I
nte
rna
t
io
na
l J
o
urna
l o
f
I
nfo
rm
a
t
ics a
nd
Co
m
m
un
ica
t
io
n T
ec
hn
o
lo
g
y
(
I
J
-
I
CT
)
Vo
l.
9
,
No
.
1
,
Ap
r
il
2
0
2
0
,
p
p
.
9
~
1
8
I
SS
N:
2252
-
8
7
7
6
,
DOI
:
1
0
.
1
1
5
9
1
/iji
ct.
v
9
i1
.
p
p
9
-
1
8
9
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//ij
ict.
ia
esco
r
e.
co
m
M
emetic
a
lg
o
rith
m for sho
rt
mess
a
g
ing
serv
ice spa
m filt
e
r using
text norma
liza
tio
n and
sema
ntic
appro
a
ch
Arno
ld Adim
a
bu
a
O
j
ug
o
1
,
Andrew
O
k
o
nji
E
bo
k
a
2
1
De
p
a
rtme
n
t
o
f
M
a
t
h
e
m
a
ti
c
s/Co
m
p
u
ter S
c
ien
c
e
,
F
e
d
e
ra
l
Un
iv
e
rsi
ty
o
f
P
e
tro
leu
m
Re
so
u
rc
e
s E
ffu
r
u
n
,
Nig
e
ria
2
De
p
a
rtme
n
t
o
f
Co
m
p
u
ter
S
c
ien
c
e
Ed
u
c
a
ti
o
n
,
F
e
d
e
ra
l
Co
ll
e
g
e
o
f
E
d
u
c
a
ti
o
n
(Tec
h
n
ica
l)
As
a
b
a
,
Nig
e
ria
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
l 1
2
,
2
0
19
R
ev
is
ed
No
v
13
,
2
0
19
Acc
ep
ted
J
an
12
,
2
0
20
To
d
a
y
’s
p
o
p
u
lari
ty
o
f
th
e
s
h
o
rt
m
e
ss
a
g
e
s
s
e
rv
ice
s
(S
M
S
)
h
a
s
c
re
a
ted
a
p
ro
p
it
io
u
s
e
n
v
iro
n
m
e
n
t
fo
r
s
p
a
m
m
in
g
to
t
h
ri
v
e
.
S
p
a
m
s
a
re
u
n
so
l
icited
a
d
v
e
rti
sin
g
,
a
d
u
l
t
-
th
e
m
e
d
o
r
in
a
p
p
ro
p
riate
c
o
n
ten
t,
p
re
m
iu
m
fra
u
d
,
sm
ish
in
g
a
n
d
m
a
lwa
re
.
Th
e
y
a
re
a
c
o
n
sta
n
t
re
m
in
d
e
r
o
f
t
h
e
n
e
e
d
fo
r
a
n
e
ffe
c
ti
v
e
sp
a
m
fil
ter.
Ho
we
v
e
r,
S
M
S
li
m
it
a
ti
o
n
s
o
f
1
6
0
-
c
h
a
rc
a
ters
a
n
d
1
4
0
-
b
y
t
e
s
siz
e
a
s
we
ll
a
s
it
s
b
e
i
n
g
rip
p
led
wi
th
sla
n
g
s,
e
m
o
ti
c
o
n
s
a
n
d
a
b
b
re
v
iat
io
n
s
fu
rt
h
e
r
in
h
i
b
it
s
e
ffe
c
ti
v
e
train
in
g
o
f
m
o
d
e
ls
to
a
id
a
c
c
u
ra
te
c
las
sifica
ti
o
n
.
Th
e
stu
d
y
p
ro
p
o
se
s
G
e
n
e
ti
c
Alg
o
r
it
h
m
Trai
n
e
d
Ba
y
e
sia
n
Ne
two
r
k
s
o
lu
t
io
n
t
h
a
t
se
e
k
s
to
n
o
rm
a
li
z
e
n
o
isy
fe
a
ts,
e
x
p
a
n
d
t
e
x
t
v
ia
u
se
o
f
lex
ico
g
ra
p
h
ic
a
n
d
se
m
a
n
ti
c
d
ictio
n
a
ries
th
a
t
u
se
s
wo
r
d
se
n
se
d
isa
m
b
ig
u
a
ti
o
n
tec
h
n
iq
u
e
t
o
train
th
e
u
n
d
e
rl
y
i
n
g
lea
rn
i
n
g
h
e
u
risti
c
s.
A
n
d
i
n
tu
r
n
,
e
ffe
c
ti
v
e
ly
h
e
lp
t
o
c
las
sify
S
M
S
in
sp
a
m
a
n
d
leg
it
ima
te
c
las
se
s.
Hy
b
rid
m
o
d
e
l
c
o
m
p
rise
s
o
f
tex
t
p
re
p
ro
c
e
ss
in
g
,
fe
a
tu
re
se
lec
ti
o
n
a
s
we
ll
a
s
train
in
g
a
n
d
c
las
sifica
ti
o
n
se
c
ti
o
n
.
S
tu
d
y
u
se
s
a
h
y
b
ri
d
G
e
n
e
ti
c
Alg
o
rit
h
m
train
e
d
Ba
y
e
sia
n
m
o
d
e
l
fo
r
wh
ich
th
e
G
A
is
u
se
d
fo
r
fe
a
tu
re
se
lec
ti
o
n
;
wh
il
e
,
t
h
e
Ba
y
e
sia
n
a
lg
o
ri
th
m
is
u
se
d
a
s
c
las
sifier
.
K
ey
w
o
r
d
s
:
B
ay
es th
eo
r
em
Me
m
etic
alg
o
r
ith
m
Sem
an
tic
p
r
o
ce
s
s
in
g
Sp
am
f
ilter
s
T
ex
t n
o
r
m
aliza
tio
n
T
ex
t p
r
o
ce
s
s
in
g
T
h
is i
s
a
n
o
p
e
n
a
c
c
e
ss
a
rticle
u
n
d
e
r th
e
CC B
Y
-
SA
li
c
e
n
se
.
C
o
r
r
e
s
p
o
nd
ing
A
uth
o
r
:
Ar
n
o
ld
Ad
im
a
b
u
a
Oju
g
o
,
Dep
ar
tm
en
t o
f
Ma
th
em
atics/C
o
m
p
u
ter
Scien
ce
,
Fed
er
al
Un
iv
er
s
ity
o
f
Petr
o
le
u
m
R
eso
u
r
ce
s
E
f
f
u
r
u
n
,
P.M
.
B
1
2
2
1
,
E
f
f
u
r
u
n
,
W
ar
r
i,
Delta
State,
Nig
er
ia
.
E
m
ail:
a
rn
o
ld
o
j
u
g
o
@g
m
a
il
.
c
o
m
1.
I
NT
RO
D
UCT
I
O
N
T
h
e
ad
v
e
n
t
o
f
s
h
o
r
t
m
ess
ag
in
g
s
er
v
ices
b
y
Neil
Pap
wo
r
th
s
i
n
ce
1
9
9
2
,
h
as
s
ee
n
g
r
ea
t
p
en
et
r
atio
n
an
d
a
tr
em
en
d
o
u
s
g
r
o
wth
r
ate
o
f
t
h
e
s
er
v
ice.
A
d
v
en
t
o
f
m
o
b
ile
p
h
o
n
es
with
en
h
a
n
ce
f
ea
t
u
r
es
h
as
co
n
tr
ib
u
ted
to
th
e
lar
g
e
s
ca
le
ad
o
p
tio
n
o
f
SMS
b
y
u
s
er
s
.
T
h
e
p
o
r
tab
il
ity
,
m
o
b
ilit
y
,
u
b
iq
u
ity
o
f
s
er
v
ices
an
d
its
lo
w
co
s
t
co
n
tin
u
es
to
p
r
o
m
o
te
tex
t
m
ess
ag
es
to
b
ec
o
m
e
th
e
m
o
s
t
u
s
ed
m
ea
n
s
o
f
elec
tr
o
n
ic
co
m
m
u
n
icatio
n
in
th
e
wo
r
ld
to
d
ay
[1
-
2]
.
Sh
o
r
t
Me
s
s
ag
e
Ser
v
ice
(
SMS)
i
s
tex
t
s
er
v
ice
co
m
p
o
n
e
n
t
o
f
p
h
o
n
es
o
r
m
o
b
ile
co
m
m
u
n
icatio
n
s
y
s
tem
s
with
s
tan
d
ar
d
ized
p
r
o
to
c
o
ls
th
at
all
o
w
th
e
ex
c
h
an
g
e
o
f
s
h
o
r
t
te
x
t
m
ess
ag
es
b
etwe
en
f
ix
ed
lin
e
o
r
m
o
b
ile
p
h
o
n
e
d
e
v
ices.
An
esti
m
ated
2
3
-
b
illi
o
n
SMS
i
s
s
en
t
d
ai
ly
wo
r
ld
wid
e
in
2
0
1
4
;
W
h
ile,
a
to
tal
o
f
8
.
3
tr
illi
o
n
SMS
was
s
en
t
wo
r
ld
wid
e
in
th
e
s
am
e
y
ea
r
with
n
et
m
ar
k
et
r
ev
e
n
u
e
o
f
o
v
er
$
1
2
8
B
illi
o
n
in
2
0
1
1
.
I
n
2
0
1
6
,
th
e
r
ev
en
u
e
was
f
o
r
ec
asted
to
b
e
o
v
er
$
1
5
3
B
illi
o
n
;
An
d
,
ev
id
en
ce
h
as
s
h
o
wn
th
at
3
.
3
9
b
illi
o
n
SMS
was
s
en
t
an
d
r
ec
eiv
ed
i
n
Nig
er
ia
alo
n
e
i
n
th
e
y
ea
r
2
0
1
3
[
1
]
.
T
h
e
in
cr
ea
s
ed
p
o
p
u
lar
ity
an
d
c
o
n
s
eq
u
en
t
p
r
o
life
r
atio
n
o
f
SMS
p
latf
o
r
m
s
,
h
as
also
s
ee
n
a
co
r
r
esp
o
n
d
i
n
g
r
is
e
in
u
n
s
o
licited
SMS
ca
lled
s
p
am
s
.
T
h
e
I
T
U
2
0
0
5
ca
m
p
aig
n
witn
ess
ed
a
r
is
e
in
th
e
u
n
s
o
licited
co
m
m
e
r
cial
ad
v
er
ts
as
s
en
t
to
m
o
b
ile
p
h
o
n
es
v
ia
SMS.
R
ec
en
t
d
r
if
t
f
r
o
m
em
ail
to
SMS
s
p
am
s
is
attr
ib
u
ted
to
th
e
av
ailab
ilit
y
o
f
ef
f
ec
tiv
e
em
ail
f
ilter
s
,
u
s
er
awa
r
en
ess
an
d
in
d
u
s
tr
y
co
llab
o
r
atio
n
[3
-
6]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
9
,
No
.
1,
Ap
r
il
20
20
:
9
–
1
8
10
Sp
am
s
ar
e
u
n
s
o
licited
elec
tr
o
n
ic
m
ess
ag
es
th
at
in
clu
d
e
,
an
d
n
o
t
lim
ited
to
,
e
m
ails
,
SMS,
Vo
ice
o
v
er
I
P (
Vo
I
P)
an
d
in
s
tan
t
m
ess
ag
in
g
f
r
o
m
ch
ats.
Sp
am
s
ar
e
u
n
s
o
licited
o
r
u
n
wan
ted
m
ess
ag
es f
r
o
m
a
s
en
d
er
,
s
en
t
in
d
is
cr
im
in
ately
with
n
o
p
r
io
r
r
elatio
n
s
h
ip
to
a
u
s
er
m
o
s
tly
f
o
r
co
m
m
er
cial
r
ea
s
o
n
s
[7
-
8]
.
SMS
Sp
am
s
r
an
g
e
s
f
r
o
m
a
d
u
lt
-
th
em
e
d
an
d
in
ap
p
r
o
p
r
iate
co
n
ten
ts
,
u
n
s
o
licited
a
d
v
er
ts
,
s
m
is
h
in
g
an
d
m
o
b
ile
m
alwa
r
e
etc.
SMS
s
p
am
s
h
av
e
s
in
ce
b
ec
o
m
e
en
o
r
m
o
u
s
ch
allen
g
e
–
ca
u
s
in
g
g
r
ea
t
lo
s
s
o
f
r
ev
en
u
e
to
I
n
ter
n
et
Ser
v
ice
Pro
v
id
er
s
,
Mo
b
ile
Netwo
r
k
Op
er
ato
r
s
an
d
u
s
er
s
in
g
e
n
er
al.
On
o
v
er
a
ll
,
s
p
am
s
g
r
ew
b
y
3
0
0
%
f
r
o
m
ju
s
t
2
0
1
1
t
o
2
0
1
2
f
r
o
m
m
illi
o
n
s
o
f
SMS
r
ec
eiv
e
d
wo
r
ld
wid
e;
An
d
3
3
.
3
%
attr
i
b
u
ted
to
s
p
am
-
r
elate
d
m
ess
ag
es
[
2
,
8
]
.
In
Nig
er
ia
alo
n
e,
an
esti
m
ated
3
3
4
,
8
5
7
,
6
8
5
SMS
s
p
am
s
wer
e
r
ec
eiv
ed
d
aily
in
2
0
1
5
.
I
m
p
ly
in
g
t
h
at
lo
ts
o
f
m
o
b
ile
p
h
o
n
e
u
s
er
s
ar
e
h
an
d
icap
p
ed
in
th
e
co
n
tr
o
l
o
f
th
e
n
u
m
b
er
o
f
s
p
am
s
th
ey
r
ec
eiv
e
[
9
,
10
]
.
B
esid
es
b
ein
g
d
is
tr
ac
tiv
e
an
d
an
n
o
y
in
g
,
u
s
er
s
n
ee
d
a
ce
r
tain
d
eg
r
ee
o
f
p
r
iv
ac
y
wit
h
th
eir
p
h
o
n
es
an
d
f
r
ee
f
r
o
m
Sp
am
an
d
v
ir
u
s
es
in
v
asio
n
s
[
1
1
-
1
4
]
.
Mo
b
ile
n
et
wo
r
k
o
p
er
at
o
r
s
ar
e
g
ea
r
ed
to
war
d
s
r
ed
u
cin
g
th
e
n
u
m
b
er
o
f
s
p
am
s
o
v
er
th
eir
n
etwo
r
k
as su
ch
f
l
o
o
d
in
g
m
ak
es th
e
SMS c
h
an
n
el
m
o
r
e
i
n
v
a
s
iv
e
an
d
less
s
ec
u
r
e
[
1
5
-
17]
.
T
h
e
tr
em
en
d
o
u
s
r
is
e
in
th
e
u
s
ag
e
o
f
SMS
is
attr
ib
u
ted
to
[
1
6
,
1
-
2]
:
a
)
T
r
u
s
t
in
SMS
ch
an
n
el:
SMS
i
s
a
p
r
iv
ate
co
m
m
u
n
icatio
n
b
etw
ee
n
two
p
ar
ties
o
n
ly
h
as
cr
ea
ted
s
o
m
e
lev
el
o
f
tr
u
s
t
an
d
ac
c
ep
tan
ce
all
o
v
er
th
e
wo
r
ld
s
u
ch
th
at
f
in
an
cial
in
s
titu
tio
n
h
as
ad
o
p
ted
its
u
s
e
in
p
ay
m
en
t
au
th
o
r
izatio
n
.
b
)
Hig
h
o
p
en
r
ate:
Av
er
ag
e
tim
e
it
tak
es
to
r
esp
o
n
d
to
SMS
is
f
aster
th
an
em
ail
an
d
v
o
ice
ca
ll
–
m
ak
in
g
it
a
p
r
ef
er
r
ed
ch
o
ice.
Statis
tic
s
h
av
e
s
h
o
wn
SMS h
as a
n
a
v
er
a
g
e
o
p
en
r
ate
o
f
9
9
% a
n
d
o
p
e
n
s
with
in
1
5
-
m
in
u
tes;
W
h
ile,
a
n
em
ail
h
as a
n
o
p
en
r
ate
o
f
2
0
-
2
5
%
an
d
o
p
e
n
wi
th
2
4
-
h
o
u
r
s
.
c)
L
o
w
co
s
t
o
f
tr
an
s
ac
tio
n
:
Av
er
ag
e
c
o
s
t
p
er
SMS
is
alm
o
s
t
n
eg
lig
ib
le,
an
d
f
r
ee
f
o
r
s
o
m
e
n
etwo
r
k
s
–
af
f
o
r
d
in
g
m
o
b
il
e
p
h
o
n
e
u
s
er
s
th
e
o
p
p
o
r
tu
n
ity
to
s
en
d
as
m
an
y
with
o
u
t
r
ec
o
u
r
s
e
to
co
s
t.
