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.
6
,
No
.
6
,
Dec
em
b
er
201
6
,
p
p
.
30
3
7
~
30
46
I
SS
N:
2
0
8
8
-
8708
,
DOI
: 1
0
.
1
1
5
9
1
/
i
j
ec
e
.
v
6i
6
.
1
0
5
7
9
3037
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ia
e
s
jo
u
r
n
a
l.c
o
m/o
n
lin
e/in
d
ex
.
p
h
p
/I
JE
C
E
Using
At
t
ribut
e
O
rien
ted
In
ducti
o
n H
ig
h
L
ev
el E
m
ergi
ng
Patter
n
(
A
O
I
-
H
E
P)
to M
ine
Frequ
ent
Patter
ns
H
a
rc
o
L
esli
e
H
endric
Sp
it
s
Wa
rna
rs
Bin
a
Nu
sa
n
tara
Un
iv
e
rsity
,
Ja
k
a
rta,
In
d
o
n
e
sia
Art
icle
I
nfo
AB
ST
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
Ma
r
2
3
,
2
0
1
6
R
ev
i
s
ed
J
u
l 1
8
,
2
0
1
6
A
cc
ep
ted
A
u
g
1
,
2
0
1
6
F
re
q
u
e
n
t
p
a
tt
e
rn
s
in
A
tt
rib
u
te
Orie
n
ted
In
d
u
c
ti
o
n
Hig
h
lev
e
l
Em
e
r
g
in
g
P
a
tt
e
r
n
(A
OI
-
HEP
),
a
re
re
c
o
g
n
ize
d
w
h
e
n
h
a
v
e
m
a
x
i
m
u
m
su
b
su
m
p
ti
o
n
targ
e
t
(su
p
e
rse
t)
in
t
o
c
o
n
t
ra
stin
g
(su
b
se
t
)
d
a
tas
e
ts
(c
o
n
tras
ti
n
g
⊂
targ
e
t)
a
n
d
h
a
v
in
g
larg
e
Hig
h
E
m
e
rg
in
g
P
a
tt
e
rn
(HE
P
)
g
ro
w
th
ra
te
a
n
d
su
p
p
o
rt
i
n
targ
e
t
d
a
tas
e
t.
HEP
F
re
q
u
e
n
t
p
a
tt
e
rn
s
h
a
d
b
e
e
n
su
c
c
e
ss
f
u
l
m
in
e
d
w
it
h
AO
I
-
HEP
u
p
o
n
4
UCI
m
a
c
h
in
e
lea
rn
in
g
d
a
tas
e
t
s
su
c
h
a
s
a
d
u
lt
,
b
re
a
st
c
a
n
c
e
r,
c
e
n
su
s
a
n
d
IP
UM
S
w
it
h
t
h
e
n
u
m
b
e
r
o
f
in
sta
n
c
e
s
o
f
4
8
8
4
2
,
5
6
9
,
2
4
5
8
2
8
5
a
n
d
2
5
6
9
3
2
re
sp
e
c
ti
v
e
l
y
a
n
d
e
a
c
h
d
a
tas
e
t
h
a
s
c
o
n
c
e
p
t
h
iera
rc
h
ies
b
u
il
t
f
ro
m
it
s
f
iv
e
c
h
o
se
n
a
tt
rib
u
tes
.
T
h
e
re
a
r
e
2
a
n
d
1
f
in
d
in
g
f
re
q
u
e
n
t
p
a
tt
e
rn
s
f
ro
m
a
d
u
lt
a
n
d
b
re
a
st
c
a
n
c
e
r
d
a
tas
e
ts
re
sp
e
c
ti
v
e
ly
,
w
h
il
e
th
e
re
is
n
o
f
re
q
u
e
n
t
p
a
tt
e
rn
f
ro
m
c
e
n
su
s
a
n
d
I
P
UMS
d
a
tas
e
ts.
T
h
e
f
in
d
in
g
HE
P
f
re
q
u
e
n
t
p
a
tt
e
rn
s
f
ro
m
a
d
u
lt
d
a
tas
e
t
a
re
a
d
u
lt
w
h
ich
h
a
v
e
g
o
v
e
rn
m
e
n
t
w
o
rk
c
las
s
w
it
h
a
n
i
n
term
e
d
iate
e
d
u
c
a
ti
o
n
(
8
0
.
5
3
%
)
a
n
d
A
m
e
ric
a
a
s
n
a
ti
v
e
c
o
u
n
try
(3
3
%
).
M
e
a
n
w
h
il
e
,
th
e
o
n
ly
1
HEP
f
re
q
u
e
n
t
p
a
tt
e
rn
f
ro
m
b
re
a
st
c
a
n
c
e
r
d
a
tas
e
t
is
b
re
a
st
c
a
n
c
e
r
wh
ich
h
a
v
e
c
lu
m
p
th
ick
n
e
ss
t
y
p
e
o
f
A
b
o
u
t
Av
e
r
Clu
m
p
w
it
h
c
e
ll
siz
e
o
f
V
e
ry
L
a
rg
e
S
ize
(3
.
5
6
%
).
F
in
d
in
g
HEP
f
re
q
u
e
n
t
p
a
tt
e
rn
s
w
it
h
A
OI
-
HEP
a
re
in
f
lu
e
n
c
e
d
b
y
lea
rn
in
g
o
n
h
ig
h
l
e
v
e
l
c
o
n
c
e
p
t
in
o
n
e
o
f
c
h
o
se
n
a
t
tri
b
u
te
a
n
d
e
x
ten
d
e
d
e
x
p
e
rime
n
t
u
p
o
n
a
d
u
lt
d
a
tas
e
t
w
h
e
re
lea
rn
o
n
m
a
rit
a
l
-
sta
tu
s
a
tt
rib
u
te s
h
o
w
e
d
th
a
t
t
h
e
re
is
n
o
f
in
d
i
n
g
f
re
q
u
e
n
t
p
a
tt
e
r
n
.
K
ey
w
o
r
d
:
A
OI
-
HE
P
Data
m
i
n
i
n
g
Fre
q
u
en
t p
atter
n
HE
P
f
r
eq
u
en
t p
atter
n
Hig
h
lev
el
e
m
er
g
in
g
p
atter
n
Co
p
y
rig
h
t
©
2
0
1
6
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
.
All
rig
h
ts
re
se
rv
e
d
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Har
co
L
esli
e
He
n
d
r
ic
Sp
its
W
ar
n
ar
s
,
B
in
a
Nu
s
an
tar
a
U
n
i
v
er
s
it
y
,
Ka
m
p
u
s
an
g
g
r
ek
,
J
l.
Keb
o
n
J
e
r
u
k
r
a
y
a
n
o
.
2
7
,
J
ak
ar
ta
B
a
r
at
-
1
1
5
3
0
,
I
n
d
o
n
esia.
E
m
ail: s
h
e
n
d
r
ic@
b
i
n
u
s
.
ed
u
1.
I
NT
RO
D
UCT
I
O
N
Fre
q
u
en
t
p
atter
n
is
a
co
m
b
i
n
a
tio
n
o
f
f
ea
t
u
r
e
p
atter
n
s
t
h
at
a
p
p
ea
r
in
d
ataset
w
ith
f
r
eq
u
e
n
c
y
n
o
t
le
s
s
th
an
a
u
s
er
-
s
p
ec
if
ied
th
r
es
h
o
ld
[
1
-
3]
an
d
th
e
f
r
eq
u
en
t
p
atter
n
s
y
n
o
n
y
m
w
it
h
lar
g
e
p
atter
n
w
as
f
ir
s
t
p
r
o
p
o
s
ed
f
o
r
m
ar
k
et
b
as
k
et
an
al
y
s
is
i
n
t
h
e
f
o
r
m
o
f
a
s
s
o
ciatio
n
r
u
les
[
4
].
W
it
h
f
r
eq
u
en
t
p
att
er
n
w
e
ca
n
h
a
v
e
s
tr
o
n
g
/s
h
ar
p
d
is
cr
i
m
in
at
io
n
p
o
w
er
w
h
er
e
h
av
e
lar
g
e
g
r
o
w
t
h
r
ate
a
n
d
s
u
p
p
o
r
t
in
tar
g
e
t
(
D2
)
d
ataset
a
n
d
o
th
er
s
u
p
p
o
r
t
in
co
n
tr
asti
n
g
(
D1
)
d
ataset
is
s
m
all
[
5
-
7
]
.
Fre
q
u
en
t
p
atter
n
s
h
av
e
b
ee
n
i
m
p
le
m
en
ted
in
ap
p
licatio
n
s
s
u
c
h
as:
c
u
s
to
m
er
tr
an
s
ac
tio
n
an
al
y
s
is
,
w
eb
m
i
n
in
g
,
s
o
f
t
w
ar
e
b
u
g
an
a
l
y
s
is
,
c
h
e
m
ical
an
d
b
io
lo
g
ical
an
al
y
s
i
s
an
d
etc
[
8
-
1
0
]
.
Fre
q
u
en
t
p
atter
n
in
A
ttrib
u
te
Or
ien
ted
I
n
d
u
c
tio
n
Hi
g
h
le
v
e
l
E
m
er
g
i
n
g
P
atter
n
(
A
OI
-
HE
P
)
,
i
s
r
ec
o
g
n
ized
w
h
e
n
h
a
v
e
m
a
x
i
m
u
m
s
u
b
s
u
m
p
tio
n
tar
g
e
t
(
s
u
p
er
s
et)
in
to
co
n
tr
a
s
tin
g
(
s
u
b
s
et)
d
atasets
(
co
n
tr
asti
n
g
⊂
tar
g
et)
an
d
h
av
in
g
lar
g
e
Hi
g
h
E
m
er
g
i
n
g
P
atter
n
(
HE
P
)
g
r
o
w
th
r
ate
an
d
s
u
p
p
o
r
t
in
tar
g
et
d
ataset
[1
1
]
.
I
n
th
e
f
ir
s
t
A
OI
-
HE
P
v
er
s
io
n
[
1
2
]
h
ad
b
ee
n
s
u
cc
ess
to
m
i
n
e:
a.
T
o
tal
Su
b
s
u
m
p
tio
n
HE
P
(
T
SHEP
)
w
h
ic
h
f
r
eq
u
e
n
t i
n
o
n
e
r
u
l
e
b
u
t le
s
s
f
r
eq
u
en
t in
a
n
o
th
er
r
u
le.
b.
Su
b
s
u
m
p
t
io
n
Ov
er
lap
p
in
g
HE
P
(
SOHE
P
)
w
h
ich
ar
e
co
m
b
i
n
atio
n
b
et
w
ee
n
s
u
b
s
u
m
p
tio
n
an
d
o
v
er
lap
p
in
g
b
et
w
ee
n
r
u
lese
ts
.
T
h
is
p
ap
er
is
co
n
tin
o
u
s
f
r
o
m
p
r
ev
io
u
s
p
ap
er
[1
1
]
w
h
er
e
m
i
n
in
g
f
r
eq
u
en
t
p
atter
n
s
w
i
th
A
OI
-
HE
P
d
o
es
n
o
t
o
n
l
y
o
n
ad
u
lt
d
atas
et
b
u
t
w
i
ll
b
e
e
x
te
n
d
ed
to
o
th
er
3
d
ataset
s
s
u
c
h
as
b
r
ea
s
t
ca
n
ce
r
,
ce
n
s
u
s
a
n
d
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
30
3
7
–
30
46
3038
I
P
UM
S
d
atasets
f
r
o
m
UC
I
Ma
ch
in
e
lear
n
in
g
[
1
3
]
.
T
h
e
ex
p
er
i
m
e
n
ts
u
p
o
n
t
h
ese
4
d
atasets
s
h
o
w
t
h
at
ad
u
lt
a
n
d
b
r
ea
s
t
ca
n
ce
r
d
ataset
s
h
av
e
f
r
eq
u
en
t
p
atter
n
s
w
h
i
le
o
n
o
t
h
e
r
h
an
d
,
ce
n
s
u
s
an
d
I
P
UM
S
d
atasets
d
o
n
o
t
h
a
v
e
f
r
eq
u
en
t
p
atter
n
s
.
I
n
p
r
ev
io
u
s
p
ap
er
[1
1
]
th
er
e
is
n
o
d
is
ti
n
ctio
n
b
et
w
ee
n
f
r
eq
u
e
n
t
p
atter
n
an
d
s
tr
o
n
g
d
is
cr
i
m
i
n
atio
n
r
u
le,
w
h
ile
i
n
th
is
p
ap
er
th
er
e
i
s
d
is
tin
ct
i
o
n
b
et
w
ee
n
f
in
d
i
n
g
f
r
eq
u
e
n
t
p
atter
n
an
d
s
tr
o
n
g
d
is
cr
i
m
i
n
atio
n
r
u
le
s
.
A
OI
-
HE
P
as
d
ata
m
i
n
in
g
tec
h
n
iq
u
e
h
a
s
o
p
p
o
r
tu
n
it
y
to
b
e
m
o
r
e
ex
p
l
o
r
ed
s
u
ch
as
m
in
i
n
g
s
i
m
ilar
p
atter
n
[1
4
]
,
in
v
er
s
e
d
is
co
v
er
y
lear
n
i
n
g
,
lear
n
in
g
m
o
r
e
th
an
2
d
atasets
,
m
u
lt
id
i
m
en
s
io
n
a
l
v
ie
w,
lear
n
in
g
o
th
er
k
n
o
w
led
g
e
r
u
le
s
an
d
s
o
o
n
[
1
5
]
.
2.
AO
I
-
H
E
P
F
RE
Q
U
E
NT
P
A
T
T
E
RN
A
L
G
O
RI
T
H
M
A
OI
-
HE
P
f
r
eq
u
e
n
t
p
atter
n
a
lg
o
r
ith
m
w
i
ll
co
n
s
is
t
2
al
g
o
r
ith
m
s
s
u
c
h
as
AOI
ch
ar
ac
t
er
is
tic
r
u
l
e
alg
o
r
ith
m
[
16
]
an
d
HE
P
f
r
eq
u
en
t p
atter
n
al
g
o
r
ith
m
a
s
s
ee
n
i
n
F
i
g
u
r
es 1
an
d
2
r
esp
ec
tiv
el
y
.
A
OI
c
h
ar
ac
ter
is
tic
r
u
le
alg
o
r
it
h
m
w
ill
b
e
r
u
n
t
w
i
ce
w
i
th
i
n
p
u
t
t
w
o
d
atasets
a
s
h
o
r
iz
o
n
tal
p
ar
titi
o
n
s
o
f
t
h
e
d
a
taset
a
n
d
a
s
u
s
u
al
l
A
OI
c
h
ar
ac
ter
is
tic
r
u
le
al
g
o
r
it
h
m
,
u
s
e
s
co
n
ce
p
t
h
ier
ar
ch
y
a
s
b
ac
k
g
r
o
u
n
d
k
n
o
w
led
g
e
f
o
r
d
ata
g
en
er
a
lizat
io
n
.
A
OI
c
h
ar
ac
ter
is
tic
r
u
le
al
g
o
r
it
h
m
w
ill
el
i
m
in
ate
d
i
s
ti
n
ct
attr
i
b
u
tes
a
n
d
tu
p
les
u
n
til
th
e
y
ar
e
less
o
r
eq
u
al
t
h
a
n
attr
ib
u
te
a
n
d
r
u
l
es
th
r
es
h
o
ld
s
r
esp
ec
tiv
el
y
[
17
]
an
d
h
a
v
e
o
u
tp
u
t
t
w
o
r
u
le
s
ets
f
o
r
ea
ch
t
w
o
in
p
u
t
d
atase
ts
.
