T
E
L
K
O
M
N
I
K
A
T
elec
o
m
m
un
ica
t
io
n,
Co
m
pu
t
ing
,
E
lect
ro
nics
a
nd
Co
ntr
o
l
Vo
l.
19
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
,
p
p
.
1
5
5
~
16
2
I
SS
N:
1
6
9
3
-
6
9
3
0
,
ac
cr
ed
ited
First Gr
ad
e
b
y
Kem
en
r
is
tek
d
i
k
ti,
Dec
r
ee
No
: 2
1
/E/KPT
/2
0
1
8
DOI
: 1
0
.
1
2
9
2
8
/TE
L
KOM
NI
K
A.
v
1
9
i1
.
1
8
0
2
4
155
J
o
ur
na
l ho
m
ep
a
g
e
:
h
ttp
:
//jo
u
r
n
a
l.u
a
d
.
a
c.
id
/in
d
ex
.
p
h
p
/TELK
OM
N
I
K
A
A
hy
brid
na
ïv
e
Ba
y
es
ba
sed o
n simila
rity mea
sure
to o
ptimize
the
mi
x
ed
-
da
ta cl
a
ss
ificatio
n
F
a
t
im
a
E
l Ba
ra
k
a
z
,
O
m
a
r
B
o
utk
ho
um
,
Abdelm
a
j
id E
l Mo
uta
o
ua
kk
il
De
p
a
rtme
n
t
o
f
c
o
m
p
u
ti
n
g
,
Laro
se
ri
Lab
o
ra
t
o
ry
,
Un
i
v
e
rsity
C
h
o
u
a
i
b
Do
u
k
k
a
li
,
El
Ja
d
id
a
,
M
o
r
o
c
c
o
Art
icle
I
nfo
AB
S
T
RAC
T
A
r
ticle
his
to
r
y:
R
ec
eiv
ed
J
u
n
3
0
,
2
0
2
0
R
ev
is
ed
Sep
8
,
2
0
2
0
Acc
ep
ted
Sep
1
6
,
2
0
2
0
In
th
is
p
a
p
e
r,
a
h
y
b
rid
m
e
t
h
o
d
h
a
s
b
e
e
n
i
n
tro
d
u
c
e
d
t
o
i
m
p
ro
v
e
t
h
e
c
las
sifica
ti
o
n
p
e
rfo
rm
a
n
c
e
o
f
n
a
ï
v
e
Ba
y
e
s
(NB)
fo
r
th
e
m
ix
e
d
d
a
tas
e
t
a
n
d
m
u
lt
i
-
c
las
s
p
r
o
b
lem
s.
Th
is
p
r
o
p
o
se
d
m
e
th
o
d
re
li
e
s
o
n
a
sim
il
a
rit
y
m
e
a
su
re
wh
ich
is
a
p
p
l
ied
t
o
p
o
rt
io
n
s
th
a
t
a
re
n
o
t
c
o
rre
c
tl
y
c
las
sified
b
y
NB
.
S
i
n
c
e
th
e
d
a
ta
c
o
n
tain
s
a
m
u
l
ti
-
v
a
l
u
e
d
sh
o
rt
tex
t
wi
th
ra
re
wo
r
d
s
th
a
t
li
m
it
th
e
NB
p
e
rfo
rm
a
n
c
e
,
we
h
a
v
e
e
m
p
l
o
y
e
d
a
n
a
d
a
p
ted
se
lec
ti
v
e
c
las
sifier
b
a
se
d
o
n
sim
il
a
rit
ies
(
CS
BS
)
c
las
sifier
to
e
x
c
e
e
d
th
e
NB
li
m
it
a
ti
o
n
s
a
n
d
in
c
lu
d
e
d
t
h
e
ra
re
wo
rd
s
in
t
h
e
c
o
m
p
u
tati
o
n
.
T
h
is ac
ti
o
n
h
a
s b
e
e
n
a
c
h
iev
e
d
b
y
tra
n
sfo
rm
in
g
th
e
fo
rm
u
la fro
m
t
h
e
p
r
o
d
u
c
t
o
f
t
h
e
p
r
o
b
a
b
i
li
ti
e
s o
f
th
e
c
a
teg
o
rica
l
v
a
riab
le t
o
it
s
su
m
we
ig
h
ted
b
y
n
u
m
e
rica
l
v
a
riab
le.
Th
e
p
r
o
p
o
se
d
a
lg
o
rit
h
m
h
a
s
b
e
e
n
e
x
p
e
rime
n
ted
o
n
c
a
rd
p
a
y
m
e
n
t
tran
sa
c
ti
o
n
d
a
ta
t
h
a
t
c
o
n
tain
s
th
e
lab
e
l
o
f
tran
sa
c
ti
o
n
s:
th
e
m
u
lt
i
-
v
a
lu
e
d
sh
o
rt
tex
t
a
n
d
t
h
e
tran
sa
c
ti
o
n
a
m
o
u
n
t.
Ba
se
d
o
n
K
-
fo
ld
c
ro
ss
v
a
li
d
a
ti
o
n
,
t
h
e
e
v
a
lu
a
ti
o
n
re
su
lt
s
c
o
n
f
irm
th
a
t
th
e
p
ro
p
o
se
d
m
e
th
o
d
a
c
h
iev
e
d
b
e
tt
e
r
re
su
lt
s
in
term
s
o
f
p
re
c
isio
n
,
re
c
a
ll
,
a
n
d
F
-
sc
o
re
c
o
m
p
a
re
d
to
NB
a
n
d
C
S
BS
c
las
sifiers
se
p
a
ra
tely
.
Be
sid
e
s,
th
e
fa
c
t
o
f
c
o
n
v
e
rti
n
g
a
p
r
o
d
u
c
t
f
o
rm
t
o
a
su
m
g
iv
e
s
m
o
re
c
h
a
n
c
e
to
ra
r
e
wo
rd
s
t
o
o
p
ti
m
ize
th
e
tex
t
c
las
sifica
ti
o
n
,
wh
ich
is
a
n
o
th
e
r
a
d
v
a
n
tag
e
o
f
th
e
p
ro
p
o
se
d
m
e
th
o
d
.
K
ey
w
o
r
d
s
:
C
S
B
S
Mix
ed
d
ata
Mu
lti
-
class
if
icatio
n
Naïv
e
B
ay
es
Sh
o
r
t te
x
t
Similar
ity
-
b
ased
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
:
Fatim
a
E
l Bar
ak
az
Dep
ar
tm
en
t o
f
co
m
p
u
tin
g
,
L
a
r
o
s
er
i L
ab
o
r
ato
r
Facu
lty
o
f
th
e
Scien
ce
s
C
h
o
u
aib
Do
u
k
k
ali
Un
iv
er
s
ity
J
ab
r
an
Kh
alil Jab
r
an
Av
e
n
u
e
B
.
P 2
9
9
-
2
4
0
0
0
,
E
l Jad
id
a,
Mo
r
o
cc
o
E
m
ail:
el.
b
ar
ak
az
.
f
atim
a@
g
m
ail.
co
m
1.
I
NT
RO
D
UCT
I
O
N
I
n
m
a
n
y
ca
s
es,
d
atasets
co
n
s
is
t o
f
b
o
th
n
u
m
er
ical
a
n
d
ca
te
g
o
r
ical
v
ar
iab
les.
Ma
n
y
class
if
ie
r
s
,
s
u
ch
as
lin
ea
r
r
eg
r
ess
io
n
,
s
u
p
p
o
r
t
v
ec
t
o
r
r
e
g
r
ess
io
n
,
an
d
k
-
n
ea
r
est
n
e
ig
h
b
o
u
r
(
KNN
)
a
r
e
well
-
d
e
f
in
ed
an
d
v
alid
ated
f
o
r
th
e
co
m
p
u
tatio
n
o
f
n
u
m
er
ical
v
ar
iab
les.
Fo
r
th
ese
alg
o
r
ith
m
s
,
it
is
ea
s
ier
to
estab
lis
h
th
e
r
elatio
n
s
b
etwe
en
a
tar
g
et
an
d
its
p
r
e
d
icto
r
s
wh
en
b
o
th
a
r
e
n
u
m
er
ical.
Ho
wev
e
r
,
th
e
n
u
m
er
ical
o
p
er
atio
n
s
ar
e
n
o
t
ap
p
licab
le
t
o
ca
teg
o
r
ical
v
a
r
iab
les,
ex
ce
p
t
i
f
it
h
as
b
ee
n
c
o
n
v
e
r
ted
to
n
u
m
er
ic
o
n
e
u
s
in
g
co
d
i
n
g
s
y
s
te
m
s
s
u
ch
as
d
u
m
m
y
co
d
in
g
,
ef
f
ec
ts
c
o
d
in
g
,
o
r
ev
en
co
n
tr
ac
t
co
d
i
n
g
[
1
,
2
]
.
An
o
th
e
r
ap
p
r
o
ac
h
is
b
ased
o
n
s
im
ilar
i
ty
an
d
d
is
s
im
ilar
ity
m
ea
s
u
r
es
b
etwe
en
ca
teg
o
r
ical
an
d
n
u
m
er
ical
v
ar
iab
les,
wh
e
r
e
th
e
d
ata
m
atr
ix
is
tr
an
s
f
o
r
m
ed
in
to
a
d
is
tan
ce
co
n
f
ig
u
r
a
tio
n
m
atr
i
x
af
ter
a
p
p
ly
in
g
s
im
ilar
o
r
d
is
s
im
ilar
f
u
n
ctio
n
s
[
3
-
5
]
.
Ho
wev
er
,
th
e
p
r
ev
i
o
u
s
ap
p
r
o
ac
h
es
in
cr
ea
s
e
th
e
n
u
m
b
er
o
f
p
r
ed
icto
r
s
wh
en
ca
teg
o
r
ical
v
ar
iab
les
ar
e
n
u
m
e
r
o
u
s
.
I
n
th
is
ca
s
e,
th
e
co
d
i
n
g
s
y
s
tem
s
p
r
o
p
o
s
ed
ad
d
itio
n
al
s
tep
s
to
r
ed
u
ce
th
e
n
u
m
b
er
o
f
p
r
ed
icto
r
s
[
6
,
7
]
.
T
h
o
u
g
h
t
h
o
s
e
ap
p
r
o
ac
h
es
d
o
n
o
t
ap
p
l
y
to
m
u
lti
-
v
alu
ed
ca
te
g
o
r
ical
v
ar
iab
l
es
th
at
co
n
tain
m
o
r
e
th
an
a
s
in
g
le
wo
r
d
,
Mik
o
lo
v
p
r
o
p
o
s
es
th
e
W
o
r
d
2
Vec
m
o
d
el
t
h
at
r
ep
r
esen
ts
th
e
tex
t
in
a
v
ec
to
r
f
o
r
m
at
an
d
s
av
es
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
1
5
5
-
16
2
156
th
e
s
y
n
tax
an
d
th
e
s
em
an
tic
m
ea
n
in
g
o
f
n
at
u
r
al
lan
g
u
a
g
e
[
8
,
9
]
.
T
h
e
W
o
r
d
2
v
ec
t
is
ap
p
licab
le
ev
en
f
o
r
a
d
is
o
r
d
er
ed
m
u
lti
-
wo
r
d
tex
t,
w
h
er
e
lin
g
u
is
tic
an
d
s
em
an
tic
r
u
les ar
e
n
o
t r
esp
ec
ted
.
I
n
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
an
d
cla
s
s
if
icatio
n
co
n
tex
t,
s
o
m
e
ap
p
r
o
ac
h
es
r
ely
in
g
o
n
s
im
ilar
ity
m
ea
s
u
r
e
class
if
icatio
n
ar
e
ap
p
ly
in
g
co
s
in
e
an
d
s
tr
in
g
s
im
ilar
ity
t
o
m
ea
s
u
r
e
th
e
d
is
tan
ce
b
etwe
en
v
ec
to
r
s
.