M
ar
k
eter
s
an
d
m
a
n
y
o
th
er
in
s
titu
tio
n
h
as
em
b
r
ac
e
b
u
lk
S
MS
a
m
ed
iu
m
f
o
r
ad
v
er
tis
in
g
an
d
in
ter
ac
t
with
cu
s
to
m
er
s
.
d
)
E
ase
an
d
C
o
n
v
en
ien
ce
o
f
tex
tin
g
e
n
ab
les
its
u
s
e
in
n
ea
r
ly
ev
er
y
en
v
ir
o
n
m
en
t
with
o
u
t
d
is
r
u
p
ti
n
g
p
eo
p
le
ar
o
u
n
d
p
h
o
n
e
u
s
er
s
;
Un
lik
e
in
v
o
ice
ca
ll,
SMS
ca
n
b
e
in
ab
s
o
lu
te
s
ilen
ce
with
o
u
t
in
co
n
v
en
ien
cin
g
p
eo
p
le
ar
o
u
n
d
.
Aid
ed
b
y
th
e
p
o
r
tab
le
s
ize
o
f
m
o
s
t
m
o
b
ile
d
ev
ices,
co
m
m
u
n
icatio
n
ca
n
b
e
d
o
n
e
al
m
o
s
t e
v
er
y
wh
er
e
an
d
a
n
y
p
o
s
itio
n
.
SMS
h
as
g
r
ea
t
b
en
ef
it
f
o
r
b
o
th
s
u
b
s
cr
ib
er
s
an
d
o
p
e
r
ato
r
s
in
d
iv
er
s
e
way
s
ce
n
ter
e
d
o
n
c
o
n
v
en
ien
ce
,
f
lex
ib
ilit
y
,
s
ea
m
less
in
teg
r
atio
n
o
f
m
ess
ag
in
g
s
er
v
ices
an
d
d
ata
ac
ce
s
s
.
Oth
er
s
m
ay
in
clu
d
e
[1
-
2]
:
(
a)
d
eliv
e
r
y
o
f
n
o
tific
atio
n
s
,
(
b
)
g
u
ar
an
tee
d
d
eliv
er
y
,
(
c
)
r
eliab
le,
lo
w
-
co
s
t
f
o
r
co
n
cise
d
ata,
(
d
)
ab
ilit
y
to
s
cr
ee
n
m
ess
ag
es
an
d
r
et
u
r
n
ca
lls
,
(
e)
in
c
r
ea
s
es
p
r
o
d
u
ctiv
ity
,
(
f
)
m
o
r
e
s
o
p
h
is
ticated
f
u
n
ctio
n
ality
p
r
o
v
i
d
es
en
h
an
ce
d
u
s
er
b
en
ef
its
,
(
g
)
d
eliv
er
y
to
m
u
lt
ip
le
u
s
er
s
at
s
am
e
tim
e,
(
h
)
ab
ilit
y
to
r
ec
eiv
e
d
iv
er
s
e
in
f
o
r
m
atio
n
,
(
i)
e
-
m
ail
g
en
er
atio
n
,
(
l)
c
r
ea
tio
n
o
f
u
s
er
g
r
o
u
p
s
,
(
m
)
in
teg
r
atio
n
with
o
th
er
d
ata
a
n
d
I
n
ter
n
et
-
b
ased
ap
p
licatio
n
,
an
d
(
n
)
in
cr
ea
s
e
in
r
ev
en
u
e
f
o
r
m
o
b
ile
n
et
wo
r
k
o
p
er
ato
r
s
(
MN
O
)
.
2.
SM
S,
SPA
M
S
AND
F
I
L
T
E
RS
T
h
e
tr
em
e
n
d
o
u
s
r
is
e
in
th
e
u
s
ag
e
o
f
SMS
is
attr
ib
u
ted
to
its
ea
s
e
o
f
u
s
e,
u
b
iq
u
ity
in
n
atu
r
e,
h
ig
h
o
p
en
r
ates,
lo
w
co
s
t
o
f
tr
an
s
ac
tio
n
an
d
in
h
er
e
n
t
tr
u
s
t
in
th
e
ch
an
n
el.
T
h
e
ea
s
e
o
f
u
s
e,
p
o
r
tab
ilit
y
,
u
b
iq
u
ity
,
lo
w
o
p
en
r
ate
an
d
lo
w
co
s
t
o
f
SMS
ar
e
m
ajo
r
f
ac
to
r
s
f
o
r
its
p
o
p
u
lar
ity
an
d
u
s
ag
e.
T
h
is
g
r
o
wth
r
ate
h
as
eq
u
ally
attr
ac
ted
s
p
a
m
m
in
g
to
th
e
ch
an
n
el.
Sp
am
m
e
r
s
ar
e
w
ell
o
r
g
an
ized
b
u
s
in
ess
es
s
ee
k
in
g
to
m
ak
e
m
o
n
ey
th
r
o
u
g
h
th
e
u
s
e
o
f
em
ail,
m
o
b
ile
(
SMS)
,
I
n
s
tan
t
m
ess
ag
e,
UseNet
n
ewsg
r
o
u
p
,
So
cial
n
e
two
r
k
an
d
in
ter
n
et
telep
h
o
n
y
ch
a
n
n
el
with
o
u
t
th
e
co
n
s
en
t
o
f
s
u
b
s
cr
ib
er
(
u
s
er
)
.
T
h
eir
m
er
ch
an
d
is
e
ar
e
u
n
s
o
licited
ad
v
er
tis
in
g
,
in
ap
p
r
o
p
r
iate
o
r
ad
u
lt
-
th
e
m
ed
co
n
ten
t,
p
r
em
iu
m
f
r
au
d
,
s
m
is
h
in
g
an
d
ev
e
n
d
is
tr
ib
u
tio
n
o
f
m
alwa
r
e
g
en
er
ally
ca
lled
s
p
am
.
SMS
s
p
am
s
ar
e
th
u
s
,
u
n
s
o
licited
an
d
u
n
wan
t
ed
m
ess
ag
es
s
en
t
to
m
o
b
ile
p
h
o
n
e
u
s
er
s
.
Sp
am
tr
en
d
is
o
n
th
e
r
is
e
an
d
its
to
ll
o
n
s
u
b
s
cr
ib
er
s
an
d
ev
en
M
NO
is
g
ettin
g
in
ten
s
iv
e
an
d
p
r
o
v
en
to
b
e
o
f
g
r
ea
t
co
n
ce
r
n
t
o
al
l
[
1
8
-
20
]
.
2
.
1
.
Sp
a
m
s
:
s
o
urce
s
a
nd
co
ns
equent
s
SMS
s
p
am
is
g
en
er
ates
f
r
o
m
v
ar
io
u
s
s
o
u
r
ce
s
;
o
n
e
o
f
th
e
ty
p
ical
s
p
am
s
o
u
r
ce
s
is
n
u
m
b
er
h
ar
v
esti
n
g
,
wh
ich
is
ca
r
r
ied
o
u
t
b
y
I
n
te
r
n
et
s
ites
o
f
f
er
in
g
“
f
r
ee
”
s
er
v
ices.
E
n
d
u
s
er
s
ca
n
also
r
ec
eiv
e
m
o
b
ile
s
p
am
f
r
o
m
th
e
f
o
llo
win
g
s
o
u
r
ce
s
[
1
2
]
:
−
Or
g
an
izatio
n
s
an
d
in
d
iv
id
u
als
th
at
p
ay
MN
O
to
d
eliv
er
SM
S
to
th
e
s
u
b
s
cr
ib
er
s
:
T
h
ey
ar
e
r
esp
o
n
s
ib
le
f
o
r
th
e
h
ig
h
est
n
u
m
b
er
o
f
s
p
am
r
ec
eiv
ed
o
n
s
u
b
s
cr
ib
er
’
s
m
o
b
il
e
p
h
o
n
es.
Alth
o
u
g
h
,
MN
Os
h
av
e
ad
o
p
ted
an
d
en
f
o
r
ce
d
u
s
e
o
f
o
p
t
-
o
u
t,
o
r
ev
e
n
o
p
t
-
in
p
r
o
ce
s
s
es f
o
r
th
e
u
s
er
to
s
to
p
r
ec
eiv
in
g
p
r
o
m
o
s
o
r
ad
s
.
−
Or
g
an
izatio
n
s
th
at
d
o
n
o
t
p
ay
f
o
r
th
e
SMS
th
at
ar
e
d
eliv
er
ed
to
th
e
s
u
b
s
cr
ib
er
s
:
th
ey
ar
e
u
s
u
ally
wo
r
s
e
an
d
co
n
s
id
er
ed
as f
r
a
u
d
b
ec
a
u
s
e
it d
am
ag
es M
NO
b
r
an
d
s
.
−
I
n
d
iv
id
u
al
o
r
ig
i
n
ated
m
ess
ag
e
s
th
at
d
is
tu
r
b
r
ec
ip
ien
ts
.
Ap
ar
t
f
r
o
m
th
e
d
i
s
tr
ac
tin
g
a
n
d
an
n
o
y
in
g
ef
f
ec
ts
o
f
s
p
am
,
th
er
e
ar
e
o
th
er
s
er
io
u
s
co
n
s
eq
u
en
ce
s
g
en
er
ated
.
T
h
e
r
e
is
th
e
is
s
u
e
o
f
co
m
p
etitio
n
f
o
r
r
eso
u
r
ce
s
b
etwe
en
m
illi
o
n
s
o
f
illeg
iti
m
ate
an
d
leg
itima
te
m
ess
ag
es
b
ein
g
tr
an
s
m
itted
.
T
h
ese
m
ess
ag
es
co
n
s
u
m
e
n
etwo
r
k
r
eso
u
r
ce
s
th
a
t
co
u
ld
h
a
v
e
o
th
er
wis
e
b
ee
n
allo
ca
ted
to
o
th
er
le
g
itima
te
s
er
v
ices
b
y
MN
O
[
1
5
]
.
Sp
am
m
in
g
ac
tiv
ities
attr
ac
ts
ex
tr
a
co
s
t
f
o
r
m
o
b
ile
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Memetic
a
lg
o
r
ith
m
fo
r
s
h
o
r
t
mess
a
g
in
g
s
ervice
s
p
a
m
filt
er
u
s
in
g
text
…
(
A
r
n
o
ld
A
d
ima
b
u
a
Oju
g
o
)
11
o
p
er
ato
r
s
t
o
ad
eq
u
ately
m
ain
t
ain
an
d
s
er
v
ice
t
h
eir
m
o
b
ile
c
o
m
m
u
n
icatio
n
in
f
r
astru
ctu
r
es
f
o
r
e
f
f
ec
tiv
e
s
er
v
ice
d
eliv
er
y
.
Als
o
f
lo
o
d
in
g
o
f
M
NO
in
f
r
astru
ctu
r
e
with
illeg
itima
te
m
ass
ag
es
ca
n
ca
u
s
e
leg
it
im
ate
u
s
er
s
to
s
u
f
f
er
d
en
ial
o
f
s
er
v
ice.
Hu
g
e
am
o
u
n
t
o
f
s
p
am
m
ess
ag
es
also
co
n
ce
r
n
s
th
e
ce
llu
la
r
ca
r
r
ier
s
as
t
h
e
m
ess
ag
es
tr
av
er
s
e
th
r
o
u
g
h
th
e
n
etwo
r
k
,
ca
u
s
i
n
g
co
n
g
esti
o
n
an
d
h
en
ce
d
e
g
r
ad
e
n
etwo
r
k
p
er
f
o
r
m
an
ce
[
1
6
]
.
Mo
b
ile
co
m
m
u
n
icatio
n
in
d
u
s
tr
ies
ar
e
also
f
ac
ed
with
th
r
ea
t
f
r
o
m
v
ir
u
s
,
T
r
o
ja
n
h
o
r
s
e,
wo
r
m
s
an
d
m
alwa
r
e
p
r
o
p
a
g
ated
b
y
s
p
am
SMS
[
1
5
]
.
Fra
u
d
u
len
t
m
ess
ag
in
g
ac
t
iv
ities
s
u
ch
as
p
h
is
h
in
g
id
en
t
ity
th
ef
t
an
d
o
th
er
f
r
au
d
r
elate
d
ac
ti
v
ities
wh
ich
wer
e
p
r
o
m
in
en
t
in
em
ail
m
ess
ag
in
g
s
er
v
ices
h
as
m
ig
r
ate
d
to
SMS
p
latf
o
r
m
[
1
7
,
18]
.
Fin
an
cial
lo
s
s
,
d
am
a
g
e
to
m
o
b
ile
u
s
er
’
s
r
ep
u
tatio
n
an
d
t
h
at
o
f
th
e
MN
O
ar
e
is
s
u
es to
b
e
co
n
s
id
er
ed
[
1
9
]
.
2
.
2
.
Sp
a
m
f
ilte
rs
SMS
s
p
am
f
ilter
s
s
h
ar
es
s
im
ilar
f
ea
tu
r
es
an
d
ch
allen
g
es
with
em
ail
s
p
am
f
ilter
s
.
T
h
ey
ar
e
b
o
th
s
ad
d
led
with
th
e
task
o
f
r
ea
l
-
tim
e
f
ilter
in
g
ef
f
icien
cy
an
d
t
h
e
o
p
tio
n
t
o
d
ec
i
d
e
b
etwe
en
clien
t
-
s
id
e
an
d
o
r
s
er
v
er
-
s
id
e
f
ilter
in
g
.
T
h
e
m
o
b
ile
s
p
ac
e
is
also
f
ac
ed
with
th
e
ch
allen
g
e
o
f
o
v
er
co
m
in
g
m
i
s
class
if
icatio
n
co
s
t
an
d
elim
in
ate
f
alse
-
p
o
s
itiv
es (
g
en
u
in
e
SMS in
co
r
r
ec
tly
class
if
ied
as sp
am
b
y
f
ilter
)
,
an
d
is
s
u
e
o
f
co
n
ce
p
t d
r
if
t
in
o
r
d
er
t
o
ev
ad
e
f
ilter
s
.
T
h
u
s
,
m
o
s
t
ex
is
tin
g
ap
p
r
o
ac
h
es
o
f
co
m
b
atin
g
SMS
s
p
am
ar
e
im
p
o
r
ted
f
r
o
m
s
u
cc
ess
f
u
l
em
ail
-
s
o
lu
ti
ons
[2
1
,
2
2
]
.
N
o
t
all
s
o
lu
tio
n
s
to
em
ai
l
s
p
am
ar
e
ap
p
licab
le
to
SMS
d
u
e
to
th
e
f
ac
t
th
at
estab
lis
h
ed
em
ail
s
p
am
f
ilter
s
ar
e
u
n
ab
le
to
tack
le
SMS
Sp
am
b
ec
au
s
e
p
e
r
f
o
r
m
an
ce
o
f
e
m
ail
s
p
am
f
ilter
s
is
s
er
io
u
s
ly
d
eg
r
ad
ed
wh
e
n
u
s
ed
to
f
ilter
SM
S
s
p
am
.
T
h
is
is
at
tr
ib
u
ted
to
its
lim
ited
1
6
0
-
ch
ar
ac
ter
o
f
1
4
0
-
b
y
tes
s
ized
m
ess
ag
es.
Als
o
,
th
ese
m
ess
ag
es
ar
e
r
if
e
with
s
lan
g
s
,
s
y
m
b
o
ls
,
em
o
tico
n
s
a
n
d
a
b
b
r
e
v
iatio
n
s
th
at
in
h
ib
it
p
r
o
p
er
class
if
icatio
n
[2
3
-
2
4
]
.
T
o
o
v
er
co
m
e
th
e
s
h
o
r
tf
all
o
f
em
ail
f
ilter
s
in
h
an
d
lin
g
SMS
s
p
am
s
u
c
ce
s
s
f
u
lly
,
a
co
m
b
in
ed
f
ilter
in
g
tech
n
i
q
u
e
t
o
r
ed
u
ce
n
o
is
e
in
SMS
an
d
ex
p
an
d
s
th
e
m
ess
ag
e
s
ize
[2
5
,
2
6
]
–
is
th
e
f
o
cu
s
o
f
th
is
r
esear
ch
.
Sp
am
f
ilter
s
ca
n
b
e
d
i
v
id
ed
i
n
to
a
n
u
m
b
e
r
o
f
b
r
o
a
d
ca
teg
o
r
ies
b
ased
o
n
t
h
e
m
eth
o
d
u
s
ed
to
f
ilter
Sp
am
.
T
h
e
y
in
clu
d
e
[2
7
]
:
lis
t
b
ased
,
ch
allen
g
e/r
esp
o
n
s
e
s
y
s
tem
,
co
n
te
n
t
b
ased
,
co
llab
o
r
ativ
e
an
d
Heu
r
is
tics
B
a
s
ed
f
ilter
s
.
2
.
3
.
Cha
lleng
e
-
re
s
po
ns
e
f
ilte
rs
T
h
is
f
ilter
f
o
r
ce
s
a
m
ess
ag
e
s
en
d
er
to
p
r
o
v
e
th
e
y
ar
e
h
u
m
an
v
ia
s
o
m
e
test
.
