Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
12
,
No.
3
,
Decem
ber
201
8
, p
p.
1
222
~
1229
IS
S
N: 25
02
-
4752, DO
I:
10
.11
591/ijeecs
.v1
2
.i
3
.pp
1222
-
1229
1222
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
An Eff
ective
a
nd
T
rustabl
e Spati
al S
ervi
ce
Recomm
endati
on
Algorith
m
f
or Spatial
Qu
ery Retr
ieval
i
n
Geo
-
So
cial N
etwork
K.
L
ak
shm
ai
ah
1
,
S.
Mur
ali Kri
shn
a
2
,
B.
Esw
ara
Red
d
y
3
1
awa
har
l
al Nehr
u
Technol
ogi
cal U
nive
rsit
y
,
H
y
d
era
bad
,
Ind
ia
2
S
V
Coll
eg
e
of
Engi
ne
eri
ng,
Ka
rak
ambadi
Ro
ad
,
T
irupati, India
3
Dept.
of
Com
pute
r
Sc
ie
nc
e & En
gine
er
ing, JNTUA
Coll
eg
e
of
En
gi
nee
r
ing, Kal
ik
i
ri
–
517234,
Ch
it
t
oor
(Dt), Indi
a
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
J
ul
29
, 2
018
Re
vised
Sep
1
0,
2018
Accepte
d
S
e
p
2
4
, 201
8
The
spatial
in
form
at
ion
(e.
g
.
,
resta
ura
n
ts/hot
el
s)
is
rel
a
te
d
with
th
e
ke
y
word(s
)
to
in
dic
a
te
th
ei
r
busi
nesses,
services
and
fe
at
ure
s.
Th
e
m
ai
n
issue
of
releva
nt
infor
m
at
ion
retrie
v
al
is
to
quer
y
an
en
ti
t
y
whi
ch
in
cl
u
des
a
set
of
spati
al que
r
y
k
e
y
words
and
h
ave t
he
sm
al
l
est
am
ount
of
in
te
r
-
obj
ec
t
d
ista
n
ce.
The
spa
ti
a
l
qu
e
rie
s
with
ke
y
w
ords
have
no
t
bee
n
ext
ensiv
ely
expl
or
ed.
Stil
l,
the
tra
di
tional
m
et
hod
w
as
foc
used
on
the
m
ult
idi
m
ens
iona
l
d
at
a
.
Previous
works
m
ostl
y
t
arg
ete
d
to
pre
di
ct
th
e
top
-
k
Ne
are
st
N
ei
ghbors
ke
y
word
quer
y
,
where
eve
r
y
k
e
y
word
should
be
equi
va
le
nt
to
the
whole
quer
y
ing
k
e
y
wo
rds.
How
eve
r,
t
he
m
ec
han
ism
does
not
conside
r
the
dens
i
t
y
of
dat
a
entities
in
the
spatial
spac
e
.
To
ov
erc
om
e
th
e
ab
ove
issues
,
An
Eff
ec
t
ive
a
nd
Trusta
ble
S
pat
i
al
S
erv
i
ce
Rec
om
m
enda
ti
o
n
(ET
SS
R)
al
gorit
hm
foc
us
es
on
the
m
ost
releva
nt
infor
m
at
ion
ret
r
ie
va
l
with
the
enha
nc
ed
a
cc
ur
acy
and
m
ini
m
al
re
trieva
l
ti
m
e
for
spat
ia
l
i
nform
at
ion
servic
es.
The
m
ai
n
goal
of
work
is
to
provide
best
spa
ti
a
l
informati
o
n
ret
ri
eva
l
with
a
n
accura
t
ene
ss
of
loc
a
ti
on
pr
edi
c
tion
and
m
ini
m
al
informati
on
ret
ri
eva
l
ti
m
e
.
The
s
y
stem
m
ini
m
iz
es
th
e
class
ifi
c
at
ion
issue
and
visual
izat
ion
pr
oble
m
for
spa
ti
a
l
informat
ion
in
Geospat
i
al
Soci
al
n
et
work
.
The
s
y
s
te
m
imp
rove
s
the
spatial
informati
on
retrie
va
l
with
an
a
cc
ura
c
y
of
loc
a
ti
on
pr
ediction
and
m
ini
m
izes
the
informat
i
on
ret
r
ie
va
l
ti
m
e
compar
e
tha
n
ex
isti
ng
m
et
hods.
Base
d
on
Expe
r
imental
esti
m
a
ti
o
ns,
proposed
ET
SS
R+KNN
enha
nc
ed
0.
48
P
(Prec
ision)
and
0.
49
R
(Rec
a
ll
)
an
d
m
ini
m
iz
ed
28
m
il
li
se
conds
quer
y
ret
r
ie
v
al
ti
m
e.
Ke
yw
or
ds:
Eff
ect
ive
and
t
ru
sta
ble
s
patia
l
s
erv
ic
e
r
ec
omm
end
at
ion
f
ram
ewo
r
k
Geo
-
S
ocial
n
et
work
Inform
at
ion
r
et
rieval
Pr
eci
sio
n
Q
ue
ry r
et
rieval
tim
e
Qu
e
ry Ret
rieva
l Tim
e (Q
RT)
Re
cal
l
Sp
at
ia
l
q
uer
y
Copyright
©
201
8
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
:
Laks
hm
ai
ah
K,
Re
search
Sc
hola
r,
Jawa
har
la
l
Ne
hru
Tec
hnol
og
i
cal
U
ni
ver
sit
y,
Hyde
rab
a
d, I
ndia
.
Em
a
il
: l
aksh
m
ai
ah.
jntu
@
gm
a
il
.co
m
1.
INTROD
U
CTION
The
s
patia
l
dat
abases
(e.
g.
,
r
est
aur
a
nts/h
otels)
are
as
so
ci
at
ed
wit
h
the
ke
yword
(s)
t
o
in
dicat
e
their
bu
si
nesses/se
r
vices/
featu
res.
An
i
nteresti
ng
issue
kn
own
as
Keyw
ord
relat
ed
in
form
at
ion
retrieval
is
to
qu
e
ry
an
ob
j
ect
whic
h
co
ntains
a
set
of
que
ry
keyw
ords
and
ha
ve
the
m
ini
m
u
m
inter
-
ob
j
ect
s
dis
ta
nce.
Sp
at
ia
l
queries
with
key
words
ha
ve
not
bee
n
e
xtensi
vely
exp
l
or
e
d.
Sti
ll
,
the
m
et
ho
ds
wer
e
f
oc
us
ed
on
the
m
ul
ti
di
m
ension
al
data
.
E
xisti
ng
w
orks
m
ai
nly
ta
rg
et
e
d
to
fin
d
t
op
-
k
Ne
arest
Nei
ghbor
s
query,
w
he
re
e
ach
node
s
hould
be
equ
ivale
nt
to
the
whole
queryi
ng
keyw
ords.
H
owev
er,
t
he
te
chn
i
que
does
not
co
ns
id
er
the
densi
ty
o
f data
obj
ect
s i
n
the
s
patia
l space.
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
An
Eff
ect
iv
e an
d
Tr
us
t
ab
le
Sp
atial Se
rvi
ce R
ecomme
nda
ti
on Al
go
rit
hm f
or S
pa
ti
al
…
(
K. La
ks
hmai
ah
)
1223
1
.
1.
Pr
ob
le
m
Ba
sed
on
the
li
te
ratur
e
stud
y,
the
resea
r
ch
w
ork
obse
rv
e
d
that
the
increasin
g
ava
il
abili
ty
and
i
m
po
rtance
of
keyword
rati
ng
in
s
patia
l
qu
e
ry
evaluati
on
f
or
the
relevan
t
s
patia
l
i
nfor
m
at
ion
retrieval
.
The
m
et
ho
d
ca
n
facil
it
at
e
people
tra
vel
no
t
only
nea
r
a
reas
bu
t
al
so
in
ci
ty
that
is
ne
w
for
us
e
rs.
F
or
inst
ance
,
the
visit
or
ar
ra
ng
em
ent
to
vis
it
a
ci
ty
req
uir
es
to
s
hoppin
g,
di
ning
an
d
a
ccom
m
od
at
ion.
It
is
po
pu
la
r
of
al
l
these
need
s
c
an
be
sat
isfie
d
with
best
i
nfor
m
at
ion
ac
cur
acy
a
nd
m
ini
m
al
tim
e
without
lo
n
g
distance
travell
ing.
