Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
23
,
No.
1
,
Ju
ly
2021
, p
p.
445
~
452
IS
S
N: 25
02
-
4752, DO
I: 10
.11
591/ijeecs
.v
23
.i
1
.
pp
445
-
452
445
Journ
al h
om
e
page
:
http:
//
ij
eecs.i
aesc
or
e.c
om
Hybrid
m
odel fo
r movi
e rec
ommendati
on sy
s
tem
u
sing co
ntent
K
-
ne
arest nei
ghbors and
restrict
ed
B
oltz
mann
machine
Day
al Kum
ar
Be
hera
1
,
M
adhab
anan
da D
as
2
,
Sub
hra
S
w
eta
nish
a
3
, P
rabira
Ku
mar
Set
h
y
4
1
,2
Schoo
l
of
Co
m
pute
r
Engi
n
ee
r
ing,
KIIT
De
emed
to
be
Univer
s
ity
,
Bhubane
sw
a
r,
Indi
a
3
Depa
rtment of
CS
E,
Tr
ide
nt
Ac
ade
m
y
of Tec
hn
olog
y
,
Bhub
ane
s
war,
Ind
ia
4
Depa
rtment of
El
e
ct
roni
cs,
Sam
bal
pur
Univ
ersity
,
J
y
oti Viha
r
,
B
urla
,
Odisha
Ind
i
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
Ma
r
1
6
, 202
1
Re
vised
Jun
3
,
2021
Accepte
d
J
un
1
7
, 202
1
One
of
the
most
comm
only
used
te
chni
qu
es
in
the
rec
o
m
m
enda
ti
on
fra
m
ework
is
co
ll
abor
at
iv
e
fi
lt
er
ing
(CF).
It
per
f
orm
s
bet
te
r
with
suffic
i
en
t
rec
ords
of
user
r
at
ing
bu
t
is
not
good
in
sparse
d
at
a
.
Cont
ent
-
bas
ed
filteri
n
g
works
well
in
th
e
sparse
dataset
as
it
finds
th
e
si
m
il
ari
t
y
b
et
wee
n
m
ovie
s
b
y
using
attributes
of
the
m
ovie
s.
RBM
is
an
ene
r
g
y
-
bas
ed
m
odel
serving
as
a
bac
kbone
of
de
e
p
le
arn
ing
and
p
erf
orm
s
well
in
rat
ing
pre
d
ic
t
ion
.
How
eve
r,
the
ra
ti
ng
pre
d
iction
is
not
pr
eferabl
e
b
y
a
singl
e
m
odel
.
Th
e
h
y
brid
m
odel
ac
hi
eve
s be
tt
e
r
r
esult
s b
y
in
te
gr
a
ti
ng
th
e
r
esult
s o
f
m
ore
tha
n
one m
odel
.
Th
is
pape
r
ana
l
y
ses
t
he
weighted
h
ybrid
CF
s
y
stem
b
y
integra
t
ing
c
onte
n
t
K
-
nea
rest
nei
ghbor
s
(KN
N)
with
restr
ic
t
ed
Bolt
zma
nn
m
ac
hine
(RB
M).
Movies
are
rec
om
m
ended
to
the
ac
ti
v
e
user
in
the
proposed
s
y
stem
b
y
integra
t
ing
the
eff
ects
of
both
c
onte
nt
-
b
ase
d
an
d
col
l
abor
a
ti
ve
f
il
te
r
ing.
Model
e
ffic
a
c
y
was
te
sted
with
Movi
eL
ens
b
enc
hm
ar
k
dataset
s.
Ke
yw
or
d
s
:
Coll
aborati
ve f
il
te
ring
Con
te
nt KN
N
Mov
ie
reco
m
m
end
at
io
n
RBM
Re
com
m
end
er
syst
e
m
This
is an
open
acc
ess arti
cl
e
un
der
the
CC
B
Y
-
SA
l
ic
ense
.
Corres
pond
in
g
Aut
h
or
:
Pr
a
bira Kum
ar S
et
hy
Dep
a
rtm
ent o
f El
ect
ro
nics
Sam
balpu
r U
niv
ersit
y
Jyoti
V
iha
r,
B
url
a,
Od
is
ha In
di
a
Em
a
il
: pr
abirse
thy.0
5@gm
ai
l.
com
1.
INTROD
U
CTION
Re
com
m
end
er
s
yst
e
m
[1]
is
prefe
rab
le
i
n
the
era
of
in
f
orm
ation
ov
e
rlo
ad
to
hel
p
us
e
rs
disco
ver
interest
ing
thi
n
gs
f
ro
m
a
n
extensi
v
e
colle
c
ti
on
of
data.
C
ollaborat
ive
filt
ering
(CF)
[2]
,
[3]
al
gorith
m
s
are
com
m
on
ly
u
sed
to cr
eat
e
re
c
omm
end
er s
yst
e
m
s b
ecause th
ey
eff
ic
ie
ntly
u
se the u
se
rs
’
ra
ti
ng
info
rm
ation
. An
it
e
m
rati
ng
m
at
rix
can
be
use
d
to
e
xplore
oth
e
r
in
visible
or
unf
ound
it
e
m
s
fo
r
sim
i
la
r
us
e
rs.
Hyb
rid
R
S
com
bin
es
CF
r
ecom
m
end
at
io
n
an
d
c
onte
nt
-
base
d
RS.
T
he
CF
-
ba
sed
rec
omm
end
at
io
n
is
ap
plica
ble
in
m
any
app
li
cat
io
n
do
m
ai
ns
[4]
,
[5]
.
Mod
el
-
base
d
Coll
aborati
ve
f
il
te
ring
ap
proa
ch
use
s
m
any
m
achine
le
ar
nin
g
[
6]
base
d
m
od
el
s
s
uch
as
Ra
nd
om
fo
rest
[7]
,
s
upport
vect
or
m
achine
(SV
M)
[8]
,
an
d
m
a
trix
facto
rizat
ion
[
9]
for
pr
e
dicti
ng
the
us
ers
’
li
ken
es
s
.
It
is
est
i
m
at
e
d
that
m
any
strea
m
ing
serv
ic
e
s
com
pan
ie
s
create
a
lot
of
rev
en
ue
by
app
ly
in
g
m
ov
ie
reco
m
m
end
at
ion
te
c
hn
i
qu
e
s.
Va
riou
s
m
od
el
s
ha
ve
bee
n
pro
po
s
ed
to
im
pr
ove
the
pro
du
ct
ivit
y o
f
the
m
ov
ie
r
ec
omm
end
at
io
n
[
10
]
fr
am
ewo
r
k.
C
F
m
e
t
h
o
d
s
a
r
e
b
r
o
a
d
l
y
d
i
v
i
d
e
d
i
n
t
o
t
w
o
c
a
t
e
g
o
r
i
e
s
:
m
e
m
o
r
y
-
b
a
s
e
d
a
n
d
m
o
d
e
l
-
b
a
s
e
d
.
I
n
a
s
p
a
r
s
e
d
a
t
a
s
e
t
,
t
h
e
m
e
m
o
r
y
-
b
a
s
e
d
m
o
d
e
l
d
o
e
s
n
o
t
p
e
r
f
o
r
m
w
e
l
l
,
u
n
l
i
k
e
t
h
e
m
o
d
e
l
-
b
a
s
e
d
C
F
.
K
-
n
e
a
r
e
s
t
n
e
i
g
h
b
o
r
s
[11
]
-
[13]
b
a
s
e
d
m
o
d
e
l
i
s
s
u
c
c
e
s
s
f
u
l
l
y
a
p
p
l
i
e
d
i
n
m
a
n
y
c
l
a
s
s
i
f
i
c
a
t
i
o
n
m
o
d
e
l
s
.
RBM
[
1
4
]
h
a
s
a
t
t
a
i
n
e
d
m
u
c
h
a
t
t
e
n
t
i
o
n
i
n
r
e
c
e
n
t
C
F
l
i
t
e
r
a
t
u
r
e
,
a
s
i
t
h
a
s
a
c
o
m
p
a
t
i
b
l
e
g
r
a
p
h
i
c
a
l
s
t
r
u
c
t
u
r
e
w
i
t
h
S
V
D
a
n
d
P
M
F
.
S
a
l
a
k
h
u
t
d
i
n
o
v
e
t
a
l
.
[
1
5
]
d
e
m
o
n
s
t
r
a
t
e
d
t
h
e
u
s
e
o
f
R
B
M
i
n
C
F
a
n
d
s
c
o
r
e
d
6
p
e
r
c
e
n
t
h
i
g
h
e
r
t
h
a
n
t
h
e
s
t
a
n
d
a
r
d
b
a
s
e
l
i
n
e
p
r
e
d
i
c
t
o
r
s
y
s
t
e
m
o
f
N
e
t
f
l
i
x
.
I
t
a
l
s
o
s
e
r
v
e
s
a
s
b
u
i
l
d
i
n
g
b
l
o
c
k
s
f
o
r
D
B
N
a
r
c
h
i
t
e
c
t
u
r
e
[
1
6
]
a
n
d
m
a
n
y
o
t
h
e
r
d
e
e
p
l
e
a
r
n
i
n
g
[
1
7
]
-
[
1
9
]
m
o
d
e
l
s
.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
445
-
452
446
C
h
u
n
c
h
u
n
L
i
a
n
d
J
u
n
L
i
[
2
0
]
h
a
v
e
u
s
e
d
d
u
a
l
c
R
B
M
t
o
a
d
d
r
e
s
s
d
u
a
l
u
s
e
r
a
n
d
i
t
e
m
p
a
t
t
e
r
n
s
.