T
h
is
f
ilter
b
lo
ck
s
u
n
d
esira
b
le
m
ess
ag
es
b
y
f
o
r
c
in
g
th
e
s
en
d
er
to
p
er
f
o
r
m
a
t
ask
b
ef
o
r
e
th
eir
m
ess
ag
e
is
d
eliv
er
ed
.
W
ith
task
s
u
cc
ess
,
th
e
m
es
s
ag
e
(
an
d
f
u
tu
r
e
m
ess
ag
es)
will
b
e
d
eliv
er
ed
to
th
e
r
ec
ip
ien
t;
W
h
ile,
f
ail
u
r
e
to
co
m
p
lete
th
e
ch
allen
g
e
af
ter
a
ce
r
tain
tim
e
p
er
io
d
,
lead
s
to
m
ess
ag
e
r
ejec
tio
n
[2
4
]
.
T
h
e
m
o
s
t c
o
m
m
o
n
c
h
allen
g
e
co
n
s
is
ts
o
f
d
is
to
r
ted
im
ag
es
an
d
te
x
t.
T
o
tr
iu
m
p
h
th
is
ch
allen
g
e,
a
u
s
er
m
u
s
t
ty
p
e
tex
t
o
r
ar
r
an
g
e
im
a
g
es
co
r
r
ec
tly
.
W
ith
ch
allen
g
e/r
esp
o
n
s
e
f
alse
p
o
s
itiv
es
ca
n
b
e
r
ed
u
ce
d
to
b
ar
est
m
in
im
u
m
.
An
o
th
er
m
er
it
o
f
th
is
ap
p
r
o
ac
h
is
in
its
lo
w
s
y
s
tem
r
eso
u
r
ce
r
eq
u
ir
e
m
en
ts
,
s
in
ce
n
o
C
PU
-
in
ten
s
i
v
e
p
atter
n
m
atch
i
n
g
is
r
eq
u
i
r
ed
.
Ho
wev
er
,
th
is
ap
p
r
o
ac
h
ca
u
s
es
m
o
r
e
p
r
o
b
le
m
s
th
an
it
s
o
lv
es.
Fo
r
in
ex
p
e
r
i
en
ce
d
o
r
v
is
u
al
h
an
d
icap
p
ed
u
s
er
s
,
th
e
ch
allen
g
es
ar
e
co
m
p
letely
u
n
s
o
lv
ab
le.
R
e
g
u
lar
u
s
er
s
ar
e
p
r
o
v
o
k
ed
b
y
t
h
e
ch
allen
g
es
an
d
c
h
o
o
s
e
n
o
t
to
d
o
s
o
s
in
ce
th
ey
v
iew
it
as
an
u
n
ac
ce
p
tab
le
i
r
r
itatio
n
.
Als
o
,
au
to
m
ate
d
e
m
ail
th
at
a
u
s
er
wo
u
ld
wa
n
t
to
r
ec
eiv
e
(
tr
a
v
el
co
n
f
ir
m
atio
n
s
,
o
n
lin
e
p
u
r
c
h
ase
r
ec
eip
ts
,
etc)
ar
e
tr
a
p
p
ed
b
y
th
is
ap
p
r
o
ac
h
an
d
n
ev
er
d
eliv
e
r
ed
[2
8
-
30
]
.
2
.
4
.
L
is
t
-
ba
s
ed
f
ilte
rs
−
B
lack
lis
t:
T
h
is
ea
r
lie
s
t
s
p
am
-
f
ilter
in
g
m
eth
o
d
s
ee
k
s
to
b
lo
ck
u
n
wan
ted
m
ess
ag
es
f
r
o
m
an
alr
ea
d
y
cr
ea
ted
lis
t
o
f
s
en
d
er
s
.
B
lack
lis
t
s
ar
e
r
ec
o
r
d
s
o
f
em
ail
ad
d
r
ess
es,
I
n
ter
n
et
Pro
to
co
l
(
I
P)
ad
d
r
e
s
s
es
an
d
p
h
o
n
e
n
u
m
b
er
s
th
at
h
av
e
b
ee
n
p
r
e
v
io
u
s
ly
u
s
ed
to
s
en
d
s
p
am
.
W
h
en
in
co
m
in
g
m
ess
ag
e
ar
r
iv
es,
s
p
am
f
ilter
ch
ec
k
s
if
I
P
,
e
m
ail
ad
d
r
ess
o
r
p
h
o
n
e
n
u
m
b
er
is
o
n
a
b
lack
li
s
t.
I
f
s
o
,
th
e
m
ess
ag
e
is
co
n
s
i
d
er
ed
s
p
am
an
d
r
ejec
ted
.
B
lack
lis
ts
en
s
u
r
e
k
n
o
wn
s
p
am
m
er
s
ca
n
n
o
t
r
ea
ch
u
s
er
s
'
in
b
o
x
es.
T
h
eir
o
n
ly
d
e
m
er
it
is
th
at
th
ey
ca
n
also
m
is
id
en
tify
leg
itima
t
e
s
en
d
er
s
as sp
am
m
er
s
[2
4
,
29
]
]
.
−
W
h
itelis
t:
T
o
b
lo
ck
s
p
am
s
,
wh
itelis
t
r
ath
er
th
an
s
p
ec
if
y
s
en
d
er
s
to
b
lo
c
k
m
ess
ag
es
f
r
o
m
,
it
s
p
ec
if
ies
wh
ich
s
en
d
er
s
to
allo
w
m
ess
ag
es
f
r
o
m
.
T
h
ese
a
d
d
r
ess
es
ar
e
s
to
r
ed
in
tr
u
s
ted
-
u
s
er
s
lis
t.
M
o
s
t
s
p
am
f
ilter
s
u
s
es
a
wh
itelis
t
alo
n
g
s
id
e
o
th
er
tech
n
iq
u
es
to
cu
t
d
o
wn
o
n
th
e
n
u
m
b
e
r
o
f
g
en
u
in
e
SMS
t
h
at
ac
cid
en
tally
g
et
f
lag
g
ed
as
s
p
am
.
A
f
ilter
th
at
u
s
es
ju
s
t
wh
itel
is
t
im
p
li
es
th
at
an
y
o
n
e
n
o
t
a
p
p
r
o
v
ed
is
au
to
m
atica
lly
b
lo
ck
ed
.
So
m
e
a
n
ti
-
s
p
am
s
u
s
e
a
wh
itelis
t
v
ar
iatio
n
ca
lled
au
to
m
atic
wh
itelis
t.
Her
e,
an
u
n
k
n
o
wn
s
en
d
er
ad
d
r
ess
is
ch
ec
k
ed
ag
ain
s
t
a
d
atab
ase;
if
th
ey
h
av
e
n
o
h
is
to
r
y
o
f
s
p
am
m
in
g
–
th
eir
m
ess
ag
e
is
d
eliv
er
ed
to
th
e
r
ec
ip
ien
t'
s
in
b
o
x
an
d
ad
d
e
d
to
th
e
wh
itelis
t
[2
4
,
2
9
]
.
−
Gr
ey
lis
t:
T
h
is
f
ilter
wo
r
k
s
wi
t
h
th
e
ass
u
m
p
tio
n
th
at
m
o
s
t
s
p
am
m
er
s
s
en
d
s
b
atch
o
f
m
ess
ag
es
o
n
ce
.
W
h
en
m
ess
ag
e
f
r
o
m
u
n
k
n
o
wn
a
d
d
r
e
s
s
is
r
ec
eiv
ed
,
it
b
lo
ck
s
an
d
r
e
v
er
t
a
f
ailu
r
e
d
eliv
e
r
y
to
th
e
s
en
d
in
g
s
er
v
e
r
.
I
f
th
e
m
ess
ag
e
is
r
esen
t,
w
h
ich
m
o
s
t
leg
itima
te
s
er
v
e
r
s
d
o
,
f
ilter
r
ec
eiv
es
it
a
n
d
a
d
d
s
th
e
ad
d
r
ess
/p
h
o
n
e
n
u
m
b
er
to
th
e
li
s
t.
Alth
o
u
g
h
o
v
er
h
ea
d
o
f
th
e
f
ilter
is
l
o
w,
its
d
em
er
it
i
s
th
e
u
n
ju
s
t
d
elay
d
eliv
er
y
ex
p
er
ien
ce
d
b
y
g
en
u
i
n
e
m
ess
ag
es to
its
r
ec
ip
ien
t
[2
4
, 2
9
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
9
,
No
.
1,
Ap
r
il
20
20
:
9
–
1
8
12
2
.
5
.
Co
nte
nt
-
ba
s
ed
f
ilte
rs
C
o
n
ten
t
-
b
ased
f
ilter
in
g
m
eth
o
d
s
ar
e
b
ased
o
n
th
e
e
v
alu
atio
n
o
f
i
n
d
iv
id
u
al
wo
r
d
s
o
r
p
h
r
ase
s
f
o
u
n
d
in
th
e
m
ail/m
ess
ag
e
to
d
eter
m
in
e
if
m
ess
ag
e
is
s
p
am
o
r
n
o
t.
T
h
is
m
eth
o
d
a
n
aly
ze
s
m
ess
ag
e
h
ea
d
er
,
s
u
b
ject
an
d
b
o
d
y
to
d
is
co
v
er
an
y
d
is
tin
ctiv
e
ch
ar
ac
ter
is
tic
[
30
]
.
T
h
ey
ar
e
f
u
r
th
er
class
if
ied
in
to
wo
r
d
-
b
ased
an
d
h
e
u
r
is
tic
f
ilter
s
.
W
o
r
d
-
b
ased
f
i
lter
s
u
s
e
a
s
et
o
f
r
u
les
to
d
etec
t
g
en
u
i
n
e
f
r
o
m
s
p
am
SMS.
Als
o
k
n
o
wn
as
r
u
le
-
f
ilter
s
,
th
ey
u
s
e
r
u
les
ab
o
u
t
ac
tu
al
wo
r
d
(
s
)
o
r
p
h
r
ase(
s
)
i
n
a
m
ess
ag
e
to
class
if
y
m
ess
ag
es
in
to
g
en
u
i
n
e
an
d
s
p
a
m
class
es.
R
u
le
f
ea
tu
r
es
in
clu
d
e
wo
r
d
t
y
p
e,
f
r
e
q
u
en
c
y
o
f
o
cc
u
r
r
en
ce
,
s
tr
u
ct
u
r
e
o
f
tex
t (
e.
g
.
f
o
n
t
s
ize,
co
l
o
u
r
etc)
,
p
r
esen
ce
o
f
m
an
y
p
er
io
d
s
b
et
wee
n
letter
s
(
e.
g
.
F.R
.
E
.
E
)
,
ex
is
ten
ce
o
f
im
ag
e,
etc.
R
u
les
ar
e
f
ilter
-
d
ep
en
d
en
t
an
d
ca
n
v
a
r
y
f
r
o
m
s
im
p
le
to
v
er
y
co
m
p
lex
.
A
d
em
e
r
it
o
f
r
u
le
-
b
ased
f
ilter
s
is
th
at:
(
a)
th
ey
ar
e
k
n
o
wled
g
e
in
ten
s
iv
e,
(
b
)
tim
e
co
n
s
u
m
in
g
p
r
o
ce
s
s
in
r
ev
iewin
g
s
p
am
m
ess
ag
es
to
d
eter
m
in
e
t
h
e
r
u
les,
an
d
(
c
)
n
ee
d
s
r
eg
u
lar
u
p
d
ate
o
f
r
u
les as sp
am
m
er
s
ch
an
g
es th
eir
tactics
[3
1
-
3
4
]
.
C
o
n
v
er
s
ely
,
h
eu
r
is
tic
-
b
ased
f
ilter
ex
am
in
es
m
ess
ag
e
co
n
ten
t
th
r
o
u
g
h
v
ar
io
u
s
alg
o
r
i
th
m
s
an
d
r
eso
u
r
ce
s
,
an
d
ass
ig
n
s
p
o
in
ts
to
wo
r
d
s
o
r
p
h
r
ases
.
W
o
r
d
s
co
m
m
o
n
ly
f
o
u
n
d
i
n
s
p
am
s
s
u
ch
as
“FR
E
E
”
o
r
"SE
X,
"
r
ec
eiv
e
h
ig
h
e
r
s
co
r
es.
T
er
m
s
co
m
m
o
n
ly
f
o
u
n
d
in
n
o
r
m
al
m
ess
a
g
es
r
ec
eiv
e
lo
we
r
s
co
r
es.
T
h
e
f
ilter
th
en
ad
d
s
u
p
t
o
tal
s
co
r
es.
I
f
th
e
m
ess
ag
e
r
ec
eiv
es
a
ce
r
tain
s
co
r
e
o
r
h
ig
h
er
(
d
ete
r
m
in
ed
b
y
a
n
ti
-
s
p
am
ap
p
licatio
n
'
s
ad
m
in
is
tr
ato
r
)
,
t
h
e
f
ilter
id
en
tifie
s
it
as
s
p
am
an
d
b
lo
ck
s
it.
Me
s
s
ag
es
with
s
co
r
e(
s
)
lo
wer
th
an
th
e
t
ar
g
et
n
u
m
b
er
ar
e
d
eliv
er
ed
to
th
e
u
s
e
[3
5
]
.
B
ay
es
ian
f
ilter
,
KNN
class
if
ier
,
Ad
aBo
o
s
t
class
if
ier
,
Gar
y
R
o
b
in
s
o
n
tech
n
iq
u
e,
Su
p
p
o
r
t
Vec
to
r
Ma
ch
i
n
e,
Neu
r
a
l
Netwo
r
k
ar
e
ex
am
p
les
[3
6
]
.
Usi
n
g
a
h
eu
r
is
tic
f
ilter
allo
ws
m
an
y
s
p
am
f
ilter
i
n
g
m
eth
o
d
s
to
b
e
u
s
ed
,
r
e
s
u
ltin
g
in
b
etter
p
er
f
o
r
m
an
ce
th
an
an
y
s
in
g
le
m
eth
o
d
b
y
its
elf
.
3.
SO
F
T
-
CO
M
P
U
T
I
NG
F
RA
M
E
WO
RK
3
.
1
.
B
a
y
esia
n
net
wo
r
k
s
Ar
e
b
ased
o
n
th
e
B
ay
esian
th
eo
r
em
o
f
co
n
d
itio
n
al
p
r
o
b
ab
il
ity
.
T
h
ey
h
av
e
b
ee
n
s
u
cc
ess
f
u
lly
ap
p
lied
to
m
an
y
d
o
m
ain
s
s
u
ch
as
m
ed
icin
e,
m
ac
h
in
e
lea
r
n
in
g
,
s
p
ee
ch
r
ec
o
g
n
itio
n
,
s
ig
n
al
p
r
o
ce
s
s
in
g
,
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
an
d
ce
llu
lar
n
etwo
r
k
s
.
T
h
e
y
a
r
e
an
attr
ac
tiv
e
m
ac
h
in
e
lear
n
i
n
g
tec
h
n
iq
u
e
th
at
r
ep
r
esen
ts
d
o
m
ain
k
n
o
wled
g
e
an
d
d
ata
in
a
n
ele
g
an
t
m
ath
em
atica
l
s
tr
u
ctu
r
e
with
s
im
p
lifie
d
v
i
s
u
al
r
ep
r
esen
tatio
n
.
B
ay
esian
n
et
s
h
o
ws
g
r
ap
h
ic
p
r
o
b
a
b
ilit
y
r
elatio
n
s
h
ip
s
b
e
twee
n
a
s
et
o
f
v
ar
iab
les
u
n
d
er
th
e
d
o
m
ain
o
f
u
n
ce
r
tain
ty
.
T
h
e
y
ar
e
u
s
u
ally
s
tr
u
ctu
r
ed
as
a
d
ir
ec
ted
ac
y
cli
c
g
r
ap
h
an
d
c
o
n
d
itio
n
al
p
r
o
b
a
b
ilit
y
tab
les
(
C
PTs).
C
PT
tab
les
r
ep
r
esen
t
p
r
o
b
ab
ilit
y
o
f
a
r
an
d
o
m
v
ar
iab
le
wh
e
r
e,
g
iv
en
th
e
o
cc
u
r
r
e
n
ce
o
f
it
s
p
ar
en
t
n
o
d
es.
W
e
ca
n
ap
p
ly
s
am
e
co
n
ce
p
t
u
al
s
tr
ateg
y
to
s
p
am
f
ilter
s
[3
7
]
.
B
ay
esian
n
et
class
if
ier
s
ar
e
b
u
ilt
b
ased
o
n
th
e
tr
ain
in
g
d
at
a.
I
ts
b
u
ild
in
g
p
r
o
ce
s
s
in
clu
d
es
s
tr
u
ctu
r
e
lear
n
in
g
,
p
ar
a
m
eter
lear
n
in
g
,
an
d
b
u
ild
in
g
p
r
o
b
ab
ilit
y
d
is
tr
i
b
u
tio
n
tab
les
f
o
r
ea
c
h
n
o
d
e
in
th
e
n
etwo
r
k
.