C
urren
t
syst
e
m
util
iz
es
the
m
ultid
im
ension
al
in
dex
e
s
for
t
op
-
k
query
retriev
al
.
I
niti
al
ly
,
it
sp
li
ts
the
high
dim
ension
al
sp
ac
es
and
in
volves
a
n
arb
it
ra
ry
set
of
us
er
-
sp
eci
fied
at
tribu
te
s.
Howev
e
r,
the
m
eth
od
is
no
t
flexible
for
al
l
possible
a
tt
ribu
te
com
bin
at
ions.
Her
e
,
there
on
e
m
ore
chall
en
geab
l
e
ta
sk
s
are
to
receive
diff
e
re
nt
ty
pe’s
ra
nk
i
ngs
for
com
bin
ed
at
trib
utes.
H
ow
e
ver,
the
m
et
ho
d
does
no
t
as
su
re
the
accu
r
acy
of
releva
nt in
for
m
at
ion
r
a
nk
i
ng. T
hese m
et
ho
ds ha
ve
l
ow ef
fici
ency
f
or
t
he
i
ncr
em
ental
q
ue
ry.
1
.
2
.
B
ackgro
und
In
[
1]
f
oc
us
ed
on
la
r
ge
-
scal
e
reco
m
m
end
er
f
ram
ewo
rks
whic
h
w
ere
ta
ke
n
be
nef
it
of
the
featur
e
s
of
the
unde
rly
ing
in
the
so
ci
al
netw
ork.
T
he
m
et
ho
d
f
oc
us
e
d
on
the
var
ie
t
y
and
unpredi
ct
abili
ty
of
the
so
ci
al
connect
io
ns
.
T
he
te
ch
nique
ta
ckled
t
he
iss
ue
s
of
data
siz
e
an
d
c
omm
un
ic
at
ion
s
pee
ds
i
n
s
ocial
grap
hs
an
d
te
ste
d
the
scal
abili
ty
of
conv
entional
rec
omm
end
er
f
ram
e
works.
I
n
[
2]
descr
ibe
d
a
syst
e
m
at
ic
m
echan
is
m
to
identify
possi
ble
sp
at
iotem
poral
prototyp
es
of
ta
sk
s
by
determ
ining
the
chall
enges
throu
gh
num
ero
us
interco
nnect
ed
m
e
tho
ds
:
util
iz
ing
kernel
de
ns
it
y
evaluati
on
f
or
sm
oo
the
d
s
ocial
m
edia
intensit
y
surf
aces;
us
in
g
eve
nt
-
unconnecte
d
s
oci
al
m
edia
po
sts
to
su
pp
or
t
m
a
p
relat
ion
ta
s
k
occurre
nce,
a
nd
regulariz
in
g
t
ask
ind
ic
at
ors
depends
on
histo
r
ic
al
var
ia
ti
on.
In
[
3]
intr
oduced
a
gr
a
ph
inv
est
igati
on
ba
sed
strat
e
gy
wh
ic
h
stud
ie
d
so
ci
al
netw
orks
with
geog
raphic
da
ta
and
dif
fere
ntiat
ed
ge
ogr
aph
ic
distance
influ
e
nces
i
n
so
ci
al
structu
re.
I
n
[
4]
ex
plaine
d
t
wo
pr
oto
c
ols
f
or
offe
rin
g
c
om
ple
te
confide
ntial
it
y
to
concern
the
SP
(
Ser
vice
Pr
ovi
der),
a
nd
conve
nient
c
onfi
den
ti
al
it
y
with
resp
e
ct
ive
cl
ie
nts.
I
n
[
5]
analy
zed
the
ge
o
-
s
ocial
co
rr
e
la
ti
on
s
a
m
on
g
LB
SN
(Lo
cat
io
n
Ba
sed
Se
rv
ic
e
N
et
work)
cl
ie
nts
at
the
ta
sk
le
vel.
The
unif
i
ed
influ
e
nce
m
et
ric,
nam
ely
glo
bal
it
erati
on
(GI
)
an
d
dynam
i
c
neig
hborh
oo
d
e
xp
a
ns
io
n
(
DN
E
)
w
orked
to
est
im
a
te
cl
ie
nt
influ
e
nce
with
ti
gh
t
hypothe
ti
cal
issue
bounds
i
n
capt
uri
ng
the
geo
-
s
ocial
cl
os
ene
s
s
of
LBSN
cl
ie
nts.
In
[
6]
address
ed
the
so
ci
o
-
s
patia
l
g
raph
wh
ic
h
de
pend
s
on
li
fe
prot
otypes,
wh
e
re
cl
ie
nts
at
ta
ch
ed
to
geog
raphical
obj
ect
s
util
iz
ing
li
fe
-
patte
r
n
e
dg
e
s.
T
he
stra
te
gy
consi
dere
d
tw
o
im
ple
m
entat
ion
s
of
s
ocio
-
sp
at
ia
l gr
a
ph st
or
a
ge. The
im
ple
m
entat
ion
u
ti
li
zed a
relat
io
na
l database
for
a v
a
riet
y
of
qu
eries an
d datas
et
.
In
[
7]
stud
ie
d
t
he
com
plexity
of
Ge
oSN
cl
ie
nts
to
m
anag
e
the
con
tri
bute
d
inform
ation
.
The
strat
egy
addresse
d
two
con
fi
den
ti
al
it
y
hazard
s
tha
t
occu
r
red
in
Geo
S
Ns:
loc
at
ion
co
nf
i
dent
ia
li
t
y
and
absen
c
e
confide
ntial
it
y.
In
[
8]
assu
re
d
f
or
fin
di
ng
the
Ci
rcle
of
Fr
ie
nds
(CoF)
siz
e
k
wer
e
an
N
P
-
hard
is
su
e
.
The
m
echan
ism
assur
ed
t
o
offe
r
a
gro
up
of
fr
ie
nds
wit
h
di
a
m
et
er.
In
[
9]
evaluated
m
any
kinds
of
f
unct
ions
for
descr
i
bing
the
pr
ic
e
a
nd
nu
m
erous
m
eth
ods
so
l
ved
th
e
CoS
K
Q
iss
ue
.
T
he
i
ncor
po
rated
pri
ce
f
un
ct
ion
con
ta
ine
d
al
l
existi
ng
p
rice
f
unct
ions
as
sp
ec
ia
l
cases,
and
t
he
unifie
d
m
eth
od
re
so
l
ved
t
he
CoS
K
Q
issue
wit
h
the
unifie
d
pr
i
ce
functi
on.
I
n
[10]
intr
oduce
d
S
ocial
-
awa
re
top
-
k
S
patia
l
Keyw
ord
(
SkS
K)
query
an
d
s
patia
l
keyw
ord
quer
y.
S
ocial
relevan
ce
feat
ur
e
util
iz
ed
to
e
nh
a
nce
t
he
se
m
antic
s
of
t
he
co
nv
e
ntio
nal
sp
at
ia
l
keyw
ord
que
ry.
In
[11]
ex
pr
e
ssed
c
orrect
an
d
est
i
m
at
ed
sol
ution
s
wh
e
n
t
he
num
ber
κ
of
query
key
word
s
is
sm
a
ll
.
Ho
we
ve
r,
w
he
n
κ
inc
r
eases,
it
becom
es
costly
.
In
[12]
desig
ne
d
a
Sp
at
ia
l
-
awa
r
e
In
te
re
st
Gro
up
(SI
G)
qu
e
ry
that
retri
eved
a
cl
ie
nt
gro
up
of
siz
e
k
wh
e
re
ev
ery
cl
ie
nt
interest
ed
in
the
query
ke
ywords
.
It
cl
ose
d
to
the Eu
cl
idea
n
sp
ace. I
n
[
13
]
su
r
veyed s
om
e
kinds of g
eo
-
te
xtu
al
ind
e
xes
. I
t diff
e
ren
ti
at
ed
the sp
at
ia
l
-
ke
ywo
r
d
qu
e
ry
pe
rfor
m
ance.
I
n
[
14]
de
sign
e
d
a
S
patia
l
In
ve
rted
Ind
ex
(SII
)
w
hich
enh
a
nce
d
the
pe
rfor
m
ance
of
top
-
k
sp
at
ia
l keyw
ord qu
e
ries. T
he i
nd
e
x
m
app
ed
a set o
f
entit
ie
s in
e
ver
y se
parat
e term
. Th
e entit
ie
s w
ere st
ored
i
n
a
diff
er
ent
m
a
nn
e
r
su
c
h
as
th
e
fili
ng
fr
e
quency
of
the
te
rm
and
retri
eved
in
m
ini
m
iz
ing
the
keywor
d
releva
nce a
nd
sp
at
ia
l pro
xim
i
ty
.
