
g
ti
m
e)
an
d
co
n
n
ec
ted
p
h
ase
(
P
r
ed
ictio
n
tim
e)
w
h
ic
h
ar
e
s
h
o
w
n
in
F
ig
u
r
e
2
.
Fig
u
r
e
2
.
C
o
m
p
u
tat
io
n
al
e
r
r
o
r
w
it
h
M
A
E
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J
E
lec&
C
o
m
p
E
n
g
I
SS
N:
2
0
8
8
-
8708
A
n
en
h
a
n
ce
d
ke
r
n
el
w
eig
h
ted
co
lla
b
o
r
a
tive
r
ec
o
mme
n
d
ed
s
ystem
to
a
llevia
te
s
p
a
r
s
ity
(
S
.
B
a
b
ee
th
a
)
453
a.
SVD
s
h
o
w
s
h
ig
h
er
r
o
r
v
alu
e
b
o
th
o
f
th
e
co
m
p
u
tat
io
n
al
p
h
a
s
es.
b.
A
lb
eit
R
B
FN
s
h
o
w
s
Nil
er
r
o
r
o
n
Mo
d
elin
g
ti
m
i
n
g
,
th
e
p
r
o
p
o
s
ed
KW
KC
alg
o
r
ith
m
also
s
h
o
w
s
n
e
g
li
g
ib
le
er
r
o
r
w
h
ich
i
s
v
er
y
clo
s
u
r
e
to
Nil.
c.
T
h
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
s
h
o
ws th
e
i
m
p
r
o
v
is
ed
r
es
u
lts
co
m
p
ar
ed
to
r
est o
f
R
B
F tec
h
n
iq
u
es
.
5
.
2
.
Chec
k
p
o
int
f
o
r
Rele
v
a
nce
in t
er
m
s
o
f
ef
f
iciency
T
h
e
p
r
ec
is
io
n
an
d
R
ec
all
p
er
ce
n
tag
e
i
s
n
ea
r
m
o
r
e
t
h
an
9
0
p
er
ce
n
tag
e,
w
h
ich
i
s
co
m
p
ar
ati
v
el
y
h
i
g
h
w
h
e
n
co
m
p
ar
ed
to
r
est
o
f
b
e
n
ch
m
ar
k
ed
p
r
i
m
ar
y
C
F
T
ec
h
n
iq
u
e
s
.
Fi
g
u
r
e
3
s
h
o
w
n
r
elev
an
ce
m
ea
s
u
r
e
u
s
i
n
g
p
r
ec
is
io
n
an
d
r
ec
a
ll.
Fig
u
r
e
3
.
R
elev
a
n
ce
m
ea
s
u
r
e
u
s
i
n
g
P
r
ec
is
io
n
an
d
R
ec
all
6.
F
UT
UR
E
WO
RK
AND
CO
NCLUS
I
O
N
On
e
o
f
th
e
b
est
w
a
y
to
m
i
n
i
n
g
t
h
e
r
eq
u
ir
ed
p
r
o
d
u
cts
in
th
e
ec
o
m
m
er
ce
w
o
r
d
is
R
ec
o
m
m
en
d
er
s
y
s
te
m
.
I
f
t
h
e
r
ec
o
m
m
e
n
d
atio
n
s
b
ased
o
n
r
ea
lit
y
o
f
t
h
e
u
s
e
r
tast
es
t
h
e
n
t
h
e
ac
cu
r
ac
y
w
il
l
b
e
u
n
b
elie
v
ab
le
.
A
lb
eit
t
h
e
p
r
i
m
ar
y
tec
h
n
iq
u
es
ar
e
b
en
ch
m
ar
k
ed
,
th
er
e
ar
e
j
e
o
p
ar
d
y
w
h
er
e
th
e
v
o
l
u
m
e
i
n
cr
ea
s
e
s
b
u
t
th
e
ac
q
u
ir
ed
u
s
er
r
ati
n
g
d
ec
r
ea
s
es.