T
h
er
e
ar
e
two
m
ajo
r
lear
n
in
g
p
r
o
ce
s
s
es
n
am
ely
:
(
a)
s
tr
u
ct
u
r
ed
lear
n
in
g
o
r
ca
s
u
al
d
is
co
v
er
y
in
w
h
ich
n
etwo
r
k
lear
n
s
th
e
s
tr
u
ctu
r
e
an
d
p
ar
a
m
eter
s
with
th
e
p
r
o
v
id
ed
in
p
u
t
d
ata.
T
h
e
ca
u
s
al
d
is
co
v
er
y
aim
s
to
lear
n
th
e
s
tr
u
ctu
r
e
an
d
lear
n
t
h
e
p
ar
a
m
eter
s
.
I
t
ac
h
iev
es
th
is
u
s
in
g
eith
er
o
f
K2
,
Hill
clim
b
in
g
an
d
T
ab
u
-
Sear
ch
;
a
n
d
(
b
)
p
r
o
b
a
b
ilit
y
d
is
tr
ib
u
tio
n
lear
n
in
g
is
ac
h
iev
ed
with
alg
o
r
ith
m
s
lik
e
B
ay
es
Net
esti
m
ato
r
,
B
MA
esti
m
ato
r
an
d
m
u
ltin
o
m
ial
esti
m
ato
r
.
On
ce
s
tr
u
ctu
r
e
lear
n
i
n
g
is
co
m
p
lete,
p
ar
am
eter
lear
n
i
n
g
co
m
p
letes
th
e
C
PT
tab
les
f
o
r
ea
ch
f
ea
tu
r
e
in
th
e
B
ay
esian
Netwo
r
k
.
T
h
e
n
etwo
r
k
d
esig
n
in
f
ig
1
is
f
o
r
d
etec
tin
g
tex
ts
in
SMS
an
d
h
elp
in
g
th
e
m
o
d
el
an
d
alg
o
r
ith
m
to
c
la
s
s
if
y
th
ese
SM
S
in
to
eith
er
o
f
g
en
u
i
n
e/leg
itima
te
an
d
s
p
am
SMS.
B
ay
esian
n
etwo
r
k
d
esig
n
n
ee
d
s
to
co
n
s
i
d
er
th
e
attr
ib
u
tes,
s
ea
r
c
h
alg
o
r
ith
m
an
d
esti
m
atio
n
alg
o
r
ith
m
s
.
T
h
u
s
,
we
u
s
e
th
e
h
ill
-
clim
b
er
s
ea
r
ch
alg
o
r
ith
m
with
f
iv
e
p
a
r
en
ts
u
s
ed
as
th
e
s
ea
r
ch
alg
o
r
i
th
m
f
o
r
th
is
n
etwo
r
k
with
s
im
p
le
esti
m
ato
r
as a
n
esti
m
ate
o
n
alg
o
r
ith
m
with
th
r
esh
o
ld
v
alu
e
“
0
.
5
”
[3
8
]
.
3.
2
.
G
enet
ic
a
lg
o
rit
h
m
(
G
A)
I
n
s
p
ir
ed
b
y
Dar
win
ia
n
ev
o
lu
tio
n
o
f
s
u
r
v
iv
al
o
f
f
ittes
t,
it
co
n
s
is
ts
o
f
a
c
h
o
s
en
p
o
p
u
latio
n
with
p
o
ten
tial
s
o
lu
tio
n
s
to
a
s
p
ec
if
i
c
task
.
E
ac
h
p
o
ten
tial
s
o
lu
tio
n
is
an
in
d
iv
i
d
u
al
f
o
r
wh
ic
h
o
p
tim
al
is
f
o
u
n
d
u
s
in
g
f
o
u
r
o
p
e
r
ato
r
s
n
a
m
ely
:
in
itialize,
s
elec
t,
cr
o
s
s
o
v
er
an
d
m
u
tat
io
n
[3
9
]
.
I
n
d
iv
id
u
als
with
g
en
es
clo
s
e
to
o
p
tim
al,
is
s
aid
to
b
e
f
it.
Fit
n
ess
f
u
n
ctio
n
d
eter
m
in
es
h
o
w
clo
s
e
a
n
in
d
iv
id
u
al
is
to
o
p
tim
al
s
o
lu
tio
n
.
[
40
-
4
2
].
T
h
e
b
asic
o
p
er
ato
r
s
f
o
r
GA
in
clu
d
e
:
−
I
n
itialize
–
I
n
d
iv
i
d
u
al
d
ata
ar
e
en
co
d
ed
i
n
to
f
o
r
m
s
s
u
itab
le
f
o
r
s
elec
tio
n
.
E
ac
h
e
n
co
d
i
n
g
s
ty
p
e
u
s
ed
h
as
its
m
er
it.
B
in
ar
y
e
n
co
d
i
n
g
s
ar
e
co
m
p
u
tatio
n
ally
m
o
r
e
ex
p
en
s
iv
e.
D
ec
im
al
en
co
d
in
g
h
as
g
r
ea
ter
d
iv
er
s
ity
in
ch
r
o
m
o
s
o
m
e
an
d
g
r
ea
ter
v
a
r
ian
ce
o
f
p
o
o
ls
g
en
er
ated
;
f
l
o
at
-
p
o
in
t
en
co
d
in
g
o
r
its
co
m
b
i
n
atio
n
is
m
o
r
e
ef
f
icien
t
th
an
b
in
ar
y
.
T
h
u
s
,
it
en
co
d
es
as
f
ix
ed
len
g
th
v
ec
t
o
r
s
f
o
r
o
n
e
o
r
m
o
r
e
p
o
o
ls
o
f
d
if
f
er
en
t
ty
p
es.
T
h
e
f
itn
ess
f
u
n
ctio
n
ev
alu
ates
h
o
w
cl
o
s
e
a
s
o
lu
tio
n
is
to
its
o
p
tim
al
–
af
ter
wh
ich
t
h
ey
ar
e
ch
o
s
en
f
o
r
r
ep
r
o
d
u
ctio
n
.
I
f
s
o
lu
tio
n
is
f
o
u
n
d
,
f
u
n
ctio
n
is
g
o
o
d
an
d
s
elec
ted
f
o
r
cr
o
s
s
o
v
er
.
T
h
e
f
itn
ess
f
u
n
ctio
n
is
th
e
o
n
ly
p
a
r
t w
ith
k
n
o
wled
g
e
o
f
t
ask
.
I
f
m
o
r
e
s
o
lu
tio
n
s
ar
e
f
o
u
n
d
,
th
e
h
i
g
h
er
its
f
itn
ess
v
alu
e.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Memetic
a
lg
o
r
ith
m
fo
r
s
h
o
r
t
mess
a
g
in
g
s
ervice
s
p
a
m
filt
er
u
s
in
g
text
…
(
A
r
n
o
ld
A
d
ima
b
u
a
Oju
g
o
)
13
−
Selectio
n
–
b
est
f
it
in
d
iv
id
u
al
s
clo
s
e
to
o
p
tim
al
ar
e
ch
o
s
en
to
m
ate.
T
h
e
lar
g
er
th
e
n
u
m
b
er
o
f
s
elec
ted
,
th
e
b
etter
th
e
ch
an
ce
s
o
f
y
iel
d
in
g
f
itter
in
d
iv
id
u
als.
T
h
is
co
n
tin
u
es
u
n
til
o
n
e
is
ch
o
s
en
,
f
r
o
m
th
e
last
two
/th
r
ee
r
em
ain
in
g
s
o
lu
tio
n
s
,
to
b
ec
o
m
e
s
elec
ted
p
ar
en
ts
t
o
n
ew
o
f
f
s
p
r
in
g
.
Selectio
n
en
s
u
r
es
th
e
f
ittes
t
in
d
iv
id
u
als
ar
e
ch
o
s
en
f
o
r
m
a
tin
g
b
u
t
also
allo
ws
f
o
r
less
f
i
t
in
d
iv
id
u
als
f
r
o
m
th
e
p
o
o
l
an
d
th
e
f
ittes
t
to
b
e
s
elec
ted
.
A
s
elec
tio
n
th
at
o
n
ly
m
ates th
e
f
ittes
t is elitis
t a
n
d
o
f
te
n
lead
s
to
co
n
v
er
g
in
g
a
t lo
ca
l o
p
tim
a.
−
C
r
o
s
s
o
v
er
en
s
u
r
es
b
est
f
it
i
n
d
iv
id
u
al
g
en
es
a
r
e
e
x
ch
an
g
e
d
to
y
ield
a
n
ew,
f
itter
p
o
o
l.
T
h
er
e
ar
e
two
cr
o
s
s
o
v
er
ty
p
e
s
(
d
ep
en
d
s
o
n
e
n
co
d
in
g
ty
p
e
u
s
ed
)
:
(
a)
s
im
p
le
cr
o
s
s
o
v
er
f
o
r
b
in
ar
y
en
c
o
d
ed
p
o
o
l.
I
t
allo
ws
s
in
g
le
-
o
r
m
u
lti
-
p
o
i
n
t
cr
o
s
s
with
all
g
en
es
f
r
o
m
a
p
ar
en
t,
a
n
d
(
b
)
ar
ith
m
etic
cr
o
s
s
o
v
er
allo
ws
n
ew
p
o
o
l
to
b
e
cr
ea
ted
b
y
ad
d
in
g
a
n
in
d
iv
i
d
u
al’
s
p
er
ce
n
ta
g
e
to
an
o
th
er
.
−
Mu
tatio
n
alter
s
ch
r
o
m
o
s
o
m
es
b
y
c
h
an
g
i
n
g
its
g
e
n
es
o
r
its
s
eq
u
en
ce
,
to
en
s
u
r
e
n
ew
p
o
o
l
co
n
v
e
r
g
es
to
g
lo
b
al
m
in
im
a
(
in
s
tead
o
f
lo
c
al
o
p
tim
a)
.
Alg
o
r
ith
m
s
to
p
s
if
o
p
tim
al
is
f
o
u
n
d
,
o
r
af
te
r
n
u
m
b
er
o
f
r
u
n
s
if
n
ew
p
o
o
ls
ar
e
cr
ea
ted
(
th
o
u
g
h
co
m
p
u
tatio
n
ally
ex
p
e
n
s
iv
e)
,
o
r
wh
en
n
o
b
etter
s
o
lu
tio
n
is
f
o
u
n
d
.
Gen
es
m
ay
ch
an
g
e
b
ased
o
n
p
r
o
b
a
b
ilit
y
o
f
m
u
tatio
n
r
ate.
Mu
tatio
n
im
p
r
o
v
es
th
e
m
u
ch
-
n
ee
d
ed
d
iv
er
s
ity
in
r
ep
r
o
d
u
ctio
n
.
C
u
ltu
r
al
GA
is
a
v
ar
ian
ts
o
f
GA
with
a
b
elief
s
p
ac
e
d
ef
in
e
a
s
th
u
s
:
(
a)
No
r
m
ativ
e
(
h
as
s
p
e
cif
i
c
v
alu
e
r
an
g
es
to
wh
ich
a
n
in
d
iv
i
d
u
al
is
b
o
u
n
d
)
,
(
b
)
Do
m
ain
(
h
as
d
ata
ab
o
u
t
task
d
o
m
ain
)
,
(
c)
T
em
p
o
r
al
(
h
as
d
ata
ab
o
u
t
e
v
en
ts
’
s
p
ac
e
is
av
ailab
le)
,
an
d
(
d
)
Sp
atial
(
h
as
to
p
o
g
r
ap
h
ical
d
ata
)
.
I
n
a
d
d
itio
n
,
an
in
f
lu
en
ce
f
u
n
ctio
n
m
ed
iates
b
etwe
en
b
elief
s
p
ac
e
a
n
d
th
e
p
o
o
l
–
to
en
s
u
r
e
an
d
alter
in
d
iv
id
u
als
in
th
e
p
o
o
l
to
co
n
f
o
r
m
to
b
elief
s
p
ac
e.
C
GA
is
ch
o
s
en
to
y
ie
ld
a
p
o
o
l
th
at
d
o
es
n
o
t
v
io
la
te
its
b
elief
s
p
ac
e
an
d
h
elp
s
r
ed
u
ce
n
u
m
b
er
o
f
p
o
s
s
ib
le
in
d
iv
id
u
als GA
g
en
er
ates till a
n
o
p
tim
u
m
is
f
o
u
n
d
[4
3
, 4
4
]
.
3.
3
.
M
o
t
iv
a
t
io
n /
s
t
a
t
e
m
ent
o
f
pro
blem
a.
Sp
am
s
h
av
e
co
n
tin
u
ed
to
s
o
a
r
with
th
e
ad
v
e
n
t
o
f
SMS.
T
h
e
alar
m
in
g
g
r
o
wth
r
ate
o
f
s
p
a
m
s
with
SMS
p
o
p
u
lar
ity
h
av
e
n
o
w
cr
ea
te
d
a
p
r
o
p
itio
u
s
en
v
ir
o
n
f
o
r
s
p
a
m
m
er
s
to
ex
p
l
o
it
s
u
b
s
cr
ib
er
s
;
T
h
u
s
,
ca
u
s
in
g
b
o
th
f
in
an
cial
lo
s
s
an
d
em
o
t
io
n
al
in
s
tab
ilit
y
as
co
n
s
eq
u
en
ce
s
to
u
s
er
s
,
co
r
p
o
r
ate
o
r
g
a
n
s
an
d
m
o
b
ile
n
etwo
r
k
o
p
er
ato
r
(
s
)
.
b.
Aca
d
em
ic
r
esear
ch
es
an
d
co
m
p
an
ies
ar
e
to
d
ay
,
f
ac
ed
with
th
e
ch
allen
g
e
o
f
d
ea
lin
g
with
SMS
s
p
am
.
A
m
ajo
r
is
s
u
e
h
as
b
ee
n
th
at
ex
is
tin
g
ap
p
r
o
ac
h
es
to
r
eso
lv
in
g
SMS
s
p
am
ar
e
im
p
o
r
ted
f
r
o
m
s
u
cc
ess
f
u
l
em
ail
an
ti
-
s
p
am
s
o
lu
tio
n
s
(
W
an
g
et
al.
,
2
0
1
0
)
.
T
h
u
s
,
ar
e
q
u
ite
u
n
ab
le
to
ef
f
ec
tiv
ely
an
d
ef
f
icien
tly
tack
le
SMS
s
p
am
s
u
cc
ess
f
u
lly
–
as
th
eir
p
e
r
f
o
r
m
an
ce
is
s
er
io
u
s
ly
h
am
p
er
ed
an
d
d
e
g
r
ad
e
d
b
y
t
h
e
p
a
r
am
etr
ic
f
ea
ts
u
s
ed
to
f
ilter
s
p
am
s
.
c.
T
h
e
f
o
r
m
u
latio
n
an
d
d
esig
n
o
f
an
ef
f
ec
tiv
e
SMS
f
ilter
h
as
co
n
tin
u
ed
t
o
s
u
f
f
er
ed
s
etb
ac
k
(
s
)
d
u
e
to
th
e
in
h
er
en
t r
ea
s
o
n
th
at
SMS
f
ilte
r
s
b
y
d
esig
n
ar
e
n
o
t
as
s
im
p
le
as
em
ail
f
ilter
s
d
u
e
to
its
lim
it
ed
s
ize
o
f
1
6
0
-
ch
ar
ac
ter
s
o
f
1
4
0
b
y
tes
s
ize
d
d
ata.
T
h
ese
am
o
n
g
s
t
o
th
er
co
n
s
tr
ain
ts
,
co
n
tin
u
e
to
cr
ea
te
r
ip
p
led
im
p
ed
im
en
t
in
s
ize
o
f
f
ea
tu
r
e
to
b
e
s
elec
ted
f
o
r
tr
ain
in
g
an
d
co
n
s
eq
u
en
tly
c
o
n
tr
ib
u
tin
g
t
o
p
o
o
r
lear
n
in
g
an
d
class
if
icatio
n
o
f
lear
n
i
n
g
a
lg
o
r
ith
m
.
d.
Fu
r
th
er
m
o
r
e
,
SMS
ar
e
r
ip
p
led
with
s
lan
g
s
,
ab
b
r
ev
iatio
n
s
,
s
y
m
b
o
ls
an
d
em
o
tico
n
s
th
at
in
h
ib
it
p
r
o
p
e
r
class
if
icatio
n
o
f
wo
r
d
s
o
r
te
x
ts
[4
5
]
.
