Int
ern
at
i
onal
Journ
al of Ele
ctrical
an
d
Co
mput
er
En
gin
eeri
ng
(IJ
E
C
E)
Vo
l.
9
, No
.
5
,
Octo
ber
201
9
, pp.
3714
~
37
19
IS
S
N:
20
88
-
8708
,
DOI: 10
.11
591/
ijece
.
v
9
i
5
.
pp3714
-
37
19
3714
Journ
al h
om
e
page
:
http:
//
ia
es
core
.c
om/
journa
ls
/i
ndex.
ph
p/IJECE
A data m
ini
ng app
ro
ach
for desir
e and in
tention to
particip
ate
in virtu
al c
ommun
ities
Öz
erk Y
avuz
1
, Adem
Ka
r
ahoca
2
, Dil
ek K
ar
ahoca
3
1
Depa
rtment
of
Com
pute
r
Engi
n
ee
ring
,
Al
ti
nbas
Univer
sit
y
,
T
urkey
2
Depa
rtment of
Software
Eng
ineeri
ng,
Eng
ine
er
i
ng
and
N
at
ura
l
S
ci
en
ce
s Fa
cul
t
y
,
Bahc
ese
h
ir
Univ
ersity
,
T
urkey
3
Depa
rtment of
Chil
d
Dev
el
opm
ent
,
Hea
lt
h
Sc
ience
s Fac
u
lty
,
Ba
hce
sehir
Univer
s
ity
,
Tu
r
key
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
hist
or
y:
Re
cei
ved
N
ov
9
, 2
01
8
Re
vised
A
pr
8
,
201
9
Accepte
d Apr
1
9
, 201
9
The
purpose
of
t
his
stud
y
is
to
in
vesti
gate
p
erf
or
m
anc
es
of
som
e
of
th
e
da
t
a
m
ini
ng
appr
oac
h
es
while
under
st
andi
ng
desire
an
d
int
ention
to
pa
rti
ci
p
ate
in
virt
ual
comm
unit
ie
s
and
it
s
a
nte
c
ede
nts.
A
rese
arc
h
m
ode
l
has
bee
n
deve
lop
ed
foll
o
wing
the
literature
rev
ie
w
an
d
the
m
odel
was
te
sted
aft
erwa
rds
.
In
re
sea
rch
p
art
of
th
e
stud
y
,
som
e
of
the
d
at
a
m
ini
ng
appr
oac
h
es
as
JRip,
Part,
OneR
Method,
Multi
lay
e
r
Per
ce
ptron
(Neur
al
Networks),
Ba
y
esia
n
N
et
wo
rks
have
b
ee
n
u
sed.
Based
on
t
he
an
aly
sis
cond
uct
ed
it
h
as
bee
n
found
out
tha
t
Mult
ila
y
e
r
Neura
l
Netw
ork
had
the
b
est
cor
re
c
t
cl
assifi
ca
t
ion
r
ate
and
lowest
RMS
E.
Ke
yw
or
d
s
:
Data m
ining
Desire a
nd
inte
ntion t
o
par
ti
ci
pate in
vi
rtual
com
m
un
it
ie
s
Ma
chine
le
a
rn
i
ng
Virtual c
omm
u
niti
es
Copyright
©
201
9
Instit
ut
e
o
f
Ad
vanc
ed
Engi
n
ee
r
ing
and
S
cienc
e
.
Al
l
rights re
serv
ed
.
Corres
pond
in
g
Aut
h
or
:
Ad
em
K
ara
hoc
a,
Dep
a
rtm
ent o
f Sof
t
war
e
E
ng
i
neer
i
ng,
Ba
hceseh
i
r Uni
ver
sit
y,
Faculty
of E
ngineerin
g, Be
sik
ta
s,
Ista
nbul, 3
4349, T
urkey.
Em
a
il
:
ade
m
.k
arahoca@
en
g.bau.ed
u.
t
r;
aka
rahoca@
gm
ai
l.
com
1.
INTROD
U
CTION
As
virt
ual
co
m
m
un
it
y
con
cept
e
m
erg
ed
duri
ng
ti
m
e,
new
def
i
niti
on
s
of
the
te
rm
fo
und
place
in
the
li
te
ratur
e.
P
or
t
er
pro
poses
a
virtu
al
com
m
un
it
y
def
init
io
n
that,
a
virtu
al
com
m
un
it
y
is
an
ag
gr
e
gatio
n
of
ind
ivi
du
al
s
or
bu
sine
ss
pa
rtner
s
who
inte
ract
around
a
sh
are
d
intere
st,
wh
e
re
the
interact
ion
is
at
le
ast
par
ti
al
ly
supporte
d
a
nd
/
or
m
ediat
ed
by
te
chnolo
gy
a
nd
gu
i
ded
by
so
m
e
protoc
ols
or
norm
s
[
1].
Plant
ap
proac
he
s
the
te
rm
fro
m
a
si
m
il
ar
per
sp
ect
ive
de
fin
ing
a
vi
rtual
c
omm
un
it
y
as
a
colle
ct
ive
gr
oup
that
com
e
tog
et
he
r
ei
ther
tem
po
ra
rily
or
per
m
anen
tl
y
through
an
el
ect
ronic
m
edium
to
enab
le
the
interact
ion
of
entit
ie
s,
ind
ivi
du
al
s
or
orga
nizat
ion
s
i
n
a
com
m
on
pr
oble
m
or
interest
sp
ace
[
2].
I
n
ad
diti
on
t
o
these,
Rhein
go
l
d
def
i
nes
a
vi
rtual
c
omm
un
it
y
as
so
ci
al
ag
gr
e
gati
on
s
that
em
erge
from
the
I
nt
ern
et
w
hen
en
ough
people car
ry on th
os
e
public discussi
ons lo
ng eno
ugh,
with
su
f
fici
ent hum
an
feeli
ng to f
orm
w
ebs
of p
e
r
so
na
l
relat
ion
s
hip
s
in
cybers
pace
[
3].
The
pur
pose
of
this
stud
y
is
to
inv
est
igate
perform
ances
of
s
om
e
of
the
data
m
ining
appr
oach
es
wh
i
le
unde
rstan
ding
desire
a
nd
inte
ntion
to
par
ti
ci
pate
in
virtu
al
com
m
un
it
ie
s
a
nd
t
he
factors
aff
ect
in
g
it
.
F
or
this
pur
po
se
a
m
od
el
has
been
dev
el
op
e
d
with
the
f
oc
us
on
desi
re
an
d
intenti
on
t
o
pa
rtic
ipate
to
vi
rtual
com
m
un
it
ie
s
a
nd
it
s
antece
de
nts.
Lat
er
f
ol
lowing
the
data
gather
i
ng
phase
and
pr
e
-
pr
ocessin
g
of
th
e
data
sever
al
data
m
ining
a
ppr
oaches
hav
e
bee
n
app
li
ed
t
o
the
data.
S
om
e
pa
rt
of
the
data
is
us
ed
f
or
tra
ini
ng
pur
po
ses
w
he
r
eas
rem
ai
nin
g
is
us
e
d
for
te
s
ti
ng
the
m
od
el
w
hich
has
be
en
form
ed
fo
ll
ow
i
ng
the
li
te
r
at
ur
e
rev
ie
w.