T
h
ese
t
w
o
r
u
le
s
ets
w
il
l
b
e
in
p
u
t
f
o
r
HE
P
f
r
eq
u
en
t
p
atter
n
alg
o
r
it
h
m
i
n
F
i
g
u
r
e
2
w
h
i
ch
ap
p
l
y
C
ar
tesi
a
n
p
r
o
d
u
ct
b
et
w
ee
n
t
h
ese
t
w
o
r
u
leset
s
a
n
d
th
e
n
o
n
f
r
eq
u
e
n
t
p
atter
n
i
n
C
ar
tesi
a
n
p
r
o
d
u
ct
r
es
u
lt
w
ill
b
e
eli
m
i
n
ated
.
Input: dataset, concept hierarchies, attribute threshold,rule threshold
Output: characteristic rule of learning task,
{
} , {
},
num_attr, |D2|,|D1|
1
For each of attribute Ai (1
i
n, where n= # of attribute
s)
in the generalized relation GR
2
{ While #_of_distinct_values_in_attribute_Ai > threshold
3
{If no higher level concept in concept hierarchy for attr Ai
4
{ remove attribute Ai }
5
Else {
substitute the value of Ai by its corresponding minimal general
ized concept}
6
Merge identical tuples
7
}
8
}
9
While #_of_tuples in GR > threshold
10
{ Selective generalize attributes
11
Merge identical tuples
12
}
Fig
u
r
e
1
.
A
OI
C
h
ar
ac
ter
is
tic
R
u
le
Alg
o
r
it
h
m
Input
:{
} , {
}, num_attr,|D2|,|D1|, GR_threshold
Output :
,|
|,(|
|/|D2|),
,|
,(|
/|D1|),HEP_GR
1
{ While(noAllANY(
))
2
{While(noAllANY(
))
3
{ SLV=0, F=0
4
for x=1 to num_attr
5
{ If(
==
[x] and
==
“ANY” ) SLV=SLV+2.1
6
If(
==
[x] and
!= “ANY” ) SLV=SLV+2
7
If(
!=
[x] and
⊂
[x] ) SLV=SLV+0.4
8
If(
!=
[x] and
⊂
[x])SLV=SLV+0.5,F++ }
9
If (SLV>=(num_attr
-
1)*0.5+0.4 and SLV<=(num_attr
-
1) *0.5+2.1 and F>=num_attr
-
1
10
HEP_GR=(|
|/|D2|)/(|
/|D1|)
11
If HEP_GR > GR_threshold
12
Print
,|
|,(|
|/|D2|),
,|
,(|
/|D1|),HEP_GR,SLV
13
}
14
}
15
}
Fig
u
r
e
2
.
HE
P
Fre
q
u
en
t P
atter
n
A
l
g
o
r
ith
m
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Usi
n
g
A
ttr
ib
u
te
Ori
en
ted
I
n
d
u
ctio
n
Hig
h
Leve
l E
merg
in
g
P
a
tter
n
.
.
.
.
(
Ha
r
co
Les
lie
Hen
d
r
i
c
S
.
W
.
)
3039
I
n
F
i
g
u
r
e
2
,
G
R
_
th
r
es
h
o
ld
h
a
s
d
ef
au
lt
b
et
w
ee
n
0
a
n
d
100
,
at
tr
ib
u
te
n
u
m
_
a
ttr
is
t
h
e
n
u
m
b
er
attr
ib
u
tes
in
r
u
le
s
ets
an
d
as
m
i
n
E
q
u
a
tio
n
1,
|
D2
|
a
n
d
|
D1
|
ar
e
to
tal
n
u
m
b
er
o
f
i
n
s
ta
n
ce
s
i
n
d
atase
t
D2
an
d
D1
r
esp
e
ctiv
el
y
as
s
h
o
w
n
i
n
E
q
u
a
tio
n
2
an
d
F
is
a
co
u
n
ter
f
o
r
AOI
-
HE
P
f
r
eq
u
e
n
t p
atter
n
s
w
h
i
ch
is
in
d
e
n
ti
f
ied
b
y
SL
V=
0
.
5
as
s
h
o
w
n
in
li
n
e
n
u
m
b
er
8
F
i
g
u
r
e
2
.
T
h
e
o
u
tp
u
t
s
f
r
o
m
HE
P
alg
o
r
it
h
m
ar
e
,
|
|,
(|
|
/
|
D2
|
)
as
s
u
p
p
o
r
t
tar
g
et
d
ataset,
,
|
,
(|
/
|
D1
|
)
as
s
u
p
p
o
r
t
co
n
tr
ast
in
g
d
ataset,
Gr
o
w
t
h
R
ate
(
HE
P
_
GR
)
an
d
SL
V
v
alu
e.
Mo
r
eo
v
er
,
lin
e
n
u
m
b
er
1
an
d
2
ar
e
u
s
ed
to
ex
clu
d
e
r
u
le
w
ith
A
NY
v
al
u
es
i
n
all
attr
ib
u
tes
in
r
u
le
s
ets
an
d
r
esp
ec
tiv
el
y
,
s
i
n
ce
r
u
l
es
w
it
h
A
NY
v
al
u
es
ar
e
le
s
s
m
ea
n
i
n
g
f
u
l
a
n
d
d
o
n
o
t
o
f
f
er
m
ea
n
i
n
g
f
u
l
i
n
ter
p
r
etatio
n
.
F
u
r
t
h
er
m
o
r
e
,
s
tate
m
e
n
t
i
n
l
in
e
n
u
m
b
er
9
i
s
u
s
ed
to
e
li
m
in
ate
n
o
n
f
r
eq
u
en
t
p
atter
n
,
w
h
er
e
E
q
u
atio
n
s
S
L
V>
=(
n
u
m
_
attr
-
1
)
*
0
.
5
+0
.
4
an
d
SL
V<
=(
n
u
m
_
a
ttr
-
1
)
*
0
.
5
+2
.
1
ar
e
r
ec
o
g
n
ized
as
m
i
n
i
m
u
m
a
n
d
m
ax
i
m
u
m
S
L
V
v
al
u
e
f
o
r
f
r
eq
u
en
t p
atter
n
.
SL
V
=
∑
(
1
)
w
h
er
e:
SL
V
=
Si
m
ilar
i
t
y
v
al
u
e
b
ased
o
n
th
e
s
i
m
i
lar
it
y
o
f
attr
ib
u
tes
h
ier
ar
ch
y
le
v
el
a
n
d
v
al
u
es
M
=
N
u
m
b
er
o
f
attr
ib
u
tes i
n
a
r
u
l
eset,
w
h
er
e
m
>
1
(
n
u
m
b
er
o
f
attr
ib
u
tes i
n
co
n
ce
p
t h
ier
ar
c
h
ie
s
-
1)
I
=
A
ttrib
u
te
p
o
s
itio
n
L
Vi
=
C
ateg
o
r
izatio
n
o
f
attr
ib
u
te
s
co
m
p
ar
is
o
n
b
ased
o
n
s
i
m
ilar
it
y
h
ier
ar
ch
y
lev
el
a
n
d
v
al
u
es
,
th
e
o
p
tio
n
s
ar
e
:
a.
I
f
h
ier
ar
c
h
y
le
v
el
is
d
i
f
f
er
e
n
t
an
d
th
e
at
tr
ib
u
te
i
n
r
u
le
o
f
r
u
l
eset
R
2
i
s
s
u
b
s
u
m
ed
b
y
th
e
a
ttrib
u
te
i
n
r
u
le
o
f
r
u
le
s
et
R
1
(
R
2
⊂
R
1
)
,
L
V=
0
.
4
.
b.
I
f
h
ier
ar
c
h
y
le
v
el
is
d
i
f
f
er
e
n
t
an
d
th
e
at
tr
ib
u
te
i
n
r
u
le
o
f
r
u
l
eset
R
1
i
s
s
u
b
s
u
m
ed
b
y
th
e
a
ttrib
u
te
i
n
r
u
le
o
f
r
u
le
s
et
R
2
(
R
1
⊂
R
2
)
,
L
V=
0
.
5
.
c.
I
f
h
ier
ar
ch
y
le
v
el
an
d
v
a
lu
e
s
a
r
e
th
e
s
a
m
e
a
n
d
th
e
attr
ib
u
tes
v
alu
e
s
ar
e
n
o
t
A
NY,
L
V=
2
.
d.
I
f
h
ier
ar
ch
y
le
v
el
an
d
v
a
lu
e
s
a
r
e
th
e
s
a
m
e
a
n
d
th
e
attr
ib
u
tes
v
alu
e
s
ar
e
ANY,
L
V=
2
.
1
.
T
h
e
f
o
u
r
ca
te
g
o
r
izatio
n
o
f
a
ttrib
u
te
co
m
p
ar
is
o
n
s
o
r
L
V
in
E
q
u
a
tio
n
1
i
s
b
ased
o
n
t
w
o
m
ai
n
ca
teg
o
r
izatio
n
s
i.e
.
s
u
b
s
u
m
p
t
io
n
(
L
V=
0
.
4
o
r
L
V=
0
.
5
)
an
d
o
v
er
lap
p
in
g
(
L
V=
2
o
r
L
V=
2
.
1
)
.
T
h
u
s
,
th
e
attr
ib
u
tes
w
i
ll
b
e
ca
teg
o
r
ized
as
s
u
b
s
u
m
p
tio
n
w
h
e
n
attr
ib
u
t
es
co
m
p
ar
is
o
n
h
as
d
i
f
f
er
en
t
h
ier
ar
ch
y
le
v
el
a
n
d
v
alu
e
(
L
V=
0
.
4
o
r
L
V=
0
.
5
)
.
On
th
e
o
t
h
er
h
a
n
d
,
th
e
at
tr
ib
u
tes
w
il
l
b
e
ca
teg
o
r
ized
o
v
er
lap
p
in
g
w
h
e
n
co
m
p
ar
is
o
n
b
et
w
ee
n
a
ttrib
u
te
s
h
as
th
e
s
a
m
e
h
ier
ar
ch
y
lev
els
a
n
d
v
a
lu
e
s
(
L
V=
2
o
r
L
V
=2
.
1
)
.
Fo
r
ea
ch
L
V
o
p
tio
n
v
alu
e
s
0
.
4
,
0
.
5
,
2
an
d
2
.
1
ar
e
u
s
er
d
ef
in
ed
n
u
m
b
er
,
w
h
er
e
o
p
tio
n
n
u
m
b
er
s
0
.
4
an
d
0
.
5
as
v
alu
es
f
o
r
s
u
b
s
u
m
p
tio
n
ca
te
g
o
r
izatio
n
(
m
in
i
m
u
m
ca
te
g
o
r
izatio
n
)
an
d
o
p
tio
n
n
u
m
b
er
s
2
an
d
2
.
1
as
v
alu
es
f
o
r
o
v
er
lap
p
in
g
ca
te
g
o
r
i
za
tio
n
(
m
ax
i
m
u
m
ca
te
g
o
r
izatio
n
)
.
L
V=
0
.
4
is
m
i
n
i
m
u
m
v
al
u
e
f
o
r
s
u
b
s
u
m
p
tio
n
ca
teg
o
r
izatio
n
an
d
i
f
r
u
le
s
et
R
2
is
s
u
b
s
u
m
ed
b
y
r
u
leset
R
1
(
R
2
⊂
R
1
)
.
3.
M
I
NING
F
RE
Q
U
E
N
T
P
AT
T
E
R
N
Fre
q
u
en
t
p
atter
n
is
a
co
m
b
i
n
a
tio
n
o
f
f
ea
t
u
r
e
p
atter
n
s
t
h
at
a
p
p
ea
r
in
d
ataset
w
ith
f
r
eq
u
e
n
c
y
n
o
t
le
s
s
th
an
a
u
s
er
-
s
p
ec
i
f
ied
t
h
r
es
h
o
ld
[
1
]
an
d
th
e
f
r
eq
u
e
n
t
p
atter
n
s
y
n
o
n
y
m
w
it
h
lar
g
e
p
atter
n
was
f
ir
s
t
p
r
o
p
o
s
ed
f
o
r
m
ar
k
et
b
ask
e
t
an
a
l
y
s
is
in
th
e
f
o
r
m
o
f
as
s
o
ciatio
n
r
u
les
[
4
]
.
Min
i
n
g
f
r
eq
u
e
n
t
p
atter
n
s
h
a
s
b
ee
n
d
o
n
e
i
n
d
ata
s
tr
ea
m
w
it
h
DS
C
L
al
g
o
r
ith
m
[1
8
]
an
d
T
o
p
-
K
C
lo
s
ed
[
19
]
.
W
ith
f
r
eq
u
en
t
p
atter
n
w
e
ca
n
h
a
v
e
s
tr
o
n
g
/
s
h
ar
p
d
is
cr
i
m
i
n
atio
n
p
o
w
er
w
h
er
e
h
av
e
lar
g
e
g
r
o
w
th
r
ate
a
n
d
s
u
p
p
o
r
t
in
tar
g
et
(
D2
)
d
ataset
an
d
o
th
er
s
u
p
p
o
r
t
in
co
n
tr
asti
n
g
(
D1
)
d
ataset
i
s
s
m
all
[
5
-
7
]
.
I
n
A
OI
-
HE
P
,
th
e
f
r
eq
u
e
n
t
p
atter
n
i
s
s
h
o
w
n
b
y
t
h
e
s
u
b
s
u
m
p
tio
n
L
V=
0
.
4
o
r
L
V=
0
.
5
an
d
as
m
en
tio
n
p
r
ev
io
u
s
l
y
w
h
e
n
L
V=
0
.
4
th
en
r
u
leset
R
2
is
s
u
b
s
u
m
ed
b
y
r
u
leset
R
1
(
R
2
⊂
R
1
)
w
h
er
e
R
2
is
s
u
b
s
et
r
u
le
an
d
R
1
is
s
u
p
er
s
et
r
u
le.
On
t
h
e
o
th
er
h
a
n
d
w
h
e
n
L
V=
0
.
5
th
en
r
u
le
s
et
R
1
is
s
u
b
s
u
m
ed
b
y
r
u
leset
R
2
(
R
1
⊂
R
2
)
w
h
er
e
R
1
is
s
u
b
s
et
r
u
le
an
d
R
2
is
s
u
p
er
s
et
r
u
le.
R
2
is
i
n
tar
g
et
(
D2
)
d
ataset
an
d
R
1
i
s
i
n
co
n
tr
asti
n
g
(
D1
)
d
ataset
(
D2
/D1
=ta
r
g
et/
co
n
tr
asti
n
g
=
R
2
/R
1
)
a
n
d
it
is
a
s
ac
co
r
d
an
ce
w
it
h
HE
P
g
r
o
w
t
h
r
ate
i
n
E
q
u
at
io
n
2
.
Su
p
er
s
et
r
u
le
is
a
f
r
eq
u
en
t p
atter
n
s
in
ce
s
u
b
s
et
r
u
le
is
p
ar
t
o
f
t
h
e
s
u
p
er
s
et
r
u
le
an
d
f
o
r
in
s
tan
ce
w
h
e
n
S
L
V
h
as
th
e
s
a
m
e
L
V
v
al
u
es
(
S
L
V=
0
.
5
+0
.
5
+0
.
5
+
0
.
5
=
2
)
th
en
ce
r
tain
l
y
th
e
n
u
m
b
er
o
f
in
s
ta
n
ce
s
in
s
u
p
er
s
et
r
u
le
is
lar
g
er
t
h
a
n
in
its
s
u
b
s
et
r
u
le.
T
h
u
s
,
th
at
i
n
s
ta
n
ce
co
n
d
itio
n
SL
V=
0
.
5
+0
.
5
+0
.
5
+0
.