T
o
o
v
e
r
co
m
e
th
ese
am
o
n
g
s
t
m
an
y
o
th
er
s
h
o
r
tf
alls
in
h
er
en
t
in
th
e
a
d
o
p
tio
n
o
f
em
ail
f
ilter
s
as
ad
ap
ted
to
h
an
d
lin
g
SMS
s
p
a
m
s
u
cc
ess
f
u
lly
,
a
h
y
b
r
id
f
ilter
in
g
tech
n
i
q
u
e
th
at
r
ed
u
ce
s
n
o
is
e
in
f
o
r
m
o
f
s
lan
g
s
,
em
o
tico
n
s
,
ab
b
r
ev
iatio
n
s
in
S
MS
as
well
a
s
ex
p
an
d
m
es
s
ag
e
s
ize
m
u
s
t
b
e
em
p
lo
y
ed
t
o
en
h
an
ce
ad
e
q
u
ate
class
if
icatio
n
.
T
h
u
s
,
o
u
r
r
esear
ch
g
o
al
(
s
)
is
to
p
r
o
p
o
s
e
a
h
y
b
r
id
d
ee
p
lear
n
in
g
n
e
u
r
al
n
et
wo
r
k
m
o
d
el
f
o
r
tex
t
n
o
r
m
aliza
tio
n
a
n
d
s
em
an
tic
ex
p
an
s
io
n
in
SMS sp
am
f
ilter
in
g
.
T
h
e
p
r
o
p
o
s
ed
m
o
d
el
p
r
o
p
e
r
ties
an
d
g
o
al
s
will in
clu
d
e:
−
Per
f
o
r
m
r
e
p
etitiv
e
task
s
with
o
u
t e
m
o
tio
n
al
d
ef
ec
ts
−
E
m
b
o
d
y
th
e
k
n
o
wled
g
e
o
f
h
u
m
an
e
x
p
er
ts
with
th
e
h
elp
o
f
s
p
ec
ial
s
o
f
twar
e
to
o
ls
,
m
a
n
ip
u
late
d
ata
t
o
s
o
lv
e
p
r
o
b
le
m
s
an
d
m
a
k
e
d
ec
i
s
io
n
s
in
th
at
d
o
m
ain
.
−
Pro
ce
s
s
es a
r
e
b
etter
f
o
r
m
alize
d
an
d
d
ef
in
ed
o
n
m
ac
h
in
es.
−
Kn
o
wled
g
eb
ase
u
p
d
ate
is
au
to
m
atic
−
Pro
ce
s
s
es a
r
e
b
etter
f
o
r
m
alize
d
an
d
d
ef
in
ed
o
n
m
ac
h
in
es.
4.
M
E
M
E
T
I
C
B
AY
E
S
I
AN
N
E
T
WO
RK
E
XP
E
RI
M
E
N
T
A
L
F
RAM
E
WO
RK
SMS
s
p
am
f
ilter
s
ca
n
h
av
e
ca
p
ac
ity
an
d
g
r
a
n
ted
ca
p
a
b
ilit
y
to
tr
an
s
cr
ib
e
em
o
tico
n
s
,
ab
b
r
e
v
iatio
n
s
a
nd
s
lan
g
s
in
to
s
tan
d
ar
d
ter
m
s
as
well
as
ex
p
an
d
m
ess
ag
e
s
ize
to
en
h
an
ce
b
etter
f
ea
tu
r
e
ex
tr
ac
tio
n
f
o
r
class
if
icatio
n
alg
o
r
ith
m
s
an
d
ap
p
r
o
ac
h
es.
T
h
e
s
tu
d
y
will
a
ls
o
s
er
v
e
to
r
ed
u
ce
o
r
th
o
g
r
ap
h
ic
er
r
o
r
f
o
u
n
d
in
SMS,
ch
at
g
r
o
u
p
s
a
n
d
an
o
t
h
er
s
o
cial
n
etwo
r
k
co
m
m
u
n
i
ca
tio
n
m
ed
i
u
m
t
h
at
im
p
e
d
es
m
ac
h
in
e
lear
n
in
g
alg
o
r
ith
m
.
T
h
is
is
b
ec
au
s
e
f
r
o
m
th
e
v
ar
i
o
u
s
ap
p
r
o
ac
h
es
a
d
o
p
ted
to
SMS
s
p
am
f
ilter
s
–
th
e
co
n
te
n
t
-
b
ased
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
9
,
No
.
1,
Ap
r
il
20
20
:
9
–
1
8
14
m
o
d
els
with
tex
t
p
r
e
-
p
r
o
ce
s
s
in
g
h
as
s
h
o
wn
to
p
er
f
o
r
m
b
etter
.
Ma
ch
in
e
tr
an
s
latio
n
(
MT
)
p
er
f
o
r
m
s
b
etter
wh
en
ap
p
lied
to
n
o
r
m
alize
d
tex
t
m
ess
ag
es
[
4
6
]
.
I
t
ca
n
co
m
b
in
ed
m
u
ltip
le
ap
p
r
o
ac
h
es
in
n
o
is
y
d
ata,
tex
t
n
o
r
m
aliza
tio
n
to
cr
ea
te
a
b
et
ter
o
u
t
p
u
t.
B
u
t,
ex
tr
ac
tin
g
o
n
ly
r
ele
v
an
t
f
ea
ts
an
d
/o
r
p
ar
am
eter
to
tr
ain
th
e
class
if
ier
h
as b
ee
n
r
ep
o
r
ted
to
co
n
tr
ib
u
te
to
th
e
ef
f
icien
c
y
o
f
SMS
s
p
am
f
ilter
s
[4
7
-
50
]
.
T
h
u
s
,
we
p
r
o
p
o
s
e
tex
t
p
r
ep
r
o
ce
s
s
in
g
SMS
s
p
am
f
ilt
er
m
o
d
el
with
th
e
ca
p
a
b
ilit
y
o
f
n
o
r
m
alizin
g
,
ex
p
an
d
in
g
t
ex
t
m
ess
ag
es
an
d
ex
tr
ac
tin
g
s
u
itab
le
f
ea
tu
r
es
as
d
ataset
in
p
u
t
p
ar
a
m
eter
s
f
o
r
tr
ain
in
g
th
e
a
d
o
p
ted
class
if
icati
o
n
alg
o
r
ith
m
an
d
m
o
d
el.
Stu
d
y
u
s
es KD
D
-
C
UP
’
9
9
d
ataset.
F
ig
u
re
1
.
P
ro
p
o
se
d
g
e
n
e
ti
c
a
l
g
o
rit
h
m
train
e
d
b
a
y
e
sia
n
n
e
two
rk
T
h
e
m
o
d
el
is
r
e
p
r
esen
ted
in
Fi
g
u
r
e
1
ex
p
lain
e
d
as th
u
s
:
a.
R
aw
tex
t r
ep
r
esen
ts
th
e
o
r
ig
in
al
tex
t f
r
o
m
t
h
e
s
en
d
er
f
o
r
n
o
r
m
aliza
tio
n
an
d
e
x
p
an
s
io
n
.
b.
T
ex
t
n
o
r
m
aliza
tio
n
em
p
lo
y
s
two
d
ictio
n
ar
ies:
(
a)
f
ir
s
t,
an
E
n
g
lis
h
d
ictio
n
a
r
y
t
o
c
h
ec
k
i
f
t
ex
t
ar
e
E
n
g
lis
h
s
o
as
to
th
en
n
o
r
m
alize
tex
t
to
its
r
o
o
t
f
o
r
m
,
a
n
d
(
b
)
s
ec
o
n
d
,
is
a
s
lan
g
d
ictio
n
ar
y
to
tr
an
s
l
ate
s
lan
g
s
in
to
E
n
g
lis
h
te
x
t.
T
h
e
b
asic
o
p
er
atio
n
o
f
th
is
s
tag
e
is
to
r
ep
lace
s
l
an
g
s
an
d
ab
b
r
ev
iatio
n
with
s
ta
n
d
ar
d
E
n
g
lis
h
wo
r
d
s
f
r
o
m
th
ese
d
ictio
n
ar
ies
.
T
h
e
Fre
elin
g
E
n
g
lis
h
d
ictio
n
a
r
y
an
d
No
s
lan
g
d
ictio
n
ar
y
a
r
e
p
r
o
p
o
s
ed
.
c.
C
o
n
ce
p
ts
g
en
er
atio
n
ar
e
s
em
an
tically
an
aly
ze
d
alr
ea
d
y
n
o
r
m
alize
d
tex
t
to
d
ed
u
ce
th
eir
co
n
ce
p
t.
T
h
e
co
n
ce
p
ts
ar
e
p
r
o
v
id
e
d
b
y
L
an
g
u
ag
e
Data
B
ase
B
ab
elNe
t r
ep
o
s
ito
r
y
.
d.
W
o
r
d
s
en
s
e
d
is
am
b
ig
u
atio
n
(
W
SD)
:
Her
e,
f
r
o
m
a
v
ar
iety
o
f
co
n
ce
p
t
g
en
e
r
ated
,
t
h
is
s
tag
e
is
u
s
ed
to
f
i
n
d
th
e
co
n
ce
p
t
th
at
is
m
o
r
e
r
ele
v
an
t
ac
co
r
d
in
g
to
th
e
c
o
n
tex
t
o
f
th
e
o
r
ig
i
n
al
m
ess
ag
e,
a
m
o
n
g
all
g
en
er
ated
co
n
ce
p
ts
r
elate
d
to
a
ce
r
tain
wo
r
d
.
I
t
e
q
u
ally
r
elies
o
n
co
n
ce
p
ts
ar
e
p
r
o
v
id
e
d
b
y
L
an
g
u
ag
e
Data
B
ase
(
L
DB
)
B
ab
elNe
t r
ep
o
s
ito
r
y
e.
T
o
k
en
is
atio
n
u
n
it:
T
o
k
en
izati
o
n
is
t
h
e
p
r
o
ce
s
s
o
f
b
r
ea
k
in
g
d
o
wn
a
tex
t
co
r
p
u
s
in
t
o
in
d
iv
id
u
al
elem
en
ts
th
at
s
er
v
e
as
in
p
u
t
f
o
r
v
ar
io
u
s
n
atu
r
al
lan
g
u
ag
e
p
r
o
ce
s
s
in
g
alg
o
r
ith
m
s
.
No
r
m
alis
ed
tex
ts
ar
e
b
r
o
k
e
n
in
to
in
d
iv
id
u
al
wo
r
d
s
an
d
s
to
p
wo
r
d
s
an
d
p
u
n
ctu
atio
n
ch
ar
ac
ter
s
ar
e
eq
u
ally
r
em
o
v
e
d
in
th
is
u
n
it.
f.
Me
r
g
in
g
R
u
le:
I
t
em
p
lo
y
s
p
a
r
am
eter
s
th
at
d
ef
in
e
th
e
co
m
b
in
atio
n
o
f
r
esu
lt
o
f
p
r
e
-
p
r
o
c
ess
in
g
(
o
r
ig
in
al
tex
t,
n
o
r
m
aliza
tio
n
an
d
d
is
a
m
b
ig
u
atio
n
s
tag
e)
.
Me
r
g
in
g
r
u
le
an
s
wer
s
th
e
q
u
esti
o
n
f
r
o
m
ea
ch
s
tag
e
as
f
o
llo
ws:
(
a)
s
h
o
u
ld
it
k
ee
p
th
e
o
r
ig
in
al
to
k
en
(
s
)
?,
(
b
)
s
h
o
u
ld
t
ex
t
n
o
r
m
aliza
tio
n
b
e
p
e
r
f
o
r
m
e
d
?,
(
c)
s
h
o
u
ld
it p
er
f
o
r
m
co
n
ce
p
ts
g
en
er
atio
n
?,
an
d
(
d
)
s
h
o
u
ld
it p
e
r
f
o
r
m
th
e
wo
r
d
s
en
s
e
d
is
am
b
ig
u
ati
o
n
?
g.
No
r
m
alize
d
a
n
d
E
x
p
a
n
d
ed
tex
t
is
a
co
m
b
in
atio
n
o
f
te
x
t
o
b
tain
ed
f
r
o
m
v
ar
i
o
u
s
o
u
tp
u
t
o
f
p
r
ef
er
r
ed
s
tag
es
o
f
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
m
o
d
el.
4
.1
.
F
e
a
t
ure
s
elec
t
io
n
,
t
ra
ini
ng
a
nd
ra
t
io
na
le
f
o
r
cho
ice
o
f
m
o
del
Nee
d
to
m
in
im
ize
th
e
n
u
m
b
e
r
o
f
f
ea
tu
r
es
as
in
p
u
t
p
ar
am
ete
r
s
f
o
r
class
if
icatio
n
–
s
in
ce
,
a
n
in
cr
ea
s
e
in
th
e
n
u
m
b
er
o
f
f
ea
tu
r
es
u
s
ed
will
ad
d
to
th
e
co
m
p
u
tati
o
n
al
co
m
p
lex
ity
o
f
th
e
s
y
s
tem
.
T
h
u
s
,
th
e
C
GA
alg
o
r
ith
m
is
u
s
ed
in
s
elec
tio
n
o
f
f
ea
tu
r
es
o
b
tain
ed
f
r
o
m
t
h
e
tex
t
p
r
e
-
p
r
o
ce
s
s
in
g
s
ec
tio
n
.
T
h
e
in
p
u
t
is
th
e
d
ataset
(
to
k
en
s
o
b
tain
ed
v
ia
t
o
k
en
izatio
n
o
f
n
o
r
m
alize
d
an
d
ex
p
an
d
ed
tex
t
f
r
o
m
tex
t
p
r
e
-
p
r
o
ce
s
s
in
g
s
ec
tio
n
)
.
T
h
e
m
o
d
el
is
m
a
d
e
u
p
o
f
th
e
f
o
llo
win
g
s
ec
tio
n
s
:
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Memetic
a
lg
o
r
ith
m
fo
r
s
h
o
r
t
mess
a
g
in
g
s
ervice
s
p
a
m
filt
er
u
s
in
g
text
…
(
A
r
n
o
ld
A
d
ima
b
u
a
Oju
g
o
)
15
−
GA
Un
it
–
y
ield
s
a
r
u
le
-
b
as
ed
,
g
e
n
etic
r
ep
r
esen
tatio
n
o
f
n
o
r
m
alize
d
a
n
d
ex
p
a
n
d
ed
test
d
ef
in
ed
.
T
h
e
alg
o
r
ith
m
th
en
in
itializes
m
o
d
el
with
a
r
an
d
o
m
p
o
p
u
latio
n
th
at
is
cr
ea
ted
an
d
s
u
b
jec
t
ed
to
r
ep
etitiv
e
ap
p
licatio
n
o
f
r
ec
o
m
b
in
atio
n
,
m
u
tatio
n
,
i
n
v
er
s
io
n
an
d
s
elec
tio
n
o
p
e
r
ato
r
s
to
im
p
r
o
v
e
th
e
g
en
er
ate
d
p
o
p
u
latio
n
f
r
o
m
th
e
o
r
ig
in
al
d
ataset.
−
E
v
alu
atio
n
Un
it
co
n
tain
s
a
f
itn
ess
f
u
n
ctio
n
th
at
m
ea
s
u
r
es
th
e
q
u
ality
o
f
r
ep
r
esen
ted
s
o
lu
ti
o
n
.
I
t
co
m
p
u
tes
o
p
tim
ality
o
f
a
s
o
lu
tio
n
b
y
co
m
p
ar
in
g
th
e
ch
r
o
m
o
s
o
m
es
ag
ain
s
t
all
o
th
er
ch
r
o
m
o
s
o
m
e
u
s
in
g
s
o
m
e
p
r
ed
ef
in
e
d
f
u
n
ctio
n
.
−
T
r
ain
in
g
Un
it:
T
r
ain
s
th
e
f
ilter
b
ased
o
n
B
ay
es
Pro
b
ab
ilit
y
T
h
eo
r
em
.
I
t u
s
es
k
n
o
wn
SMS
co
r
p
u
s
o
f
s
p
am
an
d
g
en
u
in
e
m
ess
ag
es/t
ex
ts
.
A
co
llectio
n
o
f
to
k
en
s
ap
p
ea
r
in
g
in
ea
ch
co
r
p
u
s
an
d
th
eir
to
tal
o
cc
u
r
r
en
ce
s
(
s
co
r
es)
ar
e
m
ain
tain
e
d
in
th
e
d
atab
ase
–
s
o
th
at
b
ased
o
n
th
eir
o
cc
u
r
r
en
ce
s
,
ea
c
h
s
et
o
f
s
p
am
an
d
g
en
u
in
e
d
ata
is
ass
ig
n
ed
a
cr
it
er
io
n
o
r
p
r
o
b
ab
ilit
y
s
co
r
e
f
o
r
i
ts
ca
p
ac
ity
o
f
d
eter
m
in
in
g
a
te
x
t
o
r
m
ess
ag
e
to
eith
er
b
e
a
s
p
a
m
o
r
g
en
u
i
n
e
tex
t.
4.
2
.