C
on
se
qu
e
ntly
in
a
ddit
ion
to
t
he
st
ud
ie
s
i
n
the
s
ci
entifi
c
body
of
kn
ow
le
dg
e
a
colla
borati
ve
an
d
con
t
rib
utive
da
ta
m
ining
a
ppr
oach
is
app
li
ed
to
under
sta
n
d
desir
e
and
intenti
on
to
pa
rtic
ipate
in
virtu
al
c
omm
u
niti
es.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
A data
m
i
ning
approac
h f
or
de
sire a
nd inten
ti
on
to
parti
ci
pate
in
vi
rt
ua
l c
omm
un
it
ie
s
(
Ö
zerk Y
avu
z
)
3715
2.
RESEA
R
CH MET
HO
D
Data
m
ining
c
an
be
de
fine
d
as
the
process
of
extracti
ng
hidden
patt
erns
fr
om
la
rg
e
chun
ks
of
data.
In
doin
g
this
knowle
dge
disco
ver
y,
pre
dicti
on
or
for
ecast
ing
ca
n
be
in
the
f
oc
us
of
data
m
ining
.
Wh
il
e
kn
ow
le
dg
e
disc
ov
e
ry
pro
vid
es
us
exp
li
ci
t
inf
orm
at
ion
ab
ou
t
the
cha
racteri
s
ti
cs
of
the
da
ta
set
pr
e
dicti
ve
m
odel
ing
prov
i
des
pr
e
dicti
on
s
of
fu
t
ur
e
e
ve
nts.
As
sta
te
d
by
S
i
m
ou
dis
,
data
m
ining
is
the
proces
s
of
e
xtracti
ng
va
li
d,
previ
o
usl
y
unknown
,
c
om
pr
ehe
ns
ible
and
act
ion
a
bl
e
inf
or
m
at
ion
from
la
rg
e
databases
and
us
i
ng
it
to
m
ake
bu
sine
s
s
decisi
ons
[4
]
.
Data
m
ining
bor
rows
a
ppr
oach
e
s
f
ro
m
sever
al
discipli
nes
as
sta
ti
sti
cs,
m
a
the
m
at
ic
s
or
co
m
pu
te
r
sci
ence
in
orde
r
to
fi
nd
us
e
fu
l
patte
rn
s
a
nd
kn
ow
l
edg
e
from
la
rg
e
data
set
s.
As
it
is
ind
ic
at
ed
i
n
Sh
ea
rer’s
cris
p
-
dm
m
od
el
,
a
data
m
ining
process
is
c
om
po
sed
of
busines
s
unde
rstan
ding,
data
unde
rsta
nd
i
ng,
data
preparati
on,
m
od
el
bu
il
di
ng,
te
sti
ng
/e
val
uation
a
nd
de
ploym
ent
processes
. In
t
he
f
ol
lo
wing s
ect
ion
s s
om
e o
f
the
data m
ining
a
ppr
oac
hes use
d
in a
naly
zi
ng
t
he
data set
will
b
e
introd
uced [
5].
2.1.
D
ata
gat
heri
ng
and pr
ocessin
g
As
s
uggeste
d
in
li
te
ratur
e
over
385
obse
r
vations
(
425
in
our
sam
ple
in
this
st
ud
y)
has
been
f
ound
su
f
fici
ent
f
or
the
sam
ple
siz
e
values
with
a
n
er
ror
of
5%
and
a
co
nfi
de
nce
le
vel
of
95%
(
survey
m
onkey
sit
e
-
sam
ple
size
cal
culat
or
)
.
I
n
li
te
ratur
e
us
e
d
f
or
m
ula
to
cal
culat
e
this
has
been
n=
t
2
x
(
p
x
q)/
e
2
w
her
e
n
ref
e
rs
to
sam
ple
siz
e, p
r
efe
rs
to
pr
opor
ti
on,
p
erce
ntag
e
or p
rese
nce of
the
stu
dy
cha
racte
risti
cs
(in
li
te
ratur
e
it
is
su
ggest
ed
that
wh
e
n
we
hav
e
no
pr
io
r
values
f
or
the
pr
op
or
ti
ons
to
be
est
i
m
at
ed,
we
can
us
e
p
-
a
nd
q
-
values
as
50
%.)
q=1
-
p,
e
r
efers
to
m
arg
i
n
of
er
ror;
t
=
1.96
(w
it
h
95
%
confide
nce
le
vel).
B
ase
d
on
t
hat
,
n
=
1.9
62 x 0.5
x
0.5 /
0.0
52 s
a
m
ple size
h
as
been f
ound
38
4.16 an
d r
ound
ed
to
38
5 [6
-
7
].
Scal
es
us
ed
in
the
stud
y
is
giv
e
n
in
detai
l.
Po
sit
ive
ant
ic
ipate
d
em
oti
on
s
ref
e
r
to
the
pr
e
-
f
act
ual
s
hypothesiz
e
d
to
infl
ue
nce
de
sires
to
pe
rfo
r
m
a
beh
avi
or
wh
ic
h
can
be
in
the
f
or
m
of
po
sit
ive
a
ntici
pated
e
m
otion
s
or
ne
gative
a
ntici
pa
te
d
em
otion
s
an
d
it
’s
li
kel
y
to
ex
pect
it
s
infl
uen
ce
on
virtua
l
com
m
un
it
y
par
ti
ci
patio
n
a
nd
de
sire
a
nd
i
ntentio
n
t
o
part
i
ci
pate
in
vi
rtual
c
omm
un
it
i
es
[
8
]
.
In
the
li
te
ratur
e,
it
is
pointed
ou
t
that
in
ge
ne
ral
people
are
in
a
te
nd
ency
to
exp
ect
s
om
e
return
w
hen
t
hey
sh
are
th
ei
r
know
le
dg
e
.
A
s
it
i
s
def
i
ned
by
Chiu
et
al
.
,
no
rm
of
reci
procit
y
ref
ers
to
kn
ow
l
edg
e
e
xch
a
nge
s
that
are
m
utu
al
and
pe
rceiv
ed
by
the p
a
rtie
s as
f
ai
r
an
d o
ne of
t
he
im
po
rtant
fa
ct
or
s t
hat lea
ds t
o kno
wled
ge shari
ng b
e
ha
vio
r
[
9
]
.
Perceive
d
us
e
f
uln
ess
re
fers
to
the
degree
to
wh
ic
h
a
person
belie
ves
that
us
in
g
a
pa
rtic
ular
syst
e
m
would
e
nh
a
nce
his
or
he
r
pe
r
form
ance
[
10
,
11
].
As
it
is
i
nd
ic
at
ed
by
P
or
te
r
,
in
the
te
chnolo
gy
acce
ptance
m
od
el
,
per
cei
ve
d
use
f
uln
ess
a
nd
pe
rceive
d
e
ase
of
us
e
a
re
the
belie
fs
t
hat
are
pres
um
ed
t
o
in
flue
nce
at
ti
tud
e
s
towa
rd
new
te
chnolo
gy
[
12
]
.