I
n
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
KW
KC
,
b
o
th
ac
cu
r
ac
y
a
n
d
ef
f
icien
c
y
is
p
r
o
v
en
o
n
th
e
co
m
p
ar
is
o
n
m
a
d
e
w
ith
ex
p
er
i
m
e
n
tal
s
et
u
p
w
it
h
u
s
e
r
d
ata.
C
o
m
p
ar
ativ
el
y
b
etter
r
esu
lts
s
h
o
w
n
o
n
th
e
p
r
o
p
o
s
ed
s
y
s
te
m
.
A
f
u
t
u
r
e
en
h
an
ce
m
e
n
t
is
p
la
n
n
ed
to
ex
ten
d
th
e
r
esear
ch
o
f
g
r
o
u
p
i
n
g
o
f
I
te
m
s
cl
u
s
ter
s
alo
n
g
w
it
h
U
s
er
clu
s
ter
s
to
p
r
o
v
id
e
b
etter
ac
cu
r
ac
y
.
RE
F
E
R
E
NC
E
S
[1
]
G
.
A
d
o
m
a
v
iciu
s
a
n
d
A
.
T
u
z
h
il
in
,
“
T
o
w
a
rd
th
e
Ne
x
t
G
e
n
e
ra
ti
o
n
o
f
Re
c
o
m
m
e
n
d
e
r
S
y
ste
m
s:
A
S
u
rv
e
y
o
f
th
e
S
tate
-
of
-
th
e
-
A
rt
a
n
d
P
o
ss
ib
le
Ex
ten
sio
n
s
,
”
IEE
E
T
ra
n
sa
c
ti
o
n
s
o
n
Kn
o
wled
g
e
a
n
d
Da
t
a
E
n
g
i
n
e
e
rin
g
,
v
o
l.
1
7
,
2
0
0
5
.
[2
]
G
.
A
d
o
m
a
v
iciu
s,
e
t
a
l
.
,
“
In
c
o
rp
o
ra
ti
n
g
Co
n
tex
tu
a
l
In
f
o
rm
a
ti
o
n
i
n
Re
c
o
m
m
e
n
d
e
r
S
y
ste
m
s
Us
in
g
a
M
u
lt
i
d
im
e
n
sio
n
a
l
A
p
p
ro
a
c
h
,
”
AC
M
T
ra
n
s.
I
n
fo
rm
a
ti
o
n
S
y
ste
ms
,
v
o
l.
2
3
,
2
0
0
5
.
[3
]
G
.
A
d
o
m
a
v
iciu
s
a
n
d
A
.
T
u
z
h
il
in
,
“
Ex
p
e
rt
-
Driv
e
n
V
a
li
d
a
ti
o
n
o
f
Ru
le
-
Ba
se
d
Us
e
r
M
o
d
e
ls
i
n
P
e
rso
n
a
li
z
a
ti
o
n
A
p
p
li
c
a
ti
o
n
s,”
Da
t
a
M
in
i
n
g
a
n
d
Kn
o
wled
g
e
Disc
o
v
e
ry
,
v
o
l.
5
,
p
p
.
3
3
-
5
8
,
2
0
0
1
a
.
[4
]
G
.
A
d
o
m
a
v
iciu
s
a
n
d
A
.
T
u
z
h
il
in
,
“
M
u
lt
id
im
e
n
sio
n
a
l
Re
c
o
m
m
e
n
d
e
r
S
y
ste
m
s:
A
Da
ta
W
a
r
e
h
o
u
sin
g
A
p
p
ro
a
c
h
,
”
Pro
c
.
S
e
c
o
n
d
I
n
t’l
W
o
rk
sh
o
p
El
e
c
tro
n
ic C
o
mm
e
rc
e
(
W
EL
COM
’0
1
)
,
2
0
0
1
b
.
[5
]
J.L
.
He
rlo
c
k
e
r
,
e
t
a
l
.
,
“
A
n
A
lg
o
ri
th
m
ic
F
ra
m
e
w
o
r
k
f
o
r
P
e
rf
o
r
m
in
g
Co
ll
a
b
o
ra
ti
v
e
F
il
terin
g
,
”
Pro
c
.
2
2
n
d
An
n
.
I
n
t’l
ACM
S
IGIR
Co
n
f.