Cla
s
s
if
ica
t
io
n
s
ec
t
io
n
B
ased
o
n
th
e
f
r
e
q
u
en
c
y
p
r
o
b
a
b
ilit
y
o
f
o
cc
u
r
r
en
ce
o
f
ea
ch
w
o
r
d
(
to
k
e
n
s
)
as
s
p
am
o
r
leg
iti
m
ate,
ea
ch
in
co
m
in
g
u
n
s
ee
n
n
o
r
m
alize
d
m
ess
ag
e
d
ata
is
p
r
o
c
ess
ed
an
d
class
if
ied
as
eith
er
leg
itima
te
o
r
s
p
am
b
y
th
e
B
ay
esian
clas
s
if
ier
.
I
n
th
e
ev
en
t
o
f
m
is
class
if
icatio
n
,
u
s
er
s
ca
n
r
ec
tify
th
is
class
if
icati
o
n
b
y
r
ea
d
in
g
th
e
m
ess
ag
e
an
d
r
e
-
ad
d
in
g
th
e
m
ess
ag
e
to
in
b
o
x
.
T
h
is
will
au
to
m
atica
lly
co
r
r
ec
t
an
d
u
p
d
ate
th
e
d
atab
ase
f
o
r
f
u
tu
r
e
class
if
icatio
n
.
T
h
u
s
,
m
a
k
in
g
B
ay
esian
f
ilter
s
q
u
ite
ad
a
p
tiv
e.
4.
3
.
O
utput
s
ec
t
io
n
R
esu
lt o
f
th
e
class
if
icatio
n
o
f
th
e
f
ilter
in
to
Sp
am
o
r
Ham
,
is
th
e
ex
p
ec
ted
o
u
tp
u
t o
f
th
is
u
n
i
t.
4.
3
.
E
x
perim
ent
a
l
m
o
del o
p
er
a
t
io
ns
Oju
g
o
[
4
2
]
d
escr
ib
ed
a
g
e
n
etic
alg
o
r
ith
m
tr
ai
n
ed
n
eu
r
al
n
etwo
r
k
em
p
lo
y
ed
in
ea
r
ly
d
iab
etes
d
etec
tio
n
.
GANN
is
in
itializ
ed
with
(
n
-
r
!
)
in
d
iv
id
u
al
if
-
t
h
en
,
f
u
zz
y
r
u
les
(
i.e
.
6
-
4
!
)
.
I
n
d
iv
id
u
al
f
itn
ess
is
co
m
p
u
ted
as
3
0
-
in
d
iv
id
u
als
a
r
e
s
elec
ted
v
ia
th
e
to
u
r
n
am
en
t
m
eth
o
d
to
d
e
ter
m
in
e
n
ew
p
o
o
l
an
d
s
elec
tio
n
f
o
r
m
atin
g
.
C
r
o
s
s
o
v
er
a
n
d
m
u
tatio
n
ar
e
a
p
p
lied
t
o
h
elp
n
et
le
ar
n
th
e
d
y
n
am
ic,
n
o
n
-
lin
ea
r
u
n
d
er
ly
i
n
g
f
ea
ts
o
f
in
ter
est
v
ia
m
u
ltip
o
in
t
cr
o
s
s
o
v
er
to
y
ield
n
ew
p
ar
en
ts
.
T
h
e
n
ew
p
ar
en
ts
co
n
tr
ib
u
te
to
y
ie
ld
n
ew
in
d
iv
id
u
als.
Mu
tatio
n
is
r
ea
p
p
lied
an
d
in
d
iv
id
u
als
ar
e
allo
tted
n
ew
r
an
d
o
m
v
alu
es
th
at
s
t
ill
co
n
f
o
r
m
t
o
th
e
b
elief
s
p
ac
e.
T
h
e
m
u
tatio
n
ap
p
lied
d
e
p
en
d
s
o
n
h
o
w
f
ar
C
GA
is
p
r
o
g
r
ess
ed
o
n
th
e
n
et
an
d
h
o
w
f
it
th
e
f
ittes
t
in
d
iv
id
u
al
in
th
e
p
o
o
l
(
i.e
.
f
itn
ess
o
f
t
h
e
f
itt
est
in
d
iv
id
u
al
d
iv
i
d
ed
b
y
2
)
.
New
in
d
iv
id
u
als
r
ep
lace
o
ld
with
lo
w
f
itn
ess
s
o
as
t
o
cr
ea
te
a
n
ew
p
o
o
l.
Pro
ce
s
s
co
n
tin
u
es
u
n
til
in
d
iv
id
u
al
wit
h
f
itn
ess
o
f
0
(
i.e
.
s
o
lu
tio
n
)
is
f
o
u
n
d
.
R
u
le
-
b
ased
en
co
d
ed
s
p
am
as sh
o
wn
in
T
a
b
le
1
.
Gen
er
atio
n
o
f
p
o
p
u
latio
n
f
r
o
m
p
ar
e
n
ts
as sh
o
wn
in
T
a
b
le
2
.
T
ab
le
1
.
R
u
le
-
b
ased
e
n
co
d
e
d
s
co
r
e
C
o
d
e
R
u
l
e
I
n
p
u
t
P
a
r
a
me
t
e
r
s
G
e
n
u
i
n
e
S
p
a
m
P
0
1
M
e
ss
a
g
e
S
i
z
e
0
.
5
0
0
.
5
0
P
0
2
M
e
ss
a
g
e
C
h
a
r
a
c
t
e
r
0
.
5
0
0
.
5
0
P
0
3
M
e
ss
a
g
e
F
r
o
m
0
.
5
0
0
.
5
0
P
0
4
M
e
ss
a
g
e
T
o
0
.
5
0
0
.
5
0
P
0
5
S
u
b
j
e
c
t
0
.
3
0
0
.
7
0
P
0
6
B
o
d
y
o
f
M
e
ssa
g
e
0
.
2
5
0
.
7
5
T
ab
le
2
.
1
st
a
n
d
2
nd
g
en
e
r
atio
n
o
f
p
o
p
u
latio
n
f
r
o
m
p
ar
e
n
ts
S
/
N
S
e
l
e
c
t
i
o
n
C
h
r
o
mo
so
m
e
s (B
i
n
a
r
y
0
o
r
1
)
F
i
t
n
e
ss
F
u
n
c
t
i
o
n
P
a
r
e
n
t
1
st
G
e
n
C
r
o
ss
o
v
e
r
P
a
r
e
n
t
2
n
d
G
e
n
1
50
1
1
0
0
1
0
1
a
n
d
6
1
1
0
0
01
49
2
50
1
1
0
0
1
0
2
a
n
d
5
1
1
0
0
10
50
3
50
1
1
0
0
1
0
3
a
n
d
6
1
1
0
0
01
49
4
50
1
1
0
0
1
0
4
a
n
d
5
1
1
0
0
10
50
5
30
0
1
1
1
1
0
5
a
n
d
6
0
1
1
1
01
29
6
25
0
1
1
0
0
1
6
a
n
d
5
0
1
1
0
10
26
I
n
itializatio
n
/
s
elec
tio
n
v
ia
AN
N
en
s
u
r
es
th
at
f
ir
s
t
3
-
b
elief
s
a
r
e
m
et;
m
u
tatio
n
en
s
u
r
es
f
o
u
r
t
h
b
elief
is
m
et.
I
ts
in
f
lu
e
n
ce
f
u
n
ctio
n
in
f
lu
en
ce
s
h
o
w
m
an
y
m
u
tatio
n
s
tak
e
p
lace
,
an
d
th
e
k
n
o
wled
g
e
o
f
s
o
lu
tio
n
(
h
o
w
clo
s
e
its
s
o
lu
tio
n
is
)
h
as
d
ir
ec
t
im
p
ac
t
o
n
h
o
w
alg
o
r
ith
m
is
p
r
o
ce
s
s
ed
.
Alg
o
r
ith
m
s
to
p
s
wh
en
b
est
in
d
iv
id
u
al
h
as
f
itn
ess
o
f
0
.3
.
Mo
d
el
s
to
p
s
if
s
to
p
cr
iter
i
o
n
is
m
et.
GANN
u
tili
ze
s
n
u
m
b
er
o
f
e
p
o
ch
s
to
d
eter
m
in
e
s
to
p
cr
iter
io
n
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
9
,
No
.
1,
Ap
r
il
20
20
:
9
–
1
8
16
5.
F
I
NDING
S AN
D
DI
SCUS
SI
O
N
W
ith
N
aïv
e
B
ay
es
an
d
GA
(
a
s
s
tan
d
alo
n
e
m
o
d
el)
to
b
en
ch
m
ar
k
th
e
in
tellig
e
n
t
s
y
s
tem
an
d
a
s
ce
r
tain
h
o
w
well
o
u
r
h
y
b
r
id
GABN
alg
o
r
ith
m
p
er
f
o
r
m
ed
,
we
o
b
tai
n
th
e
r
esu
lts
in
Fig
u
r
e
2
a
n
d
F
ig
u
r
e
3
r
esp
ec
tiv
ely
as
s
ee
n
b
elo
w
.
th
e
h
y
b
r
id
g
a
b
n
(
m
em
etic)
al
g
o
r
ith
m
o
u
tp
er
f
o
r
m
s
s
tan
d
al
o
n
e
n
aïv
e
b
ay
es
an
d
GA
m
o
d
el
.
Ho
wev
er
,
f
o
r
th
e
m
ea
n
p
r
o
ce
s
s
in
g
tim
e
r
eq
u
ir
e
d
to
c
o
n
v
e
r
g
e
–
it
is
f
o
u
n
d
th
at
GABN
p
er
f
o
r
m
e
d
least.
T
h
is
ca
n
b
e
attr
ib
u
ted
to
th
e
f
ac
t th
at:
(
a)
th
e
h
y
b
r
id
m
o
d
el
n
ee
d
s
to
f
ir
s
t u
s
e
GA
as p
r
e
-
p
r
o
ce
s
s
o
r
to
tr
ain
B
ay
esian
n
etwo
r
k
,
(
b
)
f
o
r
s
u
c
h
h
y
b
r
i
d
s
,
th
er
e
ar
e
alwa
y
s
s
tr
u
ctu
r
al
d
ep
en
d
e
n
cies
with
th
e
u
n
d
er
ly
in
g
h
eu
r
is
tics
em
p
lo
y
ed
/m
e
r
g
ed
an
d
co
n
f
lic
ts
in
d
ata
e
n
co
d
i
n
g
th
at
is
r
e
q
u
ir
ed
.
T
h
ese
m
u
s
t
b
e
r
eso
lv
ed
in
o
r
d
er
f
o
r
th
e
m
o
d
el
to
p
er
f
o
r
m
ap
p
r
o
p
r
iately
.
Fig
u
r
e
2
.
Mo
d
el/
h
eu
r
is
tic
ac
cu
r
ac
y
in
p
e
r
ce
n
tag
e
Fig
u
r
e
3
.
Mo
d
el/Heu
r
is
tic
co
n
v
er
g
en
ce
tim
e
i
n
s
ec
o
n
d
s
5
.
1
.
M
o
del Ev
a
lua
t
io
n
I
n
th
is
s
tu
d
y
,
ac
cu
r
ac
y
,
r
e
ca
ll,
er
r
o
r
r
ate
(
E
R
)
an
d
s
p
ec
if
icity
ar
e
u
s
ed
to
e
v
alu
ate
t
h
e
p
er
f
o
r
m
an
ce
o
f
th
e
d
etec
tio
n
m
o
d
els.
T
h
e
f
o
r
m
u
las o
f
th
e
a
b
o
v
e
c
r
iter
ia
ar
e
ca
lcu
lated
as f
o
llo
ws:
=
+
+
+
+
(1
)
=
+
(2
)
=
+
+
+
+
(3
)
=
+
(4
)
A
tr
u
e
p
o
s
itiv
e
(
T
P)
is
a
ca
s
e
(
r
u
le)
th
at
co
r
r
ec
tly
d
is
t
in
g
u
is
h
es
s
p
am
f
r
o
m
h
am
.
A
tr
u
e
n
eg
ativ
e
(
T
N)
s
h
o
ws
n
o
r
m
al
tex
t
d
ata
class
if
ied
co
r
r
ec
tly
as
n
o
r
m
al.
A
f
alse
n
eg
ativ
e
(
FN)
is
a
ca
s
e
in
wh
ich
a
tex
t
is
class
if
ied
as
n
o
r
m
al
d
ata,
an
d
a
f
alse
p
o
s
itiv
e
(
FP
)
is
a
ca
s
e
in
wh
ich
a
n
o
r
m
al
tex
t
is
cl
ass
if
ied
as
a
s
p
am
.
T
h
e
ac
cu
r
ac
y
r
ate
is
th
e
o
v
e
r
all
co
r
r
ec
t
d
etec
tio
n
ac
cu
r
ac
y
o
f
th
e
d
ataset.
E
R
r
ef
er
s
to
t
h
e
r
o
b
u
s
tn
ess
o
f
th
e
class
if
ier
,
R
ec
all
i
s
d
eg
r
ee
o
f
co
r
r
ec
tly
d
etec
ted
attac
k
ty
p
es
o
f
all
ca
s
es
class
if
ied
as
attac
k
s
;
wh
ile,
s
p
ec
if
icity
is
th
e
p
er
ce
n
tag
e
o
f
co
r
r
ec
tly
class
if
ied
d
ata.
I
n
th
e
ab
o
v
e,
h
ig
h
er
ac
cu
r
ac
y
an
d
r
ec
all
an
d
lo
wer
E
R
in
d
icate
g
o
o
d
p
er
f
o
r
m
an
ce
.
T
o
f
u
r
t
h
er
m
ea
s
u
r
e
e
f
f
ec
tiv
e
n
ess
an
d
ac
cu
r
ac
y
,
we
m
ea
s
u
r
e
th
eir
r
ate
o
f
m
is
class
if
icatio
n
an
d
co
r
r
esp
o
n
d
in
g
im
p
r
o
v
em
e
n
t
p
er
ce
n
tag
es
in
b
o
th
tr
ain
in
g
an
d
test
d
ata
s
ets
as
s
u
m
m
ar
ize
d
in
T
ab
les
3
an
d
4
r
esp
ec
tiv
ely
.
E
q
u
atio
n
s
f
o
r
m
i
s
class
if
icatio
n
r
ate
an
d
its
im
p
r
o
v
em
en
t
p
e
r
ce
n
tag
e
o
f
u
n
s
u
p
er
v
is
ed
(
B
)
m
o
d
el
ag
ain
s
t su
p
er
v
is
ed
(
A)
m
o
d
el
r
esp
ec
tiv
ely
,
is
ca
lcu
lated
as f
o
llo
ws:
T
ab
les
3
an
d
4
r
esp
ec
tiv
ely
s
h
o
ws
m
is
class
if
icatio
n
er
r
o
r
r
ate
with
Naïv
e
B
ay
es,
GA
an
d
GABN
at
2
3
.
2
%,
4
.
7
%
an
d
1
.
0
2
%
(
i.e
.
er
r
o
r
r
ate
in
f
alse
-
p
o
s
itiv
e
an
d
tr
u
e
-
n
e
g
ativ
e)
r
esp
ec
tiv
ely
;
C
o
n
s
eq
u
en
tly
,
th
ey
all
p
r
o
m
is
e
an
im
p
r
o
v
em
e
n
t r
a
te
as o
f
3
.
6
%,
4
.
0
2
% a
n
d
0
.
1
2
%
r
esp
ec
tiv
ely
.
=
.
.
(5
)
T
ab
le
3
.
Misclass
if
icatio
n
R
at
e
o
f
E
ac
h
m
o
d
el
M
o
d
e
l
C
l
a
s
si
f
i
c
a
t
i
o
n
Er
r
o
r
s
Tr
a
i
n
i
n
g
D
a
t
a
Te
st
i
n
g
D
a
t
a
N
a
ï
v
e
B
a
y
e
s
5
2
.
5
%
2
3
.
2
%
G
e
n
e
t
i
c
A
l
g
o
r
i
t
h
m
4
8
.
4
%
4
.
7
%
G
A
B
N
1
9
.
6
%
1
.
0
2
%
Evaluation Warning : The document was created with Spire.PDF for Python.
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
I
SS
N:
2252
-
8
7
7
6
Memetic
a
lg
o
r
ith
m
fo
r
s
h
o
r
t
mess
a
g
in
g
s
ervice
s
p
a
m
filt
er
u
s
in
g
text
…
(
A
r
n
o
ld
A
d
ima
b
u
a
Oju
g
o
)
17
Als
o
,
its
im
p
r
o
v
em
en
t p
e
r
ce
n
t
ag
e
is
co
m
p
u
ted
as th
u
s
:
=
(
)
−
(
)
(
)
10
0
(6
)
T
ab
le
4
.
I
m
p
r
o
v
em
en
t Per
ce
n
tag
e
M
o
d
e
l
I
mp
r
o
v
e
m
e
n
t
%
Tr
a
i
n
i
n
g
D
a
t
a
Te
st
i
n
g
D
a
t
a
N
a
ï
v
e
B
a
y
e
s
2
.
1
1
%
3
.
6
%
G
e
n
e
t
i
c
A
l
g
o
r
i
t
h
m
2
.
3
2
%
4
.