5
=2
s
h
o
w
s
t
h
at
s
u
p
er
s
et
r
u
le
(
f
r
eq
u
e
n
t
p
atter
n
)
h
as
h
ig
h
s
u
p
p
o
r
t
(
lar
g
e
p
atter
n
)
an
d
s
u
b
s
et
r
u
le
(
in
f
r
eq
u
e
n
t
p
atter
n
)
h
as
lo
w
s
u
p
p
o
r
t.
in
E
m
er
g
in
g
P
atter
n
(
E
P
)
,
p
atter
n
s
w
ill
b
e
r
ec
o
g
n
ized
as
E
P
if
h
a
v
e
h
ig
h
s
u
p
p
o
r
t (
f
r
eq
u
en
t
p
atter
n
)
in
o
n
e
clas
s
an
d
lo
w
s
u
p
p
o
r
t (
in
f
r
eq
u
en
t p
atter
n
)
in
o
th
er
o
n
e
[
3]
,
[6
].
Fro
m
f
r
eq
u
en
t
p
atter
n
s
,
w
e
ca
n
cr
ea
te
a
d
is
cr
i
m
i
n
atio
n
r
u
le
an
d
ar
e
in
ter
e
s
ted
in
m
in
i
n
g
t
h
e
f
r
eq
u
e
n
t
p
atter
n
w
ith
s
tr
o
n
g
/s
h
ar
p
d
is
cr
i
m
i
n
atio
n
p
o
w
er
.
I
n
E
P
,
th
e
s
tr
en
g
th
o
f
d
is
cr
i
m
in
at
io
n
p
o
w
er
is
e
x
p
r
ess
ed
b
y
its
lar
g
e
g
r
o
w
t
h
r
ate
an
d
s
u
p
p
o
r
t
in
tar
g
et
(
D2
)
d
ataset
[
5
-
7
]
.
T
h
is
is
ca
lled
an
ess
e
n
tial
E
m
er
g
i
n
g
P
atter
n
s
(
eE
P
)
[
6
]
.
I
n
A
OI
-
HE
P
,
th
e
s
tr
en
g
t
h
o
f
d
is
cr
i
m
in
a
tio
n
p
o
w
er
is
ex
p
r
es
s
ed
b
y
its
lar
g
e
g
r
o
wth
r
ate
an
d
s
u
p
p
o
r
t
in
tar
g
et
(
D2
)
d
ataset
as
w
ell.
C
er
tain
l
y
,
to
m
a
k
e
lar
g
e
g
r
o
wth
r
ate
ca
n
b
e
h
ap
p
en
ed
w
h
e
n
h
av
e
lar
g
e
s
u
p
p
o
r
t
in
tar
g
et
(
D2
)
d
ataset
a
n
d
lo
w
s
u
p
p
o
r
t
in
co
n
tr
asti
n
g
(
D1
)
d
ataset.
I
n
d
ee
d
,
in
E
P
,
p
atter
n
s
w
i
ll
b
e
r
ec
o
g
n
ized
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
30
3
7
–
30
46
3040
as E
P
if
h
a
v
e
h
ig
h
s
u
p
p
o
r
t in
o
n
e
class
a
n
d
lo
w
s
u
p
p
o
r
t in
o
th
er
o
n
e
[
3]
,
[6
]
.
Mo
r
eo
v
er
,
s
u
p
p
o
r
t in
co
n
tr
asti
n
g
(
D1
)
d
ataset
m
u
s
t b
e
les
s
th
a
n
s
u
p
p
o
r
t in
tar
g
et
(
D2
)
d
ataset
w
h
er
e
b
y
th
e
e
n
d
w
ill cr
ea
te
la
r
g
e
g
r
o
w
th
r
ate.
I
n
A
OI
-
HE
P
,
th
e
s
tr
e
n
g
th
o
f
d
is
cr
i
m
i
n
an
t
p
o
w
er
i
s
e
x
p
r
ess
ed
b
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
w
h
er
e
R
2
i
n
tar
g
et
(
D2
)
d
ataset
is
s
u
p
er
s
e
t
an
d
R
1
i
n
co
n
tr
asti
n
g
(
D1
)
d
ataset
is
s
u
b
s
et.
T
h
e
s
tr
e
n
g
t
h
o
f
d
is
cr
i
m
i
n
atio
n
p
o
w
er
w
it
h
s
u
b
s
u
m
p
tio
n
L
V
=0
.
5
s
h
o
w
s
t
h
at
h
a
v
e
lar
g
e
s
u
p
p
o
r
t
in
tar
g
et
(
D2
)
d
ataset
an
d
lo
w
s
u
p
p
o
r
t
in
co
n
tr
asti
n
g
(
D1
)
d
ataset,
w
h
e
r
e
b
y
t
h
e
e
n
d
w
i
ll
cr
ea
te
lar
g
e
g
r
o
w
t
h
r
ate.
T
h
u
s
,
f
o
r
d
is
cr
i
m
i
n
an
t
r
u
le
f
r
o
m
f
r
eq
u
en
t
p
atter
n
w
h
ic
h
S
L
V
v
al
u
e
w
it
h
all
s
i
m
ilar
i
t
y
s
u
b
s
u
m
p
t
io
n
L
V=
0
.
5
(
S
L
V
v
a
lu
e
w
i
th
s
i
m
i
lar
it
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
,
f
o
r
i
n
s
ta
n
ce
S
L
V=
0
.
5
+0
.
5
+0
.
5
+0
.
5
=2
)
w
i
ll
h
a
v
e
f
r
eq
u
e
n
t
p
atter
n
w
it
h
s
tr
o
n
g
d
is
cr
i
m
i
n
atio
n
p
o
w
er
.
Me
a
n
wh
ile,
t
h
er
e
i
s
S
L
V
v
al
u
e
w
it
h
n
ea
r
l
y
all
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
an
d
r
ec
o
g
n
ized
as
SL
V
v
al
u
e
w
ith
f
r
eq
u
en
t
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
.
Ho
w
e
v
er
,
SL
V
v
a
lu
e
w
it
h
f
r
eq
u
e
n
t
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
w
il
l
b
e
in
ter
ested
to
b
e
ex
p
lo
r
e
d
.
T
h
is
is
b
ec
au
s
e
w
h
e
n
t
w
o
p
ar
ts
o
f
o
b
j
ec
ts
ar
e
s
im
ilar
i
f
th
e
y
ar
e
s
i
m
i
lar
in
all
f
ea
t
u
r
es
(
f
u
ll
m
atc
h
i
n
g
s
i
m
ila
r
it
y
)
o
r
if
t
h
e
p
er
ce
n
ta
g
e
o
f
s
i
m
ilar
f
ea
t
u
r
es
is
g
r
ea
ter
t
h
an
th
e
8
0
%
[
20
]
o
r
if
th
e
y
ar
e
s
i
m
i
lar
i
n
at
leas
t 9
0
% o
f
th
e
f
ea
t
u
r
es [
2
1
]
.
Sin
ce
th
er
e
ar
e
S
L
V
v
al
u
e
w
it
h
a
ll
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
w
h
er
e
h
a
v
e
f
u
ll
s
i
m
ilar
it
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
,
th
e
n
t
h
er
e
ar
e
f
r
eq
u
en
t
p
atter
n
w
it
h
s
tr
o
n
g
d
is
c
r
i
m
i
n
atio
n
p
o
w
er
f
o
r
S
L
V
v
alu
e
w
i
th
f
r
eq
u
en
t
s
i
m
ilar
it
y
s
u
b
s
u
m
p
t
io
n
L
V=
0
.
5
at
p
er
ce
n
tag
e
v
a
lu
e
o
f
(
m
-
1
)
/m
*
1
0
0
w
h
er
e
m
as
in
E
q
u
atio
n
1
.
Si
n
ce
th
e
s
tr
en
g
th
o
f
d
is
cr
i
m
i
n
a
n
t
p
o
w
e
r
is
ex
p
r
ess
ed
b
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
an
d
f
r
eq
u
e
n
t
p
atter
n
h
as
m
i
n
i
m
u
m
an
d
m
ax
i
m
u
m
S
L
V
v
al
u
es
o
f
(
m
-
1
)
*
c+
c1
w
h
er
e
c=
0
.
5
,
c1
=
0
.
4
an
d
c=
0
.
5
,
c1
=2
.
1
th
en
(
m
-
1
)
*
0
.
5
+0
.
4
an
d
(m
-
1
)
*
0
.
5
+2
.
1
r
e
s
p
ec
tiv
el
y
.
Min
i
m
u
m
a
n
d
m
ax
i
m
u
m
SLV
v
al
u
e
f
o
r
f
r
eq
u
en
t
p
atter
n
ar
e
SL
V=
(
m
-
1
)
*
0
.
5
+0
.
4
an
d
SLV=
(
m
-
1
)
*
0
.
5
+2
.
1
s
h
o
w
th
e
f
r
eq
u
en
t
s
i
m
ilar
it
y
s
u
b
s
u
m
p
tio
n
(
L
V=
0
.
5
)
in
m
-
1
ti
m
es
at
p
er
ce
n
ta
g
e
v
alu
e
o
f
(
m
-
1
)
/
m
*
1
0
0
(
(
m
-
1
)
*
0
.
5
)
p
lu
s
0
.
4
as
m
i
n
i
m
u
m
s
u
b
s
u
m
p
tio
n
a
n
d
2
.
1
as
m
ax
i
m
u
m
o
v
er
lap
p
in
g
L
V
v
a
lu
e
ca
teg
o
r
izatio
n
r
esp
ec
tiv
el
y
.
T
h
u
s
,
m
in
i
m
u
m
an
d
m
ax
i
m
u
m
S
L
V
v
al
u
e
f
o
r
f
r
eq
u
en
t
p
atter
n
s
h
o
w
f
r
eq
u
e
n
t
s
i
m
ilar
it
y
s
u
b
s
u
m
p
tio
n
(
L
V
=0
.
5
)
at
p
er
ce
n
tag
e
v
al
u
e
o
f
(
m
-
1
)
/
m
*
1
0
0
w
h
ic
h
ex
p
r
ess
d
is
cr
i
m
i
n
atio
n
p
o
w
e
r
p
lu
s
m
i
n
i
m
u
m
s
u
b
s
u
m
p
tio
n
L
V=
0
.
4
an
d
m
ax
i
m
u
m
o
v
er
lap
p
in
g
L
V=
2
.
1
r
esp
ec
tiv
el
y
.
Fi
n
all
y
,
w
it
h
AOI
-
HE
P
w
e
ca
n
m
i
n
e
f
r
eq
u
e
n
t
p
atter
n
w
i
th
s
tr
o
n
g
d
is
cr
i
m
i
n
atio
n
p
o
w
er
i
n
o
p
tio
n
al
co
n
d
itio
n
s
:
a.
SL
V
v
al
u
e
w
it
h
f
u
ll
s
i
m
ilar
it
y
s
u
b
s
u
m
p
t
io
n
L
V=
0
.
5
.
b.
SL
V
v
al
u
e
w
ith
f
r
eq
u
en
t
s
i
m
il
ar
it
y
s
u
b
s
u
m
p
t
io
n
L
V=
0
.
5
at
p
er
ce
n
tag
e
v
al
u
e
o
f
(
m
-
1
)
/
m
*
1
0
0
w
h
er
e
m
as
in
E
q
u
atio
n
1.
Min
i
n
g
f
r
eq
u
e
n
t p
atter
n
w
it
h
t
h
at
t
w
o
o
p
tio
n
al
s
ab
o
v
e
b
et
w
e
en
f
u
l
l si
m
ilar
it
y
a
n
d
f
r
eq
u
e
n
t
s
i
m
ilar
it
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
as
m
e
n
ti
o
n
ed
ab
o
v
e
c
an
b
e
s
ee
n
in
HE
P
f
r
eq
u
en
t
p
atter
n
alg
o
r
ith
m
i
n
F
ig
u
r
e
2
b
y
u
s
i
n
g
F
attr
ib
u
te
w
h
ic
h
co
n
tr
o
l
h
o
w
m
an
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
w
h
er
e
in
d
icate
eli
m
in
a
tio
n
f
o
r
n
o
n
f
r
eq
u
en
t
p
atter
n
w
it
h
F>=
x
-
1
as s
h
o
w
n
i
n
li
n
e
n
u
m
b
er
9
HE
P
f
r
eq
u
en
t p
atter
n
alg
o
r
it
h
m
in
F
ig
u
r
e
2
.
4.
H
E
P
G
RO
W
T
H
RA
T
E
B
esid
es
eli
m
i
n
ati
n
g
p
atter
n
s
w
it
h
s
i
m
ilar
it
y
,
t
h
e
lar
g
e
n
u
m
b
er
o
f
f
r
eq
u
en
t
p
atter
n
w
i
ll
b
e
eli
m
i
n
ated
b
y
t
h
e
g
r
o
w
t
h
r
ate
f
u
n
ctio
n
}
,
{
2
1
j
i
R
R
GR
w
it
h
g
i
v
e
n
a
Gr
o
w
t
h
R
ate
t
h
r
esh
o
ld
a
n
d
t
h
er
e
is
n
o
J
u
m
p
i
n
g
Hi
g
h
lev
el
E
m
er
g
i
n
g
P
atter
n
s
(
J
HE
P
)
,
w
h
er
e
J
HE
P
is
r
elate
d
as
a
ter
m
o
f
J
E
P
.
J
E
P
is
E
P
w
i
th
s
u
p
p
o
r
t
is
0
in
o
n
e
d
ataset
an
d
m
o
r
e
th
a
n
0
i
n
t
h
e
o
th
er
d
ata
s
et
o
r
E
P
as
s
p
e
cial
t
y
p
e
o
f
E
P
w
h
ic
h
i
s
h
av
in
g
i
n
f
i
n
ite
g
r
o
w
t
h
r
ate
(
)
[
2
2
].
GR
(
X,
Y)
=
=
⁄
⁄
(
2
)
w
h
er
e:
X
=
Hig
h
lev
el
r
u
le
o
f
r
u
le
s
et
R
2
in
d
ataset
D2
.
Y
=
Hig
h
lev
el
r
u
le
o
f
r
u
le
s
et
R
1
in
d
ataset
D1
.
D2
=
Data
s
et
D2
.
D1
=
Data
s
et
D1
.
|
D2
|
=
T
o
tal
n
u
m
b
er
o
f
i
n
s
tan
ce
s
in
d
ataset
D2
.
|
D1
|
=
T
o
tal
n
u
m
b
er
o
f
i
n
s
tan
ce
s
in
d
ataset
D1
.
C
o
u
n
t R2
(
X)
=
Nu
m
b
er
o
f
h
i
g
h
le
v
el
r
u
le
X
o
f
r
u
leset
R
2
in
d
ataset
D2
.
C
o
u
n
t R1
(
Y)
=
Nu
m
b
er
o
f
h
i
g
h
le
v
el
r
u
le
Y
o
f
r
u
leset
R
1
in
d
ataset
D1
.
Su
p
p
o
r
t D
2
(
X)
=
C
o
m
p
o
s
itio
n
n
u
m
b
er
o
f
h
ig
h
lev
el
r
u
le
X
o
f
r
u
leset
R
2
in
D2
.
Su
p
p
o
r
t D
1
(
Y)
=
C
o
m
p
o
s
itio
n
n
u
m
b
er
o
f
h
ig
h
lev
el
r
u
le
Y
o
f
r
u
leset
R
1
in
D1
.
Gr
o
w
t
h
r
ate
GR
{
2
1
,
j
i
R
R
}
is
s
h
o
w
n
in
li
n
e
n
u
m
b
er
1
0
o
f
HE
P
alg
o
r
ith
m
i
n
F
i
g
u
r
e
2
is
u
s
e
d
to
d
is
cr
i
m
i
n
ate
b
et
w
ee
n
d
atasets
D2
an
d
D1
.