Oth
e
r
ap
p
r
o
ac
h
es
p
r
o
p
o
s
e
u
tter
ly
h
y
b
r
id
class
if
ier
s
d
ep
en
d
i
n
g
o
n
t
h
e
s
im
ilar
ity
-
b
ased
m
ea
s
u
r
e.
I
n
th
is
co
n
tex
t,
SB
C
a
lg
o
r
ith
m
(
s
im
ilar
ity
-
b
ased
cl
ass
if
ier
)
[
1
0
]
an
d
C
SB
S
(
s
el
ec
tiv
e
class
if
ier
b
ased
o
n
s
i
m
ilar
ities
)
ar
e
two
alg
o
r
ith
m
s
th
at
co
m
b
in
e
th
e
m
ea
s
u
r
es
o
f
eq
u
ality
,
r
eliab
ilit
y
,
an
d
d
en
s
ity
to
class
if
y
v
ec
to
r
s
.
B
o
th
class
if
ier
s
s
h
o
w
ex
ce
llen
t p
er
f
o
r
m
a
n
ce
i
n
ter
m
s
o
f
tex
t c
lass
if
icatio
n
[
1
1
,
1
2
]
.
On
th
e
o
th
er
h
a
n
d
,
n
aïv
e
B
ay
e
s
(
NB
)
is
s
till
h
ig
h
ly
u
s
ef
u
l
to
class
if
y
th
e
ca
teg
o
r
ical
an
d
n
u
m
er
ical
v
ar
iab
les
[
1
3
]
,
esp
ec
ially
co
m
p
ar
e
its
p
er
f
o
r
m
an
ce
with
o
th
er
class
if
ier
s
.
I
n
g
e
n
er
al,
i
d
en
tify
in
g
s
u
itab
le
s
im
ilar
ity
m
ea
s
u
r
es
b
etwe
en
ca
teg
o
r
ical
v
a
r
iab
les
o
r
b
e
twee
n
ca
teg
o
r
ical
an
d
n
u
m
e
r
ical
v
ar
iab
les
is
co
n
s
id
er
ed
a
co
m
p
lex
c
h
allen
g
e.
T
o
ad
d
r
ess
th
is
ch
allen
g
e,
a
h
y
b
r
id
NB
m
o
d
el
h
as
b
ee
n
c
o
n
s
tr
u
cted
u
s
in
g
an
ad
ap
ted
C
SB
S.
W
h
er
e,
th
e
ca
teg
o
r
ical
v
ar
ia
b
le
is
a
s
h
o
r
t
te
x
t,
an
d
we
ap
p
ly
to
k
en
izatio
n
an
d
s
to
p
-
w
o
r
d
s
in
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
p
h
ase.
Fo
r
c
lass
if
icatio
n
,
NB
h
as
b
ee
n
u
s
ed
to
tr
ain
o
u
r
m
o
d
el
th
at
u
s
ed
o
n
ly
th
e
ca
teg
o
r
ical
v
ar
iab
le.
An
d
f
o
r
th
e
p
o
r
tio
n
s
th
at
ar
e
p
o
o
r
ly
ex
p
lain
ed
b
y
NB
,
th
e
ad
ap
ted
C
SB
S
in
ter
v
en
ed
in
th
e
s
ec
o
n
d
p
h
ase
to
im
p
r
o
v
e
th
e
class
if
ic
atio
n
b
y
i
n
clu
d
in
g
n
u
m
er
ical
v
ar
iab
le.
T
h
e
o
r
g
an
izatio
n
o
f
th
e
p
ap
er
i
s
as
f
o
llo
ws.
Sectio
n
2
b
r
ief
ly
p
r
esen
ts
th
e
r
elate
d
wo
r
k
s
we
ad
d
r
ess
in
th
e
p
a
p
er
.
Sectio
n
3
p
r
o
v
id
e
s
d
if
f
er
en
t
m
eth
o
d
s
u
s
ed
in
th
i
s
s
tu
d
y
.
Sectio
n
4
in
tr
o
d
u
ce
s
a
d
escr
ip
tio
n
o
f
th
e
p
r
o
p
o
s
ed
h
y
b
r
id
n
aïv
e
B
ay
es
alg
o
r
ith
m
.
Sectio
n
5
s
h
o
ws
th
e
ex
p
er
im
en
tal
r
esu
lts
o
f
ap
p
l
y
in
g
alg
o
r
ith
m
s
o
n
th
e
r
ea
l c
r
ed
it c
ar
d
d
ataset.
T
h
e
last
s
ec
tio
n
p
r
esen
ts
th
e
co
n
clu
d
in
g
r
e
m
ar
k
s
2.
L
I
T
E
R
AT
U
RE
R
E
VI
E
W
2
.
1
.
Ca
t
eg
o
ric
a
l
v
a
ria
ble a
n
d si
m
ila
rit
y
m
ea
s
ures
C
ateg
o
r
ical
an
d
q
u
alitativ
e
m
u
lti
-
v
alu
ed
d
ata
h
av
e
b
ee
n
s
tu
d
i
ed
f
o
r
a
lo
n
g
tim
e
i
n
d
if
f
e
r
en
t
c
o
n
tex
ts
.
C
o
m
p
u
tin
g
s
im
ilar
ity
h
as
a
lo
n
g
h
is
to
r
y
,
s
tar
ted
with
ch
i
-
s
q
u
ar
e
in
th
e
late
1
8
0
0
s
th
at
is
f
r
eq
u
en
tly
u
s
ed
f
o
r
in
d
ep
en
d
en
ce
test
s
b
etwe
en
ca
teg
o
r
ical
v
a
r
iab
les.
Als
o
,
Pear
s
o
n
'
s
ch
i
-
s
q
u
ar
e
h
as
k
n
o
wn
m
an
y
im
p
r
o
v
em
en
ts
th
at
h
an
d
led
s
ev
er
al
d
ata
s
im
ilar
ity
ca
s
es
[
1
4
]
.
So
f
ar
,
class
ical
ca
teg
o
r
ical
d
ata
h
as
ch
an
g
ed
.
No
tab
ly
,
th
e
ca
teg
o
r
ies
n
u
m
b
er
o
f
a
q
u
alita
tiv
e
v
ar
iab
le
h
as
in
cr
ea
s
ed
to
im
p
o
r
tan
t
v
alu
es.
Als
o
,
th
e
ca
t
eg
o
r
ical
v
a
r
iab
les
s
tar
t
to
in
clu
d
e
m
u
lti
-
v
alu
ed
s
h
o
r
t
tex
t
[
1
0
]
,
s
o
m
a
n
y
lim
itat
io
n
s
ar
e
e
x
p
o
s
ed
.
Fo
r
tu
n
ately
,
d
if
f
er
e
n
t
m
eth
o
d
s
b
ased
o
n
s
im
ilar
ity
m
ea
s
u
r
es
h
av
e
b
ee
n
p
r
o
p
o
s
ed
to
o
v
er
co
m
e
th
is
ch
allen
g
e.
Ho
wev
er
,
t
h
e
p
er
f
o
r
m
an
ce
o
f
th
o
s
e
m
eth
o
d
s
d
e
p
en
d
s
lar
g
ely
o
n
d
ata
ch
ar
ac
ter
is
tics
[
1
5
]
.
Fo
r
th
e
m
ain
d
ata
ch
ar
ac
ter
i
s
tics
,
we
co
n
s
id
er
a
ca
teg
o
r
i
ca
l
d
ata
c
o
n
tain
s
N
o
b
jects,
with
p
ca
teg
o
r
ical
v
a
r
iab
les.
W
h
ile
d
en
o
tes
th
e
ℎ
v
ar
iab
le,
an
d
Ω
th
e
s
et
o
f
d
if
f
er
e
n
t
v
al
u
es
in
A
k
an
d
its
ca
r
d
in
ality
.
T
h
e
k
ey
c
h
ar
ac
ter
i
s
tics
ar
e
th
e
f
o
llo
win
g
:
−
(
)
: T
h
e
n
u
m
b
er
o
f
tim
es th
e
attr
i
b
u
te
to
tak
e
x
as a
v
alu
e
i
n
a
d
ata
s
et.
−
(
)
:
T
h
e
s
am
p
le
p
r
o
b
a
b
ilit
y
o
f
to
tak
e
x
as a
v
alu
e
i
n
a
d
ata
s
et,
an
d
it is
g
iv
en
b
y
;
(
)
=
(
)
(
1
)
−
2
(
)
:
An
o
th
er
p
r
o
b
a
b
ilit
y
f
o
r
m
u
la
o
f
to
tak
e
x
as a
v
alu
e
in
th
e
g
iv
en
d
ata
s
et,
an
d
it’s g
iv
en
b
y
;
2
(
)
=
(
)
(
(
)
−
1
)
(
−
1
)
(
2
)
I
n
g
en
er
al,
to
m
ea
s
u
r
e
a
s
im
ilar
ity
v
alu
e
b
etwe
en
two
d
ata
in
s
tan
ce
s
X
an
d
Y
b
elo
n
g
in
g
to
a
d
ata
s
et,
all
u
s
ed
m
ea
s
u
r
es r
esp
ec
t th
e
f
o
llo
win
g
f
o
r
m
:
(
,
)
=
∑
=
1
(
,
)
(
3
)
(
,
)
:
T
h
e
p
er
-
attr
i
b
u
te
s
im
ilar
ity
b
e
twee
n
two
v
alu
es f
o
r
th
e
ca
teg
o
r
ical
attr
ib
u
te
.
: T
h
e
weig
h
t a
s
s
ig
n
ed
to
th
e
a
ttrib
u
te
,
th
er
ea
f
ter
,
it is
f
ix
ed
t
o
1
/p
.
T
h
e
a
b
o
v
e
e
x
p
r
e
s
s
i
o
n
h
a
s
b
e
en
t
h
e
p
o
i
n
t
o
f
m
a
n
y
s
t
u
d
i
es
a
n
d
i
s
i
n
te
r
p
r
e
t
e
d
i
n
t
o
d
i
f
f
e
r
e
n
t
f
u
n
c
t
i
o
n
s
d
e
p
e
n
d
i
n
g
o
n
t
h
e
d
a
t
a
.
W
h
e
r
e
t
h
r
e
e
e
x
a
m
p
l
e
s
o
f
(
,
)
a
n
d
h
a
v
e
b
e
e
n
m
e
n
t
i
o
n
ed
.
S
t
a
r
ti
n
g
wi
t
h
t
h
e
s
a
m
p
l
e
o
n
e
,
t
h
e
o
v
e
r
l
a
p
m
e
a
s
u
r
e
:
i
t
c
o
u
n
t
s
t
h
e
n
u
m
b
e
r
o
f
a
t
t
r
i
b
u
t
e
s
t
h
at
m
a
t
c
h
i
n
t
h
e
t
w
o
d
a
t
a
i
n
s
t
a
n
c
e
s
,
u
s
i
n
g
t
h
e
m
e
a
s
u
r
e
(
4
)
:
(
,
)
=
{
1
=
0
≠
(
4
)
T
h
e
Go
o
d
all
4
:
m
ea
s
u
r
e:
aim
s
to
n
o
r
m
alize
th
e
s
im
ilar
ity
b
et
wee
n
two
o
b
jects,
b
ased
o
n
t
h
e
p
r
o
b
a
b
ilit
y
wh
er
e
th
e
s
im
ilar
ity
v
alu
e
o
b
s
er
v
ed
c
o
u
ld
b
e
g
en
er
ate
d
f
r
o
m
a
r
a
n
d
o
m
s
am
p
le
o
f
two
p
o
i
n
ts
[
1
6
]
.