I
n
[15]
ex
pr
es
sed
the
pri
vac
y
pr
otect
io
n
to
gu
a
ran
te
e
the
eff
ect
ive
ness
of
K
NN
qu
e
ry
processi
ng.
The
f
ram
ewo
r
k
m
anag
ed
W
i
se
D
umm
y
Sele
ct
ion
Lo
cat
ion
(
WDS
L)
al
gorithm
to
ens
ur
e
the
locat
io
n
confide
ntial
it
y.
In
[
16]
stud
ie
d
an
ef
fici
ently
pr
oces
sin
g
of
al
wa
ys
m
ov
ing
to
p
-
K
s
patia
l
keyword
(
MkSK)
qu
e
ries
over
s
patia
l
keyw
ord
in
f
or
m
at
ion
.
The
m
echan
ism
gu
aran
te
e
the
a
uthority
of
re
ported
ou
tc
om
e
s
offer
i
ng
the
c
ust
om
er
sta
ys
w
it
hin
a
reg
i
on.
In
[
17
]
e
xpla
in
ed
inte
gr
at
io
n
of
A
da
ptive
W
ei
gh
t
Ra
nk
i
ng
Po
li
cy
(AWRP)
with
intel
li
gen
t
cl
assifi
ers
(
NB
-
A
W
RP
-
DA
a
nd
J
48
-
A
WRP
-
DA)
via
dyna
m
ic
aging
fac
tor
t
o
i
m
pr
ove
cl
assifi
ers
po
wer
of
predict
io
n.
The
m
et
ho
ds
are
us
e
d
to
c
hoos
e
the
bes
t
su
bs
et
of
fe
at
ur
es
.
In
[
18]
introduc
ed
a
new
fr
a
m
ewo
r
k
cal
le
d
Fu
zzy
base
d
con
te
xtu
al
rec
om
m
end
at
ion
s
yst
e
m
fo
r
cl
assifi
cat
ion
of
c
us
t
om
er
rev
ie
ws
.
It
e
xtr
act
s
the
in
for
m
at
ion
from
t
he
re
vie
ws
ba
sed
on
the
c
onte
xt
giv
e
n
by
us
e
rs
.
In
[
19
]
stu
die
d
to
i
de
ntify
the
best
cl
assif
ie
rs
f
or
cl
ass
i
m
balanced
he
al
th
dataset
s
t
hro
ugh
a
c
os
t
-
base
d
com
par
i
so
n
of
cl
assifi
er
perf
orm
ance.
The
uneq
ual
m
isclassific
at
ion
costs
wer
e
re
pr
es
en
te
d
in
a
cost
m
at
rix,
and
c
os
t
-
be
nefi
t.
In
[
20]
des
cribe
d
a
locat
ion
base
d
a
nd
pr
e
fer
e
nce
-
aw
are
rec
omm
en
der
syst
em
to
offer
s
a
sp
eci
fic
us
er
a
set
o
f
s
patia
l i
nf
orm
ation
li
ke
r
est
au
ran
ts a
nd
shoppi
ng m
all
s in
ce
rtai
n ge
ospat
ia
l range
.
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
2
2
2
–
1
2
2
9
1224
1
.
3
.
Ob
jecti
ve
s:
The pa
per
obje
ct
ives are f
ollo
wing as:
a)
To
desig
n
a
n
E
ff
ect
ive
a
nd
T
r
us
ta
ble
S
patia
l
Ser
vice
Re
co
m
m
end
at
ion
al
gorithm
fo
r
ret
rievin
g
th
e
m
os
t
releva
nt sp
at
ia
l
infor
m
at
ion
wi
th the
best acc
ur
acy
a
nd m
ini
m
al
tim
e in G
e
o
-
S
ocial
Netw
ork.
b)
To
ap
ply
sim
ilarity
fu
nctio
n
t
o
dete
rm
ine
si
m
il
ar
us
ers
wit
h
their
inte
rest
in
sp
at
ia
l
inf
orm
ation
retrie
val
for of
fer
i
ng b
e
st spati
al
info
r
m
at
ion
r
et
rie
va
l wit
h
a
n
acc
uracy
o
f
locati
on
predict
io
n.
c)
To
reduce t
he pre
dicti
on issu
es to
visu
al
iz
e
the r
el
e
van
t i
nfor
m
at
ion
in Ge
os
pa
ti
al
So
ci
al
netw
ork.
d)
To
im
pr
ov
e
th
e
sp
at
ia
l
info
r
m
at
ion
retriev
al
with
an
accuracy
of
locat
ion
pr
e
dicti
on
and
m
ini
m
iz
e
the
inf
or
m
at
ion
r
et
rieval ti
m
e com
par
e than
e
xi
sti
ng
m
et
ho
ds.
2.
METHO
D
The
desig
n
de
s
cribes
the
c
om
plete
arch
it
ect
ur
e
for
s
patia
l
qu
e
ry
releva
nt
inf
or
m
at
ion
retrieval
with
the
best
accura
cy
and
m
ini
m
a
l
retrieval
tim
e
in
Geo
-
s
ocial
Netw
ork.
Fig
ure
1
ex
plo
re
s
the
processi
ng
ste
ps
of
propose
d
w
ork
in
detai
ls.
Her
e
,
the
desi
gn
e
x
pl
or
es
t
he
pr
e
-
proces
sing
ste
ps
,
in
volved
GSN
(
Geo
-
so
ci
al
Netw
ork)
aut
hority
an
d
al
gorithm
detai
ls
with
ps
eu
do
c
od
e
.
T
he
syst
e
m
a
lso
w
orks
to
pre
dict
the
qu
e
ry
locat
ion
base
d
cl
os
ene
ss
of
us
ers
.
T
he
te
c
hn
i
qu
e
s
de
dic
at
ed
to
offer
t
ru
sta
ble
an
d
r
el
ia
ble
sp
at
ia
l
query
serv
ic
e
i
n
G
S
N.
T
o
offer
t
he
best
s
olu
ti
on
for
a
bove
pro
blem
,
an
Eff
ec
ti
ve
an
d
Tr
us
t
able
S
patia
l
Serv
ic
e
Re
com
m
end
at
ion
al
gorithm
i
ntr
oduce
d
to
offe
r
m
os
t
rele
van
t
in
form
at
i
on
retrieval
wi
th
the
best
acc
ur
acy
and
retrie
val
m
ini
m
al
tim
e
fo
r
sp
at
ia
l
in
form
at
ion
se
rv
ic
es
.
The
m
ai
n
goal
of
a
re
searc
h
st
ud
y
is
to
offer
best
sp
at
ia
l
inf
or
m
at
ion
ret
rieval
with
a
n
acc
ur
a
cy
of
l
ocati
on
pr
e
dicti
on
an
d
m
ini
m
al
infor
m
at
ion
retriev
a
l
tim
e
.
The
te
c
hn
i
que
al
so
works
to
r
et
rieve
m
os
t
re
le
van
ce
inf
orm
at
ion
,
base
d
on
the
us
e
r
qu
e
ry
,
not
t
he
popul
arit
y
of
in
form
at
ion
or
searc
h
en
gi
ne
op
ti
m
iz
ati
on
(
SE
O)
ba
se
d
ser
vices.
Th
e
te
chn
iq
ue
al
so
w
orks
to
ra
nk
th
e
sp
at
ia
l
inform
a
ti
on
base
d
on
qu
e
ry
dem
and
.
The
syst
e
m
util
iz
es
a
'
si
m
ilarity
fu
nctio
n'
to
determ
ine
si
m
i
la
r
us
ers
with
the
ir
interest
in
s
patia
l
inf
or
m
ation
retrieval.
The
obj
ect
s
c
onnected
with
the
pri
nci
pal
qu
e
ry
keyw
ord
ob
j
ec
ts
are
to
retrieve
the
releva
nt
inform
at
ion
.
The
syst
em
red
uce
s
the
cl
assifi
cat
ion
pro
bl
e
m
to
visu
al
iz
e
the
r
el
evan
t
i
nfor
m
at
ion
in
Ge
osp
at
ia
l
So
ci
al
net
work.