I
n
t
h
e
i
r
m
o
d
e
l
r
a
t
i
n
g
m
a
t
r
i
x
r
e
f
l
e
c
t
s
a
l
a
t
e
n
t
m
a
t
r
i
x
o
f
f
e
a
t
u
r
e
s
.
T
h
e
y
u
s
e
d
s
e
p
a
r
a
t
e
R
B
M
f
o
r
e
a
c
h
u
s
e
r
,
p
r
o
p
o
s
i
n
g
a
r
e
s
t
r
i
c
t
e
d
R
B
M
(
c
R
B
M
)
r
e
c
o
m
m
e
n
d
a
t
i
o
n
f
o
r
N
e
t
f
l
i
x
m
o
v
i
e
s
.
T
h
e
C
o
n
t
e
n
t
-
b
a
s
e
d
r
e
c
o
m
m
e
n
d
a
t
i
o
n
f
o
c
u
s
e
s
p
r
i
m
a
r
i
l
y
o
n
m
o
v
i
e
d
e
s
c
r
i
p
t
i
o
n
o
r
d
e
t
a
i
l
s
,
w
h
i
l
e
C
F
f
o
c
u
s
e
s
o
n
u
s
e
r
r
a
t
i
n
g
s
f
o
r
t
h
e
m
o
v
i
e
.
I
f
r
a
t
i
n
g
i
n
f
o
r
m
a
t
i
o
n
i
s
c
o
n
s
i
d
e
r
e
d
,
i
t
i
s
p
o
s
s
i
b
l
e
t
o
c
o
l
l
e
c
t
e
x
t
r
e
m
e
l
y
s
p
a
r
s
e
d
a
t
a
a
s
f
e
w
c
o
n
s
u
m
e
r
s
c
h
o
o
s
e
t
o
g
i
v
e
t
h
e
r
a
t
i
n
g
o
f
t
h
e
p
r
o
d
u
c
t
a
f
t
e
r
p
u
r
c
h
a
s
e
.
R
B
M
,
a
l
o
n
g
w
i
t
h
C
o
n
t
e
n
t
K
N
N
,
i
s
u
s
e
d
t
o
a
d
d
r
e
s
s
t
h
i
s
i
s
s
u
e
.
F
i
s
c
h
e
r
a
n
d
I
g
e
l
[
2
1
]
d
i
s
c
u
s
s
e
d
d
i
f
f
e
r
e
n
t
R
B
M
l
e
a
r
n
i
n
g
a
l
g
o
r
i
t
h
m
s
s
u
c
h
a
s
C
D
a
n
d
p
a
r
a
l
l
e
l
t
e
m
p
e
r
i
n
g
.
T
h
e
m
o
d
i
f
i
e
d
C
D
a
l
g
o
r
i
t
h
m
c
a
l
l
e
d
‘
L
e
a
n
C
o
n
t
r
a
s
t
i
v
e
D
i
v
e
r
g
e
n
c
e
’
w
a
s
p
r
o
p
o
s
e
d
b
y
N
i
n
g
e
t
a
l
.
[
2
2
]
to
i
n
c
r
e
a
s
e
t
h
e
s
p
e
e
d
o
f
t
h
e
l
e
a
r
n
i
n
g
a
n
d
p
r
e
d
i
c
t
i
o
n
p
r
o
c
e
s
s
.
P
u
j
a
h
a
r
i
e
t
a
l
.
[
2
3
]
p
r
o
p
o
s
e
d
a
C
F
m
o
d
e
l
b
y
t
a
k
i
n
g
t
h
e
p
r
e
f
e
r
e
n
c
e
r
e
l
a
t
i
o
n
s
h
i
p
a
s
a
n
i
n
p
u
t
t
o
R
B
M
a
n
d
b
y
c
o
m
b
i
n
i
n
g
t
h
e
s
i
d
e
i
n
f
o
r
m
a
t
i
o
n
o
f
t
h
e
m
o
v
i
e
s
,
t
r
y
i
n
g
t
o
g
e
n
e
r
a
t
e
t
h
e
r
a
n
k
i
n
g
o
f
t
h
e
i
t
e
m
s
.
K
i
l
a
n
i
e
t
a
l
.
s
u
g
g
e
s
t
e
d
a
h
y
b
r
i
d
R
S
[
2
4
]
b
a
s
e
d
g
e
n
e
t
i
c
a
l
g
o
r
i
t
h
m
u
s
i
n
g
M
F
a
n
d
K
N
N
t
h
a
t
u
s
e
s
b
o
t
h
l
a
t
e
n
t
f
a
c
t
o
r
a
n
d
n
e
i
g
h
b
o
r
h
o
o
d
d
a
t
a
t
o
r
e
c
o
m
m
e
n
d
i
t
e
m
s
t
o
a
n
a
c
t
i
v
e
u
s
e
r
.
I
n
l
i
t
e
r
a
t
u
r
e
i
t
i
s
s
h
o
w
n
t
h
a
t
e
n
s
e
m
b
l
e
m
o
d
e
l
s
[
2
5
]
,
[
2
6
]
w
o
r
k
s
b
e
t
t
e
r
i
n
t
e
x
t
-
b
a
s
e
d
a
n
a
l
y
t
i
c
s
[
2
7
]
-
[
2
9
]
,
i
n
t
r
u
s
i
o
n
d
e
t
e
c
t
i
o
n
[30]
a
n
d
i
n
c
l
a
s
s
i
f
i
c
a
t
i
o
n
o
f
s
o
u
n
d
s
s
i
g
n
a
l
s
[
3
1
]
.
The
c
on
te
nt
-
ba
sed
K
N
N
w
or
ks
well
by
fin
ding
sim
i
la
rities
betwee
n
m
ov
ie
s
base
d
on
the
m
ov
ie
’s
descr
i
ptions
a
nd
oth
e
r
at
trib
ut
es
.
It
does
not
dep
e
nd
on
the
us
ers
’
rati
ng.
I
n
c
om
par
iso
n
t
o
t
he
CF
m
od
el
[32]
,
RBM
perform
s
w
el
l i
n
la
rg
e
dat
abases.
C
ontr
ibu
ti
ons
of this
work a
re s
ta
te
d
as
b
el
ow
:
Stud
yi
ng t
he
i
m
pact o
f
va
rio
us
sim
il
arity
m
easur
e
s in
u
se
r
-
base
d
a
nd it
em
-
based
co
ll
a
borati
ve
f
il
te
rin
g
.
Desig
ning
a
re
com
m
end
er
syst
e
m
to
deal
wi
th
the
s
pa
rse
da
ta
set
by
us
i
ng
the
m
ov
ie
s'
prof
il
e
s
uch
as
y
ear
of r
el
ease,
g
e
nre
,
an
d desc
ripti
on f
eat
ur
es
.
Desig
ning a
m
od
el
-
base
d rec
omm
end
er
syst
e
m
u
sin
g
Re
str
ic
te
d
Bolt
zm
an
n
Ma
c
hin
e.
Pr
op
os
in
g
a
w
ei
gh
te
d
hybr
i
d
m
od
el
base
d
on
c
on
te
nt
K
N
N
a
nd
RB
M
f
or
capt
ur
i
ng
both
co
ntent
-
ba
se
d
filt
ering
an
d hi
gh
e
r
-
orde
r
m
od
el
-
based coll
a
borati
ve fil
te
ring.
2.
RESEA
R
CH MET
HO
D
Con
si
der
U
as
the
con
s
um
er
vector
;
a
nd
M
as
the
m
ov
ie
vector
wh
e
re
p
,
q
is
the
use
r
an
d
m
ov
ie
nu
m
ber
s
,
r
espe
ct
ively
.
R
is
t
he
m
at
rix
with
the
form
p
x
q
.
r
um
,
r
̂
um
de
note
s
the
r
eal
rati
ng
of
th
e
co
nsum
er
‘
u
’
f
or
the
m
ovie
‘
m
’
and
the
exp
ect
e
d
rati
ngs.
T
he
s
im
i
la
rity
between
use
r
‘
u
’
an
d
us
er
‘
v
’
can
be
cal
culat
ed
us
in
g
(
1
)
[
33
]
.
r
̅
u
an
d
r
̅
v
re
pr
ese
nts
the
a
ve
rag
e
ra
ti
ng
of
us
er
‘
u
’
an
d
us
e
r
‘
v
’
,
r
especti
vely
.
Ra
ti
ng
of
us
er
‘
u
’
for
m
ov
ie
‘m
’
can
be
cal
c
ulate
d usin
g
(
5
)
.
sim
u
,
v
=
∑
(
r
um
−
r
̅
u
)
(
r
vm
−
r
̅
v
)
m
∈
M
uv
√
∑
(
r
um
−
r
̅
u
)
2
m
∈
M
uv
√
∑
(
r
vm
−
r
̅
v
)
2
m
∈
M
uv
(1)
M
uv
=
M
u
∩
M
v
(2)
r
̅
u
=
∑
r
um
m
∈
M
uv
|
M
uv
|
(3)
r
̅
v
=
∑
r
vm
m
∈
M
uv
|
M
uv
|
(4)
r
̂
um
=
b
um
+
∑
v
∈
N
m
k
(
u
)
si
m
u
,
v
.
(
r
vm
−
b
vm
)
∑
v
∈
N
m
k
(
u
)
si
m
u
,
v
(5)
b
um
=
µ
+
b
u
+
b
m
is
u
’
s
baseli
ne
estim
at
e
[34]
f
or
m
ov
ie
m
.