Res
e
a
rc
h
a
n
d
De
v
e
lo
p
me
n
t
in
I
n
f
o
rm
a
ti
o
n
Retrie
v
a
l
(
S
IGIR
’9
9
)
,
1
9
9
9
.
[6
]
J.L
.
He
rlo
c
k
e
r,
e
t
a
l
.
,
“
Ex
p
lain
in
g
Co
ll
a
b
o
ra
ti
v
e
F
il
terin
g
Re
c
o
m
m
e
n
d
a
ti
o
n
s,”
Pro
c
.
ACM
C
o
n
f.
C
o
mp
u
ter
S
u
p
p
o
rte
d
Co
o
p
e
ra
ti
v
e
W
o
rk
,
2
0
0
0
.
[7
]
J.L
.
He
rlo
c
k
e
r
a
n
d
J.A
.
Ko
n
sta
n
,
“
Co
n
ten
t
-
In
d
e
p
e
n
d
e
n
t
T
a
sk
F
o
c
u
se
d
Re
c
o
m
m
e
n
d
a
ti
o
n
,
”
IEE
E
In
ter
n
e
t
Co
mp
u
t
in
g
,
v
o
l
.
5
,
p
p
.
4
0
-
4
7
,
2
0
0
1
.
[8
]
M.K
.D
e
v
i
a
n
d
P
.
V
e
n
k
a
tes
h
,
“
A
n
I
m
p
ro
v
e
d
Co
ll
a
b
o
ra
ti
v
e
Re
c
o
m
m
e
n
d
e
r
S
y
st
e
m
,
”
2
0
0
9
F
irst
In
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
Ne
two
rk
s
&
Co
mm
u
n
ica
t
io
n
s
,
2
0
0
9
.
[9
]
A
.
A
n
sa
ri,
e
t
a
l
.
,
“
I
n
tern
e
t
Re
c
o
m
m
e
n
d
a
ti
o
n
s
S
y
ste
m
s,”
J
.
M
a
rk
e
ti
n
g
Res
e
a
rc
h
,
p
p
.
3
6
3
-
3
7
5
,
2
0
0
0
.
[1
0
]
C.
C.
A
g
g
a
r
wa
l,
e
t
a
l
.
,
“
Ho
rti
n
g
Ha
tch
e
s
a
n
Eg
g
:
A
Ne
w
G
r
a
p
h
-
T
h
e
o
re
ti
c
A
p
p
ro
a
c
h
to
C
o
ll
a
b
o
ra
ti
v
e
F
il
terin
g
,
”
Pro
c
.
Fi
ft
h
ACM
S
IGKD
D In
t’l
C
o
n
f.
Kn
o
wled
g
e
Disc
o
v
e
ry
a
n
d
D
a
ta
M
i
n
in
g
,
1
9
9
9
.
[1
1
]
X
iao
x
ia
S
u
n
a
n
d
Na
ss
e
r
M
.
Na
sra
b
a
d
“
T
a
s
k
-
Driv
e
n
Dic
ti
o
n
a
ry
L
e
a
rn
in
g
f
o
r
H
y
p
e
rsp
e
c
tral
I
m
a
g
e
Clas
si
f
ica
ti
o
n
W
it
h
S
tru
c
tu
re
d
S
p
a
rsity
Co
n
stra
in
ts”
,
IEE
E
T
ra
n
sa
c
ti
o
n
s
On
Ge
o
sc
ien
c
e
A
n
d
Re
m
o
te
S
e
n
sin
g
,
v
o
l.
5
3
,
n
o
.
8
,
p
p
.
4
4
5
7
-
4
4
7
1
a
u
g
u
st
2
0
1
5
.
[1
2
]
B.
V
.
A
rv
in
d
,
e
t
a
l
.
,
“
A
n
i
m
p
ro
v
ise
d
f
il
terin
g
b
a
se
d
in
t
e
ll
ig
e
n
t
re
c
o
m
m
e
n
d
a
ti
o
n
tec
h
n
iq
u
e
f
o
r
w
e
b
p
e
rso
n
a
li
z
a
ti
o
n
,”
In
d
ia
c
o
n
fer
e
n
c
e
(
INDICO
N)
2
0
1
2
,
An
n
u
a
l
IE
EE
,
2
0
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
n
t J
E
lec&
C
o
m
p
E
n
g
,
Vo
l.