T
he
pri
nc
ipal
obje
ct
al
so
helps
to
pro
po
s
ed
fr
am
ewo
r
k
to
pr
e
dict
the
m
os
t
relevan
t
and
highly
r
el
ia
ble
sp
at
ia
l
inf
or
m
at
ion
for
us
e
rs.
The
sp
at
ia
l
inf
or
m
at
ion
se
rv
ic
es
ca
n
fin
d
al
l
resta
ur
a
nt
s
in
a
giv
e
n
area;
w
her
e
ne
arest
nei
ghbo
ur
retrie
val
ca
n
on
ly
disco
ver
the
re
sta
ur
a
nt
cl
os
e
s
t
to
a
giv
e
n
ad
dr
ess
.
T
he
pro
po
s
ed
te
ch
niqu
es
util
iz
ed
t
wo
ty
pes
of
co
nc
ept
f
or
of
rankin
g
se
rvi
ces
li
ke
serv
ic
e
reco
m
m
end
at
ion
an
d
decisi
on
s
uppo
rt.
Th
e
te
chn
iq
ue
predict
s
the
ra
nking
of
sp
at
ia
l
inform
at
ion
base
d
on
us
e
r
locat
io
n.
T
he
pro
pos
ed
m
e
tho
d
util
iz
es
co
m
bin
ed
obj
ect
ra
nking
s
of
at
tribu
te
s
w
hic
h
are
colle
ct
e
d
from
diff
ere
nt
sources
.
Th
e
m
et
ho
d
m
in
i
m
iz
es
m
u
lt
ipl
e
tim
e
inp
ut
r
akin
g
issues t
o pr
e
dict t
he
m
os
t rele
van
t i
nfor
m
at
i
on
.
2
.
1.
User
If
,
a
us
e
r
nee
d
to
retrievin
g
the
inf
or
m
at
ion
wh
ic
h
is
al
read
y
store
d
in
a
centrali
zed
sto
rag
e
ser
ve
r.
First,
us
e
r
sho
uld
re
gister
th
ei
r
detai
ls
into
stora
ge
ser
ve
r.
T
hese
detai
ls
are
m
a
intai
ned
in
a
centr
a
li
zed
Database
. Onc
e u
se
r registe
re
d
thei
r deta
il
s, t
hen
t
hey ca
n
s
earch
their
que
ry in ce
ntrali
ze
d
st
or
a
ge
se
rv
e
r.
2
.
2
.
G
SN (
Ge
o
-
S
oci
al N
e
twork
)
A
uthori
t
y
The
GSN
a
uthority
can
vie
w
the
us
er
detai
ls
an
d
up
l
oad
the
s
patia
l
inf
orm
ation
with
de
scriptio
ns
.
Her
e
,
GSN
au
thorit
y
can
al
so
al
te
r
about
s
patia
l
inform
ation
w
hate
ve
r
changes
a
nd
wh
at
e
ver
in
f
orm
at
ion
he/she
re
quire
d.
It
pro
vi
des
a
chan
ce
f
or
adv
e
rtise
rs
to
ge
t
a
hu
ge
set
of
s
patia
l
inform
ation
,
w
hic
h
co
uld
po
te
ntial
ly
lea
d
to
ide
ntifyi
ng a
us
e
r
inte
res
t.
2
.
3
.
S
patial
Dataset
The
s
patia
l
-
ke
yword
se
arc
h
has
ob
ta
ine
d
s
ub
sta
ntial
at
te
ntion
f
r
om
research
so
ci
et
y.
S
om
e
pr
evi
ous
fr
am
ewo
r
ks
c
on
ce
ntrate
d
for
ret
rievin
g
i
ndivid
ual
e
ntit
ie
s
by
i
den
ti
fyi
ng
a
query
c
om
pr
isi
ng
of
a
query
locat
ion
an
d
a
set
of
qu
e
ry
keyw
ords.
E
ve
ry
retrieve
d
en
ti
ty
is
relat
ed
t
o
keyw
or
ds
re
le
van
t
to
the
sp
at
ia
l
qu
e
ry
keyw
ord
inf
or
m
at
ion
a
nd
it
is
cl
os
est
to
the
spa
ti
al
qu
e
ry.
T
he
sim
il
arity
betwe
en
file
s
is
ap
pl
ie
d
to
com
pu
te
the
relevan
ce
of
two
set
s
of
keyw
ords
.
As
it
is
pr
ob
a
bly
no
in
di
vidual
entit
y
is
relat
ed
to
al
l
qu
e
ry
keyw
ords.
Som
e
add
it
ion
al
works
ai
m
to
retrieve
m
any
entit
ie
s
wh
ic
h
tog
et
he
r
c
ov
e
r
al
l
qu
ery
key
words.
Wh
il
e
pote
ntia
ll
y
a
hu
ge
am
ount
of
entit
y
integrati
ons
ful
fill
the
req
ui
r
e
m
ent,
the
res
earch
iss
ue
is
that
the
retrieve
d
e
ntit
ie
s
m
us
t
hav
e
a
p
op
ular
s
patia
l
data
c
onnecti
on. Th
e
resea
rc
her
put
f
orwa
r
d
the
issue
to ret
rieve
entit
ie
s w
hich cov
e
r
al
l spati
al
q
uer
y keyw
or
ds
, h
a
ve
the s
m
al
le
st
a
m
ou
nt
o
f
inter
-
e
ntit
ie
s d
ist
ance and
closed
to a s
patia
l q
ue
ry locat
io
n.
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
An
Eff
ect
iv
e an
d
Tr
us
t
ab
le
Sp
atial Se
rvi
ce R
ecomme
nda
ti
on Al
go
rit
hm f
or S
pa
ti
al
…
(
K. La
ks
hmai
ah
)
1225
Figure
1.
Wo
r
kfl
ow
diagr
am
of Arc
hitec
ture Dia
gram
o
f
P
r
opos
e
d
Syst
em
2.4.
Ke
ywor
d S
e
arch
Sp
eci
fie
d
a
spa
ti
al
database
ever
y
e
ntit
y
r
el
at
ed
with
on
e
or
m
or
e
s
pa
ti
al
keyword
i
nfor
m
at
ion
.
An
e
ntit
y
wit
h
num
ero
us
s
patia
l
keywor
ds
is
change
d
to
nu
m
ero
us
entit
ie
s
placed
at
the
sa
m
e
l
ocati
on
without
loss
of
gen
e
rali
zat
ion
with
a
sing
le
disti
nct
sp
at
ia
l
keywor
d.
A
n
ETSSR
m
et
ho
d
m
anu
factu
re
s
le
ss
new
can
did
at
e
keyw
ord
co
ve
rs
wh
e
n
proc
essing
a
ca
nd
i
date
key
wor
d
cov
e
r.
The
ET
SSR
al
gorithm
al
so
processe
d
am
ount
of
can
di
date
keyw
ord
cov
e
rs
is
optim
al
the
a
m
ou
nt
of
keyword
c
overs
create
d.
Con
sec
utively
,
it
con
cl
udes
that
the
am
ount
of
keyw
ord
c
ov
e
rs
c
r
eat
ed
in
the
ETSSR
al
go
rithm
.
The
s
umm
ary i
s ind
e
pe
nd
e
nt
of
m
ai
n
qu
e
ry k
ey
w
ord
as the
inv
est
igati
on
do
e
s not ap
ply any pa
ram
et
er
on
t
he
sel
ect
ion
m
et
ho
d m
ai
n
qu
e
ry
keyw
ord.
2.5.
User P
oi
nt
of In
terests
The
o
bject
ive
of
the
i
nterf
a
c
e
is
to
offer
a
po
i
nt
of
i
nter
est
inform
at
io
n
with
,
at
le
ast
,
a
locat
ion
,
so
m
e
co
m
pu
lsory
feat
ur
es
a
nd
opti
onal
des
cripti
ons.
It
is
offer
i
ng
the
s
pa
ti
al
data;
the
el
e
m
ent
that
dev
el
ops
the
interface
util
iz
es
the
m
ap
locat
ion
databa
s
e
inform
at
ion
to
locat
e
and
exh
i
bit
a
po
int
of
interest
(
P
OI)
or
to
sel
ect
PO
I
as
route
way
point.
T
he
el
em
ent
does
not
only
offer
s
patia
l
data
search
functi
onal
it
ie
s
for
th
e
local
data
base
but
al
so
ex
presses,
ho
w
t
o
relat
e
exter
nal
key
wor
d
search
te
ch
nique
to
the
el
e
m
ent.