As
it
is
po
inted
out
in
Fish
be
in
and
Aj
ze
n’s
theor
y
of
reas
on
e
d
act
ion,
at
ti
tud
es
are
f
or
m
ed
as
a
resu
lt
of
the
belie
fs
ab
out
the
outc
om
es
of
pe
rfor
m
ing
that
act
and
e
xpect
ed
outc
om
e
s.
If
t
he
ou
tc
om
e
of
pe
rfor
m
ing
that
be
hav
i
or
s
ee
m
s
ben
efici
al
to
the
i
nd
i
vidual,
he/she
m
ay
par
ti
ci
pate
in
that
par
ti
cula
r beha
vior [
13
,
14
]
.
Early
def
i
niti
on
s
of
s
ocial
com
par
ison
t
he
or
y
date
bac
k
to
1954s
t
hat
sta
rted
with
F
est
ing
er
’s
s
oci
al
com
par
ison
th
eor
y.
As
sta
te
d
in
the
li
te
ratu
r
e
accor
ding
t
o
so
ci
al
com
par
ison
the
or
y,
t
he
re
is
a
dri
ve
w
it
hin
ind
ivi
du
al
s
to
l
ook
to
ou
tsi
de
i
m
ages
in
orde
r
to
evaluate
th
ei
r
own
opini
ons
an
d
abili
ti
es
in
the
sense
th
at
it
m
ai
nly
fo
cuses
on
e
xp
la
ini
ng
and
unde
rstan
di
ng
te
nde
ncies
of
in
div
i
du
al
s
in
evaluati
ng
a
nd
c
om
par
ing
their
own
opini
ons
and
desires
with
oth
e
rs
w
hic
h
m
a
y
le
ad
to
an
sel
f
en
hance
m
e
nt
in
ind
ivid
uals’
sel
f
i
m
ages.
As
it
is
po
inted
out
in
li
te
rat
ur
e
de
sires
pro
vid
e
the
m
otivati
on
to
decide
in
fav
or
of
act
ing
as
par
t
of
a
virtu
al
com
m
un
it
y.
Ther
e
fore
desir
e
co
ns
tr
uct
ha
s
bee
n
m
easur
e
d
with
the
help
of
qu
e
sti
on
s
a
da
pted
from
Dholokia
’s
r
es
pecti
ve
scal
e
[
15
].
As
it
is
de
fine
d
by
D
holo
kia,
W
e
-
I
ntentio
ns
co
ns
tr
uct
us
ed
in
the
m
odel
ref
e
rs
to
th
e
intenti
ons
t
o
par
ti
ci
pate
in
ge
ther
as
a
group
w
hich
is
to
be
a
func
ti
on
of
both
i
nd
i
vidual
(i.e
.,
at
ti
tud
es,
perce
ive
d
beh
a
vioral
c
ontrol,
posit
ive,
and
ne
gative
a
ntici
pated
em
otion
s)
an
d
s
oc
ia
l
determ
inants
[
15
]
.
Desire
a
nd
intenti
on
to
pa
rtic
ipate
in
vir
tual
com
m
un
it
ie
s
ref
e
rs
t
o
t
he
m
erg
e
of
we
-
intenti
on
a
nd
des
ires
of
D
hola
kia
wh
e
re
desires
pro
vid
e
the
m
otivati
on
to
de
ci
de
in
fa
vor
of
act
in
g
as
pa
rt
of
a
vi
rtual
com
m
un
it
y
a
nd
we
intenti
ons
sta
nd
f
or
the
i
ntentions
to
pa
rtic
ipate
tog
et
he
r
as
a
group,
t
o
b
e
a
functi
on
of
both
i
nd
i
vidual
(i.e.,
at
ti
tud
es,
pe
rceive
d
be
hav
i
or
al
c
ontr
ol,
posit
ive,
and
neg
at
ive
antic
ipate
d
e
m
ot
ion
s)
a
nd
so
ci
a
l
determ
inants
(
i.e.,
s
ubj
ect
ive
norm
s,
group
nor
m
s,
and
s
ocial
identit
y)
[
15
]
.
Re
s
pecti
ve
scal
es
ha
ve
bee
n
borro
wed em
pi
rical
l
y fr
om
the studie
s as s
how
n
i
n
T
a
ble
1.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
3
7
1
4
-
3
7
1
9
3716
Table
1
. Sca
le
s
us
e
d
i
n
the
stu
dy
Co
n
stru
ct
Ad
ap
ted
Fr
o
m
Po
sitiv
e anticip
ated
e
m
o
tio
n
s
Bag
o
zzi,
2
0
0
2
[
8
]
No
r
m
of
r
e
cip
rocit
y
Ch
iu
,
2
0
0
6
[
9
]
Perceived
Usef
u
ln
ess
Sh
in
,
2
0
0
8
[
10
]
Predis
p
o
sitio
n
to
Virtual Co
m
m
u
n
it
y
Usag
e
Bag
o
zzi,
2
0
0
2
[
8
]
So
cial Co
m
p
ariso
n
Ch
en
,
2
0
1
0
[
1
2
]
*
Desire A
n
d
I
n
ten
t
io
n
to Participate
I
n
Vir
tu
al
Co
m
.
Desi
res
We
-
Inten
tio
n
Dh
o
lak
ia,
2
0
0
4
[
15
]
Dh
o
lak
ia,
2
0
0
4
[
15
]
Dh
o
lak
ia,
2
0
0
4
[
15
]
*
Desire
an
d
Inten
tio
n
to
Participate
i
n
Virtual
Co
m
m
u
n
ities
is
th
e
co
m
b
in
atio
n
o
f
Desires a
n
d
W
e
Inten
tio
n
Scales
2.2.
D
ata
mi
n
ing me
thods
As
pa
rt
of
the
researc
h
c
ondu
ct
ed
seve
ral
da
ta
m
ining
ap
proach
e
s
ha
ve
be
en
ap
plied
t
o
the
data
set
.
Data
m
ining
m
et
ho
ds
ca
n
be
us
e
d
m
or
e
accuratel
y
with
data
pr
e
processin
g
ap
pr
oach
e
s
[
16
]
.
Su
c
h
as
norm
al
iz
a
ti
on
of
t
he
data
,
di
screti
zat
ion
th
e
co
nu
ti
nues
da
ta
and
et
c.
Br
ie
f
desc
riptio
ns
of
the
m
et
ho
ds
that
hav
e
b
ee
n use
d as f
ollo
w.
1)
JRi
p:
JRi
p
im
plem
ents
a
propositi
onal
r
ul
e
le
arn
e
r,
“R
e
peated
I
ncr
em
ental
Pru
ning
to
P
rod
uce
E
rro
r
Re
du
ct
io
n”
(R
IP
PER
),
as
pr
opose
d
by
Coh
e
n,
JRi
p
is
a
r
ule
le
arn
er
al
ike
in
pr
i
nciple
to
the
ru
le
le
ar
ne
r
Ri
pp
er
[
17
]
.