0
2
%
G
A
B
N
0
.
0
9
%
0
.
1
2
%
3.
CO
NCLU
SI
O
N
Fro
m
th
e
c
o
n
s
eq
u
e
n
ce
s
o
f
s
p
am
to
u
s
er
s
,
s
ev
er
al
co
n
ce
r
ted
ef
f
o
r
ts
to
d
etec
t
s
p
am
in
tr
u
s
io
n
in
v
ar
io
u
s
co
m
m
u
n
icatio
n
m
e
d
ia
h
as
p
aid
o
f
f
esp
ec
ially
in
c
o
m
b
atin
g
em
ail
s
p
am
.
Sp
am
F
ilter
s
wo
r
k
b
y
f
ir
s
t
r
ec
eiv
in
g
p
ar
t
(
o
r
all)
o
f
t
h
e
m
ess
ag
e
an
d
th
en
an
aly
zin
g
it
in
s
o
m
e
way
to
d
ec
id
e
wh
e
th
er
it
is
h
am
(
i.e
.
leg
itima
te
m
ess
ag
e)
o
r
s
p
am
.
T
h
e
p
er
f
o
r
m
a
n
ce
o
f
a
s
p
am
f
ilter
ca
n
b
e
m
ea
s
u
r
ed
b
y
th
e
n
u
m
b
er
o
f
f
alse
-
p
o
s
itiv
es
(
in
co
r
r
ec
tly
m
ar
k
ed
as
s
p
am
)
an
d
f
alse
-
n
eg
ativ
e
s
(
u
n
id
en
tifie
d
s
p
am
)
t
h
at
it
g
en
er
ates.
An
id
ea
l
s
p
am
f
ilter
will
co
r
r
ec
tly
class
if
y
all
SMS
wi
th
alm
o
s
t
ze
r
o
er
r
o
r
r
ates
o
f
f
als
e
p
o
s
itiv
e/n
eg
ativ
e
–
th
r
o
u
g
h
tr
ad
eo
f
f
s
b
etwe
en
th
e
n
u
m
b
er
o
f
f
alse p
o
s
itiv
es a
n
d
f
alse n
eg
ativ
es.
RE
F
E
R
E
NC
E
S
[1
]
Oju
g
o
,
A.
A
a
n
d
Eb
o
k
a
,
A.
O.
,
"
Co
m
p
a
ra
ti
v
e
e
v
a
l
u
a
ti
o
n
fo
r
h
i
g
h
in
telli
g
e
n
t
p
e
rfo
rm
a
n
c
e
a
d
a
p
ti
v
e
m
o
d
e
l
fo
r
sp
a
m
p
h
ish
in
g
d
e
tec
ti
o
n
,
"
Dig
i
ta
l
T
e
c
h
n
o
l
o
g
ies
,
v
o
l
.
3
,
n
o
.
1
,
p
p
.
9
-
15
,
2
0
1
8
.
[2
]
Oju
g
o
,
A.A.,
Eb
o
k
a
,
A.O.
,
"
S
i
g
n
a
tu
re
-
b
a
se
d
m
a
lwa
re
d
e
tec
ti
o
n
u
sin
g
a
p
p
ro
x
ima
te
Bo
y
e
r
M
o
o
re
s
tri
n
g
m
a
tch
in
g
a
lg
o
rit
h
m
,
"
I
n
t.
J
.
o
f
M
a
th
e
ma
t
ica
l
S
c
ien
c
e
s a
n
d
Co
mp
u
ti
n
g
,
v
o
l.
3
,
n
o
.
5
,
pp
.
49
-
62
,
2
0
1
9
.
[3
]
Tex
t
Re
q
u
e
st.
T
h
e
Co
m
p
l
e
te
Ov
e
rv
iew
o
f
Bu
sin
e
ss
Tex
ti
n
g
.
2
0
1
6
.
[we
b
]:
a
v
a
il
a
b
le
a
t
h
tt
p
s:/
/www
.
tex
treq
u
e
st.co
m
/b
l
o
g
/t
e
x
ti
n
g
-
sta
ti
stics
-
a
n
sw
e
r
-
q
u
e
stio
n
s/
[4
]
Ch
a
m
in
d
a
,
T.
,
Da
y
a
ra
tn
e
,
T.
T.
,
Am
a
ra
sin
g
h
e
,
H.
K.
N.,
Ja
y
a
k
o
d
y
,
J.
M
.
R.
S
.
,
"
Co
n
ten
t
-
b
a
se
d
h
y
b
rid
S
M
S
sp
a
m
fil
terin
g
sy
ste
m
,"
Pr
o
c
e
e
d
in
g
s
o
f
IT
RU
Res
e
a
rc
h
S
y
mp
o
siu
m
,
Un
i
v
e
rsity
o
f
M
o
ra
t
u
wa
,
p
p
.
3
1
–
35
,
2
0
1
3
.
[5
]
G
o
m
e
z
Hild
a
g
o
,
J.
M
.
,
B
u
e
n
a
g
a
Ro
d
rı
g
u
e
z
,
M
a
n
d
Co
r
ti
z
o
P
e
re
z
,
J.
C.
,
"
T
h
e
ro
le
o
f
w
o
rd
se
n
se
d
isa
m
b
ig
u
a
ti
o
n
i
n
a
u
to
m
a
ted
tex
t
c
a
teg
o
riza
ti
o
n
,"
In
Pro
c
.
o
f
t
h
e
1
0
th
N
L
DB
,
p
p
.
2
9
8
–
3
0
9
,
2
0
0
5
.
[6
]
S
h
a
h
i,
T.
B
.
,
Ya
d
a
v
A.
,
"
M
o
b
il
e
S
M
S
S
p
a
m
F
il
terin
g
fo
r
Ne
p
a
li
Tex
t
u
si
n
g
Na
ïv
e
Ba
y
e
sia
n
a
n
d
S
u
p
p
o
rt
Ve
c
to
r
M
a
c
h
in
e
,"
I
n
t.
J
.
o
f
In
tell
ig
e
n
c
e
S
c
ien
c
e
,
Co
mp
u
ter
S
c
ien
c
e
a
n
d
C
o
mm
u
n
ica
ti
o
n
s
,
v
o
l.
4
,
n
o
.
1
,
p
p
.
24
-
28
,
2
0
1
4
.
[7
]
M
u
ry
n
e
t,
S
.
,
P
i
q
u
e
ra
s
Jo
v
e
r,
R.
,
"
Ho
w
a
n
S
M
S
-
Ba
se
d
m
a
lwa
re
in
f
e
c
ti
o
n
wil
l
g
e
t
t
h
ro
tt
led
b
y
th
e
wi
re
les
s
li
n
k
,"
.
I
n
Pro
c
e
e
d
in
g
s
o
f
IEE
E
ICC
2
0
1
2
-
Co
mm
u
n
ica
ti
o
n
a
n
d
In
fo
rm
a
t
io
n
S
y
ste
ms
S
e
c
u
rity
S
y
mp
o
siu
m
(
ICC’1
2
CIS
S
)
,
Ottaw
a
,
Ca
n
a
d
a
,
2
0
1
2
.
[8
]
M
u
ry
n
e
t,
S
.
,
P
iq
u
e
ra
s
Jo
v
e
r,
R.
,
"
An
a
ly
sis
o
f
S
M
S
S
p
a
m
in
M
o
b
il
i
ty
Ne
two
r
k
s.
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
A
d
v
a
n
c
e
d
Co
mp
u
ter
S
c
ien
c
e
,
v
o
l.
1
,
n
o
.
1
p
p
.
1
-
8
,
2
0
1
1
.
[9
]
Ag
wu
.
C.
O.
,
"
T
h
e
C
o
n
se
q
u
e
n
c
e
s
o
f
M
o
b
il
e
S
p
a
m
in
Nig
e
ria
E
m
e
rg
in
g
a
n
d
E
v
o
l
v
i
n
g
M
o
b
il
e
Co
m
m
u
n
ica
ti
o
n
S
e
c
to
r
o
f
th
e
Eco
n
o
m
y
,"
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
A
d
v
a
n
c
e
d
Res
e
a
rc
h
in
Co
mp
u
ter
S
c
ien
c
e
a
n
d
S
o
f
twa
r
e
En
g
i
n
e
e
rin
g
,
v
o
l.
5
,
n
o
.
5
,
p
p
.
1
1
7
-
1
2
4
,
2
0
1
5
.
[1
0
]
M
o
b
i
le E
c
o
sy
ste
m
F
o
ru
m
,
"
2
8
%
M
o
b
i
le Co
n
su
m
e
rs Rec
e
iv
e
S
M
S
S
p
a
m
Ev
e
ry
Da
y
,
"
2
0
1
7
.
[1
1
]
Ne
e
lma
y
De
sa
i
a
n
d
M
e
e
ra
Na
r
v
e
k
a
r
,
"
No
rm
a
li
z
a
ti
o
n
o
f
No
isy
T
e
x
t
Da
ta
,"
Pro
c
e
d
ia
Co
m
p
u
ter
S
c
ien
c
e
,
v
o
l.
4
5
,
pp.
1
2
7
-
132
,
2
0
1
5
.
[1
2
]
Jia
n
g
,
N.,
Jin
,
Y.,
S
k
u
d
lark
,
A.
,
Zh
a
n
g
,
Z.
,
"
Un
d
e
rsta
n
d
i
n
g
S
M
S
sp
a
m
in
larg
e
Ce
ll
u
lar
Ne
two
rk
:
Ch
a
ra
c
teristics
,
S
trate
g
ies
a
n
d
De
fe
n
se
s,
I
n
telli
g
e
n
t
S
y
ste
m
s,
"
IEE
E
,
v
o
l.
27
,
n
o
.
6
,
p
p
.
15
-
26
,
2
0
1
1
.
[1
3
]
Zab
lo
ts
k
a
y
a
,
N.
F
ra
u
d
u
len
t
sp
a
m
.
S
e
c
u
re
li
st,
2
0
0
8
.
[1
4
]
He
d
ieh
S
a
jed
i
,
G
o
laz
in
Zarg
h
a
m
i
P
a
ra
st,
F
a
tem
e
h
Ak
b
a
ri
,
"
S
M
S
S
p
a
m
F
il
teri
n
g
Us
in
g
M
a
c
h
in
e
Lea
rn
i
n
g
Tec
h
n
iq
u
e
s: A
S
u
rv
e
y
,"
M
a
c
h
in
e
L
e
a
rn
in
g
Res
e
a
rc
h
,
v
o
l.
1
,
n
o
.
1
,
p
p
.
1
-
14
,
2
0
1
6
.
[1
5
]
Wan
g
,
C.
,
Z
h
a
n
g
,
Y.,
Ch
e
n
,
X.,
Li
u
,
Z.
,
S
h
i,
L
.
,
Ch
e
n
,
G
.
,
"
A
Be
h
a
v
io
r
-
b
a
se
d
S
M
S
An
t
i
-
S
p
a
m
S
y
ste
m
,"
IBM
J
o
u
rn
a
l
o
f
Res
e
a
rc
h
a
n
d
De
v
e
lo
p
me
n
t
,
NJ
,
USA,
v
o
l
.
54
,
n
o
.
6
,
p
p
.
6
5
1
-
6
6
6
,
2
0
1
0
.
[1
6
]
Ti
a
g
o
,
A
.
A.,
Ti
a
g
o
,
P
.
S
.
,
Ig
o
r,
S
.
,
Jo
se
,
M
a
n
d
G
o
m
e
z
Hild
a
g
o
,
J.
M
.
,
"
Tex
t
n
o
rm
a
li
z
a
ti
o
n
a
n
d
se
m
a
n
ti
c
in
d
e
x
in
g
t
o
e
n
h
a
n
c
e
In
sta
n
t
M
e
ss
a
g
in
g
a
n
d
S
M
S
sp
a
m
fil
teri
n
g
,
"
K
n
o
wled
g
e
-
B
a
se
d
S
y
ste
ms
,
v
o
l.
1
0
8
,
n
o
.
15
,
p
p
.
25
-
32
,
2
0
1
6
.
[1
7
]
AiTi
A.W
.
,
Z
h
a
n
g
,
M
.
,
Xia
o
,
J.,
S
u
,
J.
,
"
A
p
h
ra
se
-
b
a
se
d
sta
t
isti
c
a
l
m
o
d
e
l
fo
r
S
M
S
tex
t
n
o
r
m
a
li
z
a
ti
o
n
,"
In
Pro
c
e
e
d
in
g
s
o
f
t
h
e
COLING/A
CL
o
n
M
a
in
c
o
n
fer
e
n
c
e
p
o
ste
r se
ss
io
n
s
,
p
p
.
33
–
4
0
,
2
0
0
6
.
[1
8
]
Nu
ru
z
z
a
m
a
n
,
T.
M
.
,
Lee
,
C
.
,
A
b
d
u
ll
a
h
,
M
.
F
.
A.,
C
h
o
i
,
D.
,
"
S
im
p
le
S
M
S
sp
a
m
fil
teri
n
g
o
n
i
n
d
e
p
e
n
d
e
n
t
m
o
b
il
e
p
h
o
n
e
,"
J
o
u
r
n
a
l
o
f
S
e
c
u
rity a
n
d
Co
mm
u
n
ica
ti
o
n
Ne
two
rk
s
,
v
o
l
.
5
,
n
o
.
10
,
p
p
1
2
0
9
–
1
2
2
0
,
2
0
1
2
.
[1
9
]
Na
ra
y
a
n
,
A.,
S
a
x
e
n
a
,
P
.
,
"
Th
e
c
u
r
se
o
f
1
4
0
c
h
a
ra
c
ters
:
Ev
a
lu
a
ti
n
g
t
h
e
e
ffica
c
y
o
f
S
M
S
sp
a
m
d
e
tec
ti
o
n
o
n
a
n
d
ro
id
,"
In
Pro
c
e
e
d
in
g
s
o
f
th
e
1
3
t
h
AC
M
W
o
rk
sh
o
p
o
n
S
e
c
u
rity
a
n
d
Priv
a
c
y
in
S
m
a
rtp
h
o
n
e
s
a
n
d
M
o
b
il
e
De
v
ice
s
(S
PS
M
’1
3
)
,
Be
rli
n
,
G
e
rm
a
n
y
,
p
p
.
3
3
–
42
,
2
0
1
2
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
2
2
5
2
-
8
7
7
6
I
n
t J I
n
f
&
C
o
m
m
u
n
T
ec
h
n
o
l
,
Vo
l.
9
,
No
.
1,
Ap
r
il
20
20
:
9
–
1
8
18
[2
0
]
An
d
ro
u
li
d
a
k
is
I.
,
Vla
c
h
o
s
V.,
P
a
p
a
n
ik
o
lao
u
,
A.
F
IM
ES
S
:
fil
terin
g
m
o
b
i
le
e
x
tern
a
l
S
M
S
sp
a
m
.
I
n
P
ro
c
e
e
d
in
g
s
o
f
th
e
6
t
h
Ba
lk
a
n
Co
n
fe
re
n
c
e
in
I
n
fo
rm
a
ti
c
s,
ACM,
Ne
w Yo
rk
,
USA
,
p
p
.
2
2
1
-
227
,
2
0
1
3
.
[2
1
]
Ha
sib
,
S
.
,
M
o
twa
n
i
,
M
.
,
S
a
x
e
n
a
,
A.
,
"
A
n
ti
-
S
p
a
m
M
e
th
o
d
o
lo
g
ies
:
A
Co
m
p
a
ra
ti
v
e
S
t
u
d
y
.
I
n
t.
J
.
Co
mp
u
ter
S
c
i
.
In
fo
rm
a
t
io
n
T
e
c
h
n
o
l
o
g
ies
,
v
o
l.
3
,
n
o
.
6
,
p
p
.
5
3
4
1
-
5
3
4
5
,
2
0
1
2
.
[2
2
]
Tri
g
g
s R
.
,
"
Wh
a
t
is S
M
S
a
n
d
h
o
w d
o
e
s it
w
o
rk
,"
2
0
1
3
.
[2
3
]
P
ra
c
h
i,
G
.
J.,
P
a
teriy
a
,
J.R.
K.
,
"
A
S
u
r
v
e
y
o
n
Ema
il
S
p
a
m
Ty
p
e
s
a
n
d
S
p
a
m
F
i
lt
e
rin
g
Tec
h
n
iq
u
e
s
,"
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
E
n
g
i
n
e
e
rin
g
Res
e
a
rc
h
&
T
e
c
h
n
o
l
o
g
y
(IJ
ER
T
)
,
2
0
1
4
.
[2
4
]
De
lan
y
S
.
J.
Us
i
n
g
Ca
se
-
Ba
se
d
Re
a
so
n
in
g
f
o
r
S
p
a
m
F
i
lt
e
rin
g
.