The
m
et
ho
d
al
so
im
pr
oves
t
he
keyw
ord
se
arch
crit
eria
a
nd
the
li
st
of
extracte
d
outc
om
es.
The
pro
po
s
ed
desig
n
al
so
i
ntrod
uces
a
res
ol
ution
t
o
acq
uir
e
conve
ntion
P
OI
s
or
t
o
m
od
ify
con
te
nt
an
d
exp
la
natio
n
of
local
PO
I
d
y
nam
ic
ally.
2.6.
Ef
f
ecti
ve
a
n
d Trus
table
Spati
al Ser
vi
ce Recom
me
ndation
A
l
gor
ithm
An
Ef
fecti
ve
a
nd
Tr
us
ta
ble
S
patia
l
Ser
vice
Re
com
m
end
at
ion
al
gorithm
intr
oduce
d
to
f
ocus
on
the
m
os
t
relevan
t
inf
or
m
at
ion
retr
ie
val
with
t
he
best
acc
ur
acy
and
m
ini
m
al
retrieval
tim
e
fo
r
sp
at
ia
l
i
nfo
r
m
at
ion
serv
ic
es
.
The
m
echan
ism
wo
rk
s
to
retrieve
m
os
t
relevan
ce
inform
at
ion
,
dep
e
nds
on
th
e
us
er
qu
e
ry,
no
t
th
e
popula
rity
of
inf
or
m
at
ion
or
search
en
gin
e
op
ti
m
iz
ation
(
SEO)
based
s
erv
ic
es.
T
he
f
ram
ewo
rk
util
iz
es
a
'
si
m
il
ari
ty
fu
nc
ti
on
'
to
deter
m
ine
si
m
il
ar
us
ers
with
the
ir
interest
in
sp
at
ia
l
query
i
nfor
m
at
ion
ret
rieval
.
The
entit
ie
s
relat
ed
to
the
m
a
i
n
query
key
word
e
ntit
ie
s
are
t
o
retrie
ve
the
r
el
evan
t
inf
orm
at
ion
.
T
he
pri
nc
ipal
entit
y
al
so
assist
s
to
the
pr
opose
d
f
ram
ewo
r
k
to
com
pu
t
e
the
m
os
t
relev
ant
an
d
highly
reli
able
spa
ti
a
l
inf
or
m
at
ion
for use
rs.
The
s
patia
l
in
form
ation
se
r
vi
ces
can
l
ocat
e
al
l
restau
ra
nt
s,
hos
pital
an
d
bank
in
a
giv
e
n
a
rea;
wh
e
re
near
est
neig
hbor
retrie
val
can
on
ly
di
scov
e
r
the
res
ta
ur
a
nt,
hos
pital
and
bank
cl
os
est
to
a
pro
vi
de
d
address
.
Her
e
,
Geo
-
so
ci
al
Netw
ork
(
GSN)
a
uthority
can
ad
d
an
d
update
the
qu
ery
inform
at
io
n
(
QI)
fr
e
qu
e
ntly
.
G
S
N
ca
n
view
th
e
li
st
of
re
gistered
G
SN
us
e
r.
Use
r
ca
n
e
nt
er
sp
at
ia
l
qu
e
ry
(SQ
)
based
thei
r
requirem
ent.
N
ext
syst
em
pr
edict
s
us
e
r
loca
ti
on
acc
u
racy
(
LA)
t
o
rec
omm
end
the
s
pati
al
qu
e
ry
inf
orm
at
ion
serv
ic
es
(SQI
S)
.
T
he
desig
n
rem
ov
es
the
irreleva
nt
inf
or
m
at
ion
from
colle
ct
ed
inf
or
m
at
ion
sour
ce
(IS
)
.
The
m
echan
is
m
pr
ocess
t
he
qu
e
ry
to
get
th
e
m
os
t
relevant
inf
or
m
at
ion
(
RI)
from
GS
N
.
Pro
po
se
d
te
ch
ni
que
s
us
e
d
two
ty
pes
of
ra
nk
i
ng
f
un
ct
ion
li
ke
ser
vice
reco
m
m
end
at
ion
an
d
decis
ion
s
upport
te
chn
i
qu
e
to o
f
fer
m
os
t
RI.
T
he
propose
d
m
et
ho
do
l
og
y
util
iz
es
integ
rated
e
ntit
y
rankin
gs
of
featu
res
wh
ic
h
are
gat
her
e
d
f
r
om
View us
er
d
etails
Ad
d
/u
p
d
ate
sp
atial inf
o
Au
th
en
ticatio
n
Proces
s
GSN Auth
o
rit
y
Geo
-
so
cial
Netwo
rk
Ap
p
ly
ET
SSR
alg
o
rith
m
Co
llect r
elated
sp
atial inf
o
Key
wo
rd
search p
rocess
Au
th
en
ticatio
n
p
rocess
User
Visu
alize the
m
o
st
relevan
t sp
atial inf
o
Extract the
m
o
st
relevan
t sp
atial inf
o
b
ased
on
User
lo
ca
tio
n
Ap
p
ly
the
rank
in
g
f
u
n
ctio
n
Re
m
o
v
e
irr
elev
an
t
sp
atial inf
o
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
2
2
2
–
1
2
2
9
1226
diff
e
re
nt
sourc
es.
T
he
te
ch
ni
qu
e
re
du
ce
s
m
ul
ti
ple
tim
e
input
ra
nk
i
ng
issues
to
predi
ct
the
m
os
t
relevan
t
inf
or
m
at
ion
.
T
he
ps
e
udo
c
od
e
Ef
fecti
ve
a
nd
Tr
us
ta
ble
Sp
at
ia
l
Serv
ic
e
Re
c
omm
end
at
io
n
(
ETSSR)
Al
gor
it
h
m
is ex
plained
b
e
low
i
n detai
ls:
Inpu
t:
Process
sp
at
ia
l q
uer
y
(SQ)
Out
p
ut:
S
pati
al
infor
m
at
ion
retrieval
(S
IR
) wit
h
a
n
acc
ur
a
cy
o
f
locati
on
pr
e
dicti
on and
m
ini
m
al
inf
orm
at
ion
r
et
rie
va
l t
i
m
e
Proced
ure
St
ar
t
GSN a
uthority
authe
ntica
ti
on
process;
Add
a
nd
updat
e sp
at
ia
l i
nfo
r
m
at
ion
(
S
I) wit
h
la
ti
tud
e
and l
ongitu
de
in
f
orm
at
ion
;
View
s
patia
l u
ser’
s
in
form
ati
on
;
User’s a
uth
e
ntica
ti
on
process;
Detect
the
u
se
r geo
gr
a
phic
al
locat
ion
(
GL)
;
If
locat
io
n
is
det
ect
ed
Apply ETS
SR
S
patia
l Qu
e
ry (SQ
)
se
arch p
ro
ce
ss;
Fi
nd s
pa
ti
al
q
uer
y
rela
te
d
in
form
at
io
n;
I
f
Qu
e
r
y rela
te
d
in
f
orm
at
ion
(
RI
) found
Fil
te
r
the ir
releva
nt in
form
ation
from
co
ll
ect
ed
i
nform
at
ion
;
Extract
the
releva
nt in
form
ation
acc
ordi
ng to use
r q
uer
y;
Apply t
he ran
king
fun
ct
ion
base
d
cl
ose
st l
ocati
on
of u
se
r;
Dis
play
m
os
t rele
va
nt
sp
at
ia
l retrie
ve
d resu
lt
;
El
se
Re
c
omm
end
the
user t
o re
-
wr
it
e the
s
patia
l query;
El
se
Re
-
identify
t
he
g
lo
bal
posit
ion
of u
se
r;
3.
RESU
LT
S
A
ND D
I
SCUS
S
ION
3.1.
Ex
peri
ments Setu
p
The
propose
d
m
echan
ism
is
dep
l
oyed
with
In
te
l
Co
re
i
5
7
th
Gen
e
rati
on
P
ro
ces
sor
with
10
GB
RAM
,
400
GB
Me
m
or
y
a
nd
W
i
nd
ow
s
7
pro
fessi
on
al
operati
ng
syst
e
m
.
The
pro
po
se
d
m
echan
ism
is
devel
op
e
d
i
n
Java
pro
gr
am
m
ing
wit
h
J
DK
(Ja
va
D
evelo
pm
ent
K
it
)
1.8,
Net
be
ans
8.0
I
ntegr
at
e
d
Dev
el
opm
ent
En
vironm
ent,
Ap
ac
he Tom
cat
8
.