The
par
am
et
ers
b
u
an
d
b
m
ind
ic
at
e
us
er
‘
u
’
a
nd
m
ov
ie
‘
m
’
rati
ng
de
viati
ons.
Gen
e
rall
y,
sto
chasti
c
gr
a
dient
desce
nt
(
S
G
D
)
or
al
te
rn
at
ing
le
a
st
sq
ua
res
(
A
LS
)
al
go
rithm
so
lves
these
par
am
et
ers.
T
o
fi
nd
si
m
il
arit
ie
s
between
m
ov
ie
s ‘x’ an
d ‘y’
,
m
od
ifie
d
cosi
ne
-
si
m
il
arity
m
easur
e in
(6)
is
use
d.
sim
p
,
q
=
∑
(
r
ux
−
r
̅
u
)
(
r
uy
−
r
̅
u
)
u
∈
U
xy
√
∑
(
r
ux
−
r
̅
u
)
2
u
∈
U
xy
√
∑
(
r
uy
−
r
̅
u
)
2
u
∈
U
xy
(6)
The
pro
blem
with
t
his
baseli
ne
is
t
hat
to
f
ind
sim
il
arit
y
betwee
n
us
e
r
‘
u’
an
d
‘
v’,
the
rati
ngs
pro
vide
d
by
us
er
’s ‘u’ a
nd
‘
v’
are ta
ken int
o
acc
ount. It
does
not fit
w
el
l
in
s
parse
data.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Hyb
ri
d model f
or
mo
vi
e
reco
mm
e
ndatio
n sy
ste
m us
in
g
c
onte
nt
…
(
Da
y
al
Kumar B
eher
a)
447
2
.
1.
RBM
RB
M
is
an
e
ne
rg
y
-
base
d
,
pro
bab
il
ist
ic
m
od
el
with
on
ly
tw
o
la
ye
rs
of
the
neural
netw
ork
struct
ur
e:
a
hidden
la
ye
r
a
nd
a
visible
la
ye
r.
T
his
tw
o
-
la
ye
r
netw
or
k
has
the
rest
rict
ion
t
hat
t
he
re
is
no
c
on
necti
on
betwee
n
t
wo
node
s
in
the
sa
m
e
la
ye
r.
It
is
com
m
on
ly
us
ed
i
n
c
ollaborat
ive
filt
erin
g
t
o
le
arn
the
distr
ibu
ti
on
of
pro
ba
bili
ty
over
a
ra
nking
m
at
rix.
Th
is
netw
ork
is
trai
ne
d
th
rou
gh
gradie
nt
de
scent
a
nd
ba
ckw
a
r
d
pro
pag
at
io
n
,
w
her
e
eac
h
it
era
ti
on
is
hav
i
ng
a
forwar
d
pass
an
d
bac
kwar
d
pass
(
rec
onstr
uction
)
.
I
nform
at
ion
exch
a
nge
betw
een
ne
uro
ns
in
the
sa
m
e
laye
r
is
lim
it
ed
in
RB
M.
Ther
e
is
a
relat
ion
be
tween
tw
o
sep
arate
la
ye
rs
of
ne
uro
ns
.
RB
M
is
tra
ined
i
n
f
orwa
r
d
a
nd
bac
kwar
d
passes
,
an
d
t
he
in
puts
are
r
ebu
il
t
in
t
he
ba
ckw
a
r
d
pass
.
T
his
is
achieve
d
ove
r
sever
al
e
po
c
hs
befor
e
it
co
nv
e
r
ges
on
a
set
of
weig
hts
and
disto
rtio
ns
that
m
ini
m
iz
es the
reconstr
uctio
n err
or
.
It
descr
i
bes
a
di
stribu
ti
on
ove
r
V
with
a
la
ye
r
of
bin
a
ry
secret
un
it
‘
h’.
Th
e
input
la
ye
r
V
is
a
m
a
tri
x
of
k
×
m
di
m
e
ns
io
n
a
nd
=
1
,
if
us
er
rati
ng
for
the
m
ov
ie
‘m
’
is
‘k’.
The
li
ke
li
ho
od
of
eac
h
visible/
input
bin
a
ry
m
at
rix
V
c
olu
m
n
is
m
od
el
e
d
with
t
he
distri
bu
ti
on
“
So
ftm
ax.
”
Hidden
us
er
f
unct
ion
s
‘
h
’
is
li
kely
to
be
m
od
el
ed
with
s
igm
oid
f
unct
io
n
. T
he
en
e
rg
y
of the stat
e
(
,
ℎ
)
is
giv
e
n
in
(7)
[
21]
.
(
,
ℎ
)
=
−
−
ℎ
−
ℎ
(7)
In the lear
ni
ng
process
, th
e
tra
ining vect
or V
’
s
log
-
lik
el
ih
oo
d
to
w
ei
gh
t c
an
be dete
rm
ined
u
si
ng (8)
.
(
)
=
〈
ℎ
〉
−
〈
,
ℎ
〉
(8)
Wh
e
re,
〈
,
ℎ
〉
is
th
e
ex
pectat
ion
of
the
distrib
ut
ion
whose
co
m
plexit
y
is
ve
ry
hi
gh.
I
n
[
6],
CD
(Contrast
ive
D
iver
gen
ce
)
is
use
d
to
est
im
a
te
ex
pectat
ion
.
T
he
weig
ht
can
be
m
od
ifie
d
usi
ng
st
ochast
ic
ascent
giv
e
n
in
(9).
=
(
〈
ℎ
〉
−
〈
,
ℎ
〉
)
(9)
Wh
e
re,
is
the
le
arn
i
ng
rate.
To
ove
rco
m
e
the
c
om
plexity
,
we
ha
ve
use
d
1
-
ste
p
C
D
and
the
val
ues
are
updated
usi
ng
(10).
Δ
w
pq
=
α
(
〈
ϑ
p
h
q
〉
dat
a
−
〈
ϑ
p
,
h
q
〉
rec
on
stru
c
t
ion
)
(10)
In
t
he
pr
opos
e
d
m
od
el
,
a
hy
br
i
d
m
od
el
is
consi
der
e
d
with
eq
ual
weig
ht
age
f
or
both
RB
M
and
con
te
nt KN
N m
od
el
. F
igure
1 dep
ic
ts t
he
tr
ai
nin
g o
f
t
he h
ybrid
m
od
el
.
Figure
1.
The
pro
po
se
d hyb
ri
d
m
od
el
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
445
-
452
44
8
On
the
M
ov
ie
Lens
dataset
,
our
propose
d
m
od
el
is
i
m
ple
m
ented.
T
his
dat
aset
con
ta
ins
use
r
interest
s
and
m
ov
ie
pro
file
s
.
This
data
colle
ct
ion
is
t
hen
sub
j
ect
ed
t
o
pr
e
processi
ng
of
data,
i
n
w
hich
s
om
e
sa
m
ples
of
data
are
obta
in
ed.
T
he
70
-
30
te
st
div
isi
on
is
i
m
po
sed
on
t
he
pr
e
processe
d
data
.
I
n
this
ste
p
of
est
im
a
ti
on
,
the
RB
M
is
us
ed
,
and
the
RB
M
is
com
bin
ed
wi
th
the
c
on
te
nt
KNN
.
C
onte
nt
KNN
fin
ds
th
e
si
m
il
arity
between
diff
e
re
nt
m
ov
ie
s
based
on
a
tt
ribu
te
s
of
m
ov
ie
s
s
uch
as
ye
ar
of
relea
se,
ge
nr
e,
box
of
fice
hit
rat
e,
and
descr
i
ptions
of
the
m
ov
ie
.
Th
e
weigh
te
d
ave
rag
e
is
obta
ine
d
from
the
above
m
od
el
,
and
the
validat
ed
m
od
el
is use
d
to
pr
oduce a
t
op
N
r
ec
omm
end
at
io
n
f
or the acti
ve
us
er.
3.
RESU
LT
S
A
ND
DI
SCUS
S
ION
In
pyth
on
3.8
,
the
ab
ove
de
sign
e
d
al
gorithm
s
ha
ve
been
im
plem
ented,
an
d
Nu
m
Py,
pa
ndas,
sur
pr
is
e
are
the
m
ai
n
li
br
aries
use
d
f
or
im
ple
m
entat
ion
.
RB
M
net
work
is
trai
ne
d
i
n
the
f
orwa
rd
pas
s
by
fee
ding
trai
ning
inf
or
m
at
ion
on
the
vi
sible
la
ye
r
and
trai
nin
g
weig
hts
an
d
biases
betwee
n
them
durin
g
the
bac
kw
a
r
d
pass
.
T
o
ge
nerat
e
the
ou
tp
ut
of
a
ny
hidden
neur
on,
an
act
ivati
on
functi
on
recti
fied
li
ne
ar
unit
s
(
Re
LU
)
[35]
has bee
n
us
ed
.
Re
LU
is
pr
e
ferred
h
e
re as
it
does
not act
ivate
all
the
neurons
at
t
he
sam
e tim
e.
Fo
r
t
he
m
od
el
trai
ning,
eac
h
us
er
’
s
rati
ng
f
r
om
the
trai
nin
g
set
is
passe
d
as
a
batch
int
o
the
RB
M.
The
nodes
in
t
he
visi
ble
la
ye
r
ref
le
ct
the
use
r
’
s
rati
ngs
on
each
m
ov
ie
.