T
h
is
g
r
o
w
t
h
r
ate
w
h
ic
h
is
ca
lc
u
lated
u
s
i
n
g
E
q
u
atio
n
2
,
ca
n
d
ef
i
n
e
th
at
a
HE
P
is
a
r
u
le
s
et
w
h
o
s
e
s
u
p
p
o
r
t
ch
an
g
es
f
r
o
m
o
n
e
r
u
le
s
et
in
d
atase
t
D1
to
an
o
th
er
r
u
leset
in
d
ata
s
et
D2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Usi
n
g
A
ttr
ib
u
te
Ori
en
ted
I
n
d
u
ctio
n
Hig
h
Leve
l E
merg
in
g
P
a
tter
n
.
.
.
.
(
Ha
r
co
Les
lie
Hen
d
r
i
c
S
.
W
.
)
3041
I
n
o
th
er
w
o
r
d
s
,
HE
P
i
s
a
r
u
le
s
et
w
h
o
s
e
s
tr
en
g
t
h
o
f
h
ig
h
le
v
el
r
u
le
Y
o
f
r
u
lese
t
R
1
in
d
ataset
D1
c
h
a
n
g
e
s
to
h
ig
h
le
v
el
r
u
le
X
o
f
r
u
lese
t R2
in
d
ataset
D2
.
5.
AO
I
-
H
E
P
M
I
NING
F
R
E
Q
U
E
NT
P
A
T
T
E
RN
E
XP
E
RIM
E
NT
S
E
x
p
er
i
m
e
n
ts
u
s
ed
ad
u
lt,
b
r
ea
s
t
ca
n
ce
r
,
ce
n
s
u
s
a
n
d
I
P
UM
S
d
atasets
f
r
o
m
t
h
e
U
C
I
m
ac
h
i
n
e
lear
n
in
g
r
ep
o
s
ito
r
y
w
it
h
t
h
e
n
u
m
b
er
o
f
in
s
tan
ce
s
ar
e
4
8
8
4
2
,
5
6
9
,
2
4
5
8
2
8
5
an
d
2
5
6
9
3
2
r
esp
ec
tiv
el
y
[
1
3
]
.
T
h
e
p
r
o
g
r
am
s
w
er
e
r
u
n
w
ith
attr
i
b
u
te
an
d
r
u
le
th
r
es
h
o
ld
s
o
f
6
w
h
ic
h
w
er
e
ch
o
s
en
b
ased
o
n
th
e
p
r
eli
m
i
n
ar
y
ex
p
er
i
m
e
n
ts
d
o
n
e
o
n
ad
u
lt
d
ataset
s
u
c
h
th
a
t
to
g
et
m
ea
n
in
g
f
u
l
n
u
m
b
er
s
o
f
r
u
les,
a
h
ig
h
er
t
h
r
esh
o
ld
is
p
r
ef
er
ab
le
af
ter
tr
ial
ex
p
er
i
m
en
ts
.
T
h
e
ex
p
er
i
m
e
n
ts
s
h
o
w
e
d
th
at
f
r
eq
u
en
t
p
atter
n
as
r
ar
e
p
atter
n
s
an
d
ar
e
n
u
m
er
o
u
s
i
f
u
s
in
g
attr
ib
u
te
t
h
r
esh
o
ld
s
b
et
w
ee
n
4
a
n
d
6
,
an
d
r
u
les
th
r
e
s
h
o
ld
s
b
et
w
ee
n
5
an
d
1
0
.
Sin
ce
it
w
a
s
r
ar
e
to
f
in
d
f
r
eq
u
e
n
t
p
atter
n
,
w
e
d
ec
id
ed
to
u
s
e
a
b
ig
g
er
attr
ib
u
te
th
r
es
h
o
ld
o
f
6
f
o
r
ex
p
er
i
m
e
n
ts
.
Si
m
ilar
l
y
,
6
w
a
s
ch
o
s
en
f
o
r
t
h
e
r
u
le
s
t
h
r
e
s
h
o
ld
,
s
i
n
ce
6
i
s
m
ed
ian
b
et
w
ee
n
2
a
n
d
9
.
Mo
r
eo
v
er
,
w
e
o
b
tain
ed
n
u
m
er
o
u
s
f
r
eq
u
en
t p
atter
n
r
u
le
s
f
o
r
th
r
e
s
h
o
ld
s
b
et
w
ee
n
5
a
n
d
1
0
as e
x
p
ec
ted
w
h
e
n
t
h
r
es
h
o
ld
s
ar
e
b
ig
g
er
.
E
ac
h
d
ataset
h
a
s
co
n
ce
p
t
h
ier
ar
ch
ies
b
u
ilt
f
r
o
m
f
iv
e
c
h
o
s
e
n
attr
ib
u
tes
w
it
h
a
m
i
n
i
m
u
m
co
n
ce
p
t
le
v
el
o
f
th
r
ee
.
T
h
e
attr
ib
u
tes
i
n
co
n
ce
p
t
h
ier
ar
ch
ies
f
o
r
ad
u
lt
d
ataset
in
cl
u
d
e
w
o
r
k
clas
s
,
ed
u
ca
ti
o
n
,
m
ar
ital
-
s
ta
tu
s
,
o
cc
u
p
atio
n
,
an
d
n
at
iv
e
-
co
u
n
tr
y
attr
ib
u
te
s
[1
1
]
,
an
d
th
e
attr
ib
u
tes
i
n
co
n
ce
p
t
h
ier
ar
ch
ies
f
o
r
th
e
b
r
ea
s
t
ca
n
ce
r
d
ataset
co
n
tai
n
s
attr
ib
u
tes
i.
e.
clu
m
p
t
h
ic
k
n
e
s
s
,
ce
ll
s
iz
e,
ce
ll
s
h
ap
e,
b
ar
e
n
u
clei
a
n
d
n
o
r
m
al
n
u
c
leo
li
attr
ib
u
tes.
Me
a
n
w
h
ile,
class
,
m
ar
ital
s
ta
tu
s
,
m
ea
n
s
,
r
elat
1
an
d
y
ea
r
s
ch
attr
ib
u
tes,
w
e
r
e
g
iv
e
n
to
co
n
ce
p
t
h
ier
ar
ch
ie
s
f
o
r
t
h
e
C
e
n
s
u
s
d
ataset
an
d
t
h
e
attr
ib
u
tes
i
n
co
n
c
ep
t
h
ier
ar
ch
ies
f
o
r
t
h
e
I
P
UM
S
d
ataset
co
n
s
i
s
ts
o
f
r
elate
g
,
m
ar
s
t,
ed
u
cr
ec
,
m
i
g
r
at5
g
an
d
tr
an
w
o
r
k
attr
ib
u
tes.
T
ab
le
1
.
R
u
leset R
2
f
o
r
L
ea
r
n
i
n
g
Go
v
er
n
m
e
n
t Co
n
ce
p
t a
t
W
o
r
k
class
A
ttr
ib
u
te
No
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
0
1
2
3
4
5
I
n
t
e
r
me
d
i
a
t
e
A
N
Y
A
d
v
a
n
c
e
d
A
d
v
a
n
c
e
d
B
a
si
c
A
d
v
a
n
c
e
d
A
N
Y
A
N
Y
A
N
Y
A
N
Y
M
a
r
r
i
e
d
-
s
p
o
u
se
M
a
r
r
i
e
d
-
s
p
o
u
se
A
N
Y
A
N
Y
A
N
Y
A
N
Y
S
e
r
v
i
c
e
s
S
e
r
v
i
c
e
s
A
N
Y
A
m
e
r
i
c
a
A
si
a
Eu
r
o
p
e
Eu
r
o
p
e
A
n
t
a
r
t
i
c
a
3
4
5
4
7
8
6
30
17
1
1
8
0
.
5
3
%
1
8
.
3
3
%
0
0
.
7
0
%
0
0
.
4
0
%
0
0
.
0
2
%
0
0
.
0
2
%
T
ab
le
2
.
R
u
leset R
1
f
o
r
L
ea
r
n
i
n
g
No
n
Go
v
er
n
m
e
n
t
C
o
n
ce
p
t
at
W
o
r
k
class
A
tt
r
ib
u
te
No
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
0
1
2
3
4
5
7
t
h
-
8
t
h
HS
-
g
r
a
d
HS
-
g
r
a
d
A
sso
c
-
a
d
m
S
o
me
-
c
o
l
l
e
g
e
S
o
me
-
c
o
l
l
e
g
e
W
i
d
o
w
e
d
N
e
v
e
r
-
marr
i
e
d
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
M
a
r
r
i
e
d
-
s
p
o
u
se
-
a
b
se
n
t
T
o
o
l
s
A
N
Y
A
N
Y
T
o
o
l
s
A
N
Y
T
o
o
l
s
U
n
i
t
e
d
-
st
a
t
e
s
U
n
i
t
e
d
-
st
a
t
e
s
A
N
Y
U
n
i
t
e
d
-
st
a
t
e
s
U
n
i
t
e
d
-
st
a
t
e
s
U
n
i
t
e
d
-
st
a
t
e
s
1
4
5
1
2
1
0
7
.
1
4
%
2
8
.
5
7
%
3
5
.
7
1
%
0
7
.
1
4
%
1
4
.
2
9
%
0
7
.
1
4
%
T
ab
le
3
.
R
u
leset R
2
f
o
r
L
ea
r
n
i
n
g
A
b
o
u
t
A
v
er
C
lu
m
p
C
o
n
c
ep
t f
r
o
m
“
C
lu
m
p
T
h
ick
n
ess
”
Att
r
i
b
u
te
o
f
B
r
ea
s
t
C
an
ce
r
Data
s
et
No
C
e
l
l
S
i
z
e
C
e
l
l
S
h
a
p
e
B
a
r
e
N
u
c
l
e
i
N
o
r
mal
N
u
c
l
e
o
l
i
i
n
s
t
a
n
c
e
s
S
u
p
p
o
r
t
0
A
N
Y
A
N
Y
A
N
Y
A
N
Y
4
9
6
9
3
.
0
6
%
1
M
e
d
i
u
m
S
i
z
e
S
mal
l
S
h
a
p
e
A
N
Y
A
b
o
u
t
A
v
e
r
N
u
c
l
e
o
l
i
3
0
.
5
6
%
2
V
e
r
y
L
a
r
g
e
S
i
z
e
A
N
Y
A
N
Y
A
N
Y
19
3
.
5
6
%
3
M
e
d
i
u
m
S
i
z
e
L
a
r
g
e
S
h
a
p
e
A
b
o
v
e
A
v
e
r
N
u
c
l
e
i
A
N
Y
7
1
.
3
1
%
4
V
e
r
y
L
a
r
g
e
S
i
z
e
M
e
d
i
u
m
S
h
a
p
e
A
N
Y
V
e
r
y
L
a
r
g
e
N
u
c
l
e
o
l
i
3
0
.
5
6
%
5
L
a
r
g
e
S
i
z
e
V
e
r
y
L
a
r
g
e
S
h
a
p
e
V
e
r
y
L
a
r
g
e
N
u
c
l
e
i
A
N
Y
5
0
.
9
4
%
T
ab
le
4
.
R
u
leset R
1
f
o
r
L
ea
r
n
i
n
g
A
b
o
u
t
A
v
er
C
lu
m
p
C
o
n
c
ep
t f
r
o
m
“
C
lu
m
p
T
h
ick
n
ess
”
Att
r
ib
u
t
e
o
f
B
r
e
ast
C
an
ce
r
Data
s
e
t
No
C
e
l
l
S
i
z
e
C
e
l
l
S
h
a
p
e
B
a
r
e
N
u
c
l
e
i
N
o
r
mal
N
u
c
l
e
o
l
i
i
n
s
t
a
n
c
e
s
S
u
p
p
o
r
t
0
A
N
Y
A
N
Y
A
N
Y
A
N
Y
2
7
7
9
5
.
8
5
%
1
S
mal
l
S
i
z
e
L
a
r
g
e
S
h
a
p
e
V
e
r
y
L
a
r
g
e
N
u
c
l
e
i
V
e
r
y
L
a
r
g
e
N
u
c
l
e
o
l
i
1
0
.
3
5
%
2
M
e
d
i
u
m
S
i
z
e
V
e
r
y
L
a
r
g
e
S
h
a
p
e
A
N
Y
A
b
o
v
e
A
v
e
r
N
u
c
l
e
o
l
i
5
1
.
7
3
%
3
L
a
r
g
e
S
i
z
e
V
e
r
y
L
a
r
g
e
S
h
a
p
e
A
N
Y
A
N
Y
4
1
.
3
8
%
4
V
e
r
y
L
a
r
g
e
S
i
z
e
S
mal
l
S
h
a
p
e
M
e
d
i
u
m
N
u
c
l
e
i
V
e
r
y
L
a
r
g
e
N
u
c
l
e
o
l
i
1
0
.
3
5
%
5
L
a
r
g
e
S
i
z
e
S
mal
l
S
h
a
p
e
M
e
d
i
u
m
N
u
c
l
e
i
L
a
r
g
e
N
u
c
l
e
o
l
i
1
0
.
3
5
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
30
3
7
–
30
46
3042
E
ac
h
d
ataset
w
a
s
d
iv
id
ed
in
to
t
w
o
s
u
b
d
atase
ts
b
ased
o
n
le
ar
n
in
g
t
h
e
h
ig
h
le
v
el
co
n
ce
p
t
in
o
n
e
o
f
th
eir
attr
ib
u
tes.
L
ea
r
n
i
n
g
t
h
e
h
ig
h
le
v
el
co
n
ce
p
t
in
o
n
e
o
f
t
h
eir
f
iv
e
c
h
o
s
e
n
attr
ib
u
te
s
f
o
r
co
n
ce
p
t
h
ier
ar
ch
ie
s
,
m
ak
e
s
t
h
e
p
ar
a
m
eter
m
i
n
E
q
u
atio
n
1
h
a
v
e
v
al
u
e
4
,
w
h
er
e
v
alu
e
4
co
m
es
f
r
o
m
f
i
v
e
ch
o
s
en
at
tr
ib
u
tes
f
o
r
co
n
ce
p
t
h
ier
ar
ch
ies
m
i
n
u
s
1
an
d
1
is
th
e
attr
ib
u
te
f
o
r
th
e
le
ar
n
in
g
co
n
ce
p
t.
I
n
th
e
ad
u
lt
d
ataset,
w
e
lear
n
b
y
d
is
cr
i
m
i
n
ati
n
g
b
et
w
ee
n
th
e
“
g
o
v
er
n
m
en
t”
(
4
2
8
9
in
s
tan
ce
s
)
an
d
“
n
o
n
g
o
v
er
n
m
e
n
t”
(
1
4
in
s
ta
n
ce
s
)
co
n
ce
p
ts
o
f
th
e
“w
o
r
k
clas
s
”
attr
ib
u
te
[
1
4
]
in
d
atasets
D2
a
n
d
D1
r
esp
ec
tiv
el
y
.
I
n
t
h
e
b
r
ea
s
t
ca
n
ce
r
d
ataset,
w
e
lear
n
b
y
d
is
cr
i
m
i
n
ati
n
g
b
et
w
ee
n
“
ab
o
u
tav
er
clu
m
p
”
(
5
3
3
in
s
tan
ce
s
)
a
n
d
“
ab
o
v
ea
v
er
cl
u
m
p
”
(
2
8
9
in
s
tan
ce
s
)
co
n
ce
p
ts
o
f
th
e
“
cl
u
m
p
t
h
ic
k
n
ess
”
a
t
tr
ib
u
t
e
in
d
ataset
s
D2
an
d
D1
r
esp
e
ctiv
el
y
.
Me
an
w
h
i
le
C
e
n
s
u
s
d
a
taset
lear
n
s
“
g
r
ee
n
”
(
1
9
8
0
in
s
tan
ce
s
)
an
d
“
n
o
g
r
e
en
”
(
8
0
9
in
s
tan
ce
s
)
co
n
ce
p
ts
o
f
th
e
“m
ea
n
s
”
attr
ib
u
te
f
o
r
d
atasets
D2
an
d
D1
r
esp
ec
tiv
el
y
.