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
A
Hyb
r
id
n
a
ïve
B
a
ye
s
b
a
s
ed
o
n
s
imila
r
ity
mea
s
u
r
e
to
… (
F
a
tima
E
l B
a
r
a
ka
z
)
157
(
,
)
=
{
2
(
)
=
0
≠
(
5
)
2
.
2
.
B
a
nk
cus
t
o
m
er
t
r
a
ns
a
ct
io
ns
cla
s
s
if
ica
t
io
n
C
u
s
to
m
er
class
if
icat
io
n
an
d
t
ar
g
etin
g
ar
e
wid
ely
ap
p
lied
in
p
r
ac
tice.
I
n
r
ec
en
t
y
ea
r
s
,
b
an
k
s
h
av
e
in
v
ested
in
th
eir
d
ata
an
d
ap
p
li
ed
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
f
o
r
cu
s
to
m
er
id
en
tific
atio
n
,
wh
er
e
th
ey
ac
h
iev
e
d
f
r
u
i
tf
u
l r
esu
lts
.
E
s
k
in
et
a
l
.
[
1
7
]
p
r
o
p
o
s
e
th
e
u
s
e
o
f
a
r
an
d
o
m
s
am
p
lin
g
m
eth
o
d
to
im
p
r
o
v
e
th
e
s
u
p
p
o
r
t
v
ec
to
r
m
ac
h
in
e
(
SVM
)
m
o
d
el,
f
o
r
b
an
k
cu
s
to
m
e
r
ch
u
r
n
p
r
e
d
ictio
n
.
I
n
th
e
s
am
e
co
n
tex
t,
De
C
aig
n
y
et
a
l.
[
1
8
,
1
9
]
s
u
g
g
ested
a
co
m
b
i
n
atio
n
o
f
b
o
th
m
eth
o
d
s
o
f
lo
g
is
tic
r
eg
r
ess
io
n
an
d
d
ec
is
io
n
tr
ee
s
.
W
h
ile
f
o
r
f
r
a
u
d
d
etec
tio
n
,
J
u
r
g
o
v
s
k
y
et
a
l.
s
h
o
we
d
h
o
w
u
s
in
g
lo
n
g
s
h
o
r
t
-
ter
m
m
em
o
r
y
(
L
STM
)
im
p
r
o
v
es
th
e
d
etec
tio
n
ac
cu
r
ac
y
u
s
ed
th
e
R
an
d
o
m
Fo
r
est
class
if
ier
an
d
in
c
o
r
p
o
r
ated
tr
a
n
s
ac
tio
n
s
e
q
u
en
ce
s
[
2
0
]
.
Oth
er
s
f
o
c
u
s
o
n
th
e
p
r
e
-
p
r
o
ce
s
s
in
g
p
ar
t,
f
o
r
th
e
cr
ed
it
ap
p
licatio
n
s
wh
er
e
v
ar
io
u
s
in
f
o
r
m
atio
n
a
b
o
u
t
p
a
y
m
en
t
ap
p
ea
r
in
q
u
alitativ
e,
ca
teg
o
r
ica
l
attr
ib
u
tes.
In
g
e
n
er
al,
t
h
e
clas
s
if
icatio
n
o
f
cu
s
to
m
er
tr
an
s
ac
tio
n
s
co
u
ld
b
e
u
s
ed
to
ex
ten
d
a
s
y
s
tem
t
h
at
ca
n
co
m
p
u
te
s
o
cio
ec
o
lo
g
ical
im
p
ac
t
f
r
o
m
ca
te
g
o
r
ized
tr
an
s
ac
tio
n
s
,
an
d
p
r
o
v
id
e
m
o
r
e
an
aly
s
is
ab
o
u
t
t
h
e
co
m
m
u
n
ity
an
d
its
r
elatio
n
s
h
ip
with
th
e
g
eo
g
r
ap
h
ic
l
o
ca
tio
n
.
An
d
it is
u
s
ed
in
r
is
k
m
a
n
ag
e
m
en
t,
s
ec
u
r
ity
a
n
d
f
r
au
d
d
etec
t
io
n
,
o
r
co
m
m
er
cia
l d
ep
ar
tm
en
ts
b
a
n
k
to
i
d
en
tify
cu
s
to
m
er
b
eh
a
v
io
u
r
.
2
.
3
.
T
ex
t
cl
a
s
s
if
ica
t
io
n
T
ex
t
class
if
icatio
n
is
a
f
u
n
d
am
en
tal
task
in
n
atu
r
al
lan
g
u
a
g
e
p
r
o
ce
s
s
in
g
.
I
t
is
wid
ely
ap
p
lied
in
s
en
tim
en
t
an
aly
s
is
,
r
ec
o
m
m
en
d
atio
n
an
d
Fra
u
d
a
n
d
s
p
am
d
et
ec
tio
n
[
2
1
,
2
2
]
.
Ma
ch
in
e
lear
n
in
g
in
clu
d
es
m
an
y
ap
p
r
o
ac
h
es f
o
r
tex
t c
lass
if
icatio
n
as NB,
s
u
p
p
o
r
t v
ec
to
r
m
ac
h
in
e
,
an
d
o
th
er
alg
o
r
ith
m
s
.
L
a
tely
,
d
ee
p
lear
n
in
g
h
as
s
h
o
wn
an
o
v
er
-
p
er
f
o
r
m
in
g
co
m
p
a
r
ed
to
tr
a
d
itio
n
al
m
ac
h
in
e
lear
n
in
g
m
eth
o
d
s
.
An
d
th
at
is
n
o
ticed
in
th
e
k
n
o
wn
m
eth
o
d
s
b
el
o
w:
co
n
v
o
lu
tio
n
al
n
e
u
r
al
n
etwo
r
k
s
(
C
NNs)
[
2
3
]
,
r
e
cu
r
r
e
n
t
n
e
u
r
al
n
et
wo
r
k
s
(
R
NNs),
an
d
th
e
co
m
b
in
atio
n
o
f
C
NNs a
n
d
R
NNs [
2
4
]
.
Alth
o
u
g
h
th
e
g
r
ea
t
s
u
cc
ess
h
a
s
s
h
o
wn
in
p
r
o
ce
s
s
in
g
lo
n
g
s
e
n
ten
ce
s
,
it
was
n
o
t
th
e
ca
s
e
f
o
r
s
h
o
r
t
tex
t
ex
p
lain
ed
b
y
th
e
d
ata
s
p
ar
s
ity
p
r
o
b
lem
.
R
ec
en
tly
,
m
an
y
wo
r
k
s
h
av
e
b
ee
n
ap
p
ly
in
g
v
ar
io
u
s
tex
t
p
r
esen
tatio
n
m
o
d
e
ls
to
ex
tr
ac
t
m
o
r
e
in
f
o
r
m
atio
n
f
r
o
m
s
h
o
r
t
tex
t
[
2
5
,
2
6
]
.
As
m
en
tio
n
e
d
ea
r
lier
,
s
o
m
e
a
r
e
b
ase
d
o
n
f
ea
tu
r
es
f
r
o
m
m
u
ltip
le
asp
e
cts,
an
d
o
th
er
s
ar
e
b
ased
o
n
tr
a
n
s
f
o
r
m
in
g
wo
r
d
s
in
to
v
ec
to
r
s
.
Ho
wev
er
,
th
e
tex
t
r
ep
r
esen
tatio
n
s
s
till
f
ac
e
th
e
d
ata
s
p
ar
s
ity
p
r
o
b
lem
wh
en
th
e
d
ata
in
clu
d
e
m
an
y
n
ew
an
d
r
ar
e
wo
r
d
s
[
2
7
]
.
I
n
o
u
r
ca
s
e,
th
e
tex
t
in
q
u
esti
o
n
is
ca
teg
o
r
ized
as
a
s
h
o
r
t
tex
t,
wh
er
e
th
e
v
a
r
iab
le
is
v
er
y
m
u
lti
-
v
alu
ed
.
So
,
t
h
e
n
ew
an
d
r
a
r
e
wo
r
d
s
ca
u
s
e
a
s
er
io
u
s
class
if
icatio
n
p
r
o
b
lem
.
I
n
th
is
p
ap
er
,
we
p
r
o
p
o
s
e
a
h
y
b
r
id
NB
class
if
ier
b
ased
o
n
a
d
ap
ted
s
im
ilar
ity
m
ea
s
u
r
es a
p
p
lied
to
ca
r
d
tr
an
s
ac
tio
n
p
ay
m
e
n
t d
ata.
3.
RE
S
E
ARCH
M
E
T
H
O
D
3
.
1
.
N
a
ïv
e
B
a
y
es
cla
s
s
if
ier
Naiv
e
B
ay
es
i
s
a
s
u
p
er
v
is
ed
lear
n
in
g
alg
o
r
ith
m
b
ased
o
n
a
p
r
o
b
a
b
ilis
tic
clas
s
i
f
icatio
n
.
T
h
is
class
if
ier
is
ex
tr
em
ely
f
aster
co
m
p
ar
ed
to
o
th
er
m
eth
o
d
s
.
NB
aim
s
to
ca
lcu
late
th
e
j
o
in
t
p
r
o
b
ab
ilit
ies
o
f
wo
r
d
s
an
d
ca
teg
o
r
ies to
esti
m
ate
ea
ch
ca
teg
o
r
y
th
e
tex
t w
ill b
e
af
f
ec
ted
.
T
h
e
‘
Naiv
e’
ex
p
r
e
s
s
io
n
is
d
u
e
to
th
e
f
ac
t
th
e
wo
r
d
s
ar
e
in
d
ep
e
n
d
e
n
ts
.
I
n
o
th
er
wo
r
d
s
,
th
e
co
n
d
i
tio
n
al
p
r
o
b
a
b
ilit
y
o
f
a
wo
r
d
f
r
o
m
a
ca
teg
o
r
y
is
ass
u
m
ed
to
b
e
in
d
ep
en
d
en
t o
f
th
e
co
n
d
itio
n
al
p
r
o
b
ab
ilit
ies o
f
o
th
er
wo
r
d
s
f
r
o
m
th
e
s
am
e
c
ateg
o
r
y
[
2
8
]
.
3
.
2
.
CSB
S
cla
s
s
if
ier
T
h
e
C
SB
S
i
s
a
clas
s
if
ier
b
ased
o
n
s
im
ilar
ity
m
ea
s
u
r
es,
in
wh
ich
th
e
tr
ea
ted
lim
itatio
n
s
s
h
o
wn
f
o
r
s
h
o
r
t
tex
t
class
if
icatio
n
ar
e
b
ased
o
n
th
r
ee
m
ea
s
u
r
es:
eq
u
a
lity
,
r
eliab
ilit
y
,
an
d
d
en
s
ity
[
1
0
]
.
Fo
r
th
e
s
ak
e
o
f
n
o
tatio
n
,
f
o
r
a
class
C
,
we
d
i
s
tin
g
u
is
h
b
etwe
en
th
e
am
p
litu
d
e
an
d
th
e
o
wn
am
p
litu
d
e
∗
,
W
h
en
th
e
o
wn
am
p
litu
d
e
o
f
a
g
iv
e
n
attr
ib
u
te
s
er
v
es
to
p
r
ed
ict
wh
eth
er
th
is
is
r
eliab
le
r
elativ
ely
co
m
p
ar
e
to
o
t
h
er
attr
i
b
u
tes,
an
d
th
at
th
r
o
u
g
h
elim
in
atin
g
th
e
in
ter
v
als co
n
tain
i
n
g
v
al
u
es b
e
lo
n
g
in
g
to
th
e
o
th
er
class
es f
r
o
m
.