JRi
p
i
m
ple
m
ent
s
a
pr
oposi
ti
onal
ru
le
le
arn
er,
“R
epeate
d
Increm
ental
Pr
un
in
g
to
Produc
e
Error
Re
duct
io
n”
(R
IP
P
ER),
as
pro
posed
by
Cohe
n,
JRi
p
i
s
a
r
ule
le
ar
ner
al
ike
in
pri
nci
ple
to
t
he
ru
le
le
arn
er
Ri
pp
e
r
[
17
]
.
RIP
PER
ru
le
le
ar
ning
al
gorithm
is
an
e
xten
ded
ve
rsion
of
le
ar
ning
al
gorithm
IREP
(Incr
em
enta1
Re
du
ce
d
E
rror
Prun
i
ng).
It
const
ru
ct
s
a
r
ule
set
in
wh
i
ch
al
l
posit
ive
exam
ples
are
cov
e
re
d,
an
d
it
s
al
go
rithm
perform
s
eff
ic
ie
ntly
on
la
rge,
noisy
dataset
s.
Be
fo
re
buil
ding
a
r
ule
,
the
current
set
of
trai
ni
ng
e
xa
m
ples
are
par
ti
ti
on
ed
int
o
tw
o
subsets,
a
grow
i
ng
set
(
us
ua
l1y
2/3
)
an
d
a
pru
ning
set
(u
s
ual1y
1/3).
Th
e
r
ule
is
c
on
st
r
ucted
from
exa
m
ples
in
the
gr
ow
i
ng
set
.
The
r
ule
set
beg
i
ns
with
an
em
pty
ru
le
set
and
r
ules
are
ad
ded
increm
ental
l
y
to
the
ru
le
se
t
un
ti
l
no
neg
at
i
ve
exam
ples
are
cov
e
re
d.
A
fter
grow
i
ng
a
r
ul
e
fr
om
the
gr
ow
i
ng
set
,
c
onditi
on
is
delet
ed
from
the
ru
le
in
order
t
o
i
m
pr
ove the
p
e
rfor
m
ance of t
he rule set
on t
he pr
unin
g
e
xa
m
ples [
18
].
2)
PART:
T
he
P
ART
al
gorith
m
co
m
bin
es
t
wo
c
omm
on
data
m
ini
ng
str
at
egies;
the
div
ide
-
an
d
-
c
onquer
strat
egy
f
or
de
ci
sion
tree
le
a
rn
i
ng
with
the
separ
at
e
-
an
d
-
conq
uer
st
rategy
f
or
ru
le
le
a
rn
i
ng.
The
tre
e
bu
il
di
ng
al
gori
thm
sp
li
ts
a
set
of
e
xam
ples
recursively
int
o
a
par
ti
al
tree
.
The
first
ste
p
ch
oo
se
s
a
te
st
and
div
i
des
th
e
exam
ples
into
subsets.
P
A
RT
m
akes
this
cho
ic
e
i
n
ex
act
ly
the
sa
m
e
way
as
C4.
5.
The
n
the
s
ubs
et
s
are
ex
pand
ed
in
order
of
their
ave
ra
ge
e
ntr
op
y
sta
rtin
g
with
the
sm
al
l
est
.
The
reason
for
this
is
t
hat
su
bse
que
nt
subsets
will
m
os
t
li
kely
no
t
e
nd
up
be
i
ng
ex
pa
nd
e
d
a
nd
the
s
ub
s
et
with
l
ow
aver
a
ge
e
ntrop
y
is
m
or
e
li
kely
to
resu
lt
in
a
s
m
al
l
su
b
tree
an
d
t
he
refor
e
pro
du
ce
a
m
or
e
ge
ner
al
ru
le
[
19
]
.
3)
On
e
R:
O
neR,
gen
e
rates
a
on
e
-
le
vel
decisi
on
tree
t
hat
is
e
xpresse
d
in
th
e
form
of
a
set
of
r
ules
that
a
ll
te
st
on
e
par
ti
cular
at
trib
ute
.
O
neR
is
a
m
et
ho
d
that
oft
en
c
om
es
up
with
quit
e
good
r
ules
f
or
char
act
e
rizi
ng
the str
uctu
re in data
[
20
]
. Pse
udo
c
ode
for 1
R i
s as foll
ow.
Fo
r
eac
h
at
trib
ute,
Fo
r
eac
h value
of that at
trib
ut
e, m
ake a
r
ule
as foll
ows:
Count
how
of
t
en
eac
h
cl
ass
a
pp
ea
rs
Find the m
os
t f
reque
nt class
Ma
ke
the
rule
assign t
hat cla
s
s to
t
his att
rib
ut
e
-
value
.
Ca
lc
ulate
the e
rror rate
of the
ru
le
s.
Choose t
he rul
es w
it
h t
he
sm
al
le
st error rat
es [
20
].
4)
Multi
layer
Per
ceptro
n:
A
M
ul
ti
layer
Perce
ptron
is
a
ve
rsion
of
the
or
i
gina
l
per
ce
ptr
on
m
od
el
pr
op
os
e
d
by
Rosenbla
tt
in
the
1950
s
and
c
on
si
de
red
as
a
ty
pe
of
ne
ur
al
networ
ks
(R
os
e
nb
la
tt
,
1958)
.
A
pe
rcep
t
ron
(ar
ti
fici
al
neur
on)
is
a
f
unct
ion
of
se
veral
inp
ut
per
ce
ptr
on
s
w
hich
is
fo
rm
ed
as
a
com
bin
at
ion
of
in
put
wei
gh
t
s
to
t
he
hidde
n
la
ye
r
pe
rcept
rons.
As
sta
te
d
by
Ra
m
cho
un
in
li
te
rature
m
ul
ti
la
ye
r
per
ceptr
on
has o
ne
or
m
or
e h
id
den lay
ers
between
it
s inp
ut and o
ut
pu
t l
ay
ers,
the n
e
urons ar
e
orga
nized
i
n
la
ye
rs,
the
co
nnect
ions
are
al
ways
di
rected
from
inp
ut
la
ye
rs
to
outp
ut
la
ye
rs
an
d
t
he
neur
on
s
in
the
sam
e
layer
are
no
t
interc
onne
ct
ed
[
21
]
.
In
th
is
app
r
oac
h
hid
de
n
la
ye
r
is
a
fu
nctio
n
of
th
e
nodes
in
t
he pr
evio
us
lay
er,
a
nd the
outp
ut nod
e
s ar
e
a
func
ti
on
of the
node
s in
t
he hid
de
n
la
ye
r.
5)
Bayesia
n
Ne
t
work:
The
re
a
re
no
determ
i
nisti
c
ru
le
s
w
hich
al
lo
w
t
o
identify
a
s
ub
scribe
r
as
a
ri
sk
ind
ic
at
or.
G
ra
ph
ic
al
m
od
el
s
su
c
h
as
Ba
ye
sia
n
net
works
s
upply
a
gen
e
ra
l
fr
am
ewo
rk
f
or
deali
ng
with
un
ce
rtai
nly
in
a
prob
a
bili
sti
c
set
ti
ng
an
d
thus
are
well
su
it
ed
to
ta
ckle
the
pro
blem
of
pr
e
dicti
on
.