Fi
n
all
y
,
t
h
e
I
P
UM
S
d
ataset
lear
n
s
“u
n
m
ar
r
ied
”
(
1
4
0
1
2
4
in
s
tan
ce
s
)
a
n
d
“m
ar
r
ied
”
(
7
7
4
5
3
in
s
ta
n
ce
s
)
co
n
ce
p
t
s
o
f
t
h
e
“m
a
r
s
t”
attr
ib
u
te
as d
ataset
s
D2
an
d
D1
r
esp
ec
tiv
el
y
.
E
x
p
er
i
m
e
n
ts
w
er
e
ca
r
r
ied
o
u
t
b
y
a
J
av
a
an
d
test
ed
o
n
I
n
tel
(
R
)
A
to
m
(
T
M)
C
P
U
N5
5
0
(
1
.
5
0
GHz
)
w
it
h
1
.
0
0
GB
R
AM
.
T
h
e
A
OI
-
HE
P
ap
p
licatio
n
h
as
a
n
i
n
p
u
t
d
ataset
an
d
co
r
r
esp
o
n
d
in
g
co
n
ce
p
t
h
ier
ar
ch
ies
i
n
th
e
f
o
r
m
o
f
f
lat
f
i
les.
T
h
e
A
OI
-
HE
P
f
r
eq
u
e
n
t
p
atter
n
a
p
p
licatio
n
w
as
r
u
n
4
ti
m
es
as
t
h
e
n
u
m
b
er
o
f
ex
p
er
i
m
e
n
tal
d
atasets
a
n
d
w
it
h
th
e
attr
ib
u
te
an
d
r
u
le
th
r
e
s
h
o
ld
s
6
an
d
h
av
e
a
r
u
n
n
in
g
ti
m
e
o
f
ap
p
r
o
x
i
m
atel
y
3
,
3
,
4
an
d
1
3
s
ec
o
n
d
s
r
esp
ec
tiv
el
y
.
B
y
r
u
n
n
i
n
g
A
OI
-
HE
P
ap
p
licatio
n
w
it
h
in
p
u
t
ad
u
lt,
b
r
ea
s
t
ca
n
ce
r
,
ce
n
s
u
s
an
d
I
P
UM
S
d
atasets
,
w
e
h
a
v
e
r
u
lesets
R
2
an
d
R
1
w
it
h
6
tu
p
les
(
r
u
les)
ea
ch
,
in
cl
u
d
e
n
u
m
b
er
o
f
in
s
tan
ce
s
f
o
r
ea
ch
t
u
p
le
(
r
u
le)
an
d
s
u
p
p
o
r
t
f
o
r
e
ac
h
r
u
le.
E
ac
h
tab
le
h
as
f
o
u
r
at
tr
ib
u
tes
(
m
i
n
E
q
u
atio
n
1
)
w
h
ic
h
ar
e
f
r
o
m
f
i
v
e
ch
o
s
e
n
at
tr
ib
u
tes
m
i
n
u
s
1
lear
n
in
g
attr
ib
u
te.
I
n
cr
ed
ib
ly
,
th
e
ex
tr
ao
r
d
in
ar
y
r
u
n
n
in
g
t
i
m
e
o
f
1
3
s
ec
o
n
d
s
w
it
h
th
e
in
p
u
t
I
P
UM
S
d
ataset
h
ap
p
en
ed
b
ec
au
s
e
I
P
UM
S
h
as
h
u
g
e
in
s
tan
ce
s
lea
r
n
in
g
d
ata
s
et’
s
u
n
m
ar
r
ied
an
d
m
ar
r
ied
co
n
ce
p
ts
w
it
h
1
4
0
1
2
4
an
d
7
7
4
5
3
in
s
tan
ce
s
r
esp
ec
tiv
el
y
.
B
ec
au
s
e
o
f
p
ag
e
li
m
ita
tio
n
a
n
d
th
e
r
esu
lt
o
f
ex
p
er
i
m
en
t
t
h
e
n
o
n
l
y
r
u
le
s
ets
R
2
an
d
R
1
f
r
o
m
ad
u
lt
a
n
d
b
r
ea
s
t
ca
n
ce
r
d
atasets
w
h
ic
h
ar
e
s
h
o
w
n
b
et
w
ee
n
T
ab
les
1
an
d
4
.
T
h
e
r
esu
lt
s
o
f
r
u
n
n
in
g
t
h
e
A
OI
-
HE
P
f
r
eq
u
en
t
p
atter
n
ap
p
licatio
n
s
h
o
w
t
h
at
t
h
er
e
ar
e
o
n
l
y
2
a
n
d
1
f
i
n
d
in
g
f
r
eq
u
e
n
t
p
atter
n
s
f
r
o
m
ad
u
l
t
a
n
d
b
r
ea
s
t
ca
n
ce
r
d
a
tasets
w
h
ich
ar
e
s
h
o
w
n
b
et
w
ee
n
T
ab
les
5
to
1
2
an
d
1
3
r
esp
ec
tiv
el
y
,
w
h
ile
t
h
er
e
is
n
o
f
r
eq
u
e
n
t
p
atter
n
f
r
o
m
ce
n
s
u
s
an
d
I
P
UM
S
d
atasets
.
B
ased
o
n
b
et
w
e
en
T
ab
les
5
an
d
1
2
,
th
e
f
in
d
i
n
g
2
f
r
eq
u
e
n
t
p
atter
n
s
f
r
o
m
ad
u
lt d
ata
s
et
ar
e
r
u
les
n
u
m
b
er
0
an
d
1
,
in
T
ab
le
1
an
d
t
h
e
y
ar
e:
a.
A
d
u
lt
w
h
ich
h
a
v
e
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
w
it
h
a
n
in
ter
m
ed
iat
e
ed
u
ca
tio
n
(
3
4
5
4
/4
2
8
9
=8
0
.
5
3
%)
.
b.
A
d
u
lt
w
h
ich
h
a
v
e
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
w
it
h
Am
er
ica
as a
n
ati
v
e
co
u
n
tr
y
(
7
8
6
/4
2
8
9
=1
8
.
3
3
%).
T
ab
le
5
.
Fre
q
u
en
t P
att
er
n
f
o
r
Ru
le
s
ets
1
3
R
to
2
0
R
w
it
h
HE
P
GR
=
(
3
4
5
4
/4
2
8
9
)
/
(
1
/1
4
)
=
0
.
8
0
5
3
2
/0
.
0
7
1
4
3
=1
1
.
2
7
4
4
2
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
0
R
1
3
R
I
n
t
e
r
me
d
i
a
t
e
A
sso
c
-
a
d
m
A
N
Y
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
A
N
Y
T
o
o
l
s
A
N
Y
U
n
i
t
e
d
-
st
a
t
e
s
3
4
5
4
1
8
0
.
5
3
%
0
7
.
1
4
%
LV
0
.
5
0
.
5
0
.
5
0
.
5
S
L
V
=
2
1
1
.
2
7
%
T
ab
le
6
.
Fre
q
u
en
t P
att
er
n
f
o
r
Ru
le
s
ets
1
5
R
to
2
0
R
w
it
h
HE
P
GR
=
(
3
4
5
4
/4
2
8
9
)
/(
1
/1
4
)
=0
.
8
0
5
3
2
/0
.
0
7
1
4
3
=1
1
.
2
7
4
4
2
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
0
R
1
5
R
I
n
t
e
r
me
d
i
a
t
e
S
o
me
-
c
o
l
l
e
g
e
A
N
Y
M
a
r
r
i
e
d
-
s
p
o
u
se
-
a
b
se
n
t
A
N
Y
T
o
o
l
s
A
N
Y
U
n
i
t
e
d
-
st
a
t
e
s
3
4
5
4
1
8
0
.
5
3
%
0
7
.
1
4
%
LV
0
.
5
0
.
5
0
.
5
0
.
5
S
L
V
=
2
1
1
.
2
7
%
T
ab
le
7
.
Fre
q
u
en
t P
atte
r
n
f
o
r
Ru
le
s
ets
1
0
R
to
2
1
R
w
it
h
HE
P
GR
=
(
7
8
6
/4
2
8
9
)
/(
1
/
1
4
)
=
0
.
1
8
3
3
/0
.
0
7
1
4
3
=2
.
5
7
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
1
R
1
0
R
A
N
Y
7
t
h
-
8
th
A
N
Y
W
i
d
o
w
e
d
A
N
Y
T
o
o
l
s
A
m
e
r
i
c
a
U
n
i
t
e
d
-
st
a
t
e
s
7
8
6
1
1
8
.
3
3
%
0
7
.
1
4
%
LV
0
.
5
0
.
5
0
.
5
0
.
5
S
L
V
=
2
2
.
5
7
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Usi
n
g
A
ttr
ib
u
te
Ori
en
ted
I
n
d
u
ctio
n
Hig
h
Leve
l E
merg
in
g
P
a
tter
n
.
.
.
.
(
Ha
r
co
Les
lie
Hen
d
r
i
c
S
.
W
.
)
3043
T
ab
le
8
.
Fre
q
u
en
t P
atter
n
f
o
r
Ru
le
s
ets
1
3
R
to
2
1
R
w
it
h
HE
P
GR
=
(
7
8
6
/4
2
8
9
)
/(
1
/
1
4
)
=
0
.
1
8
3
3
/0
.
0
7
1
4
3
=2
.
5
7
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
1
R
1
3
R
A
N
Y
A
sso
c
-
a
d
m
A
N
Y
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
A
N
Y
T
o
o
l
s
A
m
e
r
i
c
a
U
n
i
t
e
d
-
st
a
t
e
s
7
8
6
1
1
8
.
3
3
%
0
7
.
1
4
%
LV
0
.
5
0
.
5
0
.
5
0
.
5
S
L
V
=
2
2
.
5
7
%
T
ab
le
9
.
Fre
q
u
en
t P
atter
n
f
o
r
R
u
le
s
ets
1
5
R
to
2
1
R
w
it
h
HE
P
GR
=
(
7
8
6
/4
2
8
9
)
/(
1
/
1
4
)
=
0
.
1
8
3
3
/0
.
0
7
1
4
3
=2
.
5
7
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
1
R
1
5
R
A
N
Y
S
o
me
-
C
o
l
l
e
g
e
A
N
Y
M
a
r
r
i
e
d
-
s
p
o
u
se
-
a
b
se
n
t
A
N
Y
T
o
o
l
s
A
m
e
r
i
c
a
U
n
i
t
e
d
-
st
a
t
e
s
7
8
6
1
1
8
.
3
3
%
0
7
.
1
4
%
LV
0
.
5
0
.
5
0
.
5
0
.
5
S
L
V
=
2
2
.
5
7
%
T
ab
le
1
0
.
Fre
q
u
en
t P
att
er
n
f
o
r
r
u
lesets
1
1
R
to
2
0
R
w
it
h
HE
P
GR
=
(
3
4
5
4
/4
2
8
9
)
/
(
4
/1
4
)
=0
.
8
0
5
3
2
/0
.
2
8
5
7
1
=
2
.
8
1
8
6
1
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
0
R
1
1
R
I
n
t
e
r
me
d
i
a
t
e
HS
-
G
r
a
d
A
N
Y
N
e
v
e
r
-
marr
i
e
d
A
N
Y
A
N
Y
A
N
Y
U
n
i
t
e
d
-
st
a
t
e
s
3
4
5
4
4
8
0
.
5
3
%
2
8
.
5
7
%
LV
0
.
5
0
.
5
2
.
1
0
.
5
S
L
V
=
3
.
6
2
.
8
2
%
T
ab
le
1
1
.
f
r
eq
u
en
t p
atter
n
f
o
r
r
u
leset
s
1
4
R
to
2
0
R
w
i
th
HE
P
GR
=
(
3
4
5
4
/4
2
8
9
)
/(
2
/1
4
)
=0
.
8
0
5
3
2
/0
.
1
4
2
8
6
=5
.
6
3
7
2
1
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
0
R
1
4
R
I
n
t
e
r
me
d
i
a
t
e
S
o
me
-
C
o
l
l
e
g
e
A
N
Y
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
A
N
Y
A
N
Y
A
N
Y
U
n
i
t
e
d
-
st
a
t
e
s
3
4
5
4
2
8
0
.
5
3
%
1
4
.
2
9
%
LV
0
.
5
0
.
5
2
.
1
0
.
5
S
L
V
=
3
.
6
5
.
6
4
%
T
ab
le
1
2
.
Fre
q
u
en
t P
atter
n
f
o
r
R
u
le
s
ets
1
4
R
to
2
1
R
w
it
h
HE
P
GR
=
(
7
8
6
/4
2
8
9
)
/(
2
/1
4
)
=0
.
1
8
3
3
/0
.
1
4
2
9
=1
.
2
8
R
u
l
e
se
se
t
s
Ed
u
c
a
t
i
o
n
M
a
r
i
t
a
l
O
c
c
u
p
a
t
i
o
n
C
o
u
n
t
r
y
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
1
R
1
4
R
A
N
Y
S
o
me
-
C
o
l
l
e
g
e
A
N
Y
M
a
r
r
i
e
d
-
c
i
v
-
sp
o
u
se
A
N
Y
A
N
Y
A
m
e
r
i
c
a
U
n
i
t
e
d
-
st
a
t
e
s
7
8
6
2
1
8
.
3
3
%
1
4
.
2
9
%
LV
0
.
5
0
.
5
2
.
1
0
.
5
S
L
V
=
3
.
6
1
.
2
8
%
T
ab
le
1
3
.
Fre
q
u
en
t P
atter
n
f
o
r
Ru
le
s
ets
1
4
R
to
2
2
R
w
it
h
HE
P
GR
=
(
1
9
/5
3
3
)
/(
1
/2
8
9
)
=
0
.
3
5
6
/0
.
0
3
5
=1
0
.
3
0
R
u
l
e
se
se
t
s
C
e
l
l
S
i
z
e
C
e
l
l
S
h
a
p
e
B
a
r
e
N
u
c
l
e
i
N
o
r
mal
N
u
c
l
e
o
l
i
I
n
st
a
n
c
e
s
S
u
p
p
o
r
t
2
2
R
1
4
R
V
e
r
y
L
a
r
g
e
S
i
z
e
V
e
r
y
L
a
r
g
e
S
i
z
e
A
N
Y
S
mal
l
S
h
a
p
e
A
N
Y
M
e
d
i
u
m
N
u
c
l
e
i
A
N
Y
V
e
r
y
L
a
r
g
e
N
u
c
l
e
o
l
i
19
1
3
.
5
6
%
0
.
3
5
%
LV
2
0
.
5
0
.
5
0
.
5
S
L
V
=
3
.
5
1
0
.
3
0
%
M
ea
n
w
h
ile,
b
ased
o
n
T
ab
le
1
3
,
th
e
o
n
l
y
f
i
n
d
in
g
f
r
eq
u
e
n
t
p
atter
n
f
r
o
m
b
r
ea
s
t
ca
n
ce
r
d
ata
s
et
is
r
u
l
e
n
u
m
b
er
2
in
T
ab
le
3
an
d
it
is
:
B
r
ea
s
t
ca
n
ce
r
w
h
ic
h
h
a
v
e
cl
u
m
p
t
h
ic
k
n
e
s
s
t
y
p
e
o
f
A
b
o
u
t
Av
er
C
l
u
m
p
w
i
th
ce
ll
s
ize
o
f
Ver
y
L
ar
g
e
Size
(
1
9
/5
3
3
=3
.
5
6
%).