I
nt
erconn
ect
io
n
weig
hts
are
le
a
rn
e
d
to
reconstr
uct
the
rati
ng
s
for
use
r
-
m
ov
ie
pairs
that
are
m
issin
g.
He
re,
eac
h
ind
ivi
du
al
rati
ng
is
treat
e
d
as
fiv
e
visible
nodes
,
on
e
f
or
eac
h
possible
rati
ng
value.
Ra
ti
ng
i
n
th
e
RB
M
of
five
m
ov
ie
s
ha
ving
rati
ngs
5,
NA,
3,
NA
an
d
2,
res
pecti
vely
,
is
sh
ow
n
in
F
igur
e
2.
Ra
ti
ng
of
first
m
ov
ie
is
5,
hen
ce
5
th
colum
n
is
set
to
1.
Si
m
il
arly
, co
lum
n
2
of m
ov
ie
5
is set
t
o 1.
Figure
2
.
Re
pr
esentat
ion o
f r
at
ing
in
f
or
m
at
i
on in
RB
M
Ther
e
is
a
nee
d
to
est
im
at
e
m
issi
ng
rati
ng
s
f
or
m
ov
ie
s
2
an
d
4.
T
he
m
utu
al
weig
ht
a
nd
bias
can
be
us
e
d
to
pr
e
dic
t
the
m
issi
ng
scor
e
once
the
m
od
el
is
trai
ned.
Ma
ny
sc
or
es
a
re
m
issin
g
i
n
the
real
dataset
.
In
ste
a
d
of
t
rainin
g
on
any
possible
m
ix
of
us
e
rs
a
nd
m
ov
ie
s
,
the
m
od
el
is
trai
ned
on
the
avail
able
da
ta
by
rem
ov
ing
the
m
issi
ng
sco
res
.
In
general,
us
in
g
co
ntrasti
ve
di
vergen
ce
and
Gibb’s
s
a
m
pling
,
the
m
od
el
sam
ples
the
prob
a
bili
ty
distribu
ti
on.
The
re
su
lt
ing
weig
ht
s
and
biases
a
re
re
us
e
d
f
or
oth
e
r
us
e
rs
aft
er
the
m
od
el
h
a
s
bee
n
trai
ned f
or one
us
er
.
The
nu
m
ber
of
visi
ble
node
s
is
cal
culat
ed
as
the
product
of
num
ber
of
m
ov
ie
s
and
r
at
ing
values
giv
e
n
by
t
he
us
er.
I
n
t
he
datas
et
ML100K
,
th
e
nu
m
ber
of
m
ov
ie
s
is
8211,
and
num
ber
of
us
ers
in
the
tra
in
set
is
67
1.
So,
the
m
axi
m
u
m
num
ber
of
rati
ngs
po
ssi
ble
for
the
us
er
is
82
11*5=4
1055
a
nd
the
trai
ning
rati
ng
m
at
rix
dim
ension
is
671
×
41055
.
T
he
hi
dd
e
n
no
de
r
ep
resen
ts
the
la
te
nt
var
ia
ble
.
T
he
nu
m
ber
s
of
hidde
n
node
a
re
ta
ke
n
50
f
or
the
e
xp
erim
ent.
Epo
c
hs
represe
nt
th
e
num
ber
of
it
era
ti
on
s
to
ta
ke
for
the
f
orwa
rd
a
nd
backwa
rd
pass. In
eac
h
ep
oc
h, to m
ini
m
iz
e
the error
betwee
n
act
ual and r
e
const
ru
ct
e
d
rat
ing
value,
t
he m
od
el
will
be
trai
ned
acro
ss
al
l
the
us
ers
i
n
the
tra
in
set
.
The
num
ber
of
ep
ochs
is
set
to
5
to
30
with
an
i
nc
rem
ent
of
5.
T
he
im
pa
ct
of
the
e
po
c
hs
in
the
perf
orm
ance
m
et
ric
s
is
sh
own
i
n
Figure
3
.
Th
e
le
arn
in
g
rate
c
on
t
ro
ls
how q
uickly t
he
erro
r
co
nver
ge
s and it i
s set t
o 0.001.
Bat
ch si
ze co
ntro
ls
th
e num
ber
of
use
rs
to
b
e
pro
ce
ssed
at
a
t
i
m
e,
and
it
is
set
to
10
0.
On
ce
the
m
od
el
is
trai
ned
,
the
updated
w
ei
gh
t
an
d
bias
es
are
us
ed
to
get
the
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Hyb
ri
d model f
or
mo
vi
e
reco
mm
e
ndatio
n sy
ste
m us
in
g
c
onte
nt
…
(
Da
y
al
Kumar B
eher
a)
449
reco
m
m
end
at
ion
f
or
the
act
i
ve
us
er
.
T
he
s
of
tm
ax
act
ivati
on
f
un
ct
i
on
is
us
e
d
in
t
he
vis
ible
la
ye
r
to
ge
t
the
pro
bab
il
it
y sco
re th
at
i
nd
ic
at
e
d
the
li
keness
of the
us
e
r
f
or
a pro
duct
.
R
BM
is
co
m
bin
ed
with
Con
t
ent
KNN
to
co
ns
ide
r
the
ad
va
ntages
of
both
con
te
nt
-
based
filt
ering
an
d
m
od
el
-
based
c
ollaborat
ive
filt
ering
m
od
el
.
Con
te
nt
K
N
N
works
well
ev
en
for
ne
w
use
rs
a
nd
RB
M
works
well
for
la
r
ge
us
er
rati
ng
s.
The
pe
rform
ance
of
va
r
iou
s
e
ns
em
b
le
m
od
el
s
ha
s
bee
n
eval
uated
on
Mov
ie
Le
ns1M
and
10
0K
dat
aset
s
down
l
oa
ded
f
r
om
the
gr
ou
plens.o
rg
si
te
.
70
%
data
is
us
ed
f
or
trai
ni
ng
the
m
od
el
an
d 3
0%
d
at
a is
us
e
d t
o
te
st t
he
m
odel
.
F
i
g
u
r
e
4
d
e
p
i
c
t
s
t
h
e
r
e
s
u
l
t
s
o
f
U
s
e
r
-
b
a
s
e
d
a
n
d
I
t
e
m
-
b
a
s
e
d
C
F
u
n
d
e
r
v
a
r
i
o
u
s
s
i
m
i
l
a
r
i
t
y
m
e
a
s
u
r
e
s
s
u
c
h
a
s
c
o
s
i
n
e
,
p
e
a
r
s
o
n
,
a
n
d
m
e
a
n
s
q
u
a
r
e
d
d
i
f
f
e
r
e
n
c
e
(
M
S
D
)
.
I
t
e
m
-
b
a
s
e
d
C
F
p
e
r
f
o
r
m
s
b
e
t
t
e
r
u
n
d
e
r
M
S
D
s
i
m
i
l
a
r
i
t
y
.
I
n
C
o
n
t
e
n
t
K
N
N
,
b
y
f
i
n
d
i
n
g
t
h
e
s
i
m
i
l
a
r
i
t
y
b
a
s
e
d
o
n
y
e
a
r
o
f
r
e
l
e
a
s
e
a
n
d
g
e
n
r
e
,
R
M
S
E
s
c
o
r
e
s
i
n
M
L
1
0
0
K
d
a
t
a
s
e
t
a
r
e
1
.
0
6
2
4
a
n
d
1
.
0
5
4
8
,
r
e
s
p
e
c
t
i
v
e
l
y
.
I
t
i
n
d
i
c
a
t
e
s
t
h
a
t
g
e
n
r
e
p
l
a
y
s
a
m
a
j
o
r
r
o
l
e
i
n
r
e
c
o
m
m
e
n
d
i
n
g
t
h
e
m
o
v
i
e
s
.
A
c
c
u
r
a
c
y
m
e
a
s
u
r
e
i
s
n
o
t
c
o
n
s
i
d
e
r
e
d
h
e
r
e
,
a
s
t
h
e
m
o
d
e
l
i
s
n
o
t
a
c
l
a
s
s
i
f
i
c
a
t
i
o
n
m
o
d
e
l
.
Re
su
lt
s
of
c
ombinin
g
RB
M
w
it
h
Con
te
nt
K
NN
a
re
s
hown
in
Table
1.
w
1
an
d
w2
rep
r
esent
wei
ght
assigne
d
t
o
RB
M
and
co
ntent
KNN,
res
pecti
vely
.
A
series
of
e
xperim
ents
are
perform
ed
to
see
the
im
p
act
of
w1
a
nd
w2.
T
o
ge
ner
at
e
the
rati
ng
pr
e
dicti
on
of
the
ta
r
ge
t
us
er
-
m
ov
ie
pa
ir,
a
weig
hted
aver
a
ge
is
co
m
pu
te
d
by
co
ns
ide
rin
g
w1
a
nd
w
2
weig
hts
of
the
m
od
el
s.
Ro
ot
m
ean
square
error
(RMSE)
,
m
ean
abs
olu
t
e
error
(MAE
)
[
36
]
s
cor
es
of
the
hy
br
id
al
go
rith
m
s
are
sh
ow
n
in
Table
1.
A
lowe
r
value
i
ndic
at
es
bette
r
resu
lt
s.
ML1M,
ML1
00K
re
prese
nts
Mov
ie
Le
ns
on
e
m
i
ll
ion
and
100K
rati
ngs,
re
sp
ect
ively
.
For
con
te
nt
K
N
N,
the
K
value
is
set
to
40.