0.3,
a
nd M
YSQL
5.5 dat
a
base.
3.1.
1.
D
ata
Fo
r
e
xperim
ental
evaluati
on
s
,
pro
po
se
d
m
e
chan
ism
sel
ect
sp
at
ia
l
do
m
ain
three
dif
fer
e
nt
dataset
s
nam
ely
Ho
s
pital
dataset
with
250
record
s,
R
est
aur
a
nt
datas
e
t
200
rec
ords
and
Ba
nk
data
set
with
25
0
re
cord
s
is coll
ect
e
d wit
h nam
e, locati
on
,
a
nd se
rv
ic
es
d
esc
riptio
ns
de
ta
il
s.
3.2.
Simul
at
io
n R
es
ult
In
t
he
sect
io
n,
an
Ef
fecti
ve
&
T
ru
sta
ble
Sp
at
ia
l
Ser
vic
e
Re
com
m
end
at
ion
(ETS
SR)
al
gorithm
inv
est
igate
s
th
e
m
at
he
m
a
ti
ca
l
struct
ur
e
of
the
rele
va
nt
i
nfor
m
at
ion
retrieval
with
th
e
best
accurac
y
li
ke
pr
eci
sio
n
(
P)
,
recall
(R)
an
d
r
et
rieval
m
ini
m
al
tim
e
fo
r
sp
a
ti
al
inform
at
io
n
ser
vices
in
Geo
-
s
ocial
Network.
Her
e
,
it
il
lustrate
s
pr
eci
sio
n,
r
ecal
l,
and
que
r
y
retrieval
tim
e
(Q
RT
)
to
co
m
pu
te
the
per
f
or
m
ance
of
propose
d
ETSSR
m
et
hod.
It
i
den
ti
fie
d,
how
ca
n
ac
hie
ve
t
he
best
acc
ur
acy
with
m
in
i
m
al
qu
ery
retr
ie
val
tim
e
for
s
patia
l
inf
or
m
at
ion
r
et
rieval in
G
e
o
-
S
ocial
N
et
w
ork
(G
S
N
).
3.
2
.
1.
Preci
sio
n (
P
)
The
preci
sio
n
exp
la
in
the
a
gr
eem
ent
of
a
set
of
retriev
ed
res
ults
a
m
ong
them
sel
ves.
Pr
eci
sio
n
represe
nts
the
dev
ia
ti
on
of
set
of
retrie
ve
d
res
ults
from
the
arit
hm
eti
c
m
ean
of
th
e
set
.
The
pre
ci
sion
con
ce
ntrate
d
on
the
i
den
ti
fic
at
ion
a
nd
el
im
inati
on
of
syst
e
m
at
ic
err
or
s
.
Pr
eci
sio
n
ca
n
be
c
om
pu
te
d
as
in
E
quat
ion (
1).
=
+
(1)
Wh
e
re,
T
P
is t
r
uly p
os
it
ive a
nd F
P
is
a f
al
se
po
sit
ive
.
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
An
Eff
ect
iv
e an
d
Tr
us
t
ab
le
Sp
atial Se
rvi
ce R
ecomme
nda
ti
on Al
go
rit
hm f
or S
pa
ti
al
…
(
K. La
ks
hmai
ah
)
1227
3.
2
.
2.
Rec
all
Re
cal
l
exp
ress
ed
as
the
total
nu
m
ber
of
rel
evan
t
in
form
ation
w
hich
is
extracte
d
based
on
s
ear
c
h
qu
e
ries
a
nd
di
vid
e
d
by
the
t
otal
num
ber
of
avail
a
ble
rel
evan
t
i
nfor
m
ation
.
Re
cal
l
is
the
rati
o
of
th
e
total
a
m
ou
nt
of
rel
evan
t
searc
he
d
inf
or
m
at
ion
to
the
total
nu
m
ber
avail
a
ble
releva
nt
reco
r
ds
in
cent
rali
zed
database
. I
t i
s
descr
i
bed in
E
qu
at
io
n (
2
).
N
P
p
F
T
T
c
a
l
l
Re
(2)
Wh
e
re,
T
P
is t
r
uly p
os
it
ive a
nd F
N
is a
f
al
se
neg
at
ive
.
3.
2
.
3.
Quer
y Retrie
va
l T
im
e (QRT)
The
pr
opos
e
d
ETSSR
m
eth
od
il
lustrate
s
a
m
at
he
m
ati
cal
m
od
el
for
qu
e
ry
retrie
val
tim
e
i
n
E
quat
ion
(
3)
.
The
pro
po
se
d
te
chn
iq
ue
is
c
om
pu
te
d
qu
e
ry
retrieval
ti
m
e
base
d
on
ti
m
e
ta
ken
by
ce
ntrali
ze
d
serv
e
r
t
o proce
ss
the
us
e
r req
uested
query.
Qu
e
ry Ret
rieva
l Tim
e (Q
RT)
is cal
culat
ed
a
s:
(3)
Wh
e
re,
T
CD
is
a total
num
ber
of sp
at
ia
l rec
or
d
a
nd T
AR
is av
erag
e
r
et
rie
val
tim
e
fo
r
a
quer
y processi
ng of
sp
at
ia
l use
r.
Table
1
il
lustr
at
es
the
preci
sion
(P),
recall
(
R)
an
d
Qu
e
ry
Re
trie
val
Tim
e
(Q
RT
)
in
m
i
llisec
onds
for
Ho
s
pital
,
Re
sta
ur
a
nt
an
d
Ba
nk
sp
at
ia
l
dat
aset
.
The
pro
pose
d
ETS
SR
m
et
ho
d
dem
on
strat
es
thei
r
aver
a
ge
values
f
or
res
pecti
ve
c
onstr
ai
nt
with
t
he
r
especti
ve
data
set
.
He
re,
pro
po
s
ed
Ef
fecti
ve
&
Tr
us
ta
ble
Sp
at
ia
l
Ser
vice
Re
co
m
m
end
at
ion
(
ETSSR)
al
gor
it
h
m
is
est
i
mate
d
with
pr
e
vious
m
et
ho
dolo
gies
su
c
h
as
Mos
t
Pr
efe
rr
e
d
Ca
te
gory
base
d
Re
com
m
end
at
io
n
(MPCR
)
[
20]
,
P
ref
e
ren
ce
-
ba
sed
C
ollab
or
a
ti
ve
Fil
te
rin
g
(
PCF)
[20]
and
L
oca
ti
on
-
base
d
Col
la
borati
ve
Fil
te
rin
g
(LCF
)
[
20
]
m
echan
ism
s
wh
ic
h
res
ults
are
disp
la
ye
d
in
Table
1.
Table
s
hows
t
he
pr
eci
si
on
;
re
cal
l
and
query
retrieval
ti
m
e
f
or
H
os
pital
,
B
ank
an
d
Re
sta
ur
a
nt
s
patia
l
dataset
s.
Acc
or
ding
to
ta
ble
1
ou
tc
om
es,
it
no
ti
ced
ETS
SR
perform
ed
well
on
Ho
s
pital
,
Ba
nk
a
nd
Re
sta
ur
a
nt
sp
at
ia
l
dataset
.
Finall
y,
the
arti
cl
e
cl
aims
t
hat
the
pro
pos
ed
ETSSR
is
the
best
ap
proa
ch
f
or
ov
e
rall
sp
at
ia
l
da
ta
set
s.
The
t
echn
i
qu
e
is
int
egr
at
e
d
with
K
NN
(
K
-
Nea
res
t
Neighbor)
cl
assifi
er
to
pred
ic
t
to
retrieve
qu
e
ry
locat
ion
an
d us
er acc
ur
acy
.
Table
1.
Pr
eci
s
ion
(P), Recal
l
(R), an
d Q
uer
y
Ret
rieval Tim
e (
QRT)
i
n
m
illi
seconds fo
r H
os
pital
, Resta
urant
and Ban
k
s
pa
ti
al
d
at
aset
Lear
n
in
g
Alg
o
rithm
s
Ho
sp
ital
Res
tau
rant
Ban
k
P
R
QRT
P
R
QRT
P
R
QRT
MPCR
0
.39
0
.29
190
0
.5
0
.15
250
0
.28
0
.5
40
PCF
0
.43
0
.41
620
0
.6
0
.24
750
0
.35
0
.7
110
LCF
0
.19
0
.13
50
0
.2
0
.1
110
0
.21
0
.15
30
ET
SSR
0
.99
0
.98
42
0
.88
0
.91
50
0
.97
0
.95
15
Figure
2.