T
h
e
t
w
o
o
f
ad
u
lt
d
atase
t’
s
f
r
e
q
u
en
t
p
atter
n
s
ar
e
t
h
e
h
i
g
h
e
s
t
s
co
r
e
r
u
les
w
it
h
3
4
5
4
an
d
7
8
6
in
s
tan
ce
s
in
T
ab
le
1
,
w
h
ile
t
h
e
o
n
l
y
o
n
e
b
r
ea
s
t c
an
ce
r
d
ataset
’
s
f
r
eq
u
e
n
t p
atter
n
i
s
t
h
e
s
ec
o
n
d
h
ig
h
es
t sco
r
e
r
u
le
w
it
h
1
9
in
s
ta
n
ce
s
w
h
ic
h
ar
e
m
u
ch
d
i
f
f
er
en
t
w
i
th
t
h
e
f
i
r
s
t
r
u
le
w
i
th
4
9
6
in
s
tan
ce
s
i
n
T
ab
le
3
.
Ho
w
e
v
er
,
th
is
b
r
ea
s
t
ca
n
ce
r
’
s
f
r
eq
u
e
n
t p
atter
n
f
u
lf
il
l o
f
AOI
-
HE
P
f
r
eq
u
e
n
t p
atter
n
w
h
er
e
h
a
v
in
g
:
a.
Ma
x
i
m
u
m
s
u
b
s
u
m
p
tio
n
tar
g
et
(
s
u
p
er
s
et)
i
n
to
co
n
tr
a
s
ti
n
g
(
s
u
b
s
et)
d
atasets
(
co
n
tr
ast
in
g
⊂
tar
g
et)
T
ab
le
1
3
s
h
o
w
s
t
h
at
r
u
le
2
2
R
as
t
ar
g
et
(
s
u
p
er
s
et)
d
ataset
h
as
m
ax
i
m
u
m
s
u
b
s
u
m
p
tio
n
(
⊂
)
i
n
t
o
r
u
le
1
4
R
as
co
n
tr
asti
n
g
(
s
u
b
s
et)
d
ataset
wh
ich
i
s
s
h
o
w
ed
w
ith
m
a
x
i
m
u
m
L
V=
0
.
5
an
d
SLV
v
al
u
e
is
3
.
5
.
b.
L
ar
g
e
HE
P
f
r
eq
u
e
n
t
p
atter
n
g
r
o
w
th
r
ate
a
n
d
s
u
p
p
o
r
t
in
tar
g
et
d
ataset
[1
1
].
T
ab
le
1
3
s
h
o
w
s
t
h
at
f
r
eq
u
e
n
t
p
atter
n
h
as
lar
g
e
HE
P
f
r
eq
u
en
t
p
atter
n
1
0
.
3
0
%
(
1
9
/5
3
3
)
/(
1
/2
8
9
)
=0
.
3
5
6
/0
.
0
3
5
=1
0
.
3
0
%)
an
d
lar
g
e
s
u
p
p
o
r
t
r
u
le
2
2
R
as
tar
g
et
(
s
u
p
er
s
e
t)
d
ataset,
w
h
er
e
s
u
p
p
o
r
t
2
2
R
as
ta
r
g
et
(
s
u
p
er
s
et)
d
ataset
(
3
.
5
6
%)
is
lar
g
e
th
an
1
4
R
as c
o
n
tr
ast
in
g
(
s
u
b
s
et)
d
ataset
(
0
.
3
5
%).
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
30
3
7
–
30
46
3044
T
h
e
r
esu
lts
r
u
n
n
in
g
o
f
AOI
-
H
E
P
f
r
eq
u
en
t p
atter
n
ap
p
licatio
n
u
p
o
n
ad
u
lt d
ataset
ca
n
b
e
s
e
en
b
et
w
ee
n
T
ab
les 5
an
d
1
2
w
h
er
e:
a.
T
h
er
e
ar
e
5
SL
V
v
al
u
e
f
r
eq
u
en
t
p
atter
n
s
w
i
th
f
u
ll
s
i
m
ilar
i
t
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
as
s
h
o
w
n
b
et
w
ee
n
tab
les 5
an
d
9
.
b.
T
h
er
e
ar
e
3
SL
V
v
al
u
e
f
r
eq
u
en
t
p
atter
n
s
w
it
h
f
r
eq
u
e
n
t
s
i
m
ilar
it
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
at
p
er
ce
n
tag
e
v
alu
e
o
f
(
m
-
1
)
/
m
*
1
0
0
w
h
er
e
m
as i
n
E
q
u
atio
n
1
,
as s
h
o
w
n
b
et
w
ee
n
T
ab
les 1
0
an
d
1
2
.
Me
an
w
h
ile,
T
h
e
r
esu
lt
s
r
u
n
n
i
n
g
o
f
A
OI
-
HE
P
f
r
eq
u
en
t p
atter
n
ap
p
licatio
n
u
p
o
n
b
r
ea
s
t
ca
n
c
er
d
ataset
ca
n
b
e
s
ee
n
i
n
T
ab
le
1
3
w
h
er
e
: T
h
er
e
ar
e
1
SL
V
v
al
u
e
f
r
eq
u
en
t p
atter
n
s
w
i
th
f
r
eq
u
e
n
t si
m
i
lar
it
y
s
u
b
s
u
m
p
tio
n
L
V=
0
.
5
at
p
er
ce
n
tag
e
v
a
lu
e
o
f
(
m
-
1
)
/
m
*
1
0
0
w
h
er
e
m
a
s
in
E
q
u
atio
n
1.
B
ased
o
n
f
in
d
i
n
g
f
r
eq
u
en
t
p
atter
n
s
b
et
w
ee
n
T
ab
les
5
an
d
1
3
,
th
e
s
tr
o
n
g
d
is
cr
i
m
i
n
atio
n
r
u
le
ca
n
b
e
f
o
r
m
u
lated
:
1.
T
h
er
e
ar
e
1
1
.
2
7
4
4
g
r
o
w
t
h
r
ate
s
ad
u
lt
d
ataset
w
it
h
8
0
.
5
3
% f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
(
w
it
h
an
i
n
ter
m
ed
iate
ed
u
ca
tio
n
)
an
d
7
.
1
4
%
in
f
r
eq
u
e
n
t
p
atter
n
in
n
o
n
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
(
w
it
h
a
s
s
o
c
-
ad
m
ed
u
ca
tio
n
,
m
ar
r
ied
-
civ
-
s
p
o
u
s
e
m
ar
ital stat
u
s
,
to
o
ls
o
cc
u
p
atio
n
an
d
f
r
o
m
th
e
U
n
ited
State
s
)
.
2.
T
h
er
e
ar
e
1
1
.
2
7
4
4
g
r
o
w
t
h
r
ate
s
ad
u
lt
d
ataset
w
it
h
8
0
.
5
3
% f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
(
w
it
h
an
i
n
ter
m
ed
iate
ed
u
ca
tio
n
)
a
n
d
7
.
1
4
% in
f
r
eq
u
en
t p
atter
n
i
n
n
o
n
g
o
v
er
n
m
e
n
t
w
o
r
k
c
lass
(
w
ith
s
o
m
e
co
lle
g
e
ed
u
ca
tio
n
,
m
ar
r
ied
-
s
p
o
u
s
e
-
ab
s
en
t
m
ar
i
tal
s
tat
u
s
,
to
o
ls
o
cc
u
p
atio
n
an
d
f
r
o
m
th
e
U
n
ited
Stat
es).
3.
T
h
er
e
ar
e
2
.
5
7
g
r
o
w
t
h
r
ates
a
d
u
lt
d
ataset
w
i
th
1
8
.
3
3
%
f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
e
n
t
wo
r
k
class
(
w
it
h
a
n
Am
er
ica
a
s
n
ati
v
e
co
u
n
tr
y
)
an
d
7
.
1
4
%
in
f
r
eq
u
e
n
t
p
atter
n
i
n
n
o
n
g
o
v
er
n
m
e
n
t
w
o
r
k
cl
ass
(
w
i
th
7
th
-
8
th
ed
u
ca
tio
n
,
w
id
o
w
ed
m
ar
ital st
atu
s
,
to
o
ls
o
cc
u
p
atio
n
a
n
d
f
r
o
m
th
e
U
n
ited
Sta
tes).
4.
T
h
er
e
ar
e
2
.
5
7
g
r
o
w
t
h
r
ates
a
d
u
lt
d
ataset
w
i
th
1
8
.
3
3
%
f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
e
n
t
wo
r
k
class
(
w
it
h
a
n
Am
er
ica
a
s
n
ati
v
e
co
u
n
tr
y
)
a
n
d
7
.
1
4
%
i
n
f
r
eq
u
e
n
t
p
atter
n
i
n
n
o
n
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
(
w
it
h
a
s
s
o
c
-
ad
m
ed
u
ca
tio
n
,
m
ar
r
ied
-
civ
-
s
p
o
u
s
e
m
ar
ital stat
u
s
,
to
o
ls
o
cc
u
p
atio
n
an
d
f
r
o
m
th
e
U
n
ited
State
s
)
.
5.
T
h
er
e
ar
e
2
.
5
7
g
r
o
w
t
h
r
ates
a
d
u
lt
d
ataset
w
i
th
1
8
.
3
3
%
f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
e
n
t
wo
r
k
class
(
w
it
h
a
n
Am
er
ica
as
n
ati
v
e
co
u
n
tr
y
)
an
d
7
.
1
4
%
in
f
r
eq
u
en
t
p
atter
n
in
n
o
n
g
o
v
er
n
m
e
n
t
w
o
r
k
cla
s
s
(
w
it
h
s
o
m
e
-
co
lleg
e
ed
u
ca
tio
n
,
m
ar
r
ied
-
s
p
o
u
s
e
-
ab
s
en
t
m
ar
ital
s
tat
u
s
,
to
o
ls
o
cc
u
p
atio
n
a
n
d
f
r
o
m
t
h
e
Un
ited
States
)
.
6.
T
h
er
e
ar
e
2
.
8
1
8
6
1
g
r
o
w
t
h
r
ate
s
ad
u
lt
d
ataset
w
it
h
8
0
.
5
3
%
f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
(
w
it
h
an
in
ter
m
ed
iate
ed
u
ca
tio
n
)
an
d
2
8
.
5
7
%
in
f
r
eq
u
e
n
t
p
atter
n
i
n
n
o
n
g
o
v
er
n
m
e
n
t
w
o
r
k
clas
s
(
w
it
h
HS
-
Gr
a
d
ed
u
ca
tio
n
,
Nev
er
-
m
ar
r
ied
m
ar
ital stat
u
s
a
n
d
f
r
o
m
t
h
e
Un
ited
States
)
.
7.
T
h
er
e
ar
e
5
.
6
3
7
2
1
g
r
o
w
t
h
r
ate
s
ad
u
lt
d
at
aset
w
it
h
8
0
.
5
3
% f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
en
t
w
o
r
k
cla
s
s
(
w
it
h
an
i
n
ter
m
ed
iate
ed
u
ca
t
io
n
)
a
n
d
1
4
.
2
8
%
in
f
r
eq
u
en
t
p
atter
n
i
n
n
o
n
g
o
v
er
n
m
e
n
t
w
o
r
k
c
l
ass
(
w
it
h
s
o
m
e
co
lleg
e
ed
u
ca
tio
n
,
m
ar
r
ied
-
ci
v
-
s
p
o
u
s
e
m
ar
ital
s
tat
u
s
a
n
d
f
r
o
m
th
e
U
n
ited
Sta
tes).
8.
T
h
er
e
ar
e
1
.
2
8
g
r
o
w
t
h
r
ates
a
d
u
lt
d
ataset
w
i
th
1
8
.
3
3
%
f
r
eq
u
en
t
p
atter
n
i
n
g
o
v
er
n
m
e
n
t
wo
r
k
class
(
w
it
h
a
n
Am
er
ica
as
n
ati
v
e
co
u
n
tr
y
)
a
n
d
1
4
.
2
9
%
in
f
r
eq
u
e
n
t
p
atter
n
in
n
o
n
g
o
v
er
n
m
en
t
w
o
r
k
c
la
s
s
(
w
ith
s
o
m
e
-
co
lleg
e
ed
u
ca
tio
n
,
m
ar
r
ied
-
ci
v
-
s
p
o
u
s
e
m
ar
ital
s
tat
u
s
a
n
d
f
r
o
m
th
e
U
n
it
ed
Sta
tes).
9.
T
h
er
e
ar
e
1
0
.
3
0
g
r
o
w
th
r
ate
s
b
r
ea
s
t
ca
n
ce
r
d
ataset
w
it
h
3
.
5
6
%
f
r
eq
u
en
t
p
atter
n
i
n
clu
m
p
th
ic
k
n
e
s
s
t
y
p
e
o
f
A
b
o
u
t
Av
er
C
lu
m
p
(
w
it
h
ce
ll
s
i
ze
o
f
Ver
y
L
ar
g
eS
ize)
an
d
0
.
3
5
% in
f
r
eq
u
e
n
t p
atter
n
i
n
c
lu
m
p
th
ic
k
n
e
s
s
t
y
p
e
o
f
A
b
o
v
e
Av
er
C
l
u
m
p
(
w
it
h
ce
ll
s
ize
o
f
Ver
y
L
ar
g
eSize,
ce
ll
s
h
ap
e
o
f
S
m
a
llS
h
ap
e,
B
ar
e
Nu
clei
o
f
Me
d
iu
m
Nu
cle
i a
n
d
No
r
m
al
N
u
cleo
li o
f
Ver
y
L
ar
g
eN
u
cleo
li)
.
Fin
all
y
,
e
x
p
er
i
m
e
n
t
s
s
h
o
w
ed
t
h
at
ad
u
lt
d
ataset
w
h
ich
lear
n
o
n
w
o
r
k
clas
s
attr
ib
u
te
ar
e
i
n
t
er
esti
n
g
to
m
i
n
e
s
in
ce
h
a
v
i
n
g
f
o
u
r
f
r
eq
u
e
n
t
p
atter
n
s
w
h
ic
h
ar
e
r
ec
o
g
n
iz
ed
as
s
tr
o
n
g
d
is
cr
i
m
in
a
tio
n
r
u
les.
Di
s
cr
i
m
i
n
ati
n
g
r
u
les
b
et
w
ee
n
T
ab
les
5
an
d
1
3
s
h
o
w
as
s
tr
o
n
g
d
is
cr
i
m
i
n
atin
g
p
o
w
er
w
h
er
e
t
h
e
y
h
a
v
e
lar
g
e
g
r
o
w
t
h
r
ate
s
(
b
et
w
ee
n
1
.
2
8
an
d
1
1
.
2
7
7
4
)
an
d
s
u
p
p
o
r
ts
in
tar
g
et
(
D2
)
d
atasets
(
b
et
w
ee
n
3
.
5
6
%
an
d
8
0
.
5
3
%).
Mo
r
eo
v
er
,
th
e
y
h
a
v
e
s
m
all
s
u
p
p
o
r
ts
i
n
co
n
tr
asti
n
g
(
D1
)
d
ataset
b
et
w
ee
n
0
.
3
5
% a
n
d
2
8
.
5
7
%
w
h
er
e
ea
ch
o
f
th
e
s
u
p
p
o
r
t in
co
n
tr
asti
n
g
(
D1
)
d
ataset
is
le
s
s
th
an
t
h
e
s
u
p
p
o
r
t in
tar
g
et
(
D2
)
d
ataset.
6.
AO
I
-
H
E
P
J
UST
I
F
I
CA
T
I
O
N
Sin
ce
A
OI
-
HE
P
w
as
p
r
o
p
o
s
ed
b
ased
o
n
p
r
ev
io
u
s
d
ata
m
i
n
i
n
g
tec
h
n
iq
u
es
s
u
c
h
as
A
ttrib
u
te
o
r
ien
ted
I
n
d
u
ctio
n
(
AOI
)
an
d
E
m
er
g
i
n
g
P
atter
n
(
E
P
)
th
en
AOI
-
HE
P
w
ill
b
e
d
is
t
in
g
u
is
h
ed
w
it
h
A
OI
a
n
d
E
P
.