10
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
0
:
4
4
7
-
454
454
[1
3
]
Ilk
e
rBa
y
r
a
m
“
S
p
a
rsit
y
W
it
h
in
a
n
d
A
c
ro
ss
Ov
e
rlap
p
in
g
G
ro
u
p
s”
,
IEE
E
S
ig
n
a
l
P
r
o
c
e
ss
in
g
L
e
tt
e
rs,
v
o
l.
2
5
,
n
o
.
2
,
p
p
.
2
8
8
–
2
9
2
f
e
b
ru
a
ry
2
0
1
8
.
[1
4
]
M
.
P
ry
o
r,
“
T
h
e
e
ff
e
c
ts
o
f
sin
g
u
lar
v
a
lu
e
d
e
c
o
m
p
o
siti
o
n
o
n
c
o
ll
a
b
o
ra
ti
v
e
f
il
terin
g
,
”
Co
mp
u
ter
S
c
ien
c
e
De
p
a
rt.,
Da
rtmo
u
t
h
C
o
ll
e
g
e
,
H
a
n
o
v
e
r,
NH,
T
e
c
h
.
Re
p
.
PC
S
-
T
R
9
8
-
3
3
8
,
1
9
9
8
.
[1
5
]
M.K
.D
e
v
i
a
n
d
P
.
V
e
n
k
a
tes
h
,
“
S
m
o
o
th
e
n
in
g
a
p
p
r
o
a
c
h
to
a
ll
e
v
iate
th
e
m
e
a
g
e
r
ra
ti
n
g
p
ro
b
lem
in
c
o
ll
a
b
o
ra
ti
v
e
Re
c
o
m
m
e
n
d
e
r
S
y
ste
m
,”
Fu
tu
re
g
e
n
e
ra
ti
o
n
c
o
m
p
u
ter
sy
ste
ms
,
v
o
l.
29
,
p
p
.
2
6
2
-
2
7
0
,
2
0
1
3
.
[1
6
]
X
ian
g
ro
n
g
Zen
g
a
n
d
M
a
rio
A
.
T.
F
ig
u
e
ired
o
“
Ro
b
u
st
S
p
a
rsity
An
d
C
lu
ste
rin
g
Re
g
u
lariz
a
ti
o
n
F
o
r
Re
g
re
ss
io
n
,
”
2
2
n
d
E
u
ro
p
e
a
n
S
ig
n
a
l
P
ro
c
e
ss
in
g
Co
n
f
e
re
n
c
e
(EUS
IP
CO)
,
p
p
1
7
7
6
–
1
7
8
0
.
[1
7
]
R.
O.
Du
d
a
,
e
t
a
l
.
,
“
P
a
tt
e
rn
Clas
sifica
ti
o
n
,”
Jo
h
n
W
il
e
y
&
S
o
n
s,
2
0
0
1
.
[1
8
]
M
.
K.D
e
vi
a
n
d
P
.
V
e
n
k
a
tes
h
,
“
I
DSS
:
a
n
i
n
telli
g
e
n
t
d
e
c
isi
o
n
su
p
p
o
rt
sy
ste
m
f
o
r
e
-
p
u
rc
h
a
sin
g
u
sin
g
CBR
a
n
d
CF
,
”
In
t.
J
.
Ag
e
n
t
-
Or
ien
ted
S
o
f
tw
a
re
En
g
in
e
e
rin
g
.
[1
9
]
Yo
u
c
h
u
n
J
.
,
e
t
a
l
.
,
“
A
u
to
m
a
ti
o
n
De
p
a
rtme
n
t
X
iam
e
n
Un
iv
e
rsit
y
„M
issin
g
V
a
lu
e
P
re
d
icti
o
n
Us
in
g
Co
-
c
lu
ste
rin
g
a
n
d
RBF
f
o
r
Co
ll
a
b
o
ra
ti
v
e
F
il
terin
g
,
”
IEE
E
2
0
1
5
I
n
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
C
lo
u
d
C
o
mp
u
ti
n
g
a
n
d
Bi
g
Da
ta
,
2
0
1
5
.