E
very
gr
a
ph
of
a
Ba
y
esi
an
ne
tw
ork
cod
e
s
a
cl
ass
of
pr
ob
a
bili
ty
distribu
ti
ons
.
T
he
nodes
of
th
at
gr
a
ph
c
om
ply
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
A data
m
i
ning
approac
h f
or
de
sire a
nd inten
ti
on
to
parti
ci
pate
in
vi
rt
ua
l c
omm
un
it
ie
s
(
Ö
zerk Y
avu
z
)
3717
with
the
var
ia
bles
of
the
prob
le
m
do
m
ai
n.
Arr
ow
s
betw
een
no
des
denote
al
lowe
d
(c
ausal)
relat
ions
betwee
n
the
va
riab
le
s.
These
de
pende
ncies
are
qua
ntifie
d
by
co
ndit
ion
al
distrib
utio
ns
f
or
e
ve
ry
no
de
giv
e
n
it
s
par
e
nt
s
[
22
]
.
A
Ba
ye
sia
n
netw
ork
B
ov
e
r
a
set
of
var
ia
bles
U
is
a
netw
ork
str
uc
ture
Bs,
wh
ic
h
is
directed
acy
cl
ic
gr
ap
h
(DAG)
ov
e
r
U
a
nd
set
of
pr
obabili
ty
ta
bles
Bp
={
p
(
u|pa
(u)
)|u
Є
U}
w
he
r
e
pa(u) i
s the
set
of p
a
ren
ts
of
u i
n
Bs.
A
Ba
ye
s
ia
n
net
wor
k
re
pr
ese
nts
pro
ba
bili
ty
d
ist
ribu
ti
on
s
[2
3,
24
].
3.
FIN
DINGS
Re
li
abili
t
y
of
the
co
ns
tr
ucts
hav
e
bee
n
re
-
a
ssessed
an
d
re
-
eval
uated
c
on
siderin
g
sug
ge
ste
d
lo
wer
lim
it
of
Cr
onb
ach’
s
al
pha
in
li
te
ratur
e.
As
it
is
sho
wn
in
T
able
2
with
the
sam
ple
siz
e
of
425
it
has
bee
n
see
n
that
al
l
Cronba
ch
al
ph
a v
al
ue
s
fo
r
the
re
sp
ec
ti
ve
con
str
ucts
hav
e
a
val
ue
of
hig
he
r
than
.
70,
in
othe
r
w
ords
al
l
the
co
ns
tr
ucts
us
e
d
in
t
he
re
s
earch
m
od
el
a
r
e
sta
ti
sti
cally
reli
able
and
can
be
reg
a
r
ded
a
s
reli
able
c
on
st
ru
ct
s
of the
researc
h m
od
el
[
25
]
. Fr
om
this r
easo
n,
Table
2.
Rel
ia
bi
li
t
y
m
easur
es
of the s
cal
es
Ite
m
s
Cro
n
b
ach Alp
h
a
Po
sitiv
e anticip
ate
d
e
m
o
tio
n
s
7
,91
8
No
r
m
of
r
e
cip
rocit
y
2
,79
8
Perceived
Usef
u
ln
ess
3
,88
6
Predis
p
o
sitio
n
to
Virtual Co
m
m
u
n
it
y
Usag
e
4
,85
7
So
cial Co
m
p
ariso
n
3
,86
9
*
Desire a
n
d
I
n
ten
ti
o
n
to Participate V
irtual Co
m
m
u
n
itie
s
5
,92
1
In
this
stu
dy,
ben
c
hm
ark
in
g
of
the
al
gorith
m
s
of
JRi
p,
Pa
rt,
OneR
Me
tho
d,
Mult
il
ay
er
Perce
ptr
on
,
Ba
ye
sia
n
Networks
ha
ve
be
en
pe
rfor
m
ed
.
In
te
sti
ng
th
e
researc
h
m
od
el
with
eac
h
of
the
data
m
ining
appr
oach
es
66
per
ce
nt
of
t
he
data
has
bee
n
us
ed
for
the
tr
ai
nin
g
w
her
eas
rem
a
ining
part
of
the
data
se
t
has
been
us
e
d
f
or
the
te
sti
ng
of
the
m
od
el
.
Am
on
g
diff
e
r
ent
data
m
ini
ng
a
ppr
oac
hes
JRi
p
ha
d
the
values
(RMSE=
0.2
91
3;
P
recisi
on
=
N/A;
C
orrect
Cl
assifi
cat
ion
Ra
te
=90
.
90%;
I
ncorr
ect
Cl
a
ssific
at
ion
Ra
t
e=9.
09
;
Tru
e
P
os
it
ive
Ra
te
=0.
90
9
a
nd False
P
os
it
ive Rat
e=0.
909).
Part
ha
d
t
he
va
l
ues
(RMSE
=0
.264
;
Pr
eci
sio
n=
0.923;
C
orrec
t
Cl
assifi
cat
ion
Ra
te
=91.60%;
In
c
orrect
Cl
assifi
cat
ion
Ra
te
=8.
39;
Tr
ue
Posi
ti
ve
Ra
te
=0.
91
6
an
d
False
Po
sit
ive
Ra
te
=0.
83
9)
.
On
eR
had
the
values
(RMSE=
0.3
01
5;
P
recisi
on
=
N/A;
C
orrect
Cl
assifi
cat
ion
Ra
te
=90
.
90%;
I
ncorr
ec
t
Cl
a
ssific
at
ion
Ra
t
e=9.
09
;
Tru
e
P
os
it
ive
Ra
te
=0.
90
9
a
nd False
P
os
it
ive Rat
e=0.
909).
Mult
il
ay
er
Percep
tr
on
ha
d
the
val
ues
(R
MSE=
0.2
476;
Pr
eci
sio
n=0.921
;
C
orrect
Cl
assifi
cat
io
n
Ra
te
=93
.
007%
;
In
c
orrect
Cl
assifi
cat
ion
Ra
t
e=6.
99
;
T
rue
P
os
it
ive
Ra
te
=0.930
a
nd
False
Po
sit
ive
Ra
te
=
0.561
)
and
finall
y
Ba
ye
sia
n
Netw
orks
had
the
values
(RMSE
=0.287
3;
Pr
ec
isi
on
=
0.876;
Correct
Cl
assifi
cat
io
n
Ra
te
=89
.
51%;
In
c
orrect
Cl
assifi
cat
ion
Ra
te
=10
.
49;
Tru
e
Po
sit
ive
Ra
te
=0.
89
5
and
False
Po
sit
iv
e
Ra
te
=0.
70
3)
.
Pr
eci
sio
n
valu
es
of
JRi
p
an
d
On
eR
m
et
ho
d
co
uld
not
been
cal
culat
e
d
sin
ce
propo
r
ti
on
of
instances
tr
uly
cl
assifi
ed
of
a
cl
ass
div
ide
d
by
the
total
instances
cl
assifi
e
d
in
that
cl
ass
hav
e
bee
n
cal
c
ulate
d
unde
fine
d
in
t
he
c
onfu
si
on
m
at
rix.