Sin
ce
A
OI
-
HE
P
is
co
m
b
i
n
atio
n
b
et
w
ee
n
t
w
o
d
ata
m
i
n
i
n
g
tech
n
iq
u
es
s
u
ch
as
A
OI
a
n
d
E
P
,
th
e
n
A
OI
-
HE
P
is
b
et
ter
th
an
t
h
ese
t
w
o
d
ata
m
i
n
i
n
g
te
ch
n
iq
u
es.
Ob
v
io
u
s
l
y
,
A
OI
-
HE
P
is
p
er
f
ec
t
s
in
ce
its
m
i
x
t
u
r
e
o
f
s
tr
en
g
t
h
o
f
th
es
e
t
w
o
d
ata
m
i
n
i
n
g
tec
h
n
iq
u
es.
T
a
b
le
1
4
s
h
o
w
s
t
h
e
p
er
f
o
r
m
an
ce
m
etr
ic
w
it
h
n
u
m
b
er
o
f
r
u
les
r
es
u
lted
an
d
p
r
o
ce
s
s
in
g
t
i
m
e
a
m
o
n
g
A
OI
-
HE
P
,
A
OI
an
d
E
P
.
I
n
n
u
m
b
er
o
f
r
u
le
s
r
es
u
lted
,
T
ab
le
1
4
s
h
o
w
s
A
OI
-
HE
P
h
as
s
u
p
er
io
r
it
y
r
at
h
er
t
h
a
n
AOI
an
d
E
P
w
h
er
e
A
OI
-
HE
P
h
as
a
f
e
w
n
u
m
b
er
o
f
r
u
le
s
r
es
u
lted
w
h
il
s
t
A
OI
an
d
E
P
h
av
e
i
n
ter
m
ed
iat
e
an
d
m
a
n
y
n
u
m
b
er
o
f
r
u
les
r
es
u
lted
r
esp
ec
ti
v
el
y
.
A
OI
-
HE
P
h
a
s
s
u
p
er
io
r
it
y
w
it
h
a
f
e
w
n
u
m
b
er
o
f
r
u
l
es
r
esu
lted
b
ec
au
s
e
A
OI
-
HE
P
ap
p
lies
ca
r
tesi
an
p
r
o
d
u
ct
b
etw
ee
n
r
u
leset
s
o
u
t
p
u
t
f
r
o
m
A
OI
ch
ar
ac
ter
is
tic
r
u
le
alg
o
r
ith
m
,
as
m
en
tio
n
ed
i
n
Sectio
n
5
.
Mo
r
eo
v
er
,
th
e
ca
r
tesi
a
n
p
r
o
d
u
ct
ar
e
eli
m
i
n
ated
w
it
h
f
r
eq
u
e
n
t
p
att
er
n
.
Me
an
w
h
ile,
E
P
Evaluation Warning : The document was created with Spire.PDF for Python.
I
J
E
C
E
I
SS
N:
2
0
8
8
-
8708
Usi
n
g
A
ttr
ib
u
te
Ori
en
ted
I
n
d
u
ctio
n
Hig
h
Leve
l E
merg
in
g
P
a
tter
n
.
.
.
.
(
Ha
r
co
Les
lie
Hen
d
r
i
c
S
.
W
.
)
3045
h
as
w
ea
k
n
e
s
s
i
n
m
a
n
y
n
u
m
b
er
o
f
r
u
les
r
es
u
lted
s
i
n
ce
E
P
d
ea
ls
w
it
h
lo
w
lev
e
l
d
ata
w
h
i
ch
h
av
e
m
an
y
lo
w
lev
el
r
u
le
s
.
A
OI
-
HE
P
an
d
A
O
I
u
s
e
co
n
ce
p
t
h
ier
ar
ch
y
to
g
e
n
er
alize
f
r
o
m
lo
w
lev
e
l
d
ata
in
to
h
i
g
h
le
v
el
d
ata,
an
d
as
a
r
es
u
lt
A
OI
-
HE
P
an
d
A
OI
m
i
n
i
n
g
h
i
g
h
lev
e
l
r
u
le
s
w
h
ic
h
ar
e
les
s
th
a
n
lo
w
le
v
e
l
r
u
les.
T
h
u
s
,
A
OI
-
HE
P
h
as
a
f
e
w
n
u
m
b
er
o
f
r
u
l
es
r
esu
lted
b
ec
au
s
e
A
OI
-
HE
P
m
in
i
n
g
h
i
g
h
le
v
el
r
u
les
w
h
ic
h
ar
e
less
th
an
lo
w
lev
el
r
u
le
s
,
ap
p
lies
ca
r
tesi
an
p
r
o
d
u
ct
an
d
eli
m
i
n
ates i
t b
y
d
eter
m
in
in
g
t
y
p
e
o
f
HE
P
.
Ho
w
e
v
er
,
in
ti
m
e
to
p
r
o
ce
s
s
as
s
h
o
w
n
i
n
T
ab
le
1
4
,
A
OI
-
HE
P
h
as
m
ed
iu
m
cla
s
s
i
f
icatio
n
s
i
n
ce
A
OI
-
HE
P
ap
p
lies
ca
r
tesi
an
p
r
o
d
u
ct
b
et
w
ee
n
r
u
le
s
ets
o
u
tp
u
t
f
r
o
m
AOI
ch
ar
ac
ter
is
tic
r
u
le
al
g
o
r
ith
m
.
P
er
f
o
r
m
an
ce
m
etr
ic
i
n
T
ab
le
1
4
s
h
o
w
s
AO
I
-
HE
P
an
d
A
OI
h
av
e
b
etter
p
er
f
o
r
m
an
ce
in
ti
m
e
to
p
r
o
ce
s
s
ag
ai
n
s
t
E
P
,
s
in
ce
b
o
th
o
f
th
e
m
d
ea
l
w
it
h
h
ig
h
l
ev
el
d
ata.
Sin
ce
E
P
d
ea
ls
w
it
h
lo
w
lev
el
d
ata
w
h
ic
h
h
a
v
e
m
an
y
lo
w
le
v
el
r
u
le
s
th
en
E
P
h
a
s
w
ea
k
n
es
s
w
it
h
s
lo
w
p
er
f
o
r
m
an
ce
i
n
ti
m
e
to
p
r
o
ce
s
s
,
w
h
ile
A
OI
-
HE
P
an
d
A
OI
u
s
e
co
n
ce
p
t
h
ier
ar
ch
y
to
g
e
n
er
alize
f
r
o
m
l
o
w
le
v
el
d
ata
i
n
to
h
i
g
h
le
v
el
d
ata
w
h
er
e
h
i
g
h
le
v
el
d
ata
h
a
v
e
les
s
d
ata
r
ath
e
r
th
an
lo
w
le
v
el
d
ata.
Ob
v
io
u
s
l
y
,
ti
m
e
to
p
r
o
ce
s
s
h
ig
h
le
v
el
d
ata
w
ill
h
a
v
e
b
etter
p
er
f
o
r
m
a
n
ce
s
i
n
ce
d
ea
l
w
it
h
less
d
ata
an
d
th
e
o
th
er
h
an
d
,
t
i
m
e
to
p
r
o
ce
s
s
lo
w
le
v
el
d
ata
w
il
l
h
a
v
e
s
lo
w
p
er
f
o
r
m
a
n
ce
s
i
n
ce
d
ea
l
w
it
h
h
u
g
e
d
ata.
R
ath
er
th
a
n
A
OI
,
A
OI
-
H
E
P
h
as
lo
w
er
p
er
f
o
r
m
a
n
ce
in
t
i
m
e
to
p
r
o
ce
s
s
,
s
in
ce
A
OI
-
HE
P
ap
p
lies
ca
r
tesi
an
p
r
o
d
u
ct
b
etw
ee
n
r
u
le
s
ets
o
u
tp
u
t
f
r
o
m
AOI
ch
ar
ac
ter
is
tic
r
u
l
e
alg
o
r
ith
m
,
an
d
ca
r
tesi
a
n
p
r
o
d
u
ct
ar
e
eli
m
i
n
ated
w
it
h
f
r
eq
u
en
t p
atter
n
s
.
T
ab
le
1
4
.
P
er
f
o
r
m
an
ce
Me
tr
ic
Am
o
n
g
A
OI
-
HE
P
,
A
OI
an
d
E
P
A
O
I
-
H
EP
A
O
I
EP
N
u
mb
e
r
o
f
r
u
l
e
s re
su
l
t
e
d
P
r
o
c
e
ssi
n
g
t
i
me
F
e
w
M
e
d
i
u
m
I
n
t
e
r
me
d
i
a
t
e
F
a
st
e
st
M
a
n
y
S
l
o
w
7.
CO
NCLU
SI
O
N
Min
i
n
g
HE
P
f
r
eq
u
e
n
t
p
atter
n
s
w
it
h
A
OI
-
HE
P
ar
e
in
f
l
u
e
n
c
ed
b
y
lear
n
i
n
g
o
n
h
i
g
h
lev
e
l
co
n
ce
p
t
in
o
n
e
o
f
ch
o
s
e
n
attr
ib
u
te
a
n
d
ex
ten
d
ed
ex
p
er
i
m
en
t
u
p
o
n
ad
u
lt
d
ataset
w
h
er
e
lear
n
o
n
m
ar
ital
-
s
tat
u
s
attr
ib
u
te
s
h
o
w
ed
t
h
at
t
h
er
e
is
n
o
f
i
n
d
in
g
f
r
eq
u
e
n
t
p
atter
n
.
T
h
e
r
esear
ch
f
o
r
m
i
n
i
n
g
HE
P
f
r
e
q
u
en
t
p
atter
n
s
w
it
h
A
OI
-
HE
P
is
in
ter
e
s
ted
to
b
e
ex
ten
d
ed
w
h
er
e
m
i
n
in
g
HE
P
f
r
eq
u
en
t
p
atter
n
s
ca
n
b
e
d
o
n
e
b
y
s
ea
r
ch
i
n
g
o
n
ev
er
y
ea
c
h
attr
ib
u
te
i
n
d
atase
t
f
o
r
f
i
n
d
in
g
p
o
s
s
ib
le
f
r
eq
u
en
t
p
atter
n
s
.
Mo
r
eo
v
er
,
s
in
ce
t
h
e
r
e
ar
e
m
o
r
e
th
a
n
2
co
n
ce
p
ts
i
n
h
ig
h
le
v
el
at
tr
ib
u
te
co
n
ce
p
t,
t
h
en
m
i
n
i
n
g
HE
P
f
r
eq
u
en
t
p
atter
n
s
n
ee
d
to
b
e
ex
te
n
d
ed
to
d
is
cr
i
m
i
n
ate
m
o
r
e
th
a
n
2
r
u
lesets
.
F
u
r
t
h
er
m
o
r
e,
th
e
e
x
p
er
i
m
en
ts
s
h
o
w
ed
t
h
at
t
h
er
e
ar
e
ca
n
d
id
ate
HE
P
f
r
eq
u
en
t p
atter
n
s
i
n
ce
n
s
u
s
d
at
aset in
r
e
v
er
s
e
co
n
d
itio
n
,
t
h
e
n
m
i
n
in
g
HE
P
f
r
eq
u
e
n
t p
atter
n
s
h
o
u
ld
b
e
ex
te
n
d
ed
to
m
in
i
n
g
in
v
er
s
e
p
atter
n
s
.
T
h
is
r
esear
c
h
s
h
o
u
ld
n
ee
d
m
o
r
e
ex
ten
d
ed
r
esear
ch
a
n
d
ex
p
er
i
m
en
ts
i
n
o
r
d
er
to
f
i
n
d
j
u
s
tif
icat
io
n
o
f
t
h
is
m
in
in
g
ap
p
r
o
ac
h
w
it
h
o
th
er
f
r
eq
u
e
n
t
p
atter
n
alg
o
r
ith
m
s
,
t
h
e
in
p
u
t
d
ata
s
ets
s
h
o
u
ld
b
e
ap
p
lied
to
o
th
er
f
r
eq
u
en
t
p
att
er
n
alg
o
r
it
h
m
in
o
r
d
er
to
f
i
n
d
th
e
d
i
f
f
er
e
n
ce
s
i
n
ter
m
o
f
p
er
f
o
r
m
a
n
ce
,
t
y
p
e
a
n
d
k
in
d
o
f
p
atter
n
s
,
ad
v
a
n
ta
g
es a
n
d
d
is
ad
v
an
tag
e
s
.
ACK
NO
WL
E
D
G
E
M
E
NT
S
T
h
is
r
esear
ch
is
s
u
p
p
o
r
ted
u
n
d
er
P
r
o
g
r
a
m
o
f
r
esear
ch
i
n
ce
n
ti
v
e
o
f
n
atio
n
al
i
n
n
o
v
atio
n
s
y
s
te
m
(
SIN
AS)
f
r
o
m
Min
i
s
tr
y
o
f
R
e
s
ea
r
ch
,
T
ec
h
n
o
lo
g
y
an
d
Hi
g
h
er
E
d
u
ca
tio
n
o
f
th
e
R
ep
u
b
lic
o
f
I
n
d
o
n
esia,
d
ec
r
ee
n
u
m
b
er
1
4
7
/M/Kp
/I
V/2
0
1
5
,
R
esear
ch
co
d
e:
R
D
-
2015
-
0020.
RE
F
E
R
E
NC
E
S
[1
]
J.
Ha
n
,
e
t
a
l
.
,
“
F
re
q
u
e
n
t
p
a
tt
e
r
n
m
in
in
g
:
c
u
rre
n
t
sta
tu
s
a
n
d
f
u
t
u
r
e
d
irec
ti
o
n
s,”
Da
t
a
M
in
Kn
o
wl
Disc
,
v
ol
/i
ss
u
e
:
15
(
1
)
,
p
p
.
5
5
-
8
6
,
2
0
0
7
.
[2
]
J.
Ha
n
,
e
t
a
l.
,
“
M
in
in
g
F
re
q
u
e
n
t
P
a
tt
e
r
n
s
w
it
h
o
u
t
Ca
n
d
i
d
a
te
Ge
n
e
ra
ti
o
n
:
A
F
re
q
u
e
n
t
-
P
a
tt
e
rn
T
re
e
Ap
p
r
o
a
c
h
,
”
Da
ta
M
in
.
Kn
o
wl.
Disc
o
v
.
,
v
ol
/i
ss
u
e
:
8
(
1
)
,
p
p
.
5
3
-
8
7
,
2
0
0
4
.
[3
]
R.
P
o
d
ra
z
a
a
n
d
K.
T
o
m
a
sz
e
w
sk
i,
“
KT
D
A
:
E
m
e
r
g
in
g
P
a
tt
e
rn
s
Ba
se
d
Da
ta
A
n
a
ly
sis
S
y
ste
m
,
”
i
n
Pro
c
e
e
d
in
g
s o
f
XX
I
Fa
ll
M
e
e
ti
n
g
o
f
Po
li
sh
I
n
fo
rm
a
ti
o
n
Pro
c
e
ss
in
g
S
o
c
iety
,
p
p
.
2
1
3
-
2
2
1
,
2
0
0
5
.
[4
]
R.
A
g
r
a
w
a
l,
e
t
a
l.
,
“
M
in
i
n
g
a
ss
o
c
iatio
n
r
u
les
b
e
tw
e
e
n
se
ts
o
f
i
tem
s
in
larg
e
d
a
tab
a
se
s,”
AC
M
S
IGM
OD
Rec
,
v
ol
/i
ss
u
e
:
22
(
2
)
,
p
p
.
2
0
7
-
2
1
6
,
1
9
9
3
.
[5
]
K.
Ra
m
a
m
o
h
a
n
a
ra
o
,
e
t
a
l.