I
n
C
SB
S
clas
s
if
ier
[
1
1
]
,
eq
u
ality
is
m
ea
s
u
r
ed
b
y
th
e
n
u
m
b
e
r
o
f
o
b
jects
s
h
ar
in
g
th
e
s
am
e
v
a
lu
es
p
e
r
attr
ib
u
te.
T
h
e
h
ig
h
er
th
e
m
ea
s
u
r
e
is
,
th
e
m
o
r
e
th
e
v
al
u
es
in
d
icate
th
e
m
em
b
er
s
h
ip
to
th
e
cl
ass
.
Ho
wev
er
,
th
e
o
wn
am
p
l
itu
d
e
in
d
icate
s
th
e
r
e
liab
ly
o
f
th
e
attr
ib
u
te.
At
th
e
s
am
e
tim
e,
an
in
s
tan
ce
is
m
o
r
e
lik
ely
to
b
elo
n
g
to
a
class
wh
en
th
e
attr
ib
u
te
v
alu
e
is
in
clu
d
ed
i
n
its
o
wn
a
m
p
litu
d
e.
W
h
ile
th
e
d
en
s
ity
o
f
th
e
m
em
b
er
s
h
ip
o
f
a
n
in
s
tan
ce
to
a
class
C
is
m
ea
s
u
r
ed
u
s
in
g
th
e
(
6
)
:
=
1
∑
=
1
×
+
∗
−
(
,
̅
̅
̅
)
+
+
(
6
)
wh
er
e
: M
is
th
e
n
u
m
b
er
o
f
attr
ib
u
tes.
N
: T
h
e
n
u
m
b
er
o
f
in
s
tan
ce
s
.
: T
h
e
co
ef
f
icien
t
o
f
r
eliab
ilit
y
o
n
to
p
r
ed
ict
C
.
: T
h
e
n
u
m
b
er
o
f
in
s
tan
ce
s
th
at
tak
e
th
e
v
alu
e
o
f
p
r
o
ce
s
s
ed
in
s
tan
ce
o
n
attr
ib
u
te
ℎ
p
er
C
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
1
5
5
-
16
2
158
∗
: T
h
e
o
wn
am
p
litu
d
e
o
f
C
p
er
attr
ib
u
te
ℎ
.
: T
h
e
s
im
p
le
am
p
litu
d
e
o
f
C
p
er
attr
ib
u
te
ℎ
.
̅
:
T
h
e
ce
n
ter
o
f
C
p
er
attr
ib
u
te
ℎ
.
: T
h
e
n
u
m
b
er
o
f
in
s
tan
ce
s
th
at
tak
e
th
e
v
alu
e
o
f
p
r
o
ce
s
s
ed
in
s
tan
ce
o
n
attr
ib
u
te
ℎ
.
:
A
v
er
y
s
m
all
p
o
s
itiv
e
v
alu
e.
Fin
ally
,
th
e
class
o
f
a
g
iv
en
in
s
tan
ce
is
th
e
o
n
e
h
av
i
n
g
th
e
h
i
g
h
est m
em
b
er
s
h
ip
m
ea
s
u
r
e
us
in
g
(
7
)
:
=
(
7
)
4.
T
H
E
P
RO
P
O
SE
D
M
E
T
H
O
D
T
h
e
p
r
im
ar
y
p
u
r
p
o
s
e
o
f
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
is
to
p
r
o
v
i
d
e
a
n
ew
h
y
b
r
id
alg
o
r
ith
m
th
at
p
er
f
o
r
m
s
b
etter
f
o
r
m
ix
e
d
d
ata.
T
h
is
alg
o
r
ith
m
co
m
b
in
es
th
e
in
d
iv
id
u
al
s
tr
en
g
th
s
o
f
NB
f
o
r
tex
t
ap
p
licatio
n
an
d
C
SB
S.
I
t
m
itig
ates
th
e
d
is
ad
v
an
tag
es
o
f
th
e
two
m
eth
o
d
s
k
n
o
win
g
t
h
at
th
e
p
er
f
o
r
m
an
ce
o
f
NB
m
o
v
es
d
o
wn
wh
er
e
th
e
n
u
m
b
er
o
f
r
ar
e
wo
r
d
s
g
o
es u
p
.
B
esid
es,
it h
as n
u
m
er
o
u
s
ad
v
an
tag
es th
at
ca
n
b
e
d
escr
ib
ed
as f
o
llo
ws:
−
B
y
co
m
b
in
in
g
a
p
r
o
b
ab
ilis
tic
alg
o
r
ith
m
with
an
alg
o
r
ith
m
b
ased
o
n
d
is
tan
ce
an
d
d
en
s
ity
,
th
e
m
o
d
el
elim
i
n
ates th
e
p
r
o
b
a
b
ilis
tic
p
r
o
p
er
ty
o
f
th
e
p
r
o
p
o
s
ed
m
eth
o
d
.
−
T
h
e
co
m
p
u
tatio
n
c
o
m
p
lex
ity
i
s
lo
wer
co
m
p
ar
e
d
to
NB
m
o
d
e
l
as
th
e
p
r
o
p
o
s
ed
class
if
ier
tu
r
n
ed
th
e
p
r
o
d
u
ct
f
o
r
m
in
t
o
a
s
u
m
f
o
r
m
.
−
T
h
e
im
p
ac
t
o
f
r
a
r
e
wo
r
d
s
n
u
m
b
er
ca
n
n
o
t
b
e
ig
n
o
r
ed
s
in
ce
it
b
ec
o
m
es
an
o
p
t
im
izer
o
f
class
if
icatio
n
p
er
f
o
r
m
an
ce
.
−
T
h
e
C
SB
S c
o
n
tain
s
a
n
o
r
m
alize
d
d
is
tan
ce
,
wh
ich
is
b
etter
f
o
r
n
u
m
er
ical
v
ar
iab
les ap
p
licatio
n
s
.
−
I
m
p
lem
en
tatio
n
is
m
o
r
e
s
im
p
l
e
an
d
ea
s
ier
.
T
h
e
co
m
m
u
n
icate
d
a
d
v
an
tag
e
s
co
u
ld
b
e
n
o
ticed
th
r
o
u
g
h
th
e
alg
o
r
ith
m
’
s
d
escr
ip
tio
n
as
s
h
o
wn
i
n
Fig
u
r
e
1
.
T
h
e
p
r
o
ce
s
s
s
h
o
ws
th
e
m
ain
s
tep
s
to
ex
ce
ed
th
e
co
n
s
tr
ain
t
d
u
e
to
NB
f
ail
to
class
if
y
a
p
ar
ticu
lar
in
s
tan
ce
,
an
d
th
e
co
m
b
in
atio
n
with
th
e
a
d
ap
ted
C
SB
S
in
a
s
p
ec
if
ic
s
tag
e.
T
o
illu
s
tr
ate
t
h
e
lo
g
ic
o
f
o
u
r
p
r
o
p
o
s
e
d
m
o
d
el,
Fig
u
r
e
2
r
ep
r
esen
ts
th
e
d
ea
lin
g
o
f
d
if
f
er
e
n
t
co
m
p
o
n
e
n
ts
at
ea
ch
lev
el.
T
h
e
tr
ial
s
’
n
u
m
b
er
is
b
ased
o
n
th
e
v
alu
e
o
f
K.
Fo
r
ea
ch
tr
ial
th
e
NB
cla
s
s
if
ie
s
th
e
tex
t
in
s
t
an
ce
s
b
ased
o
n
th
e
o
cc
u
r
r
en
c
e
o
f
wo
r
d
s
an
d
th
e
p
r
o
b
a
b
ilit
ies
o
f
b
elo
n
g
i
n
g
.
Ho
wev
er
,
an
d
d
u
e
to
th
e
h
i
g
h
n
u
m
b
er
o
f
r
ar
e
wo
r
d
s
,
th
e
NB
af
f
ec
ts
an
im
p
o
r
tan
t
p
o
r
tio
n
to
th
e
wr
o
n
g
class
.
B
y
ad
d
in
g
th
e
weig
h
t
o
f
t
h
e
n
u
m
e
r
ical
d
im
en
s
io
n
,
th
e
ad
ap
te
d
C
SB
S
tr
ies
to
m
ak
e
th
e
class
if
icatio
n
b
etter
an
d
p
r
o
m
o
te
th
e
p
o
s
itio
n
o
f
ea
ch
wo
r
d
in
th
e
d
ataset.
Fig
u
r
e
1
.
T
h
e
p
r
o
p
o
s
ed
al
g
o
r
it
h
m
Fig
u
r
e
2
.
I
ll
u
s
tr
atio
n
o
f
d
if
f
er
e
n
t stag
es o
f
p
r
o
p
o
s
ed
alg
o
r
ith
m
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
A
Hyb
r
id
n
a
ïve
B
a
ye
s
b
a
s
ed
o
n
s
imila
r
ity
mea
s
u
r
e
to
… (
F
a
tima
E
l B
a
r
a
ka
z
)
159
5.
RE
SUL
T
S
A
ND
A
NAL
Y
SI
S
5
.
1
.
E
x
perim
ent
s
5
.
1
.
1
.
Da
t
a
des
cr
iptio
n a
nd
prepa
ra
t
io
n
T
h
e
aim
o
f
o
u
r
p
r
o
p
o
s
ed
s
o
lu
t
io
n
is
to
ef
f
ec
tiv
el
y
h
a
n
d
le
m
i
x
ed
d
ata
f
o
r
ca
r
d
tr
a
n
s
ac
tio
n
s
p
ay
m
en
t
class
if
icatio
n
p
r
o
b
lem
s
.
T
h
e
d
ataset
illu
s
tr
atio
n
co
n
tain
s
1
3
1
2
in
s
tan
ce
s
an
d
two
v
ar
iab
les.
T
h
e
f
ir
s
t
v
ar
iab
le
is
a
ca
teg
o
r
ical
v
ar
iab
le
th
at
d
escr
ib
es th
e
tr
an
s
ac
tio
n
lab
els.
T
h
e
s
ec
o
n
d
is
th
e
n
u
m
er
ic
v
a
r
iab
le
th
at
co
n
s
is
ts
o
f
t
h
e
am
o
u
n
t
ass
o
ciate
d
with
ea
ch
o
p
er
atio
n
.
W
e
ex
tr
ac
t
ed
th
e
d
ata
f
r
o
m
a
p
e
r
s
o
n
al
ac
co
u
n
t
cr
ea
te
d
in
Mo
r
o
cc
an
b
an
k
te
r
r
ito
r
y
th
at
we
aim
to
class
if
y
th
em
in
to
f
o
u
r
class
es.
Ob
s
er
v
in
g
o
u
r
d
ataset,
th
e
ca
t
eg
o
r
ical
v
ar
iab
le
is
an
u
n
s
tr
u
c
tu
r
ed
tex
t
an
d
d
o
es
n
o
t
s
tr
ictly
r
esp
ec
t
th
e
s
y
n
tax
o
r
th
e
s
em
an
tic
m
ea
n
in
g
o
f
n
atu
r
al
la
n
g
u
a
g
e
(
E
n
g
lis
h
,
Fre
n
ch
.
.
.
)
,
o
r
an
y
ab
b
r
ev
iatio
n
r
u
les.
Or
eith
er
th
e
em
p
lace
m
e
n
t
o
f
a
wo
r
d
in
a
s
en
ten
ce
d
o
es
n
o
t
h
av
e
an
y
im
p
o
r
tan
ce
.