Am
ong
al
l
the
al
gor
it
h
m
s,
m
ulti
layer
pe
rce
ptr
on
had
the
m
os
t
correct
cl
assifi
cat
ion
r
at
e
with
93
.
00
7
pe
rcen
t,
a
go
od
tr
ue
posit
iv
e
rate
of
0.9
30
and
a
preci
sio
n
0.
921.
Part
m
et
ho
d
had
a
co
rr
ect
c
la
ssific
at
ion
ra
te
of
91.60
pe
r
cent,
true
po
sit
ive
rate
of
0.9
16
a
nd
a
preci
sion
va
lue
of
0.
923.
Mult
il
ay
er
perce
ptr
on
ha
d
th
e
lowest
RM
S
E
with
a
val
ue
of
0.2
4.
C
omparis
on
of
data
m
ining
m
et
hods
us
e
d
can
be
see
n
i
n Table
3.
Table
3.
C
om
par
iso
n of
data
m
ining
m
et
ho
ds use
d
Metho
d
RMSE
Precisio
n
Co
rr
ectly
Clas
sif
ied
%
Inco
rr
ectl
y
Clas
sif
ied
%
Tr
u
e Pos
itiv
e
Rate
False Po
sitiv
e
Rate
JR
ip
0
.29
N/A
9
0
.90
9
,09
0
,90
0
,90
Part
0
.26
0
,92
9
1
.60
8
,39
0
,91
0
,83
On
eR Metho
d
0
.30
N/A
9
0
.90
9
,09
0
,90
0
,90
Multilaye
r
Pe
rcept
ron
0
.24
0
,92
9
3
.00
6
,99
0
,93
0
,56
Bayesian
N
etwo
rk
s
0
.28
0
,87
8
9
.51
1
0
,49
0
,89
0
,70
4.
DISCU
SSI
ON A
ND CON
C
LUSIO
N
In
this
stu
dy,
we
inv
est
iga
te
d
the
facto
r
s
beh
i
nd
desire
an
d
intenti
on
to
par
ti
ci
pa
te
in
virtu
al
com
m
un
it
ie
s
fo
ll
ow
i
ng
a
n
in
te
ns
ive
li
te
ratu
re
re
vie
w.
T
hi
s
is
la
te
r
fo
ll
owed
with
the
m
od
el
form
ati
on
a
nd
app
ly
in
g
the
da
ta
m
ining
te
chn
i
qu
e
s
as
sug
gested
in
li
te
ra
ture.
I
n
the
a
na
ly
sis
par
t
of
th
e
stud
y
we
exa
m
ined
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2088
-
8708
In
t J
Elec
&
C
om
p
En
g,
V
ol.
9
, N
o.
5
,
Oct
ober
201
9
:
3
7
1
4
-
3
7
1
9
3718
relat
ion
s
hip
of
po
sit
ive
a
ntici
pated
em
otion
s
,
norm
of
recip
ro
ci
ty
,
so
ci
al
c
om
par
ison,
pre
disposit
ion
t
owar
ds
virtu
al
c
omm
un
it
y
us
age
a
nd
pe
rceive
d
us
ef
uln
e
ss
with
desire
an
d
intenti
on
to
pa
rtic
ipate
in
virtu
al
com
m
un
it
ie
s.
In
doin
g
s
o
we
trai
ned
t
he
m
o
del
us
in
g
66
pe
rcen
t
of
the
da
ta
of
trai
ni
ng
of
the
m
od
el
w
her
ea
s
rem
ai
nin
g
pa
rt
for
the
test
ing
of the m
od
el
for
eac
h
a
ppro
a
c
h.
Data
m
ining
c
an
be
de
fine
d
as
the
process
of
e
xtracti
ng
hidden
patte
rns
fr
om
la
rg
e
chun
ks
of
data.
In
doin
g
this
knowle
dge
disc
ov
e
ry,
pr
e
dicti
on
or
f
oreca
sti
ng
can
be
i
n
t
he
f
oc
us
of
da
ta
m
ining
.
J
rip
,
pa
rt,
on
e
r
m
et
ho
d,
Mult
il
ay
er P
erceptr
on
(Ne
ur
a
l Netwo
r
ks), a
nd
Bay
esi
an
N
et
work
s
hav
e
be
en
ch
os
en
a
s the d
at
a
m
ining
te
ch
niq
ue
s
in
orde
r
to
exam
ine
desire
an
d
intent
ion
to
par
ti
ci
pa
te
in
virtu
al
com
m
un
it
ie
s
f
or
thi
s
pur
po
se
.
Am
on
g
t
hem
JRi
p
is
a
ru
le
le
ar
ner
al
ike
in
pri
nciple
to
th
e
ru
le
le
ar
ner
Ri
pp
er
[
17
]
.
T
he
pa
rt
al
gorithm
co
m
bin
es
t
wo
c
omm
on
data
m
ining
strat
egies;
t
he
di
vid
e
a
nd
conq
uer
st
rategy
f
or
decisi
on
tree
le
arn
in
g
with
t
he
se
par
at
e
a
nd
co
nque
r
strat
egy
f
or
r
ule
le
a
rn
i
ng.
O
ner
ge
ner
at
es
a one
l
evel
decisi
on
tr
ee
that
is
expresse
d
in
the
form
of
a
set
of
r
ules
th
at
al
l
te
st
on
e
par
ti
cula
r
at
tribu
te
.
A
Mult
il
ay
er
Perce
ptr
on
is
a
ver
si
on
of
the
or
i
gin
al
pe
rce
pt
ron
m
od
el
propose
d
by
Ros
enb
la
tt
in
the
1950s
a
nd
c
on
sidere
d
as
a
ty
pe
of
neural
netw
orks
[
26
]
.
A
perce
ptr
on
(a
rtific
ia
l
neuron)
is
a
functi
on
of
sever
al
i
nput
per
ce
ptr
ons
w
hich
is
form
ed
as
a
com
bin
at
ion
of
input
wei
gh
ts
to
the
hidden
la
ye
r
per
ce
ptr
on
s
w
hich
le
a
d
them
to
the
ou
t
put
la
ye
r.
Finaly
gr
a
phic
al
m
od
el
s
su
c
h
as
ba
ye
sia
n
netw
or
ks
s
upply
a
ge
ner
al
fr
am
ewo
r
k
for
deali
ng
wit
h
un
ce
rtai
nly i
n a p
roba
bili
sti
c sett
ing
a
nd th
us are
well
s
uited to
tackle
t
he pr
oble
m
o
f
pre
dicti
on
.
In
this
stu
dy,
we
ha
ve
m
e
t
ou
r
ob
j
ect
ives
of
evaluati
ng
an
d
inv
est
i
gating
the
per
f
or
m
ances
of
dif
fer
e
nt
data
m
ining
te
chn
i
qu
e
s
for
th
e
data
set
that
is
being
us
e
d
to
unde
rstan
d
de
sire
and
inte
nt
ion
to
par
ti
ci
pa
te
in
virtu
al
c
omm
u
niti
es.
In
a
ddit
ion
to
the
st
udie
s
in
the
sci
entifi
c
body
of
knowle
dge
a
colla
borati
ve
and
con
t
rib
utive
da
ta
m
ining
ap
proac
h
is
app
l
ie
d
to
unde
rst
and
desire
a
nd
intenti
on
to
par
ti
ci
pate
in
virtu
al
com
m
un
it
ie
s.
