Indonesi
an
Journa
l
of El
ect
ri
cal Engineer
ing
an
d
Comp
ut
er
Scie
nce
Vo
l.
9
, No
.
1
,
J
an
ua
ry
201
8
, p
p.
146
~
151
IS
S
N:
25
02
-
4752
,
DOI: 10
.11
591/
ijeecs
.
v9.i
1
.
pp
146
-
151
146
Journ
al h
om
e
page
:
http:
//
ia
es
core.c
om/j
ourn
als/i
ndex.
ph
p/ij
eecs
Determi
nin
g Hotspots of
Ro
ad Ac
cidents Usin
g Sp
atia
l
An
alys
i
s
S.Sari
f
ah
Rad
iah Shari
ff
*
1
,
Ha
m
dan A
bd
ul M
aad
2
,
Nu
r
sy
az
a N
arsuh
a Abd
ul H
alim
3
,
Z
uraida
h
Der
as
it
4
1
Malay
s
ia Insti
t
ute
of
Tr
ansport (MITRAN
S),
Univer
siti
T
eknologi
MA
RA Shah
Alam,
Ma
lay
sia
2,4
Cent
re
for
Sta
t
isti
cs
and
De
ci
si
on
Scie
n
ce Studi
es,
Facu
lty
of
C
om
pute
r
&
M
at
h
emati
c
al Sci
en
ces
,
Univer
sit
i
Te
knologi MA
RA Shah
Alam, M
al
a
y
s
ia
3
Digi
Tele
communic
a
ti
ons,
Shah
Alam,
Ma
lay
sia
Art
ic
le
In
f
o
ABSTR
A
CT
Art
ic
le
history:
Re
cei
ved
A
ug
11
, 201
7
Re
vised
N
ov
1
0
, 2
01
7
Accepte
d
N
ov
2
7
, 201
7
Road
acci
den
ts
cont
inuousl
y
b
ec
om
e
a
m
aj
or
proble
m
in
Malay
si
a
and
conse
quentl
y
caus
e
loss
of
li
fe
or
prope
rt
y
.
Due
to
tha
t,
m
an
y
r
oad
ac
c
ide
nt
dat
a
hav
e
be
en
col
lect
ed
b
y
h
ighwa
y
concess
iona
ri
es
or
build
–
oper
ate
–
tra
nsfer
oper
atin
g
companie
s
in
t
he
count
r
y
m
ea
n
t
for
coming
up
with
prope
r
count
er
m
ea
sur
e
s.
Sever
al
anal
yses
ca
n
be
done
on
the
a
cc
um
ul
at
ed
d
at
a
in
orde
r
to
improve
roa
d
safe
t
y
.
In
thi
s
stud
y
th
e
r
epor
te
d
ro
ad
acc
ide
nts
ca
s
es
in
North
South
Expre
ss
wa
y
(
NS
E)
from
Sun
gai
Pet
ani
to
Bukit
L
anj
an
duri
ng
2011
to
2014
per
iod
is
an
aly
z
ed.
Th
e
ai
m
is
to
det
ermine
whethe
r
th
e
pat
t
ern
is
cl
uster
ed
a
t
c
ertain
area
and
to
ide
nt
if
y
spati
a
l
p
at
t
ern
o
f
hot
spots
ac
ross
thi
s
lon
gest
cont
rol
le
d
-
ac
c
ess
expr
essw
a
y
in
Ma
lay
si
a
as
h
otspot
re
pre
sents
th
e
l
oca
t
ion
of
the
r
oad
which
is
c
onsidere
d
high
risk
and
the
proba
bil
i
t
y
of
t
ra
ffic
acci
den
ts
in
re
l
at
ion
to
the
l
eve
l
of
risk
in
th
e
surrounding
areas.
As
no
m
et
h
odolog
y
for
id
e
nti
f
y
ing
ho
tspot
has
bee
n
agr
ee
d
g
loball
y
y
et
;
h
ence
th
is
stud
y
h
el
ped
det
ermining
t
he
suita
b
le
princ
iples
and
t
ec
hniqu
es
for
d
et
ermina
ti
on
of
the
ho
tspot
on
Malay
si
an
highwa
y
s
.