I
t
co
u
ld
b
e
ca
teg
o
r
ized
as
a
n
o
r
m
al
ca
teg
o
r
ical
d
im
en
s
i
o
n
with
f
ew
v
alu
es,
o
th
er
ca
s
es
co
n
tain
m
u
lti
-
v
alu
es,
f
u
r
th
er
,
an
d
it
m
ay
also
b
e
class
ed
as sh
o
r
t te
x
t
.
I
n
T
ab
le
1
,
ea
ch
ca
s
e
h
as b
ee
n
p
r
esen
te
d
with
s
o
m
e
s
elec
ted
in
s
tan
ce
s
.
T
h
e
p
r
e
p
ar
atio
n
o
f
s
u
ch
d
ata
im
p
o
s
es
th
r
ee
p
a
r
ts
:
to
k
en
izat
io
n
,
r
e
m
o
v
al
o
f
s
to
p
wo
r
d
s
,
t
h
en
th
e
co
n
s
tr
u
ctio
n
o
f
th
e
b
ag
o
f
wo
r
d
s
.
T
o
to
k
en
ize
th
e
tex
t
o
f
th
e
ca
teg
o
r
ical
v
ar
iab
le,
s
tr
in
g
s
o
f
tex
t
h
av
e
b
ee
n
s
p
lit
in
to
wo
r
d
s
,
we
m
o
v
ed
,
an
d
th
e
s
to
p
wo
r
d
s
h
av
e
b
ee
n
id
en
ti
f
ied
.
Fo
r
ex
am
p
le:
th
e,
an
d
,
o
r
.
.
.
Sto
p
wo
r
d
s
ca
n
also
b
e
a
s
p
ec
if
i
ed
lis
t
o
f
ex
p
r
e
s
s
io
n
s
,
f
o
r
ex
am
p
le,
tak
in
g
t
h
e
lab
el:
“Su
p
er
m
ar
k
et
E
L
J
ADI
DA”
,
th
e
ex
p
r
ess
io
n
“E
L
J
ADI
DA”
wh
ich
is
a
n
a
m
e
o
f
a
Mo
r
o
cc
an
city
,
h
as
n
o
s
en
s
e
in
o
u
r
p
r
o
p
o
s
ed
m
o
d
e
l,
s
o
o
u
r
lis
t
o
f
s
to
p
wo
r
d
s
co
m
b
in
e
th
e
s
tan
d
ar
d
s
to
p
wo
r
d
s
in
Fre
n
ch
an
d
E
n
g
li
s
h
lan
g
u
ag
es
lis
t
an
d
th
e
lis
t
o
f
all
Mo
r
o
cc
an
cities.
Fin
ally
,
th
e
b
ag
o
f
wo
r
d
s
h
as
b
ee
n
co
n
s
tr
u
cted
as
a
m
atr
ix
.
T
h
is
o
n
e
h
elp
s
th
e
class
if
ier
to
t
r
ain
o
n
th
e
d
ata
an
d
r
ec
o
v
er
s
th
e
s
ig
n
if
ica
n
t te
r
m
s
o
f
ea
ch
class
.
T
ab
le
1
.
Dif
f
e
r
en
t c
ases
s
elec
t
ed
f
r
o
m
p
ay
m
en
t tr
an
s
ac
tio
n
t
ex
t v
ar
iab
le
C
a
se
P
a
y
m
e
n
t
t
r
a
n
s
a
c
t
i
o
n
t
e
x
t
C
o
mm
e
n
t
S
t
a
n
d
a
r
d
C
a
t
e
g
o
r
i
c
a
l
d
i
me
n
si
o
n
“
A
c
h
a
t
Y
V
ES R
O
C
H
ER
M
A
R
O
C
“
“
A
c
h
a
t
v
i
a
W
W
W
.
A
LI
EX
P
R
ESS
.
C
O
M
“
“
P
a
y
U
B
E
R
M
A
R
O
C
E
-
C
O
M
b
i
l
l
”
Ea
c
h
i
n
s
t
a
n
c
e
b
e
l
o
n
g
s
t
o
d
i
f
f
e
r
e
n
t
c
l
a
sses,
a
n
d
i
t
a
p
p
e
a
r
s i
n
o
n
e
f
o
r
m f
o
r
t
h
e
w
h
o
l
e
d
a
t
a
s
e
t
.
M
u
l
t
i
-
v
a
l
u
e
c
a
t
e
g
o
r
i
c
a
l
d
i
me
n
si
o
n
“
A
c
h
a
t
M
a
r
j
a
n
e
mar
k
e
t
A
l
i
n
a
”
“
A
c
h
a
t
M
a
r
j
a
n
e
B
i
g
d
i
l
”
“
P
a
y
M
a
r
j
a
n
e
b
i
l
l
”
A
l
l
i
n
s
t
a
n
c
e
s
b
e
l
o
n
g
t
o
sa
me
c
l
a
s
ses,
h
o
w
e
v
e
r
t
h
e
t
h
i
r
d
o
n
e
w
i
l
l
b
e
mi
s
c
l
a
ssi
f
i
e
d
b
a
s
e
d
u
s
i
n
g
N
B
.
S
h
o
r
t
t
e
x
t
“
B
i
l
l
L’
A
R
B
R
E
D
E
ZO
E”
“
F
a
c
t
u
r
e
K
I
N
A
N
I
C
H
A
U
S
S
U
R
ES”
“
G
R
A
S
S
A
V
O
Y
E
M
o
l
a
y
Y
o
u
ss
e
f”
Th
e
r
a
r
e
w
o
r
d
s
a
r
e
h
i
g
h
l
y
r
e
p
r
e
s
e
n
t
e
d
i
n
t
h
i
s
samp
l
e
,
t
h
e
o
n
l
y
k
e
y
w
o
r
d
s
a
r
e
“
b
i
l
l
”
a
n
d
“
f
a
c
t
u
r
e
”
,
a
n
d
t
h
e
b
o
t
h
a
r
e
n
o
t
e
n
o
u
g
h
t
o
a
f
f
e
c
t
a
c
o
r
r
e
c
t
c
l
a
ss
i
f
i
c
a
t
i
o
n
w
i
t
h
N
B
.
5
.
1
.
2
.
E
x
perim
ent
a
l
pro
ce
d
ures
T
o
ev
al
u
ate
th
e
p
r
o
p
o
s
ed
alg
o
r
ith
m
,
we
t
r
ain
with
t
h
r
ee
m
o
d
e
ls
.
T
h
e
f
ir
s
t
is
NB
,
wh
ich
was
ap
p
lied
to
th
e
ca
teg
o
r
ical
v
a
r
iab
le
to
av
o
id
th
e
o
v
er
lap
p
in
g
o
f
th
e
n
u
m
er
ical
v
ar
ia
b
le.
T
h
e
s
ec
o
n
d
m
o
d
el
u
s
ed
th
e
ad
ap
ted
C
SB
S
o
n
b
o
t
h
ca
teg
o
r
ical
an
d
n
u
m
er
ical
v
ar
iab
les.
T
h
e
last
o
n
e
i
n
tr
o
d
u
ce
d
o
u
r
p
r
o
p
o
s
ed
m
o
d
el
t
h
at
co
m
b
in
es
th
e
NB
an
d
th
e
ad
a
p
ted
C
SB
S
alg
o
r
ith
m
.
T
o
ad
j
u
s
t
th
e
C
S
B
S
(
cited
in
(
6)
)
to
th
e
s
tr
u
ctu
r
e
o
f
th
e
d
ataset.
T
h
e
ad
ap
ted
C
SB
S i
s
g
iv
en
in
t
h
e
(
8
)
:
(
X
)
=
+
∗
−
(
,
̅
̅
̅
)
+
+
×
1
∑
′
=
1
(
8
)
wh
er
e
:
in
d
icate
th
e
f
r
eq
u
en
cy
o
f
th
e
wo
r
d
p
er
class
C
.
t: u
s
ed
to
in
d
ex
th
e
p
a
r
am
eter
s
o
f
th
e
n
u
m
er
ical
attr
ib
u
te.
M
′
:
Nu
m
b
er
o
f
wo
r
d
s
o
f
th
e
ca
te
g
o
r
ical
v
ar
ia
b
le
Fo
r
a
r
ea
s
o
n
ab
le
co
m
p
ar
is
o
n
,
we
o
r
g
an
ized
t
h
e
d
ataset
in
to
d
if
f
er
en
t
s
u
b
s
et
s
izes,
n
=2
8
0
,
5
6
0
,
8
4
0
,
an
d
1
1
2
0
,
r
esp
ec
tiv
ely
,
wh
ich
ar
e
s
elec
ted
ea
ch
tim
e
ar
b
i
tr
ar
ily
f
r
o
m
o
u
r
d
ataset
o
f
1
3
1
2
in
s
tan
ce
s
.
T
h
e
K
-
Fo
ld
C
r
o
s
s
-
v
alid
atio
n
s
am
p
lin
g
m
eth
o
d
is
f
r
e
q
u
en
tly
u
s
ed
to
e
v
alu
ate
m
o
d
els
i
n
m
ac
h
in
e
lear
n
in
g
a
n
d
d
ata
m
in
in
g
.
T
h
e
d
ataset
is
s
eg
m
en
ted
r
an
d
o
m
ly
in
to
K
s
eg
m
en
ts
,
wh
er
e
ea
ch
s
eg
m
en
t
is
r
eta
in
ed
o
n
ce
,
an
d
th
e
class
if
ier
is
lear
n
ed
o
n
th
e
o
t
h
er
K
-
1
s
eg
m
en
ts
.
I
n
o
u
r
ca
s
e,
K
will tak
e
4
,
7
,
a
n
d
1
0
,
r
esp
e
ctiv
ely
.
T
h
er
ef
o
r
e,
th
e
lear
n
in
g
p
r
o
ce
d
u
r
e
is
p
er
f
o
r
m
ed
K
tim
es
o
n
e
ac
h
d
if
f
e
r
en
t
s
u
b
s
et.
T
h
e
o
v
e
r
all
p
er
f
o
r
m
an
ce
is
ev
alu
ated
in
ter
m
s
o
f
r
ec
all,
p
r
ec
is
io
n
,
an
d
F
-
m
ea
s
u
r
e:
Pr
e
c
ision
=
+
(
9
)
R
e
c
a
l
l
=
+
(
1
0
)
F
_
s
c
or
e
=
2
×
×
+
(
1
1
)
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
1
5
5
-
16
2
160
wh
er
e
: FN
is
th
e
n
u
m
b
er
o
f
f
a
ls
e
n
eg
ativ
es.
FP
is
th
e
n
u
m
b
er
o
f
f
alse p
o
s
itiv
es.
T
P is
th
e
n
u
m
b
er
o
f
tr
u
e
p
o
s
iti
v
es.
Th
e
ca
lcu
latio
n
o
f
th
o
s
e
two
f
a
cto
r
s
in
a
m
u
lti
-
class
class
if
ier
s
itu
atio
n
r
eq
u
est th
e
n
o
tio
n
s
b
elo
w:
C
las
s
if
ied
C
=A
ctu
al
11
…
1
…
1
Th
e
co
n
f
u
s
io
n
elem
en
ts
f
o
r
ea
ch
class
ar
e
g
iv
en
b
y
:
=
;
=
∑
=
1
−
(
1
2
,
1
3
)
=
∑
=
1
−
(
1
4
)
=
∑
∑
=
1
=
1
−
−
−
(
1
5
)
5
.
2
.
E
x
perim
ent
s
re
s
ults
:
T
h
e
p
e
r
f
o
r
m
an
ce
ev
alu
atio
n
o
f
o
u
r
h
y
b
r
id
m
o
d
el
co
n
s
t
r
u
cted
u
s
in
g
K
-
f
o
ld
cr
o
s
s
-
v
alid
atio
n
in
tr
o
d
u
ce
d
in
th
e
s
ec
tio
n
ab
o
v
e.