Pr
eci
sion f
or Hos
pital
, Resta
ur
a
nt
and Ban
k datas
et
AR
CD
T
T
Q
R
T
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.
12
, N
o.
3
,
Dece
m
ber
2
01
8
:
1
2
2
2
–
1
2
2
9
1228
Figure
3.
Re
cal
l for
Ho
s
pital
, R
est
aur
a
nt and
Ban
k datase
t
Figure
4.
Q
uery
Ret
rieval Tim
e fo
r
Hos
pital
, Resta
ur
a
nt a
nd Ban
k datase
t
Accor
ding
to
Figure
2
to
4
obser
vatio
ns
for
Hospita
l,
Re
sta
ur
a
nt
a
nd
Ba
nk
s
patia
l
dataset
s,
the
propose
d
ETSSR+
KNN
appr
oach
com
pu
te
s
preci
sio
n,
recall
a
nd
query
pr
ocessin
g
tim
e
fo
r
ide
nt
ify
ing
the
ef
fecti
ven
e
ss
of
te
ch
nique
.
The
pro
pose
d
ETSSR+
KNN
is
com
p
uted
with
MPC
R,
P
CF
an
d
LCF
e
xisti
ng
appr
oach
be
ha
lf
of
pr
eci
sio
n,
recall
an
d
query
proce
ssin
g
tim
e.
PCF
is
the
nea
rest
c
om
pet
it
or
on
preci
sio
n
and
recall
co
nst
raints.
T
he
m
et
ho
d
is
util
iz
ed
to
ret
rieve
web
base
d
qu
ery
inf
or
m
at
ion
.
Howe
ver
,
it
fail
s
to
pre
dict
an
d
c
at
egorized
the
re
trie
ved
que
ry
of
m
ulti
ple
ty
pe
inf
or
m
at
ion
from
diff
ere
nt
so
urces
.
LCF
is
the
cl
os
est
m
e
tho
d
beh
al
f
of
on
qu
e
ry
retrie
val
tim
e
(Q
RT)
c
on
st
raint.
LCF
m
et
ho
d
is
use
d
to
ap
ply
ra
nki
ng
functi
on
to
re
trie
ve
releva
nt
resu
lt
first
f
ro
m
m
ult
iple
ty
pe
of
query
inform
ation
database
.
H
oweve
r,
the
te
ch
nique
consum
e
m
or
e
tim
e
fo
r
quer
y
retrieval
with
le
ss
l
ocati
on
predict
io
n
acc
ur
acy
.
The
pro
po
s
e
d
ETSSR+
KNN pr
ovide enha
nc
ed
0.4
8
P (
P
rec
isi
on
)
a
nd
0.4
9 R
(
Re
cal
l) an
d
m
ini
m
iz
ed
28
m
illi
secon
ds
query
retrieval
ti
m
e.
Finall
y,
the
pa
per
an
nounce
s
the
pro
pose
d
ETS
SR+K
NN
ap
proac
h
perf
or
m
s
best
on
ever
y
par
am
et
er &
r
e
sp
ect
ive i
nput
const
raints.
4.
CONCL
US
I
O
N
An
Ef
fecti
ve
a
nd
Tr
us
ta
ble
S
patia
l
Serv
ic
e
Re
com
m
end
at
ion
Fr
am
ewo
r
k
is
pr
ese
nted
to
const
ernat
e
on
the
m
os
t
r
el
evan
t
in
f
or
m
at
ion
ret
rieval
with
the
acc
ur
at
e
ness
a
nd
m
ini
m
al
retrie
val
tim
e
fo
r
s
patia
l
inf
or
m
at
ion
serv
ic
es.
T
he
m
a
j
or
ai
m
of
te
ch
nique
is
to
off
er
best
sp
at
ia
l
inf
or
m
at
ion
retrieval
with
loc
at
ion
pr
e
dicti
on
acc
ur
acy
an
d
m
ini
m
al
inform
at
io
n
retrieval
ti
m
e.
In
Ge
ospat
ia
l
So
ci
al
netwo
r
k,
the
cl
assif
ic
at
ion
and
visu
al
iz
at
ion
iss
ues
are
m
ini
m
iz
ed
for
the
sp
at
ia
l
query
releva
nt
inform
ation
retri
eval.
The
pro
po
s
ed
fr
am
ewo
r
k
is
highly
ded
ic
at
ed
to
offer
m
os
t
relevan
t
i
nfor
m
at
ion
to
us
ers.
T
he
s
patia
l
inform
at
ion
serv
i
ces
can
locat
e
al
l
restau
ran
ts,
ho
sp
it
al
and
ba
nk
relat
ed
in
for
m
at
ion
in
a
giv
en
a
rea;
wh
e
re
nea
rest
neig
hbour
retrieval
ca
n
only
disc
ov
e
r
t
he
hos
pital
an
d
bank
relat
ed
in
form
ation
,
cl
ose
st
to
a
gi
ven
a
ddress.
T
wo k
ind
s
o
f
con
ce
pts
util
iz
ed
to
the
pro
po
s
ed
f
ram
ewo
r
k
su
c
h
as
r
ank
i
ng
se
rv
ic
e
s
li
ke
serv
ic
e
reco
m
m
end
at
ion
a
nd
decisi
on
support.
T
he
te
ch
nique
predict
s
the
ran
ki
ng
of
s
patia
l
inform
ation
de
pends
on
us
e
r
lo
cat
ion
.
The
pro
pose
d
ETSSR+
KNN
pro
vid
e
en
ha
nc
ed
0.4
8
P
(P
r
eci
sion
)
a
nd
0.4
9
R
(Recal
l)
and
m
i
ni
m
iz
e
d
28
m
illi
secon
ds
query
retrieval
t
i
m
e.
Finall
y,
the
pap
e
r
a
nnounce
s
the
pro
pose
d
ETS
SR
a
ppr
oach
pe
rform
s
best
on ev
e
ry
par
a
m
et
er &
r
es
pec
ti
ve
input c
ons
trai
nts.
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
An
Eff
ect
iv
e an
d
Tr
us
t
ab
le
Sp
atial Se
rvi
ce R
ecomme
nda
ti
on Al
go
rit
hm f
or S
pa
ti
al
…
(
K. La
ks
hmai
ah
)
1229
In
f
uture,
the
work
ca
n
be
e
xten
ded
t
o
pro
cess
sp
at
ia
l
query
inform
at
io
n
in
cl
oud
e
nv
i
ro
nm
ent
with
data
an
d
loca
ti
on
pri
vacy
because
;
secu
r
it
y
is
chall
eng
eable
ta
s
k
in
Geo
-
S
ocial
Netw
ork
during
data
con
t
rib
ution an
d retrie
val.
REFERE
NCE
S
[1]
Ei
rin
aki
,
M.
,
Ga
o,
J.,
Varl
amis,
I.
,
&
Tser
p
es,
K.,
"Rec
om
m
ende
r
S
y
stems
for
La
rge
-
Sc
al
e
So
ci
a
l
Networks
:
A
rev
ie
w of
ch
al
l
e
nges
and
solu
ti
o
ns"
,
Fut
ure
Gen
erati
on
Comput
e
r Sy
stems,
vo
l. 7
8,
pp
.
413
–
418
,
2018.
[2]
Gao,
Y.,
W
ang,
S.,
Padm
ana
bha
n,
A.,
Yin,
J.,
&
Cao,
G.,
"M
appi
ng
spati
ote
m
por
al
pat
t
ern
s
of
eve
nts
using
socia
l
m
edi
a:
a
c
ase
stud
y
of
infl
u
enz
a
tre
nds
",
Int
ernati
onal
Journal
of
Geographic
al
Information
Sci
e
nce
,
vol
.
32,
no.
3
,
pp.
425
-
449
,
20
18.
[3]
Scel
l
at
o,
S
.
,
Ma
scolo,
C.
,
Mus
ole
si,
M.
,
&
Lato
ra,
V.
,
"D
ista
nc
e
Matters:
Geo
-
socia
l
Me
tri
cs
f
or
Online
Soci
a
l
Networks",
In
W
OSN
,
pp.
1
-
9,
20
10.
[4]
Masce
tti
,
S
.