Fig
ur
e
3
an
d
Table
2
il
lust
rate
the
RM
SE
scor
e
of
va
rio
us
al
gorithm
s
i
n
dif
fer
e
nt
ep
oc
hs
of
trai
ning RBM
. T
he
a
ver
a
ge
v
a
lue of RM
SE
is cal
culat
ed
i
n t
he
Ra
nd
om
m
od
el
f
or
dif
fer
e
nt epoc
hs
.
F
i
g
u
r
e
3
.
R
M
S
E
s
c
o
r
e
o
f
v
a
r
i
o
u
s
m
o
d
e
l
s
o
n
M
L
1
M
d
a
t
a
F
i
g
u
r
e
4
.
R
M
S
E
s
c
o
r
e
o
f
U
s
e
r
-
b
a
s
e
d
a
n
d
I
t
e
m
-
b
a
s
e
d
C
F
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
445
-
452
450
Table
1
.
Perfor
m
anc
e
m
et
r
ic
s o
f
RBM
and
H
y
br
id
m
odel
s on
MovieLens da
t
a
RMSE
MAE
Dataset
ML1
00K
ML1
M
ML1
00K
ML
1
M
Ran
d
o
m
1.
5067
1
.46
1
5
1
.20
1
.16
RBM
1
.21
2
6
1
.19
8
5
0
.96
0
.86
Co
n
ten
tKNN
1
.05
4
8
1
.03
2
2
0
.74
0
.72
RB
M+Co
n
ten
tKNN
1
.11
5
5
1
.06
5
2
0.
93
0.
76
RB
M
+ Ran
d
o
m
1
.32
5
4
1
.25
7
8
1
.05
0
.98
Table
2
.
RMS
E
s
cor
e
of
v
ari
ous
m
odel
s on
ML1
M da
ta
Epo
ch
s
5
10
15
20
25
30
Ran
d
o
m
1
.48
2
1
1
.45
4
7
1
.46
1
5
1
.45
0
5
1
.45
0
5
1
.45
0
5
RBM
1
.25
1
3
1
.24
6
6
1
.19
8
5
1
.19
8
5
1
.19
8
5
1
.19
8
5
Hy
b
rid
(RB
M
+
C
o
n
ten
tKNN)
1
.18
1
2
1
.14
0
6
1
.06
5
2
1
.06
5
2
1
.06
5
2
1
.06
5
2
4.
CONCL
US
I
O
N
Coll
aborati
ve
f
il
te
ring
a
nd
c
on
te
nt
-
based
r
ecom
m
end
at
io
n
a
re
the
po
pula
r
te
c
hn
i
qu
e
s
in
filt
erin
g
m
ov
ie
s
fr
om
a
la
rg
e
c
ollec
ti
on
.
In
C
F,
the
m
ean
square
di
ff
e
ren
ce
pe
rform
s
bette
r
as
com
par
ed
to
c
os
ine
a
nd
Pears
on
sim
il
a
rity
m
easur
es.
In
c
on
te
nt
K
N
N,
genre
-
base
d
cal
culat
ion
hi
gh
ly
im
pa
ct
s
t
he
rec
omm
end
at
ion
.
RB
M
fits
well
in
m
od
el
-
ba
se
d
c
ollaborat
ive
filt
ering
an
d
con
te
nt
K
N
N
in
ver
y
s
par
se
dataset
.
T
he
goal
of
this
pap
e
r
is
to
us
e
a
weig
hte
d
hy
br
id
m
od
e
l
to
see
bo
t
h
RB
M
and
c
onte
nt
K
NN’s
com
bin
e
d
im
pact
on
the
m
ov
ie
reco
m
mend
at
io
n.
Bot
h
the
m
od
el
s
are
assigne
d
with
eq
ual
weig
htage.
The
e
ff
ect
of
the
hybri
d
m
od
el
on
rati
ng
data
from
Mov
ie
Lens
100
K
a
nd
1
M
is
th
oro
ug
hly
inv
est
i
gated.
Ta
ble
1
disp
la
ys
the
RM
S
E
an
d
MAE
sco
re
of
diff
e
re
nt
m
od
e
ls.
RB
M
with
c
on
te
nt
K
NN
i
s
sh
ow
n
to
ha
ve
bette
r
eff
ic
ie
ncy.
Ra
ndom
m
ov
ie
reco
m
m
end
at
ion
is
the
w
orst
per
f
orm
er.
In
the
fu
t
ur
e
,
RB
M
and
c
onte
nt
KNN
can
be
exten
ded
t
o
th
e
fiel
ds
of tea
chin
g, m
us
ic
, ne
ws,
a
nd
o
the
r recom
m
end
at
io
ns.
REFERE
NCE
S
[1]
V.
Agarwal
and
A.
Vijay
alaksh
m
i,
“
Rec
om
m
en
der
s
y
stem
for
surplus
stock
cl
e
ara
nc
e,”
Inte
rna
ti
onal
Journal
o
f
El
e
ct
rica
l
and
C
omputer
Engi
n
e
ering
,
vo
l. 9, no. 5, pp. 3813
–
382
1,
2019
,
doi
:
10
.
11591/i
jece
.
v9i5
.
pp3813
-
3821.
[2]
N.
S.
A.
Rahma
n,
L
.
Hand
a
y
ani,
M.
S.
Othm
an,
W
.
M.
Al
-
R
ah
m
i,
S.
Kasim
,
a
nd
T.
Sutikno,
“
Socia
l
m
ed
ia
fo
r
col
l
abor
ative
learni
ng,
”
In
te
rn
ati
onal
Journal
of
El
e
ct
rica
l
and
Compute
r
Engi
ne
ering
,
vol.
10,
no.
1,
pp.
1070
–
1078
,
2020,
doi
:
10
.
11
591/i
jece
.
v10i1
.
pp1070
-
1078.
[3]
S.
Babe
e
tha,
B.
Murugana
ntha
m
,
S.
Gane
sh
Ku
m
ar,
and
A.
Mur
ugan,
“
An
enha
n
ce
d
ker
n
el
we
ig
hte
d
co
ll
abor
at
iv
e
rec
om
m
ende
d
s
y
stem
to
al
l
evi
a
te
sparsit
y
,
”
In
te
rn
ati
onal
Journal
of
El
e
ct
ri
cal
and
Computer
Enginee
ring
,
vo
l.
10
,
no.
1
,
pp
.
447
–
4
54
,
2020
,
doi
:
10
.
11591/i
j
ece.
v10
i1.
pp447
-
454.
[4]
M.
Jali
l
i,
S.
Ah
m
adi
an,
M.
I
zadi,
P.
Morad
i,
a
nd
M.
Sale
hi
,
“
Eva
lu
at
ing
Co
llabora
t
ive
Fil
te
ri
ng
Rec
om
m
ende
r
Algorit
hm
s: A
Surve
y
,
”
IE
EE A
c
ce
ss
,
vol
.
6
,
pp
.
74003
–
74024,
2
018,
doi
:
10
.
110
9/ACCESS
.
2018.
2883742
.
[5]
D.
K.
Behe
ra,
M
.
Das,
and
S.
Swe
ta
n
isha,
“
A
Res
ea
rch
on
Coll
ab
ora
ti
v
e
Filt
er
ing
Based
Movie
Rec
om
m
enda
ti
ons:
From
Neighbor
hood
to
Dee
p
Le
a
rning
Based
S
y
s
t
em,”
Int
.
J
.
R
ec
e
nt
Technol.
Eng
.
,
vol. 8,
no
.
4,
pp
.
10809
–
10814,
2019,
doi
:
10
.
35
940/i
jrte.
d43
62
.
118419.
[6]
M.
N.
Meqda
d,
R.
Al
-
Akam
,
and
S.
Kadr
y
,
“
New
pre
dic
ti
on
m
ethod
for
dat
a
sprea
ding
in
social
net
works
base
d
o
n
m
ac
hine
learni
n
g
al
gorit
hm
,
”
T
E
L
K
O
M
N
I
K
A
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
,
C
o
m
p
u
t
i
n
g
,
E
l
e
c
t
r
o
n
i
c
s
a
n
d
C
o
n
t
r
o
l
,
vol
.
18,
no.
6
,
pp.
3331
-
3338
,
2020,
doi
:
10
.
12
928/t
el
kom
nik
a.v18i6.
16300.
[7]
Z
.
R
u
s
t
a
m
a
n
d
G
.
S
.
S
a
r
a
g
i
h
,
“
P
r
e
d
i
c
t
i
o
n
s
c
h
i
z
o
p
h
r
e
n
i
a
u
s
i
n
g
r
a
n
d
o
m
f
o
r
e
s
t
,
”
T
E
L
K
O
M
N
I
K
A
T
e
l
e
c
o
m
m
u
n
i
c
a
t
i
o
n
,
C
o
m
p
u
t
i
n
g
,
E
l
e
c
t
r
o
n
i
c
s
a
n
d
C
o
n
t
r
o
l
,
v
o
l
.
1
8
,
n
o
.
3
,
p
p
.
1
4
3
3
–
1
4
3
8
,
2
0
2
0
,
d
o
i
:
1
0
.
1
2
9
2
8
/
T
E
L
K
O
M
N
I
K
A
.
v
1
8
i
3
.
1
4
8
3
7
.
[8]
N.
P.
Shetty
,
J.
Shetty
,
R
.
Narul
a,
and
K.