Sin
ce
th
e
p
ar
am
eter
K
to
o
k
d
if
f
er
e
n
t
v
alu
es,
we
co
m
p
u
te
th
e
m
o
d
el
o
n
3
0
tr
ials
f
o
r
ea
ch
s
am
p
le
s
ize.
T
h
e
r
esu
lts
f
o
r
t
h
e
th
r
ee
class
if
ier
s
NB
,
ad
ap
ted
C
SB
S,
an
d
th
e
p
r
o
p
o
s
ed
m
et
h
o
d
ar
e
r
ep
o
r
ted
in
T
ab
le
2.
T
h
e
im
p
r
o
v
e
m
en
ts
o
f
th
e
h
y
b
r
id
m
eth
o
d
in
ter
m
s
o
f
th
e
d
if
f
er
e
n
t
m
ea
s
u
r
es
r
ef
er
at
f
ir
s
t
to
th
e
p
er
f
o
r
m
an
ce
o
f
n
a
ïv
e
B
ay
es
o
n
th
e
d
ataset,
th
e
n
at
s
ec
o
n
d
to
th
e
ad
d
in
g
o
f
th
e
ad
ap
ted
C
SB
S
p
er
f
o
r
m
an
ce
ap
p
lied
to
th
e
p
a
r
titi
o
n
s
p
o
o
r
ly
class
if
ied
.
Fu
r
th
er
m
o
r
e
,
th
e
n
o
tab
le
r
o
le
o
f
t
h
e
ad
ap
ted
C
SB
S
co
u
ld
n
o
t
b
e
d
en
ied
,
s
in
ce
it
k
ep
t
an
ex
ce
llen
t
h
ar
m
o
n
ic
m
e
an
b
etwe
en
th
e
r
ec
all
an
d
th
e
p
r
ec
is
io
n
f
o
r
ea
c
h
d
if
f
er
en
t
s
im
u
lat
io
n
.
An
d
b
etter
,
wh
en
it
is
co
m
b
in
ed
with
NB
p
er
f
o
r
m
an
ce
.
T
o
p
r
esen
t
th
e
p
r
o
g
r
ess
o
f
o
u
r
class
if
ier
in
ter
m
m
u
lti
-
class
if
icatio
n
im
p
r
o
v
em
en
t,
we
s
elec
ted
f
o
r
K=
1
0
f
o
u
r
tr
ials
r
an
d
o
m
ly
ap
p
lied
o
n
a
s
am
p
le
o
f
n
=2
8
0
.
An
d
b
ased
o
n
T
ab
le
3
,
wh
ich
d
escr
ib
e
s
th
e
r
ec
all,
p
r
ec
is
io
n
,
a
n
d
F
-
s
co
r
e
v
alu
es,
th
e
p
r
o
p
o
s
ed
m
eth
o
d
o
u
t
p
er
f
o
r
m
e
d
f
o
r
t
h
e
th
r
ee
e
v
alu
atio
n
i
n
d
i
ca
to
r
s
.
T
ab
le
2
.
T
h
e
re
su
lt
s
o
f
t
h
e
d
iffer
e
n
t
c
las
sifier fo
r
d
iffere
n
t
K
v
a
lu
e
,
b
a
se
d
o
n
3
0
tri
a
ls o
n
a
v
e
ra
g
e
N
a
i
v
e
B
a
y
e
s
A
d
a
p
t
e
d
C
S
B
S
Th
e
p
r
o
p
o
s
e
d
mo
d
e
l
S
a
mp
l
e
s
i
z
e
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
F
-
sco
r
e
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
F
-
sco
r
e
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
F
-
sco
r
e
K
=
4
2
8
0
0
.
6
3
0
.
7
6
0
.
6
2
0
.
7
8
0
.
7
9
0
.
8
3
0
.
7
9
0
.
8
9
0
.
8
9
5
6
0
0
.
6
1
0
.
7
3
0
.
6
2
0
.
7
5
0
.
8
2
0
.
7
9
0
.
7
8
0
.
8
9
0
.
8
6
8
4
0
0
.
7
2
0
.
7
1
0
.
7
1
0
.
8
3
0
.
8
9
0
.
7
7
0
.
8
8
0
.
9
3
0
.
8
6
1
1
2
0
0
.
7
6
0
.
6
8
0
.
7
2
0
.
7
6
0
.
7
5
0
.
7
2
0
.
8
9
0
.
8
9
0
.
9
4
K
=
7
2
8
0
0
.
7
1
0
.
7
5
0
.
6
4
0
.
7
8
0
.
8
4
0
.
7
5
0
.
8
4
0
.
9
2
0
.
9
3
5
6
0
0
.
7
8
0
.
6
9
0
.
6
2
0
.
8
4
0
.
7
4
0
.
6
4
0
.
8
0
.
8
8
0
.
8
5
8
4
0
0
.
6
3
0
.
7
9
0
.
7
2
0
.
6
5
0
.
8
7
0
.
6
3
0
.
8
3
0
.
9
4
0
.
8
3
1
1
2
0
0
.
6
7
0
.
7
1
0
.
7
4
0
.
7
4
0
.
8
5
0
.
7
1
0
.
9
8
0
.
9
1
0
.
8
9
K
=
1
0
2
8
0
0
.
6
0
.
6
1
0
.
6
2
0
.
7
7
0
.
8
9
0
.
6
2
0
.
8
8
0
.
8
0
.
8
8
5
6
0
0
.
7
0
.
6
0
.
6
2
0
.
8
3
0
.
8
1
0
.
7
3
0
.
8
4
0
.
9
7
0
.
8
8
4
0
0
.
7
6
0
.
8
0
.
7
1
0
.
7
4
0
.
7
4
0
.
6
6
0
.
7
7
0
.
9
6
0
.
8
9
1
1
2
0
0
.
7
8
0
.
6
7
0
.
7
2
0
.
7
2
0
.
8
4
0
.
7
7
0
.
9
0
.
8
8
0
.
9
4
T
ab
le
3
.
T
h
e
r
esu
lts
o
f
p
r
ec
is
io
n
,
r
ec
al
l,
a
n
d
F
-
s
co
r
e
p
er
tr
ia
l a
n
d
p
er
m
eth
o
d
M
e
t
h
o
d
R
e
c
a
l
l
P
r
e
c
i
s
i
o
n
F
-
S
c
o
r
e
Tr
i
a
l
.
1
1
N
a
i
v
e
B
a
y
e
s
0
.
7
8
0
.
8
9
0
.
8
3
2
A
d
a
p
t
e
d
C
S
B
S
0
.
7
4
0
.
7
6
0
.
7
5
3
P
r
o
p
o
se
d
m
e
t
h
o
d
0
.
8
9
0
.
9
4
0
.
9
1
Tr
i
a
l
.
2
4
N
a
i
v
e
B
a
y
e
s
0
.
9
0
.
8
5
0
.
8
8
5
A
d
a
p
t
e
d
C
S
B
S
0
.
7
8
0
.
7
7
0
.
7
8
6
P
r
o
p
o
se
d
m
e
t
h
o
d
0
.
9
4
0
.
9
1
0
.
9
2
Tr
i
a
l
.
3
7
N
a
i
v
e
B
a
y
e
s
0
.
8
7
0
.
9
0
.
8
8
8
A
d
a
p
t
e
d
C
S
B
S
0
.
8
0
.
7
4
0
.
7
7
9
P
r
o
p
o
se
d
m
e
t
h
o
d
0
.
9
3
0
.
9
4
0
.
9
4
Tr
i
a
l
.
4
10
N
a
i
v
e
B
a
y
e
s
0
.
8
3
0
.
8
5
0
.
8
4
11
A
d
a
p
t
e
d
C
S
B
S
0
.
7
7
0
.
7
5
0
.
7
6
12
P
r
o
p
o
se
d
m
e
t
h
o
d
0
.
9
0
.
9
3
0
.
9
1
E
v
en
m
o
r
e,
th
e
h
y
b
r
id
m
et
h
o
d
g
u
ar
a
n
tees a
g
o
o
d
ef
f
icien
c
y
i
n
ter
m
s
o
f
th
e
o
n
e
class
class
if
icatio
n
p
er
f
o
r
m
an
ce
,
s
o
we
h
av
e:
Evaluation Warning : The document was created with Spire.PDF for Python.
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
A
Hyb
r
id
n
a
ïve
B
a
ye
s
b
a
s
ed
o
n
s
imila
r
ity
mea
s
u
r
e
to
… (
F
a
tima
E
l B
a
r
a
ka
z
)
161
Pre
cisi
o
n
(
C
NB)
<
Pre
cisi
o
n
(
C
The
poposed
method
)
An
d
:
R
ec
all
(
C
NB
)
<
R
ec
all(
C
The
poposed
met
ho
d
)
T
o
v
is
u
alize
th
is
,
en
h
an
ce
,
a
d
em
o
n
s
tr
atio
n
with
a
co
n
f
u
s
io
n
m
atr
ix
is
r
ec
o
m
m
en
d
ed
.
Fig
u
r
e
3
illu
s
tr
ates
th
e
co
n
f
u
s
io
n
m
at
r
ix
o
f
d
i
f
f
er
en
t
s
elec
ted
tr
ials
p
er
m
eth
o
d
.
Mo
v
in
g
f
r
o
m
NB
to
ad
ap
ted
C
SB
S
to
th
e
p
r
o
p
o
s
e
d
m
eth
o
d
f
o
r
ea
ch
tr
ial,
th
e
n
u
m
b
er
s
in
th
e
co
n
f
u
s
io
n
m
at
r
ix
in
cr
ea
s
ed
wh
er
e
th
e
n
u
m
b
e
r
s
o
u
ts
id
e
d
ec
r
ea
s
ed
,
wh
ich
p
r
o
v
es
th
e
p
r
o
g
r
ess
o
f
o
n
e
-
class
class
if
icatio
n
.
W
e
a
ls
o
n
o
te
th
at
th
e
T
r
u
e
Po
s
itiv
e
in
tab
les
(
3
)
,
(
6
)
,
(
9
)
,
a
n
d
(
1
2
)
ar
e
b
etter
th
an
its
eq
u
iv
alen
t
in
tab
les
(
2
)
,
(
5
)
,
(
8
)
,
an
d
(
1
1
)
.
T
h
is
r
esu
lt
h
ig
h
lig
h
ts
th
e
f
ac
t
o
f
h
o
w
th
e
h
y
b
r
id
m
et
h
o
d
w
o
r
k
s
s
ig
n
i
f
ican
tly
b
etter
f
o
r
t
h
e
r
ar
e
wo
r
d
s
an
d
ac
h
ie
v
ed
e
x
ce
llen
t r
esu
lts
f
o
r
b
o
th
m
ix
ed
d
ata
class
if
icatio
n
an
d
tex
t
cla
s
s
if
icatio
n
.
I
n
g
en
er
al,
th
e
NB
s
h
o
ws
g
o
o
d
r
esu
lts
co
m
p
a
r
in
g
to
th
e
r
esu
lts
o
f
C
S
B
S.
Ho
wev
er
,
th
e
co
m
b
in
a
tio
n
o
f
b
o
th
ac
h
iev
ed
m
ea
n
in
g
f
u
l c
lass
if
icatio
n
p
r
o
g
r
ess
.
Fig
u
r
e
3.
T
h
e
co
n
f
u
s
io
n
m
atr
ices o
f
f
o
u
r
tr
ials
wer
e
r
an
d
o
m
ly
s
elec
ted
to
ex
p
lain
t
h
e
r
esu
l
t o
f
T
ab
le
3
6.