Ba
sed
on
the
r
esults,
m
ulti
la
y
er
pe
rce
ptr
on
ha
d
the
m
os
t
co
rr
ect
cl
assifi
cat
ion
rate
with
93.
007
per
ce
nt,
a
good
true
posit
ive
rate
of
0.9
30
and
a
preci
sio
n
0
.
921.
Pa
rt
m
e
thod
ha
d
a
corr
ect
cl
assifi
cat
i
on
r
at
e
of
91.60
p
e
rce
nt,
true
posit
ive
rate
of
0.9
16
and
a
preci
sio
n
value
of
0.
923
.
Mult
il
ay
er
per
ce
ptr
on
ha
d
th
e
lowest
RM
SE
with
a
value
of
0.
24.
Ba
sed
on
the
high
co
rr
ect
cl
assifi
ca
ti
on
rate
an
d
low
RM
SE
m
e
asur
e
,
m
ul
ti
la
ye
r
perce
ptr
on
(
ne
ur
a
l
netw
ork
)
ca
n
be
c
onside
r
ed
as
an
ef
fe
ct
ive
m
et
ho
d
and
ca
n
be
use
d
i
n
unde
rstan
ding
desire a
n
d
inte
ntion t
o pa
rtic
ipate
in virt
ual
com
m
un
it
ie
s a
nd it
s an
te
ce
de
nts.
REFERE
NCE
S
[1]
Porter,
C.
E.
,
“
A
ty
po
log
y
of
virt
ual
comm
u
nit
ie
s:
A
m
ult
i‐
disci
pli
n
ar
y
fou
ndat
ion
for
fu
t
ure
rese
ar
ch,”
Journal
of
Computer‐Me
dia
te
d
Comm
unic
ati
on
,
1
0(1)
Article
3
,
2
004.
[2]
Plant
,
R.
,
“
Onlin
e
comm
unit
ie
s
,”
Te
chno
log
y
in
Societ
y
.
26,
pp.
5
1
-
65
,
2004
.
[3]
Rhei
ngold
,
H.
,
“
The
v
irt
ua
l com
m
unity
.
US
A:
MIT
Press
,
”
2000
.
[4]
Sim
oudis,
E.
“
Rea
l
ity
Ch
ec
k
for D
at
a
Min
ing,”
I
EE
E
EXPE
RT
,
1
1(5),
26
-
33
,
199
6.
[5]
Shear
er,
C.
,
“
Th
e
CRIS
P
-
DM
mode
l:
th
e
n
ew
bl
uepr
int
for
data
m
ini
ng,
”
Journal
of
Data
Wareh
ousing
,
vol
.
5
,
pp.
13
-
22
,
2000
.
[6]
Águila
,
R.
D
.
M.,
Ramírez
,
G.A.
,
“
Serie
s:
basic
stat
isti
cs
for
bus
y
cl
in
icians,”
Al
le
rgol
Imm
un
opathol
,
42
(5)
,
pp.
485
-
492
,
20
13
.
[7]
Yau
C.
,
“
R
tu
tor
ia
l
with
b
a
y
esi
an
statistic
s using
openbugs,
”
2013
.
[8]
Bagoz
z
i,
R.
P.
&
Dholaki
a
U.
M.,
“
Inte
nti
on
al
socia
l
a
ct
ion
i
n
virt
ual
communit
ie
s
,
”
Journa
l
of
Inte
rac
tiv
e
Marke
ti
ng
,
16
(2
),
pp
.
2
-
21
,
2002
.
[9]
Chiu,
C.
M.
,
Hs
u
M.,
W
ang,
E.,
“
Understa
nding
knowledge
sharing
in
virt
u
al
comm
unit
ie
s:
an
int
egr
a
ti
on
of
socia
l
ca
p
ital an
d
social cognitiv
e
th
eor
ie
s
,
”
Dec
i
sion S
upport
Sys
te
ms
,
42
(3), pp.
1872
-
1888,
200
6.
[10]
Shin,
D.
H.,
“
Understa
nding
pu
rch
asing
beh
aviors
in
a
virt
ual
ec
onom
y
:
Consum
er
beha
vior
i
nvolvi
ng
virt
u
al
cur
ren
c
y
in
W
eb
2.
0
comm
unit
ies
,
”
Int
erac
t
ing
w
it
h
compute
rs
,
2
0(4
-
5),
433
-
446
,
2008.
[11]
Davis,
F.D.
,
“
Perc
e
ive
d
use
fuln
ess,
per
c
ei
v
ed
e
ase
of
use
and
user
a
cc
ep
ta
n
c
e
of
info
rm
at
io
n
te
c
hnolog
y
,
”
MIS
Q
uarterly
,
13
(3), pp. 319
–
340,
1989
.
[12]
Porter,
E
.
&
Do
nthu,
N.
,
“
Us
ing
the
technolog
y
a
ccept
an
ce
m
odel
to
exp
lain
how
at
ti
tud
es
de
te
rm
ine
in
te
rne
t
usage
:
th
e
role
of
per
ceive
d
ac
c
ess
bar
rie
rs
and
demograph
ic
s,”
Journal
o
f
Busine
ss
Re
s
earc
h
,
59
(9)
,
pp.
999
-
1007
,
2
006.
[13]
Fis
hbei
n,
M.
,
M
anf
red
o,
M.J.
,
“
A
the
or
y
of
beh
a
vior
ch
ange
influenci
ng
hum
an
beha
vior
:
th
eor
y
and
appl
i
cations
in
re
creat
ion
,
tou
rism
and
na
tura
l
resourc
es
m
ana
g
ement,
”
Champa
ign,
I
ll
ino
is:
Sag
amor
e
Publ
ishin
g
,
1992
.
[14]
Ajze
n,
I.
&
Fis
hbei
n,
M
.
,
“
Understa
nding
attitudes
and
pre
d
icting
soci
al
b
ehaviour
,
”
Englew
ood
Cli
f
fs,
N
J
:
Prent
i
ce
Hal
l,
1
980.
[15]
Dholaki
a
,
U.M.,
Bagoz
zi,
R.
P.
&
Pear
ob,
L
.
K.
,
“
A
socia
l
infl
ue
nce
m
odel
of
co
nsum
er
par
ti
ci
p
a
ti
on
in
ne
twork
and
sm
al
l
group
base
d
v
irt
ua
l
c
om
m
unit
ie
s,”
In
te
rnational
Jour
nal
of
Re
search
in
Marke
ti
ng
,
2
1,
pp.
241
-
263
,
2004.
[16]
Al
-
Ta
i
e,
M.
Z.,
Kadr
y
,
S.,
Luc
a
s,
J.P.,
“
Online
Data
Preproc
essing:
A
Case
Stud
y
Approa
ch,”
Inte
rnation
a
l
Journal
of
Elec
t
rical
and
Computer
Eng
ine
ering
(
IJE
CE)
,
9(4),
2
019
[17]
Cohen,
W
.