T
ando
na,
“
Com
par
ison
stud
y
of
m
ac
h
ine
l
ea
rning
class
ifi
ers
to
d
et
e
c
t
anomali
es,
”
Inter
nati
onal
Journal
of
El
e
ct
rica
l
and
Computer
Engi
nee
ring
,
vol.
10,
no.
5,
pp.
5
445
–
5452,
20
20
,
doi:
10
.
11591/IJ
ECE
.
V10I5.PP
5
445
-
5452.
[9]
M.
Vlac
hos,
C.
Dunner,
R.
He
c
kel
,
V.
G
.
Vass
il
ia
dis
,
T.
Parn
el
l
,
and
K.
Atasu
,
“
Address
ing
Inte
rpre
t
abi
l
ity
an
d
Cold
-
Start
in
Matri
x
Fa
ct
ori
zati
on
for
Rec
om
m
ende
r
S
y
s
te
m
s,”
IEE
E
Tr
ans.
K
nowl.
Data
Eng
.
,
vol
.
3
1
,
no.
7
,
pp.
1253
–
1266
,
2019,
doi
:
10
.
11
09/T
KD
E.
2018.
2829521.
[10]
R.
Saga
and
Y.
Duan,
"A
ppar
e
l
Goods
Rec
om
m
ende
r
S
y
stem
Based
on
Im
age
Shape
Fea
ture
s
Ext
ra
cted
b
y
a
CNN
,
"
2018
IE
EE
In
te
rnationa
l
Confe
ren
ce
on
Syste
ms
,
Man
,
and
Cyb
erne
tic
s
(
SMC
)
,
2018,
pp.
2365
-
23
69
,
doi:
10
.
1109/SM
C.
2018.
00406
.
[11]
E.
G.
Nihad
,
E
.
N.
El Mokhta
r,
Z.
Abde
lha
m
id,
and
A.
A.
Moha
m
m
ed,
“
Hy
brid
appr
oac
h
of
th
e
f
uzzy
C
-
m
ea
ns
a
nd
the
K
-
nea
rest
n
ei
ghbors
m
et
hods
during
the
ret
ri
eve
phase
of
d
y
namic
c
ase
base
d
rea
soning
for
per
s
onal
iz
e
d
foll
ow
-
up
of
le
a
rne
rs
in
rea
l
ti
m
e,
”
Int
ernati
onal
Journal
of
El
ectric
al
and
Computer
Engi
ne
erin
g
,
vol.
9,
no.
6
,
pp.
4939
–
4950
,
2019,
doi
:
10
.
11
591/i
jece
.
v9i6
.
p
p4939
-
4950.
[12]
M
.
A
l
z
a
q
e
b
a
h
,
N
.
A
l
r
e
f
a
i
,
E
.
A
.
E
.
A
h
m
e
d
,
S
.
J
a
w
a
r
n
e
h
,
a
n
d
M
.
K
.
A
l
s
m
a
d
i
,
“
N
e
i
g
h
b
o
r
h
o
o
d
s
e
a
r
c
h
m
e
t
h
o
d
s
w
i
t
h
m
o
t
h
o
p
t
i
m
i
z
a
t
i
o
n
a
l
g
o
r
i
t
h
m
a
s
a
w
r
a
p
p
e
r
m
e
t
h
o
d
f
o
r
f
e
a
t
u
r
e
s
e
l
e
c
t
i
o
n
p
r
o
b
l
e
m
s
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
0
,
n
o
.
4
,
p
p
.
3
6
7
2
–
3
6
8
4
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
0
i
4
.
p
p
3
6
7
2
-
3
6
8
4
.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Hyb
ri
d model f
or
mo
vi
e
reco
mm
e
ndatio
n sy
ste
m us
in
g
c
onte
nt
…
(
Da
y
al
Kumar B
eher
a)
451
[13]
S.
T.
Ahm
ed,
R
.
Al
-
Ham
dani
,
an
d
M.
S.
Croock
,
“
Enha
nce
m
ent
o
f
student
per
for
m
anc
e
pr
edi
c
ti
o
n
using
m
odifi
e
d
K
-
nea
rest
ne
igh
bor,
”
TEL
KOMNIKA
Tele
com
municat
ion
,
Computing,
Elec
tronic
s
and
Control
,
vol.
18
,
no.
4,
pp.
1777
–
1783
,
2020,
doi
:
10
.
12
928/T
EL
KO
MN
IKA
.
V18I4.13849.
[14]
D.
K.
Behe
r
a,
M
.
Das,
S.
Sw
et
an
isha,
and
B
.
Nai
k,
“
Coll
abor
a
ti
v
e
fil
t
eri
ng
using
restr
icted
bol
tz
m
ann
m
ac
hine
and
fuz
z
y
C
-
m
ea
ns,”
in
Adv
an
ces
in
Inte
ll
ig
en
t
Sys
te
ms
and
Computing
,
2018,
vol.
71
0,
pp.
723
–
731,
doi:
10
.
1007/97
8
-
981
-
10
-
7871
-
2_69.
[15]
R
.
S
a
l
a
k
h
u
t
d
i
n
o
v
,
A
.
M
n
i
h
,
a
n
d
G
.
H
i
n
t
o
n
,
“
R
e
s
t
r
i
c
t
e
d
B
o
l
t
z
m
a
n
n
M
a
c
h
i
n
e
s
f
o
r
C
o
l
l
a
b
o
r
a
t
i
v
e
F
i
l
t
e
r
i
n
g
,
”
i
n
P
r
o
c
e
e
d
i
n
g
s
o
f
t
h
e
2
4
t
h
I
n
t
e
r
n
a
t
i
o
n
a
l
C
o
n
f
e
r
-
e
n
c
e
o
n
M
a
c
h
i
n
e
L
e
a
r
n
i
n
g
,
2
0
0
7
,
p
p
.
7
9
1
–
7
9
8
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
n
o
n
r
w
a
.
2
0
0
8
.
0
7
.
0
0
8
.
[16]
J.
Li
u
,
X.
Song,
Z.
Chen,
and
J.
Ma,
“
Neura
l
fa
shion
expe
r
ts:
I
know
how
to
m
ake
the
comple
m
ent
ar
y
c
lot
hin
g
m
at
chi
ng,
”
N
euroc
omputing
,
vol
.
359,
pp.
249
–
26
3,
2019
,
doi
:
10
.
1016/j
.
n
euc
o
m
.
2
019.
05.
081
.
[17]
H.
M.
Ariz
a,
H.
H.
Martí
n
ez,
a
nd
L.
A
.
Gavir
i
a
Roa
,
“
Rec
ogn
it
ion
s
y
stem
for
fac
i
al
expr
essio
n
b
y
proc
essing
images
with
de
e
p
le
arn
ing
neur
a
l
net
work,
”
Tel
k
omnika
(
Tele
co
mm
unic
ati
on
Co
mput.
Elec
tron.
Control.
,
vo
l.
17
,
no.
6
,
pp
.
2975
–
29
82,
2019
,
doi
: 10.12928/
T
EL
K
OM
NIK
A.v17i
6.
12948.
[18]
Q.
A.
Abed,
O.
M.
Fadhil
,
and
W
.
L.
Al
-
Yasee
n,
“
Data
m
ini
n
g
in
web
per
sonaliza
t
ion
using
the
ble
nded
de
ep
le
arn
ing
m
odel
,
”
Indone
sian
J
ournal
of
El
e
c
tric
al
Eng
ineeri
ng
and
Computer
Scienc
e
,
v
ol.
20,
no
.
3,
pp.
1507
-
1512
,
2020,
doi
:
10
.
11
591/i
jeec
s.v20
.
i
3.
pp1507
-
1512.
[19]
H.
As
il
and
J.
Baghe
rz
ade
h
,
“
A
new
appr
oa
c
hto
image
cl
ass
ifi
c
at
ion
b
ase
d
on
ade
ep
m
ult
iclass
AdaBoostin
g
ense
m
ble
,
”
Inter
nati
onal
Journ
al
of
Elec
tric
al
&
C
omputer
Engi
nee
ring
,
vol
.
10,
no.
5
,
pp.
4
872
–
4880,
2020
,
doi:
10
.
11591/ij
ec
e
.
v10i5
.
pp487
2
-
4880.
[20]
C.
Li
and
J.
Li
,
"Colla
bora
t
ive
fil
tering
base
d
on
dual
condi
tional
restr
i
ct
ed
Bolt
zmann
m
achine
s,"
2017
36th
Chine
se
Con
trol
Confe
ren
ce (
CCC)
,
2017,
pp
.
10
871
-
10874,
d
oi
: 10.23919/
ChiCC
.
2017.
8029090
.
[21]
A.
Fis
che
r
and
C.
Ige
l
,
“
Trainin
g
restr
icted
Bol
t
zmann
m
ac
hine
s
:
An
int
roduc
ti
o
n,
”
Pa
tt
ern
Re
c
ognit
ion
,
vol.
47
,
no.
1
,
pp
.
25
–
39
,
2014,
doi: 10.
10
16/j
.
p
at
cog
.
2013
.
05.
025
.
[22]
L.
Ning,
R.
Pit
tman,
and
X.
Shen,
“
LCD:
A
Fast
Contra
stive
Diver
gence
Based
Algorit
h
m
for
Restri
ct
e
d
Bolt
zmann
Ma
c
hine
,
”
N
eural
Ne
tworks
,
vol
.
108
,
pp.
399
–
410,
20
18,
doi
:
10
.
1016
/j
.
n
eune
t
.
2018.
0
8.
018.
[23]
A
.
P
u
j
a
h
a
r
i
a
n
d
D
.