[2
0
]
X
.
S
u
a
n
d
T
.
M
.
Kh
o
sh
g
o
f
t
aar
,“
Re
v
ie
w
A
rti
c
le
A
S
u
rv
e
y
o
f
C
o
ll
a
b
o
ra
ti
v
e
F
il
terin
g
Tec
h
n
iq
u
e
s
,”
Ad
v
a
n
c
e
s
in
Arti
fi
c
ia
l
In
telli
g
e
n
c
e
,
2
0
0
9
.
[2
1
]
M
.
D.
Bu
h
m
a
n
n
,
“
A
p
p
ro
x
im
a
ti
o
n
a
n
d
In
terp
o
lati
o
n
w
it
h
Ra
d
ial
F
u
n
c
ti
o
n
s,”
M
u
lt
iva
ri
a
te
Ap
p
r
o
x
ima
ti
o
n
a
n
d
Ap
p
li
c
a
ti
o
n
s
,
i
n
N.
Dy
n
,
e
t
a
l
.
,
Ca
m
b
rid
g
e
U
n
iv
.
P
re
ss
,
2
0
0
1
.
[2
2
]
D.
Bil
lsu
s
a
n
d
M
.
P
a
z
z
a
n
i,
“
Us
e
r
M
o
d
e
li
n
g
f
o
r
A
d
a
p
ti
v
e
Ne
w
s
A
c
c
e
ss
,
”
Us
e
r
M
o
d
e
li
n
g
a
n
d
Us
e
r
-
Ad
a
p
te
d
In
ter
a
c
ti
o
n
,
v
o
l
.
1
0
,
2
3
,
p
p
.
1
4
7
-
1
8
0
,
2
0
0
0
.
[2
3
]
S
.
Na
g
a
lak
sh
m
i,
e
t
a
l
.
,
“
M
a
th
e
m
a
ti
c
a
l
A
p
p
ro
x
ima
ti
o
n
f
o
r
M
o
d
e
l
b
a
se
d
Re
c
o
m
m
e
n
d
e
r
S
y
ste
m
,
”
P
re
s
e
n
ted
a
t
Pro
c
e
e
d
in
g
s
o
f
t
h
e
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
S
u
p
p
ly Ch
a
in
M
a
n
a
g
e
me
n
t
a
n
d
In
fo
rm
a
t
io
n
S
y
ste
ms
,
2
0
0
8
.
[2
4
]
J.L
.
He
rlo
c
k
e
r,
e
t
a
l
.
,
“
Ev
a
lu
a
ti
n
g
Co
ll
a
b
o
ra
ti
v
e
F
il
terin
g
Re
c
o
m
m
e
n
d
e
r
S
y
ste
m
s,”
AC
M
T
ra
n
s.
In
fo
rm
a
ti
o
n
S
y
ste
ms
,
v
o
l.
2
2
,
p
p
.
5
-
5
3
,
2
0
0
4
.
[2
5
]
M.
De
sh
p
a
n
d
e
a
n
d
G
.
Ka
r
y
p
is,
“
Ite
m
-
Ba
se
d
T
o
p
-
N
Re
c
o
m
m
e
n
d
a
ti
o
n
A
lg
o
rit
h
m
s,”
ACM
T
ra
n
s.
In
f
o
rm
a
ti
o
n
S
y
ste
ms
,
v
o
l.
2
2
,
p
p
.
1
4
3
-
1
7
7
,
2
0
0
4
.
[2
6
]
N.
L
it
tl
e
sto
n
e
a
n
d
M
.
W
a
r
m
u
th
,
“
T
h
e
W
e
i
g
h
ted
M
a
jo
rit
y
A
lg
o
ri
th
m
,
”
In
fo
rm
a
ti
o
n
a
n
d
C
o
mp
u
ta
ti
o
n
,
v
o
l.
1
0
8
,
p
p
.
2
1
2
2
6
1
,
1
9
9
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.