CO
NCLU
SI
O
N
T
h
e
m
ain
o
b
jectiv
e
o
f
t
h
is
co
n
tr
ib
u
tio
n
is
to
d
ea
l
with
th
e
cla
s
s
if
icatio
n
o
f
m
i
x
ed
d
ata
th
at
i
n
clu
d
e
a
m
u
lti
-
v
alu
e
d
s
h
o
r
t
te
x
t
v
a
r
ia
b
le.
W
e
in
tr
o
d
u
ce
d
a
h
y
b
r
id
n
aïv
e
B
ay
es
th
at
is
b
ased
o
n
s
im
ilar
ity
m
ea
s
u
r
es
t
o
ef
f
ec
tiv
ely
p
r
o
ce
s
s
b
o
th
ca
teg
o
r
ical
an
d
n
u
m
e
r
ical
v
ar
iab
les.
I
n
th
e
p
r
o
p
o
s
ed
m
eth
o
d
,
th
e
n
ai
v
e
B
ay
es
p
r
ed
icts
th
e
p
o
r
tio
n
o
f
th
e
tar
g
et
o
n
l
y
e
x
p
lain
ed
b
y
th
e
ca
teg
o
r
ical
v
a
r
iab
le,
an
d
t
h
e
r
e
m
ain
in
g
p
a
r
t
is
p
r
ed
icted
u
s
in
g
th
e
ad
ap
te
d
C
SB
S
th
at
p
r
o
v
id
es
g
o
o
d
class
if
icatio
n
u
s
in
g
n
u
m
er
ical
v
ar
iab
les
.
T
h
e
p
r
o
p
o
s
ed
s
o
lu
tio
n
co
m
b
in
es
NB
with
an
a
d
ap
ted
C
SB
S.
T
h
e
h
y
b
r
id
m
o
d
el
was
co
m
p
ar
ed
to
th
e
n
aïv
e
B
ay
es,
an
d
th
e
ad
ap
te
d
C
S
B
S
s
ep
ar
ately
.
T
h
e
ex
p
er
i
m
en
ts
wer
e
p
er
f
o
r
m
ed
u
s
in
g
t
h
e
ca
r
d
tr
a
n
s
ac
tio
n
s
p
ay
m
e
n
t
d
ata
th
at
co
n
tain
s
a
m
u
lti
-
v
alu
ed
s
h
o
r
t
tex
t
v
ar
iab
le
an
d
n
u
m
er
ical
v
ar
iab
le.
T
h
e
s
o
lu
tio
n
h
as
ac
h
iev
ed
s
ig
n
if
ican
t
p
r
o
g
r
ess
i
n
ter
m
s
o
f
r
ec
all,
p
r
ec
is
io
n
,
a
n
d
F
-
m
ea
s
u
r
e.
Fu
r
th
er
m
o
r
e,
it
d
e
als
well
with
r
ar
e
wo
r
d
s
is
s
u
es
,
an
d
also
im
p
r
o
v
es
th
e
class
if
icatio
n
o
f
th
e
m
o
d
el
.
T
h
is
wo
r
k
is
lim
ited
b
ec
au
s
e
it
h
as
n
o
t
b
ee
n
a
p
p
lied
to
d
if
f
er
en
t
k
n
o
wn
d
ataset
y
et.
Ho
w
ev
er
,
it
was
p
r
o
p
o
s
ed
to
h
an
d
le
th
e
c
lass
if
icatio
n
o
f
s
h
o
r
t
tex
t
u
s
in
g
m
u
lti
-
v
alu
e
d
v
ar
iab
les,
ap
p
lied
to
a
r
ea
l
ca
s
e
p
r
o
b
lem
:
ca
r
d
tr
an
s
ac
tio
n
p
ay
m
en
t
class
if
icatio
n
.
T
h
is
s
tu
d
y
co
u
l
d
b
e
ex
ten
d
ed
o
n
m
an
y
m
ix
ed
d
atasets
in
a
d
if
f
er
en
t
f
ield
in
o
r
d
er
t
o
o
p
tim
ize
th
e
class
if
icatio
n
o
f
ca
teg
o
r
ical
d
im
e
n
s
io
n
s
.
I
n
f
u
tu
r
e
wo
r
k
,
th
e
d
im
en
s
io
n
ality
o
f
v
ec
to
r
-
tex
t su
p
p
o
r
ted
b
y
o
u
r
m
eth
o
d
will b
e
in
v
esti
g
ated
wh
ile
m
ain
tain
i
n
g
its
s
im
p
licity
.
ACK
NO
WL
E
DG
E
M
E
NT
S
T
h
is
s
tu
d
y
was
s
u
p
p
o
r
ted
b
y
th
e
R
esear
ch
team
at
I
NDAT
AC
O
R
E
,
a
co
m
p
an
y
o
f
a
r
tific
ial
in
tellig
en
ce
s
o
lu
tio
n
s
.
RE
F
E
R
E
NC
E
S
[1
]
Co
h
e
n
,
“
Ap
p
li
e
d
M
u
lt
i
p
le
Re
g
r
e
ss
io
n
/Co
rre
lati
o
n
A
n
a
ly
sis
fo
r
th
e
Be
h
a
v
i
o
ra
l
S
c
ien
c
e
s,”
Ama
z
o
n
W
a
re
h
o
u
se
,
Fu
lf
il
le
d
b
y
Ama
zo
n
,
2
0
1
3
.
[2
]
C.
M
.
Cu
a
d
ra
s,
C.
Are
a
n
s,
a
n
d
J.
F
o
rti
a
n
a
,
“
S
o
m
e
c
o
m
p
u
tat
io
n
a
l
a
sp
e
c
ts
o
f
a
d
istan
c
e
—
b
a
se
d
m
o
d
e
l
fo
r
p
re
d
ictio
n
,
”
C
o
mm
u
n
ic
a
ti
o
n
s i
n
S
ta
ti
stics
-
S
imu
l
a
ti
o
n
a
n
d
C
o
mp
u
t
a
ti
o
n
,
v
o
l
.
2
5
,
n
o
.
3
,
p
p
.
5
9
3
-
6
0
9
,
1
9
9
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.
I
SS
N
:
1
6
9
3
-
6
9
3
0
T
E
L
KOM
NI
KA
T
elec
o
m
m
u
n
C
o
m
p
u
t E
l Co
n
tr
o
l
,
Vo
l.
19
,
No
.
1
,
Feb
r
u
ar
y
2
0
2
1
:
1
5
5
-
16
2
162
[3
]
C.
Cu
a
d
ra
s a
n
d
C.
Are
n
a
s,
“
A
d
is
tan
c
e
-
b
a
se
d
re
g
re
ss
io
n
m
o
d
e
l
f
o
r
p
re
d
ictio
n
wit
h
m
ix
e
d
d
a
ta,”
C
o
m
mu
n
ica
t
io
n
s in
S
ta
ti
st
ics
-
T
h
e
o
ry
a
n
d
M
e
th
o
d
s
,
v
o
l.
1
9
,
n
o
.
6
,
p
p
.
2
2
6
1
-
2
2
7
9
,
1
9
9
0
.
[4
]
E.
B.
D.
Va
l,
M
.
M
.
C.
Bielsa
,
a
n
d
J.
F
o
r
ti
a
n
a
,
“
S
e
lec
ti
o
n
o
f
P
re
d
icto
rs
in
Dista
n
c
e
-
Ba
se
d
Re
g
r
e
ss
io
n
,
”
Co
mm
u
n
ica
ti
o
n
s in
S
ta
ti
st
ics
-
S
im
u
la
t
io
n
a
n
d
Co
mp
u
ta
ti
o
n
,
v
o
l.
3
6
,
n
o
.
1
,
p
p
.
8
7
-
9
8
,
2
0
0
7
.
[5
]
M
.
Yu
a
n
a
n
d
Y.
L
in
,
“
M
o
d
e
l
se
l
e
c
ti
o
n
a
n
d
e
stim
a
ti
o
n
in
re
g
re
ss
i
o
n
with
g
ro
u
p
e
d
v
a
riab
les
,
”
J
o
u
rn
a
l
o
f
th
e
Ro
y
a
l
S
ta
ti
st
ica
l
S
o
c
iety
:
S
e
rie
s B
(
S
ta
ti
stica
l
M
e
t
h
o
d
o
l
o
g
y
)
,
v
o
l.
6
8
,
n
o
.
1
,
p
p
.
4
9
-
6
7
,
2
0
0
6
.
[6
]
L.
M
e
ier,
S
.
V.
D.
G
e
e
r,
a
n
d
P
.
B
ü
h
lma
n
n
,
“
T
h
e
g
ro
u
p
las
so
f
o
r
lo
g
isti
c
re
g
re
ss
io
n
,
”
J
o
u
rn
a
l
o
f
th
e
Ro
y
a
l
S
t
a
ti
stica
l
S
o
c
iety
:
S
e
rie
s B
(
S
ta
t
isti
c
a
l
M
e
t
h
o
d
o
l
o
g
y
)
,
v
o
l.
7
0
,
n
o
.
1
,
p
p
.
5
3
-
7
1
,
2
0
0
8
.
[7
]
V.
K.
Ay
y
a
d
e
v
a
ra
,
“
Wo
r
d
2
v
e
c
,
”
Pro
M
a
c
h
i
n
e
L
e
a
rn
in
g
Al
g
o
rit
h
m
s
,
p
p
.
1
6
7
–
1
7
8
,
2
0
1
8
.
[8
]
A.
Ne
e
lak
a
n
tan
,
J.
S
h
a
n
k
a
r,
A.
P
a
ss
o
s,
a
n
d
A.
M
c
c
a
ll
u
m
,
“
Ef
ficie
n
t
No
n
-
p
a
ra
m
e
tri
c
Esti
m
a
ti
o
n
o
f
M
u
lt
ip
le
Emb
e
d
d
i
n
g
s
p
e
r
W
o
rd
i
n
Ve
c
to
r
S
p
a
c
e
,
”
Pro
c
e
e
d
in
g
s
o
f
th
e
2
0
1
4
Co
n
fer
e
n
c
e
o
n
Em
p
iri
c
a
l
M
e
t
h
o
d
s
i
n
N
a
t
u
ra
l
L
a
n
g
u
a
g
e
Pr
o
c
e
ss
in
g
(E
M
NL
P)
,
2
0
1
4
.
[9
]
P.
Ve
n
k
a
tes
wa
ri,
P
.
Um
a
m
a
h
e
sw
a
ri,
K.
Ra
jes
h
,
J.
G
lo
ry
Th
e
p
h
o
ra
l
,
“
G
e
n
e
b
a
se
d
Dise
a
se
P
re
d
ictio
n
u
si
n
g
P
a
tt
e
r
n
S
imilarity
b
a
se
d
Clas
sifica
ti
o
n
,
”
I
n
ter
n
a
ti
o
n
a
l
J
o
u
rn
a
l
o
f
I
n
n
o
v
a
ti
v
e
T
e
c
h
n
o
lo
g
y
a
n
d
Exp
l
o
rin
g
En
g
i
n
e
e
rin
g
Reg
u
la
r
Iss
u
e
,
v
o
l.
8
,
n
o
.
1
1
,
p
p
.
3
2
2
3
-
3
2
2
7
,
2
0
1
9
.
[1
0
]
A.
S
k
a
b
a
r,
“
Dire
c
ti
o
n
-
of
-
C
h
a
n
g
e
F
in
a
n
c
ial
Ti
m
e
S
e
ries
F
o
re
c
a
stin
g
u
si
n
g
a
S
imilarity
-
Ba
se
d
Clas
sifica
ti
o
n
M
o
d
e
l,
”
J
o
u
rn
a
l
o
f
F
o
re
c
a
stin
g
,
v
o
l.