S
.
S
i
s
o
d
i
a
,
“
M
o
d
e
l
i
n
g
S
i
d
e
I
n
f
o
r
m
a
t
i
o
n
i
n
P
r
e
f
e
r
e
n
c
e
R
e
l
a
t
i
o
n
b
a
s
e
d
R
e
s
t
r
i
c
t
e
d
B
o
l
t
z
m
a
n
n
M
a
c
h
i
n
e
f
o
r
r
e
c
o
m
m
e
n
d
e
r
s
y
s
t
e
m
s
,
”
I
n
f
o
r
m
a
t
i
o
n
S
c
i
e
n
c
e
s
,
v
o
l
.
4
9
0
,
p
p
.
1
2
6
–
1
4
5
,
2
0
1
9
,
d
o
i
:
1
0
.
1
0
1
6
/
j
.
i
n
s
.
2
0
1
9
.
0
3
.
0
6
4
.
[24]
Y.
Kila
ni,
A.
F.
Otoom
,
A.
Alsarha
n,
and
M.
Al
m
aay
ah,
“
A
gene
tic
al
gori
thms
-
base
d
h
y
br
id
recom
m
ende
r
sy
ste
m
of
m
at
ri
x
fac
toriza
t
ion
and
nei
gh
borhood
-
base
d
t
ec
hniqu
es,
”
J
our
nal
of
Computational
Sci
en
ce
,
vol.
28,
pp.
78
–
93,
2018,
doi
:
10
.
10
16/j
.
jo
cs.
2018
.
0
8.
007.
[25]
Z.
M.
Alg
elal,
E.
A.
G.
Aldha
her
,
D.
N.
Abd
ul
-
W
adood,
and
R.
H.
A
.
Al
-
Saghee
r
,
“
Botnet
det
e
ct
i
on
using
ense
m
ble
class
ifi
ers
of
net
work
flow,
”
Int
ernational
Journal
of
El
e
ct
rica
l
and
Computer
Engi
n
ee
ring
,
vo
l.
10
,
no.
3
,
pp
.
2543
–
2550,
2019
,
doi
: 10.11591/
i
je
c
e.
v
10i3.
pp2543
-
25
50.
[26]
A.
C.
Alhadi
,
A.
Dera
m
an,
M.
M.
A.
Jali
l,
W
.
N.
J.
W
.
Yu
ss
of
,
and
R.
Mohem
ad,
“
A
computat
iona
l
an
aly
sis
of
short
sente
nce
s
base
d
on
ense
m
ble
sim
il
ari
t
y
m
odel
,
”
Inter
nati
onal
Journal
of
El
e
ct
ri
cal
and
Computer
Engi
ne
ering
,
vol
.
9
,
no
.
6
,
pp
.
53
86
–
5394,
2019
,
doi:
10
.
11591/ij
ec
e
.
v9i6
.
pp5386
-
5394.
[27]
A.
Alrum
ai
h,
A.
Al
-
Sabbagh,
R.
Alsaba
h,
H.
Kharrufa
,
and
J.
Baldw
in,
“
Senti
m
ent
ana
l
y
sis
of
comm
ent
s
in
social
m
edi
a,
”
Inte
rnat
ional
Journal
o
f
El
e
ct
rica
l
&
C
omputer
Engi
ne
ering
(
20
88
-
8708)
,
vol.
10,
no.
6,
pp.
5917
–
592
2
,
2020,
doi
:
10
.
11
591/i
jece
.
v10i6
.
pp5917
-
5922.
[28]
K
.
A
r
u
n
a
n
d
A
.
S
r
i
n
a
g
e
s
h
,
“
M
u
l
t
i
-
l
i
n
g
u
a
l
T
w
i
t
t
e
r
s
e
n
t
i
m
e
n
t
a
n
a
l
y
s
i
s
u
s
i
n
g
m
a
c
h
i
n
e
l
e
a
r
n
i
n
g
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
i
c
a
l
&
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
0
,
n
o
.
6
,
p
p
.
5
9
9
2
–
6
0
0
0
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
0
i
6
.
p
p
5
9
9
2
-
6
0
0
0
.
[29]
A.
Kum
ar,
J.
M.
Chatterjee
,
and
V.
G
.
Día
z
,
“
A
novel
h
y
b
rid
appr
o
ac
h
o
f
SV
M
combined
with
NLP
an
d
proba
bil
ist
ic
neu
ral
net
work
for
email
phishing,”
Inte
rnational
J
ournal
of
El
ec
tr
ic
al
and
Computer
Engi
ne
ering
,
vol.
10
,
no
.
1
,
pp
.
486
–
493
,
2020
,
doi: 10.
11591
/i
j
ec
e
.
v10i1
.
pp486
-
493.
[30]
P
.
I
.
P
r
i
y
a
d
a
r
s
i
n
i
a
n
d
G
.
A
n
u
r
a
d
h
a
,
“
A
n
o
v
e
l
e
n
s
e
m
b
l
e
m
o
d
e
l
i
n
g
f
o
r
i
n
t
r
u
s
i
o
n
d
e
t
e
c
t
i
o
n
s
y
s
t
e
m
,
”
I
n
t
e
r
n
a
t
i
o
n
a
l
J
o
u
r
n
a
l
o
f
E
l
e
c
t
r
i
c
a
l
a
n
d
C
o
m
p
u
t
e
r
E
n
g
i
n
e
e
r
i
n
g
,
v
o
l
.
1
0
,
n
o
.
2
,
p
p
.
1
9
6
3
–
1
9
7
1
,
2
0
2
0
,
d
o
i
:
1
0
.
1
1
5
9
1
/
i
j
e
c
e
.
v
1
0
i
2
.
p
p
1
9
6
3
-
1
9
7
1
.
[31]
Z.
Ne
il
i
,
M
.
Fe
z
ari
,
and
A.
R
edjati,
“
EL
M
and
K
-
nn
m
ac
hine
le
arn
ing
in
cl
assifi
ca
t
ion
of
Brea
th
sounds
signal
s,”
Inte
rnational
Jo
urnal
of
El
ectri
cal
&
Computer
Engi
nee
ring
(
2088
-
8708)
,
vol.
10,
no.
4,
pp.
3
528
–
3536,
2020,
doi:
10
.
11591/ij
ec
e
.
v10i4
.
pp352
8
-
3536.
[32]
D.
K.
Beh
era,
M.
Das,
and
S.
Sw
et
ani
sha,
“
Predicting
users’
pre
fer
en
ce
s
for
m
ovie
rec
om
m
ende
r
s
y
st
em
using
restr
icted
Bolt
z
m
ann
m
ac
hine
,
”
in
Adv
ance
s
in
Inte
llige
n
t
System
s
and
Co
mputing
,
2019,
vol.
7
11,
pp.
759
–
769,
doi:
10
.
1007/97
8
-
981
-
10
-
8055
-
5_67.
[33]
X.
Bai
,
M.
W
a
ng,
I.
L
ee,
Z.
Y
ang,
X.
Kong,
a
nd
F.
Xia,
“
Sci
ent
ific
pap
er
re
c
om
m
enda
ti
on:
A
surve
y
,
”
IEEE
Ac
c
ess
,
vol
.
7
,
p
p.
9324
–
9339
,
2
019,
doi
:
10
.
110
9/ACCESS
.
2018.
2890388.
[34]
Y.
Koren
and
J.
Sill
,
“
Coll
abor
ative
filteri
ng
on
o
rdina
l
user
fe
edb
ac
k,
”
Tw
ent
y
-
thir
d
int
ernati
onal
joi
nt
con
fe
ren
c
e
on
artif
i
ci
a
l
in
te
l
li
gen
ce
,
pp.
302
2
–
3026,
2013
.
[35]
Q.
W
ang,
B
.
Pe
ng,
X.
Shi
,
T.
S
hang,
and
M.
Shang,
“
DCCR:
Dee
p
Col
la
bora
t
ive
Conjun
ct
iv
e
Rec
om
m
ende
r
f
or
Rat
ing
Predictio
n,
”
I
EEE
A
ccess
,
vol
.
7
,
pp
.
6018
6
–
60198,
2019
,
doi:
10
.
1109/AC
CESS
.
2019.
291
5531.
[36]
A.
A.
Ojugo
an
d
D.
O.
Otakor
e
,
“
Com
puta
ti
ona
l
sol
uti
on
of
n
etw
orks
ver
sus
cl
uster
grouping
f
or
socia
l
n
et
wor
k
cont
a
ct
re
commende
r
s
y
st
em,”
I
nte
rnational
Jou
rnal
of
Information
and
Comm
un
ic
ati
on
Te
chnol
o
gy
,
vol
.
9,
no.
3,
p
p
.
185
-
194
,
20
20,
doi
:
10
.
1159
1/i
jict
.
v9i3.
pp18
5
-
194.
Evaluation Warning : The document was created with Spire.PDF for Python.
IS
S
N
:
2502
-
4752
Ind
on
esi
a
n
J
E
le
c Eng &
Co
m
p
Sci,
Vo
l.
23
, N
o.
1
,
Ju
ly
2021
:
445
-
452
452
BIOGR
AP
HI
ES OF
A
UTH
OR
S
Day
al
Kumar
Beher
a
has
ob
t
ai
ned
B
.
E
.
degr
ee
with
honours
in
Inform
at
ion
Te
chno
log
y
from
Nati
onal
I
nstit
ute
of
Sc
ie
n
ce
and
Te
chno
lo
g
y
,
Berh
ampur,
Odisha
in
the
y
e
ar
2006
and
complet
ed
M.