,
“
Eff
i
c
ien
t
M
in
i
n
g
o
f
Co
n
tras
t
P
a
tt
e
rn
s
a
n
d
T
h
e
ir
A
p
p
li
c
a
ti
o
n
s
to
Cla
ss
if
ic
a
ti
o
n
,
”
i
n
Pro
c
e
e
d
in
g
s
o
f
th
e
3
r
d
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
In
telli
g
e
n
t
S
e
n
sin
g
a
n
d
In
f
o
rm
a
ti
o
n
Pr
o
c
e
ss
in
g
(
ICIS
IP
'
0
5
)
,
IEE
E
Co
m
p
u
ter
S
o
c
iety
,
p
p
.
3
9
-
4
7
,
2
0
0
5
.
[6
]
H.
F
a
n
a
n
d
K.
Ra
m
a
m
o
h
a
n
a
ra
o
,
“
A
Ba
y
e
sia
n
a
p
p
ro
a
c
h
to
u
se
e
m
e
rg
in
g
p
a
tt
e
rn
s
f
o
r
c
las
si
f
ica
ti
o
n
,
”
i
n
Pro
c
e
e
d
in
g
s
o
f
t
h
e
1
4
th
Au
stra
la
si
a
n
d
a
t
a
b
a
se
c
o
n
fer
e
n
c
e
(
ADC
'
0
3
)
,
p
p
.
3
9
-
4
8
,
2
0
0
3
.
[7
]
G
.
Do
n
g
a
n
d
J.
L
i,
“
E
ff
icie
n
t
m
i
n
in
g
o
f
e
m
e
r
g
in
g
p
a
tt
e
rn
s:
d
isc
o
v
e
rin
g
tren
d
s
a
n
d
d
if
fe
re
n
c
e
s,”
i
n
Pro
c
e
e
d
in
g
s
o
f
th
e
5
t
h
ACM
S
IGKD
D i
n
ter
n
a
ti
o
n
a
l
c
o
n
fer
e
n
c
e
o
n
Kn
o
wled
g
e
d
is
c
o
v
e
ry
a
n
d
d
a
t
a
mi
n
in
g
,
p
p
.
4
3
-
5
2
,
1
9
9
9
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
0
8
8
-
8708
I
J
E
C
E
Vo
l.
6
,
No
.
6
,
Dec
em
b
er
2
0
1
6
:
30
3
7
–
30
46
3046
[8
]
C.
C.
Ag
g
a
r
w
a
l,
“
A
n
in
tro
d
u
c
ti
o
n
to
F
re
q
u
e
n
t
P
a
tt
e
rn
M
i
n
in
g
,
”
F
re
q
u
e
n
t
Pa
t
ter
n
M
in
in
g
,
C.
C.
Ag
g
a
r
wa
l
a
n
d
J.
Ha
n
(e
d
s.),
S
p
ri
n
g
e
r,
p
p
.
1
-
1
7
,
2
0
1
4
.
[9
]
C.
C.
Ag
g
a
r
w
a
l,
e
t
a
l.
,
“
F
re
q
u
e
n
t
p
a
tt
e
rn
m
in
in
g
A
lg
o
rit
h
m
:
A
S
u
rv
e
y
,
”
Fre
q
u
e
n
t
Pa
tt
e
rn
M
in
i
n
g
,
C.
C.
Ag
g
a
r
w
a
l
a
n
d
J.
Ha
n
(e
d
s.),
S
p
ri
n
g
e
r,
p
p
.
1
9
-
6
4
,
2
0
1
4
.
[1
0
]
A
.
Zi
m
e
k
,
e
t
a
l.
,
“
F
re
q
u
e
n
t
P
a
t
tern
M
i
n
in
g
A
lg
o
rit
h
m
f
o
r
Da
t
a
c
lu
ste
rin
g
,
”
Fre
q
u
e
n
t
Pa
tt
e
rn
M
in
in
g
,
C.
C.
Ag
g
a
r
wa
l
a
n
d
J.
Ha
n
(e
d
s.
)
,
S
p
rin
g
e
r,
p
p
.
4
0
3
-
4
2
3
,
2
0
1
4
.
[1
1
]
S
.
W
a
rn
a
rs,
“
M
in
in
g
F
re
q
u
e
n
t
P
a
tt
e
rn
w
it
h
A
tt
rib
u
te
Orie
n
ted
In
d
u
c
ti
o
n
Hig
h
lev
e
l
Em
e
rg
in
g
P
a
tt
e
rn
(A
OI
-
HEP
)
,
”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
IEE
E
th
e
2
n
d
In
ter
n
a
t
io
n
a
l
C
o
n
fer
e
n
c
e
o
n
I
n
fo
rm
a
ti
o
n
a
n
d
C
o
mm
u
n
ica
ti
o
n
T
e
c
h
n
o
lo
g
y
(
IEE
E
ICo
ICT
2
0
1
4
),
Ba
n
d
u
n
g
,
I
n
d
o
n
e
s
ia
,
p
p
.
1
4
4
-
1
4
9
,
2
8
-
3
0
M
a
y
2
0
1
4
.
[1
2
]
S
.
W
a
rn
a
rs,
“
A
tt
rib
u
te
Orie
n
te
d
In
d
u
c
t
io
n
o
f
Hig
h
-
lev
e
l
Em
e
rg
in
g
P
a
tt
e
rn
s,”
in
Pro
c
e
e
d
i
n
g
s
o
f
th
e
IE
EE
In
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
Gr
a
n
u
la
r
C
o
mp
u
ti
n
g
(
IEE
E
Gr
C),
Ha
n
g
zh
o
u
,
Ch
i
n
a
,
p
p
.
5
2
5
–
5
3
0
,
1
1
-
1
3
A
u
g
u
s
t
2
0
1
2
.
[1
3
]
A
.
F
ra
n
k
a
n
d
A
.
A
su
n
c
io
n
,
“
UCI
M
a
c
h
in
e
L
e
a
rn
in
g
Re
p
o
sito
ry
,”
Irv
in
e
,
C
A
,
Un
iv
e
rsit
y
o
f
Ca
li
f
o
rn
ia,
S
c
h
o
o
l
o
f
In
f
o
rm
a
ti
o
n
a
n
d
Co
m
p
u
ter S
c
ie
n
c
e
,
2
0
1
0
.
[
h
t
tp
:/
/arc
h
iv
e
.
ics
.
u
c
i.
e
d
u
/m
l]
.
[1
4
]
S
.
W
a
rn
a
rs,
“
M
in
i
n
g
F
re
q
u
e
n
t
a
n
d
S
im
il
a
r
P
a
tt
e
r
n
s
w
it
h
A
tt
rib
u
te Orien
ted
In
d
u
c
ti
o
n
Hig
h
L
e
v
e
l
Eme
rg
in
g
P
a
tt
e
rn
(A
OI
-
HEP
)
Da
ta
M
in
in
g
T
e
c
h
n
iq
u
e
,
”
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
Eme
rg
in
g
T
e
c
h
n
o
l
o
g
ies
in
Co
m
p
u
t
a
ti
o
n
a
l
a
n
d
Ap
p
li
e
d
S
c
ie
n
c
e
s (
IJ
ET
CAS
)
,
v
o
l
/
issu
e
:
3
(
11
)
,
p
p
.
2
6
6
-
2
7
6
,
2
0
1
4
.
[1
5
]
S
.
W
a
rn
a
rs,
“
A
tt
rib
u
te
Orie
n
ted
In
d
u
c
ti
o
n
Hig
h
L
e
v
e
l
E
m
e
r
g
in
g
P
a
tt
e
rn
(A
OI
-
HEP
)
f
u
tu
re
re
se
a
r
c
h
,
”
i
n
Pro
c
e
e
d
in
g
s
o
f
IEE
E
th
e
8
t
h
In
te
rn
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
In
f
o
rm
a
ti
o
n
&
Co
mm
u
n
ica
ti
o
n
T
e
c
h
n
o
l
o
g
y
a
n
d
S
y
ste
ms
(
ICT
S
),
S
u
r
a
b
a
y
a
,
I
n
d
o
n
e
si
a
,
p
p
.
1
3
-
1
8
,
2
4
-
2
5
S
e
p
tem
b
e
r
2
0
1
4
.
[1
6
]
J.
Ha
n
,
e
t
a
l.
,
“
Kn
o
w
led
g
e
d
isc
o
v
e
r
y
in
d
a
tab
a
se
s:
A
n
a
tt
rib
u
ted
a
p
p
ro
a
c
h
,
”
i
n
Pro
c
e
e
d
i
n
g
o
f
th
e
1
8
t
h
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Ver
y
L
a
r
g
e
Da
ta
Ba
se
s
,
p
p
.
5
4
7
-
5
5
9
,
1
9
9
2
.
[1
7
]
Y.
Ca
i,
e
t
a
l.
,
“
A
n
a
tt
rib
u
te
-
o
ri
e
n
ted
a
p
p
r
o
a
c
h
f
o
r
lea
rn
in
g
c
las
sif
ic
a
ti
o
n
ru
l
e
s
f
ro
m
re
latio
n
a
l
d
a
tab
a
se
s,”
i
n
Pro
c
e
e
d
in
g
s
o
f
6
th
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Da
t
a
E
n
g
i
n
e
e
rin
g
,
p
p
.
2
8
1
-
2
8
8
,
1
9
9
0
.
[1
8
]
Z.
C
.
S
e
n
g
,
e
t
a
l
.
,
“
F
re
q
u
e
n
t
it
e
m
se
ts
m
in
in
g
b
a
se
d
o
n
c
o
n
c
e
p
t
latti
c
e
a
n
d
slid
in
g
w
in
d
o
w
s
,
”
T
e
lko
mn
ika
,
v
o
l
/
issu
e
:
11
(
8
)
,
p
p
.
4
7
8
0
-
4
7
8
7
,
A
u
g
u
st 2
0
1
3
.
[1
9
]
M
.
Yim
in
,
e
t
a
l.
,
“
A
n
e
ff
icie
n
t
a
lg
o
rit
h
m
f
o
r
m
in
in
g
T
o
p
-
k
c
lo
se
d
f
re
q
u
e
n
t
it
e
m
se
ts
o
v
e
r
d
a
ta
stre
a
m
s
o
v
e
r
d
a
ta
stre
a
m
s
,
”
T
e
lko
mn
ika
,
v
o
l
/i
ss
u
e
:
11
(
7
)
,
p
p
.
3
7
5
9
-
3
7
6
6
,
Ju
ly
2
0
1
3
.
[2
0
]
R.
Da
n
g
e
r,
e
t
a
l
.
,
“
Ob
jec
tm
in
e
r:
A
n
e
w
a
p
p
ro
a
c
h
f
o
r
M
in
i
n
g
Co
m
p
lex
o
b
jec
ts,
”
i
n
Pro
c
e
e
d
in
g
s
o
f
t
h
e
6
t
h
in
ter
n
a
t
io
n
a
l
c
o
n
fer
e
n
c
e
o
n
E
n
ter
p
rise
In
fo
rm
a
ti
o
n
S
y
ste
ms
(
ICEIS
’0
4
)
,
p
p
.
4
2
-
4
7
,
2
0
0
4
.
[2
1
]
A.
Y.
R
.
G
o
n
z
a
lez
,
e
t
a
l.
,
“
M
in
in
g
F
re
q
u
e
n
t
S
im
il
a
r
P
a
tt
e
rn
s
o
n
M
ix
e
d
Da
ta,”
i
n
Pro
c
e
e
d
i
n
g
s
o
f
th
e
1
3
t
h
Ib
e
ro
a
me
ric
a
n
c
o
n
g
re
ss
o
n
P
a
tt
e
rn
Rec
o
g
n
it
i
o
n
:
Pr
o
g
re
ss
in
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
Ima
g
e
An
a
lys
is
a
n
d
Ap
p
li
c
a
ti
o
n
s(
CIAR
P
'
0
8
)
,
p
p
.
1
3
6
-
1
4
4
,
2
0
0
8
.
[2
2
]
J.
L
i,
e
t
a
l.
,
“
In
sta
n
c
e
-
b
a
se
d
c
las
sif
ica
ti
o
n
b
y
Eme
rg
in
g
P
a
tt
e
rn
s
,
”
i
n
p
ro
c
e
e
d
in
g
o
f
t
h
e
4
t
h
E
u
ro
p
e
a
n
Co
n
fer
e
n
c
e
o
n
Pri
n
c
ip
les
o
f
Da
ta
M
i
n
in
g
a
n
d
Kn
o
wle
d
g
e
Disc
o
v
e
ry
(
PKDD’0
0
)
,
p
p
.
1
9
1
-
2
0
0
,
2
0
0
0
.
B
I
O
G
RAP
H
Y
O
F
AUTHO
R
He
a
d
o
f
In
f
o
rm
a
ti
o
n
sy
ste
m
c
o
n
c
e
n
tratio
n
a
t
D
o
c
to
r
o
f
Co
m
p
u
ter
S
c
ien
c
e
,
Bi
n
a
Nu
sa
n
tara
u
n
iv
e
rsity
(
ww
w
.
d
c
s.b
in
u
s.ac
.
id
)
.
Ha
v
in
g
b
a
c
h
e
lo
r
d
e
g
re
e
in
Co
m
p
u
ter
S
c
ien
c
e
b
e
t
w
e
e
n
Ju
l
y
1
9
9
1
-
A
p
ril
1
9
9
5
w
it
h
in
f
o
rm
a
ti
o
n
sy
ste
m
to
p
ic,
m
a
ste
r
d
e
g
re
e
in
In
f
o
rm
a
ti
o
n
T
e
c
h
n
o
l
o
g
y
f
ro
m
u
n
iv
e
rsity
o
f
In
d
o
n
e
sia
b
e
tw
e
e
n
Ju
ly
2
0
0
4
-
J
an
u
a
ry
2
0
0
7
w
it
h
d
a
ta
w
a
re
h
o
u
se
th
e
sis
to
p
ic.
Be
twe
e
n
S
e
p
t
e
m
b
e
r
2
0
0
8
-
D
ec
e
m
b
e
r
2
0
1
2
d
id
P
h
D
c
o
m
p
u
ter
sc
ien
c
e
a
t
th
e
M
a
n
c
h
e
ste
r
M
e
tro
p
o
li
tan
u
n
i
v
e
rsity
,
Un
it
e
d
Kin
g
d
o
m
w
it
h
d
a
ta
m
in
in
g
th
e
sis
to
p
ic.
Ha
v
e
b
e
e
n
IT
lec
tu
re
r
sin
c
e
1
9
9
5
a
n
d
h
a
v
e
In
d
o
n
e
sia
n
n
a
ti
o
n
a
l
a
c
a
d
e
m
i
c
p
o
siti
o
n
ra
n
k
(Je
n
jan
g
jab
a
tan
a
k
a
d
e
m
ik
)
A
s
so
c
iate
p
ro
f
e
ss
o
r
(
L
e
k
to
rKe
p
a
la,
5
5
0
p
o
i
n
ts)
sin
c
e
2
0
0
7
.
I
h
a
v
e
re
se
a
rc
h
in
tere
st
o
n
f
i
e
ld
su
c
h
a
s
Big
Da
t
a
,
P
a
ra
ll
e
l
c
o
m
p
u
ti
n
g
,
Da
ta
M
in
in
g
,
M
a
c
h
i
n
e
L
e
a
rn
in
g
,
in
telli
g
e
n
t
a
p
p
li
c
a
ti
o
n
a
n
d
so
o
n
.
M
y
p
u
b
li
c
a
ti
o
n
s
c
a
n
b
e
re
a
c
h
e
d
a
t
h
tt
p
s:/
/www
.
re
se
a
rc
h
g
a
te.n
e
t/
p
ro
fil
e
/Ha
rc
o
_
L
e
slie_
He
n
d
ric_
S
p
it
s_
W
a
rn
a
rs2
.
Evaluation Warning : The document was created with Spire.PDF for Python.