“
Fast
eff
e
ct
iv
e
ru
le
induction.
In
A.
Prieditis
and
S.
Russ
el
l
(ed
s.),
”
Proceedi
ng
s
of
th
e
12
t
h
Inte
rnational
Co
nfe
renc
e
on
Mac
hine
Learning
,
L
ake
Ta
ho
e, CA,
pp.
115
–
123,
199
5.
Evaluation Warning : The document was created with Spire.PDF for Python.
In
t J
Elec
&
C
om
p
En
g
IS
S
N:
20
88
-
8708
A data
m
i
ning
approac
h f
or
de
sire a
nd inten
ti
on
to
parti
ci
pate
in
vi
rt
ua
l c
omm
un
it
ie
s
(
Ö
zerk Y
avu
z
)
3719
[18]
Sasaki
M.
,
Kit
a K.,
“
Rule ba
sed
te
xt
cate
gor
iz
a
tion using
hi
era
r
c
hic
a
l
c
ategori
es
,
”
IE
EE
.
[19]
Bla
ckmore,
K.
&
Boss
om
ai
er,
T.
R.
J.
.
,
“
Com
par
ison
of
See5
a
nd
J48.PA
RT
a
lgori
thms
for
m
issing
per
sons
profil
ing
,
”
pp.
3
37
-
342,
2002
.
[20]
Frank
E.
and
W
it
te
n
I.
H.
“
Data
Mining:
Pr
ac
t
ic
a
l
Mac
hin
e
Le
arn
ing
Too
l
s
and
Te
chni
q
ues
with
Java
Im
ple
m
e
nta
ti
ons
,
”
Morgan
Kauf
mann P
ubli
shers
:
San
Fran
ci
sco
,
CA,
2000.
[21]
Ramchoun,
H.
r
.
,
Idr
issi,
M.
m
.
,
Ghanou,
Y.
y
.
,
&
Ettaouil,
M
.
m
.
“
New
Modeli
ng
of
Multi
lay
er
Perc
ept
ron
Archi
tectur
e
Op
ti
m
iz
ation
with
Regul
ar
iz
a
ti
on:
An
Applic
at
ion
to
Patt
ern
Cl
assific
a
ti
o
n,
”
IAE
N
G
Inte
rnational
Journal
of
Computer
Sc
ie
nc
e,
44(
3),
261
-
269
,
201
7.
[22]
Ta
nigu
chi
M.,
Haft
M.,
Hollm´en
J.,
and
Tre
sp
V,
“
Fraud
det
ec
ti
on
in
comm
unic
a
ti
ons
net
wor
ks
using
neur
al
and
proba
bilis
ti
c
m
et
hods,”
In
Proce
ed
ings
of
the
1998
IEE
E
Int
ernati
onal
Confe
r
enc
e
on
A
coustics
,
Speech
and
Signal
Proce
ss
in
g
,
(ICAS
SP
’98)
,
Volum
e
II
,
pp
.
1241
–
1244,
199
8.
[23]
Boucka
er
t,
R.
R.
,
“
Ba
y
esia
n
n
et
work
cl
assifi
e
rs
in
W
eka
,
”
(W
orking
pape
r
serie
s.
Univer
sit
y
of
W
ai
kat
o
,
Depa
rtment
o
f
C
om
pute
r
Scie
n
ce.
No.
14/2004)
.
Ham
il
ton,
N
ew Zeala
nd:
Univer
sit
y
of
W
a
ika
to
,
2004.
[24]
Burhanuddi
n,
M
.
A.,
Ism
ai
l,
R.
,
Iz
za
imah
,
N.,
Moham
m
ed,
A.
A
-
J.,
Za
ino
l,
N.,
“
Anal
y
sis
of
M
obil
e
Servi
c
e
Providers
Perform
anc
e
Us
ing
N
ai
ve
Ba
y
es
Dat
a
Mining
Te
chn
ique
,
”
Inte
rnati
onal
Journal
of
El
ec
tri
cal
an
d
Computer
Engi
n
ee
ring (
IJE
C
E)
,
8(6),
pp
.
5153
-
51
61,
2018
.
[25]
Hair
,
J
.
,
Black,
W
.
,
Bab
in, B. &
Anderson,
R.
,
“
Multi
var
i
ate
d
ata
an
aly
sis.
NJ
:
Prentice
H
al
l
,
”
20
10.
[26]
Rosenbla
tt,
F.
,
&
Cornell
Aero
naut
i
ca
l
La
bor
ator
y
,
“
The
p
erc
e
ptron:
A
the
o
r
y
of
stat
ist
ical
s
epa
rab
il
i
t
y
in
cogni
ti
v
e
s
y
stem
s (Proje
ct Para
)
,
”
Buffa
lo, N.
Y:
Cornel
l
Aerona
u
ti
c
al
La
bor
at
or
y
,
1958.
BIOGR
AP
H
I
ES
OF
A
UTH
ORS
Öz
er
k
Yav
u
z
h
olds
a
PhD
in
B
uss
ine
ss
Ad
m
ini
strat
ion
and
Ms
c
.
in
Com
pute
r
E
ngine
er
ing.
He
is
int
ere
sted
in
software
engi
ne
eri
ng,
computer
engi
ne
eri
ng,
da
t
a
m
ini
ng,
virt
ua
l
comm
unit
ie
s,
m
ark
et
ing
and
m
ana
gement.
He
has
pub
li
shed
articles
in
v
arious
fie
lds.
He
h
as
interna
t
ional
working
expe
ri
e
nce
in
sev
eral
co
untri
es
and
cur
re
ntly
working
in
Alti
nbas
Univer
sit
y
,
Com
pute
r
Engi
ne
eri
ng
D
ep
art
m
ent
in
Ist
anb
ul.
Ad
em
Kar
ahoca
holds
a
PhD
in
Software
Eng
ine
er
ing.
He
is
i
nte
rest
ed
in
hu
m
an
–
computer
int
er
ac
t
ion,
web
-
base
d
educ
a
ti
o
n
sy
st
ems
,
dat
a
m
ini
ng,
big
dat
a
and
m
ana
gement
informati
on
s
y
stems
.
He
has
publi
shed
a
rticl
es
at
pr
esti
gious
journa
ls
about
dat
a
m
ini
ng
app
li
c
at
ions
and
business i
nform
at
ion
s
y
s
te
m
s in
h
ea
l
th, t
ourism
an
d
educat
ion
.
Dile
k
Kar
ahoc
a
is
a
socia
l
a
nthropol
ogist
.
Holding
a
PhD
in
Com
pute
r
Educ
a
ti
on
and
Instruc
ti
on
al
Tec
hnologi
es,
she
is
int
e
reste
d
in
hu
m
an
computer
in
te
ra
ct
ion
,
web
b
ase
d
edu
cation
s
y
stems
,
and
b
l
ende
d
l
ea
rn
ing
m
et
hodologi
es.
She
has
publi
sh
ed
seve
r
al
art
i
cles
about
use
of
informati
on
s
y
st
ems
in
he
al
th
,
to
urism
,
and educa
ti
o
n.
Evaluation Warning : The document was created with Spire.PDF for Python.