Tech.
,
from
Coll
e
ge
of
Engi
ne
ering
and
Te
chno
l
og
y
,
Bhubane
s
war,
in
2012.
Curre
ntly
he
is
pursuing
his
Ph.
D.
at
KIIT
Uni
ver
sit
y
,
Odisha.
He
has
b
ee
n
working
as
an
As
st.
Profess
or
in
the
depa
rtment
of
CS
E
at
Sili
c
on
Instit
ute
of
Te
chnol
og
y
,
Bhu
bane
sw
ar.
He
is
havi
ng
th
irt
e
e
n
y
ea
rs
of
t
eachi
ng
expe
ri
ence
a
nd
ca
rr
y
ing
out
rese
arc
h
in
va
rio
us
aspe
c
ts
of
Rec
om
m
ende
r
S
y
stem,
Mac
hin
e
Le
arn
ing,
Dat
a
Scie
nc
e,
IoT
and
Remote
Sensing.
He
is
havi
ng
m
an
y
pu
bli
c
at
ions
in
var
ious
journals
an
d
conf
er
ences
a
nd
guide
d
m
an
y
B.
T
ec
h
.
an
d
M.T
ec
h
.
project
s
in
his
are
a
of
int
ere
sts.
He
is
a
li
fe
m
ember
of
India
n
Societ
y
f
or
Te
chnica
l
Educ
a
ti
on
(ISTE)
and
IAENG
.
was
the
Edi
tor
-
in
-
Chie
f
of
the
IEE
E
Tr
ansa
ctio
ns
on
Pow
er
El
e
ct
roni
cs
from
2006
to
2012.
Madhab
ana
nd
a
Das
has
bee
n
working
as
a
Senior
Profess
or
in
School
o
f
Com
pute
r
Engi
ne
eri
ng,
KI
IT
Dee
m
ed
to
be
Univer
sit
y
,
Bh
ubane
sw
ar,
Odis
ha.
He
is
havi
ng
26
y
ea
rs
of
te
a
chi
ng
exp
eri
e
nce
and
13
y
ears
of
industry
e
xper
ie
n
ce.
His
rese
arc
h
in
te
r
ests
enc
om
pass
Com
puta
ti
onal
I
nte
lligen
ce,
Soft
Com
puti
ng,
Artifi
c
ial
Inte
l
li
ge
nce
and
patter
n
rec
ognition.
He
is
havi
ng
a
la
rge
num
ber
of
rese
arc
h
pub
li
c
at
ions
in
var
i
ous
int
ern
ationa
l
conf
er
enc
e
proc
ee
d
ings a
nd
journa
ls
and
gu
i
ded
m
an
y
M.
Tec
h.
and
Ph.D.
Sch
ola
rs i
n
his
areas of
in
te
r
est.
Su
b
hra
Sw
eta
nis
ha
has
bee
n
working
an
A
ss
ista
nt
Professor
with
the
Depa
rtment
of
Com
pute
r
Scie
n
ce
and
Engi
ne
ering
in
Tri
den
t
Ac
ade
m
y
of
T
ec
hn
olog
y
Bhub
ane
s
war,
Odisha,
India
.
She
has
r
ec
e
ive
d
M.T
ec
h
.
degr
e
e
in
Com
pute
r
Scie
nc
e
and
Engi
ne
eri
n
g
from
KII
T
Univer
sit
y
in
2
009
a
nd
compl
et
ed
B
.
E
.
from
Utkal
Univ
ersi
t
y
in
2005
.
Cur
ren
tly
she
is
working
as
a
Ph.D.
schol
ar
in
K
IIT
Univer
sit
y
,
India
.
Her
cur
re
nt
rese
ar
ch
inter
ests
inc
lud
e
Mac
hine
Learni
ng,
Dat
a
Scie
n
ce
and
Remote
Sensing.
She
is
havi
ng
fourt
ee
n
y
e
ars
of
profe
ss
iona
l
e
x
per
ie
n
ce
and
p
ubli
shed
m
an
y
rese
arc
h
pap
e
rs
in
var
ious
journa
ls
and
conf
ere
n
ce
s.
Sh
e
is
a
li
fe
m
em
ber
of
India
n
Societ
y
for
T
ec
h
nic
a
l
Educ
a
ti
on
(ISTE)
and
IAENG
.
P.
K
.
Sethy
cu
rr
ent
l
y
working
as
As
sistant
Profess
or
in
Depa
rtment
of
El
e
ct
roni
cs
,
Sam
bal
pur
Univer
sit
y
sinc
e
2013.
He
has
8
y
e
ars
of
t
ea
ch
in
g,
rese
arc
h
&
ad
m
ini
strat
ive
exp
eri
en
ce
and
4
y
e
ars
of
Industr
y
expe
r
ie
nc
e.
Previousl
y
he
worked
as
Eng
in
ee
r
in
Doordars
han,
P
rasha
r
Bhara
t
i
since
20
09
to
2013.
He
has
rec
ei
v
ed
his
Ph.D.
and
M.
Te
ch
degr
ee
fro
m
Sa
m
bal
pur
Univer
sit
y
and
IIT
(ISM
)
Dhanba
d
respe
c
ti
ve
l
y
.
His
r
ese
arc
h
are
a
is
image
proc
essing,
m
ac
hine
l
ea
rn
in
g
and
dee
p
lear
ning.
He
has
pu
bli
shed
60
rese
a
rch
pape
r
in
d
if
fer
ent
r
epute
journa
l
and
conf
ere
nc
e.
In
add
it
i
on,
he
has
two
pat
en
ts.
He
is
al
so
edi
torial
boar
d
m
ember
of
Inte
rna
ti
ona
l
Journal
of
El
e
ct
ri
cal
and
Com
pute
r
Engi
ne
eri
ng.
Re
ce
nt
l
y
h
e
has
rec
ei
v
ed
“
InSc
Young
Achie
ver
Aw
ard
”
for
the
rese
arc
h
p
ape
r
“
Dete
c
ti
on
of
co
r
onavi
rus
(COV
ID
-
19)
base
d
on
Dee
p
Fe
at
ur
e
s
and
Support
V
ec
tor
Mac
hin
e,
o
rga
nized
b
y
Insti
tut
e
of
Schol
ars,
Ministr
y
of
MS
ME,
Governm
ent
of
India.
He
is
Senior
Me
m
ber
of
IEE
E
.
He
is
the
fr
equent
rev
i
ewe
r
of
Journal
of
Com
puta
ti
onal
and
The
ore
ti
c
al
N
anosc
ie
n
ce,
Journal
of
Int
el
l
ig
ent
&
Fuz
z
y
S
y
stems
,
Indon
esia
n
Journal
o
f
El
e
ct
r
ic
a
l
En
gine
er
ing
and
I
nform
at
ic
s,
TELKO
MN
IK
A
Te
l
ec
om
m
unic
ation,
Com
puti
ng
,
El
e
ct
ron
ic
s
an
d
Control,
Karb
al
a
Inte
rn
at
ion
al
Journal
of
Modern
Scie
n
c
e,
Journa
l
of
Am
bie
nt
Intelli
genc
e
and
Hu
m
ani
ze
d
Com
p
uti
ng,
Scie
n
ti
fi
c
Afric
an,
ACM
Tra
nsac
ti
ons
on
Com
puta
ti
ona
l
Biol
og
y
and
B
i
oinformati
cs,
C
urre
nt
Medi
cal
Im
agi
ng,
Journa
l
of
Com
puta
ti
o
nal
and
Th
eor
etical
Nanosci
en
c
e,
Concurr
ent
E
ngine
er
ing:
Resea
rch
and
A
ppli
c
at
ions,
Info
rm
at
ion
S
ec
urity
Journal:
A
Glob
al
Perspec
t
ive
,
J
ourna
l
of
X
-
Ra
y
Sc
ie
nc
e
and
Te
chnol
og
y
,
As
ia
n
Journal
of
Medical
Princi
p
le
s
and
Cli
nical
Pra
ct
i
ce
,
I
EE
E
Acc
ess,
Chaos,
Solit
ons
&
Frac
t
al
s,
Spanish
Journal
of
Agric
ul
t
ura
l
Research,
Com
pute
rs
in
Biol
og
y
and
Medic
in
es
,
Mul
ti
m
edi
a
Tool
s
and
Appli
ca
t
io
ns,
Inte
rn
at
ion
a
l
Journal
of
TROPICA
L
DI
SEAS
E
&
Hea
l
th,
Journal
of
El
e
ct
roni
c
Im
agi
ng,
Int
ern
ation
al
Journal
of
Speec
h
Te
chno
l
og
y
,
Springer
,
AI
and
Societ
y
,
Com
pute
r
and
El
e
ct
ron
ic
s
in
Agric
ult
ur
e,
Journal
of
Food
Proce
ss
E
ngine
er
ing,
Journa
l
of
Superc
om
puti
ng,
B
iomedic
a
l
Signa
l
Proce
ss
ing
&
Control
,
Journ
al
of
Applie
d
Re
sea
rch
on
M
edicinal
and
Arom
at
i
c
Plant
s
,
Scie
nti
f
ic
Repor
t,
In
te
rn
at
ion
al
J
ourna
l
of
S
y
st
e
m
As
suranc
e
E
ngine
er
ing
and
Mana
gement,
W
ire
le
ss
Personal
Com
m
unic
a
t
i
ons.
Evaluation Warning : The document was created with Spire.PDF for Python.