,
Freni,
D.,
Be
tt
in
i,
C.
,
W
ang,
X.
S.,
&
Jajod
ia
,
S.,
"P
rivac
y
in
geo
-
soci
al
netw
orks:
proxim
i
t
y
noti
ficat
ion
with
untruste
d
serv
ice
providers
and
cur
ious
buddie
s"
,
The
VLDB
Jou
rnal
—
the
Inte
rn
ati
onal
Journal
on
Ve
ry
Lar
ge
D
ata
Bases
,
vol
.
2
0,
no
.
4
,
pp
.
541
-
566,
2011
.
[5]
Zha
ng,
C.
,
Shou,
L.,
Chen,
K.
,
Chen,
G.,
&
B
ei
,
Y.
,
"Eva
lu
at
i
ng
geo
-
socia
l
in
flue
nc
e
in
location
-
base
d
social
net
works
",
In
Proce
ed
ings
of
the
21st
ACM
int
ernati
ona
l
conf
ere
n
ce
o
n
Information
and
knowl
ed
ge
management
,
ACM
,
pp.
1442
-
14
51,
20
12
.
[6]
Do
y
tsher
,
Y.
,
Galon,
B
.
,
&
Kanz
a,
Y.
,
"Q
u
er
y
ing
g
eo
-
soci
al
d
at
a
b
y
br
id
ging
spat
ia
l
ne
t
works
and
soci
al
net
works
",
In
P
roce
edi
ngs
of
t
he
2nd
ACM
SIGSPATIAL
Inte
rnational
Workshop
on
Locati
on
Based
Social
Net
works,
pp
.
39
-
46,
2010
.
[7]
Freni,
D.,
Ru
iz
,
C.
V
.
,
Masce
tt
i
,
S.,
Bet
t
ini,
C.
,
&
Jensen,
C.
S.
,
"P
rese
rving
locati
on
and
abse
nc
e
privac
y
in
geo
-
socia
l
ne
tworks"
,
In
Proceedi
n
gs
of
the
19th
ACM
int
ernational
conf
ere
n
ce
on
Information
and
knowl
ed
ge
management
,
pp
.
309
-
318
,
ACM
,
2010
.
[8]
Li
u,
W
.
,
Sun,
W
.
,
Chen
,
C.
,
Huang,
Y.,
Jing,
Y.,
&
Chen,
K.,
"Circ
l
e
of
frie
nd
quer
y
in
geo
-
social
net
wor
ks"
,
In
Inte
rnat
ional
Confe
renc
e
on
Database
System
s for
Adv
an
ce
d
Applications
,
pp
.
126
-
137
,
2012
.
[9]
H.
K.
H.
Chan
,
C.
Long
and
R.
C.
W
.
W
ong
,
“
On
Gene
raliz
ing
Coll
ec
t
ive
Spat
ia
l
Ke
y
wor
d
Queri
es”
,
IE
E
E
Tra
nsac
ti
ons on
Know
le
dge
and
Data
Engi
ne
eri
n
g.
pp
.
1
-
18
,
201
8.
[10]
W
u,
D.,
L
i,
Y.,
Choi,
B
.
,
&
X
u,
J.,
"S
oci
al
-
a
ware
top
-
k
spa
tial
ke
y
word
se
ar
ch",
In
Mobile
Data
Manage
ment
(
MDM
)
,
2014
IE
EE
15
th
Int
ernat
ional
Conf
ere
nc
e,
vo
l. 1, pp
.
235
-
244,
2014
.
[11]
Yao,
B.
,
T
ang,
M.,
&
Li
,
F.
,
“
Multi
-
appr
oxim
at
e
-
k
e
y
word
rou
ti
ng
in
GIS
dat
a”
,
In
Pro
ce
ed
ing
s
of
the
19th
ACM
SIGSPATIAL
Int
ernati
onal
Confer
enc
e
on
Ad
vances i
n
Geographi
c
Information
Sy
stems,
pp.
201
-
2
10,
2011
.
[12]
Li
,
Y.,
W
u,
D.
,
Xu,
J.
,
Cho
i
,
B.
,
&
Su,
W
.
,
"S
pat
ial
-
awa
r
e
int
er
est
group
quer
i
es
in
lo
c
at
ion
-
b
ase
d
soci
al
net
works
",
Data
&
Knowle
dge
E
ngine
ering
,
vol
.
92
,
pp
.
20
-
38
,
2
014.
[13]
Chen,
L
.
,
Cong
,
G.,
Jensen,
C.
S.,
&
W
u,
D.,
"S
pat
ia
l
k
e
y
wor
d
quer
y
pro
ce
ss
ing:
an
exp
eri
m
ent
a
l
evalua
t
ion
"
In
Proceedi
ngs
of
th
e
VLDB
En
dowment,
vo
l. 6, no. 3, pp. 217
-
2
28,
2013
.
[14]
Rocha
,
J.
B.
,
Gkorgkas,
O.,
Jonass
en,
S.,
&
Nørvåg,
K.,
"Effici
ent
proc
essing
of
top
-
k
spati
al
k
e
y
word
quer
ie
s"
,
In
Inte
rnat
ional
Symposium on
S
pati
al
and
Temp
oral
Databases,
pp.
205
-
222
,
20
11.
[15]
Alra
hhal,
M.
S.,
Khem
akhe
m
,
M.,
&
Jam
bi,
K.,
"A
gent
-
Based
Sy
st
em
for
Eff
ic
ie
n
t
kNN
Quer
y
Proce
ss
ing
wit
h
Com
pre
hensive
Privacy
Prote
ct
ion",
(
IJA
CS
A)
Inte
rnationa
l
Journal
of
Adv
anc
ed
Computer
Scienc
e
and
Appl
ic
a
ti
ons,
vo
l.
9
,
no
.
1
,
2018
.
[16]
W
u,
D.,
Yi
u,
M.
L.
,
Jensen
,
C.
S.,
&
Cong,
G.,
"Effi
c
ie
n
t
cont
inuousl
y
m
oving
top
-
k
spa
ti
al
k
e
y
word
qu
e
r
y
proc
essing",
In
Data
Engi
n
ee
rin
g
(
ICDE)
,
IEE
E 27th
Int
ernati
on
al
Conf
ere
nce,
p
p.
541
-
552
,
201
1.
[17]
Olanr
ewa
ju
,
R.
F.,
&
Azm
an,
A
.
W
.
Inte
ll
ig
ent
Coopera
t
ive
Adapti
v
e
W
ei
ght
Ranki
ng
Policy
v
i
a
d
y
namic
ag
ing
base
d
on
NB
and
J48
cl
assifie
rs.
Indone
sian
Jour
nal
of
El
e
ct
rica
l
Engi
ne
ering
and
Informatic
s
(
IJ
EE
I)
,
2017;
5
(4)
:
357
-
365.
[18]
Sultha
na
,
R.
,
&
Ramasam
y
,
S.
Conte
xt
Based
Cla
ss
ifi
c
at
ion
of
Revi
ews
Us
ing
As
socia
ti
on
Ru
le
Mining,
Fuz
z
y
Logi
cs
and
Onto
log
y
.
Bul
l
et
in
of
El
e
ct
rica
l
Eng
in
ee
ring a
nd
Infor
matic
s
,
2017
;
6
(
3):
250
-
255.
[19]
Rao,
R.
R.
,
&
M
akki
th
a
y
a,
K.
L
e
arn
ing
from
a
Cla
ss
Im
bal
anc
ed
Public
Hea
lt
h
D
at
ase
t:
a
Cost
-
ba
sed
Com
par
iso
n
of
Cla
ss
ifier
Per
form
anc
e.
Int
ernati
onal Journal of
E
le
c
tric
al
and
Computer
Eng
i
nee
ring
,
2017;
7
(4):
2215
-
2222
.
[20]
Bao,
J.
,
Zhe
ng
,
Y.,
&
Mokbel
,
M.
F.,
“
Lo
ca
t
io
n
-
base
d
and
pre
fer
ence
-
awa
r
e
r
e
comm
enda
ti
on
u
sing
sparse
geo
-
socia
l
ne
tworki
ng
dat
a
”,
In
Proce
ed
ings
of
t
he
20th
international
con
fe
re
n
ce
on
advan
c
es
in
geographic
inf
orm
ati
on
syst
ems
,
ACM
,
pp
.
199
-
208,
2012
.
Evaluation Warning : The document was created with Spire.PDF for Python.