3
2
,
n
o
.
5
,
p
p
.
4
0
9
-
4
2
2
,
2
0
1
3
.
[1
1
]
W.
Ch
e
rif,
A
.
M
a
d
a
n
i,
a
n
d
M
.
K
issi,
“
A
No
v
e
l
S
imilari
ty
-
Ba
se
d
Alg
o
rit
h
m
fo
r
S
u
p
e
rv
ise
d
Bi
n
a
ry
Clas
sifica
ti
o
n
:
S
a
n
d
a
lwo
o
d
O
d
o
r
Ap
p
li
c
a
ti
o
n
,
”
S
S
RN
El
e
c
tro
n
ic Jo
u
rn
a
l
,
2
0
1
8
.
[1
2
]
S
.
Ch
e
n
,
G
.
I.
Web
b
,
L
.
Li
u
,
a
n
d
X.
M
a
,
“
A
n
o
v
e
l
se
lec
ti
v
e
n
a
ï
v
e
Ba
y
e
s
a
lg
o
rit
h
m
,
”
K
n
o
wled
g
e
-
Ba
se
d
S
y
ste
ms
,
v
o
l.
1
9
2
,
2
0
2
0
.
[1
3
]
Z.
E.
Ra
sjid
a
n
d
R.
S
e
ti
a
wa
n
,
“
P
e
rfo
rm
a
n
c
e
Co
m
p
a
riso
n
a
n
d
Op
ti
m
iza
ti
o
n
o
f
Tex
t
Do
c
u
m
e
n
t
Clas
sifica
ti
o
n
u
sin
g
k
-
NN
a
n
d
Na
ïv
e
Ba
y
e
s Clas
sifica
ti
o
n
Tec
h
n
iq
u
e
s,”
Pro
c
e
d
i
a
Co
m
p
u
ter
S
c
ien
c
e
,
v
o
l.
1
1
6
,
p
p
.
1
0
7
-
1
1
2
,
2
0
1
7
.
[1
4
]
B.
C.
Bro
o
k
e
s
a
n
d
H
.
Cra
m
e
r,
“
Th
e
El
e
m
e
n
ts
o
f
P
r
o
b
a
b
il
it
y
T
h
e
o
ry
a
n
d
S
o
m
e
o
f
Its
Ap
p
li
c
a
ti
o
n
s,”
T
h
e
M
a
t
h
e
ma
ti
c
a
l
Ga
ze
tt
e
,
v
o
l.
4
0
,
n
o
.
3
3
2
,
p
.
1
5
3
,
1
9
5
6
.
[1
5
]
Z.
Š
u
lc
a
n
d
H.
Ře
z
a
n
k
o
v
á
,
“
Ev
a
lu
a
ti
o
n
o
f
Re
c
e
n
t
S
imilarity
M
e
a
su
r
e
s
fo
r
Ca
teg
o
rica
l
Da
ta,”
I
n
ter
n
a
t
i
o
n
a
l
S
c
ie
n
ti
fi
c
Co
n
fer
e
n
c
e
,
2
0
1
4
.
[1
6
]
D.
W.
G
o
o
d
a
ll
,
“
A Ne
w S
imilarit
y
In
d
e
x
Ba
se
d
o
n
P
ro
b
a
b
il
i
ty
,
”
B
i
o
me
trics
,
v
o
l
.
2
2
,
n
o
.
4
,
p
p
.
8
8
2
-
9
0
7
,
1
9
6
6
.
[1
7
]
E.
Es
k
in
,
A.
Arn
o
ld
,
M
.
P
re
ra
u
,
L.
P
o
rt
n
o
y
,
a
n
d
S
.
S
t
o
lf
o
,
“
A
G
e
o
m
e
tri
c
F
ra
m
e
wo
rk
f
o
r
Un
su
p
e
r
v
ise
d
A
n
o
m
a
ly
De
tec
ti
o
n
,
”
Ad
v
a
n
c
e
s in
I
n
f
o
rm
a
t
io
n
S
e
c
u
rity A
p
p
li
c
a
ti
o
n
s o
f
Da
t
a
M
in
i
n
g
in
C
o
mp
u
ter
S
e
c
u
rity
,
p
p
.
7
7
-
1
0
1
,
2
0
0
2
.
[1
8
]
A.
D.
Ca
ig
n
y
,
K.
C
o
u
ss
e
m
e
n
t,
a
n
d
K.
W.
D.
B
o
c
k
,
“
A
n
e
w
h
y
b
rid
c
las
sifica
ti
o
n
a
lg
o
rit
h
m
fo
r
c
u
sto
m
e
r
c
h
u
rn
p
re
d
ictio
n
b
a
se
d
o
n
lo
g
isti
c
re
g
r
e
ss
io
n
a
n
d
d
e
c
isio
n
tree
s,”
Eu
ro
p
e
a
n
J
o
u
r
n
a
l
o
f
O
p
e
ra
ti
o
n
a
l
Res
e
a
rc
h
,
v
o
l.
2
6
9
,
n
o
.
2
,
p
p
.
7
6
0
–
7
7
2
,
2
0
1
8
.
[1
9
]
G
.
Nie
,
W.
Ro
we
,
L.
Zh
a
n
g
,
Y.
T
ian
,
a
n
d
Y.
S
h
i,
“
Cre
d
i
t
c
a
rd
c
h
u
r
n
fo
re
c
a
stin
g
b
y
l
o
g
ist
ic
re
g
re
ss
io
n
a
n
d
d
e
c
isio
n
tree
,
”
Exp
e
rt S
y
ste
ms
wit
h
A
p
p
li
c
a
ti
o
n
s
,
v
o
l
.
3
8
,
n
o
.
1
2
,
p
p
.
1
5
2
7
3
-
1
5
2
8
5
,
2
0
1
1
.
[2
0
]
J.
Ju
rg
o
v
sk
y
,
M
.
G
ra
n
it
z
e
r,
K.
Zi
e
g
ler,
S
.
Ca
lab
re
tt
o
,
P
.
-
E.
P
o
r
ti
e
r,
L.
He
-
G
u
e
lt
o
n
,
a
n
d
O.
Ca
e
len
,
“
S
e
q
u
e
n
c
e
c
las
sifica
ti
o
n
fo
r
c
re
d
i
t
-
c
a
rd
fra
u
d
d
e
tec
ti
o
n
,
”
Exp
e
rt S
y
ste
ms
wit
h
Ap
p
li
c
a
ti
o
n
s
,
v
o
l
.
1
0
0
,
p
p
.
2
3
4
-
2
4
5
,
2
0
1
8
.
[2
1
]
Y.
S
u
,
Y.
Hu
a
n
g
,
a
n
d
C.
-
C.
J.
Ku
o
,
“
Eff
icie
n
t
Te
x
t
Clas
sifica
ti
o
n
Us
in
g
Tree
-
stru
c
tu
re
d
M
u
lt
i
-
l
in
e
a
r
P
rin
c
ip
a
l
Co
m
p
o
n
e
n
t
A
n
a
ly
sis,”
a
rX
iv.o
rg
,
2
4
-
F
e
b
-
2
0
1
8
.
[
2
2
]
H
.
K
a
u
d
e
r
e
r
a
n
d
H
.
-
J
.
M
u
c
h
a
,
“
S
u
p
e
r
v
i
s
e
d
L
e
a
r
n
i
n
g
w
i
t
h
Q
u
a
l
i
t
a
t
i
v
e
a
n
d
M
i
x
e
d
A
t
t
r
i
b
u
t
e
s
,
”
C
l
a
s
s
i
f
i
c
a
t
i
o
n
,
D
a
t
a
A
n
a
l
y
s
i
s
,
a
n
d
D
a
t
a
H
i
g
h
w
a
y
s
S
t
u
d
i
e
s
i
n
C
l
a
s
s
i
f
i
c
a
t
i
o
n
,
D
a
t
a
A
n
a
l
y
s
i
s
,
a
n
d
K
n
o
w
l
e
d
g
e
O
r
g
a
n
i
z
a
t
i
o
n
,
p
p
.
3
7
4
-
3
8
2
,
1
9
9
8
.
[2
3
]
H.
Ch
e
n
,
M
.
S
u
n
,
C
.
Tu
,
Y.
Li
n
,
a
n
d
Z.
Li
u
,
“
Ne
u
ra
l
S
e
n
ti
m
e
n
t
Clas
sifica
ti
o
n
with
Us
e
r
a
n
d
P
ro
d
u
c
t
Atten
ti
o
n
,
”
Pro
c
e
e
d
i
n
g
s
o
f
t
h
e
2
0
1
6
C
o
n
fer
e
n
c
e
o
n
Em
p
irica
l
M
e
t
h
o
d
s
in
N
a
t
u
ra
l
L
a
n
g
u
a
g
e
Pro
c
e
ss
in
g
,
2
0
1
6
.
[2
4
]
Y.
S
u
,
Y.
Hu
a
n
g
,
a
n
d
C.
-
C.
J.
Ku
o
,
“
Eff
icie
n
t
Te
x
t
Clas
sifica
ti
o
n
Us
in
g
Tree
-
stru
c
tu
re
d
M
u
lt
i
-
l
in
e
a
r
P
rin
c
ip
a
l
Co
m
p
o
n
e
n
t
A
n
a
ly
sis,”
a
rX
iv.o
rg
,
2
4
-
F
e
b
-
2
0
1
8
.
[2
5
]
W.
Hu
a
,
Z
.
Wa
n
g
,
H.
Wan
g
,
K.
Zh
e
n
g
,
a
n
d
X.
Zh
o
u
,
“
S
h
o
rt
tex
t
u
n
d
e
rsta
n
d
i
n
g
th
r
o
u
g
h
le
x
ica
l
-
se
m
a
n
ti
c
a
n
a
ly
sis,”
2
0
1
5
IE
EE
3
1
st I
n
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
D
a
ta
En
g
in
e
e
rin
g
,
p
p
.
4
9
5
–
5
0
6
,
2
0
1
5
.
[2
6
]
Y.
S
u
,
R.
Li
n
,
a
n
d
C.
-
C.
J.
Ku
o
,
“
Tree
-
stru
c
tu
re
d
m
u
lt
i
-
sta
g
e
p
rin
c
ip
a
l
c
o
m
p
o
n
e
n
t
a
n
a
l
y
sis
(T
M
P
C
A):
th
e
o
r
y
a
n
d
a
p
p
li
c
a
ti
o
n
s,”
a
rXiv
.
o
rg
,
2
0
1
8
.
[2
7
]
R.
M
a
li
k
,
“
Lea
rn
in
g
a
c
las
sifica
ti
o
n
m
o
d
e
l
f
o
r
se
g
m
e
n
tatio
n
,
”
Pro
c
e
e
d
in
g
s
Nin
t
h
IEE
E
I
n
ter
n
a
t
io
n
a
l
Co
n
fer
e
n
c
e
o
n
Co
mp
u
ter
V
isio
n
,
2
0
0
3
.
[2
8
]
N.
S
h
a
rm
a
,
M
.
S
i
n
g
h
,
“
M
o
d
if
y
i
n
g
Na
iv
e
Ba
y
e
s
c
las
sifier
fo
r
m
u
l
ti
n
o
m
ial
tex
t
c
las
sifica
ti
o
n
,
”
2
0
1
6
In
ter
n
a
ti
o
n
a
l
Co
n
fer
e
n
c
e
o
n
Rec
e
n
t
A
d
v
a
n
c
e
s
a
n
d
In
n
o
v
a
ti
o
n
s in
En
g
i
n
e
e
rin
g
(I
CRA
IE)
,
p
p
.
1
-
7
,
2
0
1
6
.
Evaluation Warning : The document was created with Spire.PDF for Python.