P
u
b
l
ish
e
d
P
h
D
T
h
e
sis
su
b
m
it
te
d
t
o
th
e
D
u
b
l
i
n
In
stit
u
te
o
f
Tec
h
n
o
l
o
g
y
i
n
f
u
lfi
l
lme
n
t
o
f
t
h
e
re
q
u
irem
e
n
ts
fo
r
t
h
e
d
e
g
re
e
o
f
Do
c
to
r
o
f
P
h
il
o
so
p
h
y
S
c
h
o
o
l
o
f
Co
m
p
u
ti
n
g
,
Du
b
li
n
In
stit
u
te
o
f
Tec
h
n
o
l
o
g
y
.
[we
b
]:
a
v
a
il
a
b
le
a
t
h
tt
p
s:/
/p
d
fs.se
m
a
n
ti
c
sc
h
o
lar.o
r
g
/c
9
3
4
/9
d
fe
8
0
c
7
6
2
2
4
9
b
d
b
0
3
0
1
8
5
c
4
8
1
6
5
3
c
f
b
2
b
a
6
.
p
d
f
[2
5
]
Co
rm
a
c
k
,
G
.
V.,
G
o
m
e
z
Hid
a
lg
o
,
J.
M
.
,
P
u
e
rtas
S
a
n
z
,
E.
,
"
S
p
a
m
F
i
lt
e
rin
g
fo
r
S
h
o
rt
M
e
ss
a
g
e
s
,"
i
n
Pr
o
c
.
o
f
t
h
e
1
6
t
h
ACM
CIKM
,
p
p
.
3
1
3
–
3
2
0
,
2
0
0
7
.
[2
6
]
Xu
,
Q.,
E
v
a
n
,
W.
X.
,
Qia
n
g
,
Y.,
Jia
c
h
u
n
,
D.,
Jie
p
in
g
,
Z.
,
"
S
M
S
S
p
a
m
De
tec
ti
o
n
Us
in
g
No
n
-
Co
n
te
n
t
F
e
a
tu
re
s
,"
in
IEE
E
In
t
e
ll
ig
e
n
t
S
y
ste
ms
,
v
o
l.
27
,
n
o
.
6
,
p
p
.
44
-
51
,
2
0
1
2
.
[2
7
]
Uy
sa
l,
A.
K.,
G
u
n
a
l,
S
.
,
Erg
in
,
S
.
,
&
S
o
ra
G
u
n
a
l,
E.
,
"
Th
e
imp
a
c
t
o
f
fe
a
tu
re
e
x
trac
ti
o
n
a
n
d
se
lec
ti
o
n
o
n
S
M
S
sp
a
m
fil
terin
g
,"
J
o
u
rn
a
l
E
lek
tro
n
ika
IR
El
e
k
tro
tec
h
n
ika
,
v
o
l.
19
,
n
o
.
5
,
p
p
.
67
–
72
,
2
0
1
3
.
[2
8
]
S
a
tt
e
rfield
,
B.
1
0
sp
a
m
fil
teri
n
g
m
e
th
o
d
s.
[we
b
]
a
v
a
il
a
b
le at
h
tt
p
:
//
ww
w.t
e
c
h
so
u
p
c
a
n
a
d
a
.
c
a
/en
/l
e
a
rn
in
g
_
c
e
n
ter/1
0
_
sfm
_
e
x
p
lai
n
e
d
[2
9
]
Co
o
k
,
D.
,
"
Ca
tch
i
n
g
S
p
a
m
b
e
fo
r
e
it
a
rriv
e
s:
Do
m
a
in
S
p
e
c
ifi
c
Dy
n
a
m
ic
Blac
k
li
sts,
Au
stra
li
a
n
C
o
mp
u
ter
S
o
c
iety
,
ACM,
2
0
0
6
.
[3
0
]
Ha
n
,
B.
,
Co
o
k
,
P
.
,
Ba
l
d
win
,
T
.
Lex
ica
l
No
rm
a
li
sa
ti
o
n
o
f
S
h
o
rt
Tex
t
M
e
ss
a
g
e
s.
ACM
Tran
sa
c
ti
o
n
o
n
I
n
telli
g
e
n
t
S
y
ste
m
s a
n
d
Tec
h
n
o
l
o
g
y
,
2
0
1
1
.
Article
A
[3
1
]
P
e
rlro
th
,
N.
S
p
a
m
In
v
a
d
e
s
la
st
Re
fu
g
e
,
th
e
Ce
ll
p
h
o
n
e
.
Ne
w
Yo
rk
Ti
m
e
s,
2
0
1
2
.
[we
b
]:
a
v
a
il
a
b
le
a
t
h
tt
p
:
//
p
re
v
iew
.
ti
n
y
u
rl
.
c
o
m
/7
n
wv
m
3
g
[3
2
]
Hu
a
n
g
,
D.,
G
a
n
,
Z.
,
C
h
o
w,
T.
W
.
S
.
,
"
En
h
a
n
c
e
d
fe
a
tu
re
se
lec
ti
o
n
m
o
d
e
ls
u
si
n
g
g
ra
d
ien
t
-
b
a
se
d
a
n
d
p
o
in
t
i
n
jec
ti
o
n
tec
h
n
iq
u
e
s,"
Ne
u
r
o
c
o
mp
u
ti
n
g
,
v
o
l.
7
1
,
p
p
.
3
1
1
4
–
3
1
2
3
,
2
0
0
8
.
[3
3
]
G
h
e
y
a
s
I.
A.,
S
m
it
h
,
L
.
S
.
,
"
F
e
a
tu
r
e
su
b
se
t
se
lec
ti
o
n
i
n
larg
e
d
ime
n
sio
n
a
l
it
y
d
o
m
a
in
s,
"
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
v
o
l.
4
3
,
p
p
.
5
–
1
3
,
2
0
1
0
.
[3
4
]
Va
fa
ie,
H.,
De
-
Jo
n
g
,
K.
G
e
n
e
ti
c
Alg
o
rit
h
m
a
s
a
To
o
l
fo
r
F
e
a
tu
r
e
S
e
lec
ti
o
n
in
M
a
c
h
in
e
Lea
rn
i
n
g
,
1
9
9
7
.
[we
b
]
re
se
a
rc
h
g
a
te.n
e
t/
p
u
b
li
c
a
ti
o
n
/2
7
2
2
3
5
3
_
G
e
n
e
ti
c
_
Alg
o
ri
th
m
s_
a
s_
T
o
o
l_
fo
r
_
F
e
a
tu
re
_
S
e
lec
ti
o
n
_
in
_
M
a
c
h
in
e
_
Lea
rn
in
g
[3
5
]
S
u
n
g
-
S
a
m
,
H.,
Wa
n
h
e
e
,
L.
,
M
y
u
n
g
-
M
o
o
k
,
H.,
"
Th
e
F
e
a
tu
re
S
e
l
e
c
ti
o
n
M
e
th
o
d
b
a
se
d
o
n
G
e
n
e
ti
c
Alg
o
rit
h
m
fo
r
Eff
icie
n
t
o
f
Te
x
t
Cl
u
ste
rin
g
a
n
d
Tex
t
Clas
sifica
ti
o
n
,
"
In
t.
J
o
u
rn
a
l
o
n
A
d
v
a
n
c
e
S
o
ft
C
o
mp
u
ti
n
g
A
p
p
li
c
a
ti
o
n
,
v
o
l
.
7
,
n
o
.
1
,
p
p
.
23
-
40
,
2
0
1
5
.
[3
6
]
Ca
th
e
rin
e
,
K.
,
F
ra
n
ç
o
is,
Y.,
G
é
ra
l
d
in
e
,
D.
,
"
N
o
rm
a
li
z
in
g
S
M
S
:
a
re
two
m
e
tap
h
o
rs
b
e
tt
e
r
th
a
n
o
n
e
,"
i
n
Pr
o
c
e
e
d
in
g
s
o
f
2
2
n
d
I
n
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
C
o
mp
u
ta
ti
o
n
a
l
L
i
n
g
u
isti
c
s
,
p
p
4
4
1
–
4
4
8
,
2
0
0
8
.
[3
7
]
Oju
g
o
,
A
.
A
a
n
d
Eb
o
k
a
,
A.O.
,
"
M
o
d
e
li
n
g
t
h
e
c
o
m
p
u
tati
o
n
a
l
so
l
u
ti
o
n
fo
r
m
a
rk
e
t
b
a
sk
e
t
a
ss
o
c
iat
iv
e
r
u
le
m
i
n
in
g
a
p
p
ro
a
c
h
e
s u
si
n
g
n
e
u
ra
l
n
e
two
r
k
,
"
Dig
it
a
l
T
e
c
h
n
o
lo
g
ies
,
v
o
l.
3
,
n
o
.
1
,
p
p
.
1
-
8
,
2
0
1
8
.
[3
8
]
Oju
g
o
,
A.A.
,
E
b
o
k
a
,
A.O.
,
"
I
n
v
e
n
to
ry
m
a
n
a
g
e
m
e
n
t
a
n
d
p
re
d
icti
o
n
u
sin
g
m
a
rk
e
t
b
a
sk
e
t
a
n
a
ly
sis
a
ss
o
c
iativ
e
r
u
l
e
m
in
in
g
:
m
e
m
e
ti
c
a
lg
o
rit
h
m
a
p
p
ro
a
c
h
,
"
In
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
In
fo
rm
a
t
ics
a
n
d
C
o
mm
u
n
ica
ti
o
n
T
e
c
h
n
o
l
o
g
y
(IJ
-
ICT
)
,
v
o
l.
8
,
n
o
.
3
,
2
0
1
9
.
[3
9
]
Oju
g
o
,
A.A.,
A
.
Eb
o
k
a
.
,
E.
O
k
o
n
ta.,
R.
Yo
r
o
.
,
F
.
Ag
h
wa
re
.
,
"
G
e
n
e
ti
c
a
lg
o
rit
h
m
ru
le
-
b
a
se
d
i
n
t
ru
sio
n
d
e
tec
ti
o
n
sy
ste
m
,
"
J
o
u
r
n
a
l
o
f
Eme
rg
in
g
T
r
e
n
d
s in
Co
mp
u
ti
n
g
In
f
o
rm
a
ti
o
n
S
y
ste
m
,
v
o
l.
3
,
n
o
.
8
,
pp
.
1
1
8
2
-
1
1
9
4
,
2
0
1
2
.
[4
0
]
Oju
g
o
,
A.A.,
Y
o
ro
,
R
.
E.
,
"
Co
m
p
u
tati
o
n
a
l
i
n
telli
g
e
n
c
e
in
st
o
c
h
a
stic
so
lu
ti
o
n
f
o
r
T
o
ro
i
d
a
l
N
-
q
u
e
e
n
,
"
I
n
telli
g
e
n
c
e
Co
mp
u
t
in
g
a
n
d
Ap
p
li
c
a
t
io
n
s
,
v
o
l.
2
,
n
o
.
1
,
p
p
.
46
-
56
,
2
0
1
3
.
[4
1
]
Oju
g
o
,
A.A.,
Emu
d
ian
u
g
h
e
,
J.,
Yo
ro
,
R.
E
.
,
Ok
o
n
ta,
E.
O.
,
Eb
o
k
a
,
A.O.
,
"
A
h
y
b
r
id
n
e
u
ra
l
n
e
two
rk
g
ra
v
it
a
ti
o
n
a
l
se
a
rc
h
a
lg
o
rit
h
m
f
o
r
ra
i
n
fa
ll
r
u
n
o
ff
m
o
d
e
li
n
g
a
n
d
sim
u
lati
o
n
in
h
y
d
ro
l
o
g
y
,
"
Pro
g
re
ss
in
In
telli
g
e
n
c
e
Co
mp
u
ti
n
g
a
n
d
Ap
p
li
c
a
ti
o
n
s
,
vo
l.
2
,
n
o
.
1
,
p
p
.
22
-
3
3
,
2
0
1
3
.
[4
2
]
Oju
g
o
,
A.A.,
D.
Oy
e
m
a
d
e
.
,
Y
o
r
o
,
R.
E
.
,
E
b
o
k
a
,
A.O.,
Ye
ro
k
u
n
,
M
.
O.,
Ug
b
o
h
,
E.
,
"
A
c
o
m
p
a
ra
ti
v
e
e
v
o
l
u
ti
o
n
a
ry
m
o
d
e
ls f
o
r
s
o
lv
i
n
g
S
u
d
o
k
u
,
"
Au
t
o
ma
ti
o
n
,
C
o
n
tr
o
l
&
In
telli
g
e
n
t
S
y
st
e
ms
,
v
o
l.
1
,
n
o
.
5
,
p
p
.
1
1
3
-
1
2
0
,
2
0
1
3
.
[4
3
]
Re
y
n
o
l
d
s,
R
.
,
"
In
tro
d
u
c
ti
o
n
to
c
u
lt
u
ra
l
a
lg
o
rit
h
m
s,
"
T
ra
n
s
a
c
ti
o
n
o
n
Evo
lu
ti
o
n
a
ry
Pro
g
r
a
mm
in
g
(IE
EE
)
,
p
p
.
1
3
1
-
1
3
9
,
1
9
9
4
.
[4
4
]
P
e
re
z
,
M
a
n
d
M
a
rwa
la,
T.
,
"
S
t
o
c
h
a
stic
o
p
ti
m
iza
ti
o
n
a
p
p
r
o
a
c
h
e
s
fo
r
so
l
v
in
g
S
u
d
o
k
u
,
"
in
Pr
o
c
e
e
d
in
g
o
f
IEE
E
Co
n
g
re
ss
o
n
Evo
lu
ti
o
n
a
ry
Co
m
p
u
ti
n
g
,
p
p
2
5
6
–
2
7
9
,
2
0
1
1
.
[4
5
]
G
o
m
e
z
Hid
a
lg
o
,
J
.
M
,
Brin
g
a
s,
G
.
C.
,
S
a
n
z
,
E
.
P
.
,
G
ra
c
ia,
F
.
C.
,
"
Co
n
ten
t
-
b
a
se
d
S
M
S
S
p
a
m
fil
terin
g
,
"
i
n
Pro
c
e
e
d
in
g
s
o
f
2
0
0
6
ACM
S
y
mp
o
siu
m o
n
Do
c
u
me
n
t
En
g
i
n
e
e
rin
g
,
p
p
1
0
7
-
1
1
4
,
2
0
0
6
.
[4
6
]
46
Va
fa
ie,
H.,
De
-
Jo
n
g
,
K.
"
G
e
n
e
ti
c
Alg
o
ri
th
m
a
s a
To
o
l
f
o
r
F
e
a
tu
r
e
S
e
lec
ti
o
n
,
"
in
M
a
c
h
i
n
e
L
e
a
rn
i
n
g
,
1
9
9
7
,
[4
7
]
G
h
e
y
a
s
I.
A.,
S
m
it
h
,
L
.
S
.
,
"
F
e
a
tu
r
e
su
b
se
t
se
lec
ti
o
n
i
n
larg
e
d
ime
n
sio
n
a
li
t
y
d
o
m
a
in
s,
"
Pa
tt
e
rn
Rec
o
g
n
it
io
n
,
v
o
l.
43
,
p
p
.
5
–
13
,
2
0
1
0
.
[4
8
]
Oju
g
o
,
A.A.,
A
.
Eb
o
k
a
.
,
R.
Yo
ro
.
,
M
.
Ye
ro
k
u
n
.
,
F
.
N.
Ef
o
z
ia.
,
"
Hy
b
r
id
m
o
d
e
l
fo
r
e
a
rly
d
ia
b
e
tes
d
iag
n
o
sis,
"
M
a
t
h
e
ma
ti
c
s a
n
d
Co
mp
u
ter
s in
S
c
ien
c
e
&
In
d
u
stry
,
v
o
l
.
50
,
pp
.
207
-
2
1
7
,
2
0
1
5
.
[4
9
]
P
h
il
li
p
,
K.
,
Hie
u
,
H.
M
.
,
"
Op
e
n
s
o
u
rc
e
to
o
l
k
it
f
o
r
sta
ti
stica
l
m
a
c
h
in
e
tran
sla
ti
o
n
,"
T
e
c
h
n
ica
l
re
p
o
rt,
An
n
u
a
l
M
e
e
ti
n
g
o
f
t
h
e
Asso
c
ia
t
io
n
fo
r C
o
mp
u
ta
ti
o
n
a
l
L
in
g
u
isti
c
s (A
CL
)
,
2
0
0
7
.
[5
0
]
S
e
th
i,
G
.
,
Bh
o
o
tn
a
,
V.
,
"
S
M
S
S
p
a
m
F
il
terin
g
Ap
p
li
c
a
ti
o
n
u
sin
g
An
d
r
o
id
,"
I
n
ter
n
a
ti
o
n
a
l
J
o
u
r
n
a
l
fo
r
C
o
mp
u
ter
S
c
ien
c
e
a
n
d
In
fo
rm
a
t
io
n
T
e
c
h
n
o
l
o
g
ies
,
v
o
l
.
5
,
n
o
.
3
,
p
p.
1
4
2
4
-
1
4
2
6
,
2
0
1
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.