Two
spati
al
an
aly
sis
te
chn
ique
s
were
appl
ied,
Nea
re
s
t
Neighbor
hood
Hier
arc
h
ical
(
NN
H)
Cluste
ri
ng
and
Spa
ti
a
l
T
emporal
Cluste
ring
,
using
CrimeStat
®
and
visual
izing
in
ArcGIS
™
software
to
ca
l
cul
a
te
the
co
nce
ntr
at
ion
of
t
he
in
ci
den
ts
an
d
the
re
sults
ar
e
compar
ed
base
d
on
the
ir
accura
c
ie
s.
Result
s
ide
nti
fi
ed
seve
r
al
hotspots
and
show
ed
tha
t
they
var
ie
d
in
n
um
ber
and
loc
a
t
ions,
depe
nd
ing
on
the
ir
p
ara
m
et
er
v
al
ues
.
Further
an
aly
s
is
on
select
ed
hot
spot
locati
on
show
ed
tha
t
Spatial
T
empora
l
Cluste
ring
(ST
AC)
has
a
highe
r
accurac
y
i
ndex
compare
d
to
Nea
re
s
t
Neighbor
Hier
a
rc
hical
Cluste
r
i
ng
(NN
H).
Se
ver
al
re
comm
enda
ti
ons
on
count
er
m
ea
sure
s ha
ve al
so b
ee
n
proposed
base
d
on
the details r
es
ult
s
.
Ke
yw
or
d
s
:
Pr
e
dicti
ve
Acc
ur
acy
Road
A
cci
den
t
s Hotsp
ots
Sp
at
ia
l A
naly
sis
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
:
S.S
ari
fah Ra
di
ah
S
ha
riff
,
Ma
la
ysi
a In
sti
tute o
f
T
ra
ns
po
rt
(MI
TRA
NS
)
,
Un
i
ver
sit
i Te
knol
og
i M
ARA
Sh
a
h Alam
,
Ma
la
ysi
a
.
Em
a
il
:
rad
ia
h@t
m
sk
.u
itm
.ed
u.m
y
1.
INTROD
U
CTION
On
e
of
t
he
m
ajo
r
cause
s
of
de
at
h
in
Ma
la
ysi
a
is
due
to
roa
d
acci
den
ts.
W
it
h
the
ra
pid
de
velo
pm
ent
in
Ma
la
ysi
a’s
econom
y,
the
issue
of
r
oa
d
safety
beco
m
e
s
increasin
gly
i
m
po
rtant.
A
cc
ordin
g
to
Ma
la
ysi
an
In
sti
tute
of
Ro
ad
Sa
fety
Re
se
arch
(MIR
OS),
in
2014
t
he
tot
al
nu
m
ber
of
r
oad
acci
de
nts
was
476,1
96
w
it
h
the
nu
m
ber
of
ro
a
d
death
s
is
6,676
cases
seri
ou
sly
in
j
ured
i
s
4,
43
2
cases
and
sli
ghtl
y
inj
ure
d
is
8,
59
8
cases.
Ma
la
ysi
a
has
been
ranke
d
the
8
th
in
fatal
it
ie
s
of
ro
a
d
cr
ashes
in
the
Mortal
it
y
fr
om
Road
Cras
hes
in
193
Countries
Re
port.
Ma
la
ysi
a
Road
Tra
nsp
ort
s
De
par
tm
ent
(RT
D)
re
port
ed
th
at
in
2014,
Ma
la
ysi
a
lo
st
RM
nin
e
bill
ion
du
e
to
road
death
s
as
sud
de
n
de
at
h
m
a
y
resu
lt
in
lost
of
nat
ural
assest
.
On
a
ver
a
ge,
m
or
e
than
t
e
n
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c
En
g
&
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Determi
ning
H
otspots
of R
oad
Acci
den
ts
Us
ing
Spatial
An
alysis
(
S.S
ar
if
ah R
adia
h Sh
ar
i
ff
)
147
bill
ion
rin
gg
it
has
l
os
ses
due
to
r
oa
d
acci
de
nts
e
ver
y
ye
ar
.
The
hi
gh
est
fa
ta
li
ti
es
by
the
age
gro
up
are
a
m
ong
young
per
s
on
aged
16
to
25
ye
ars
old
[
1].
A
stu
dy
[2
]
on
Ma
la
ysi
a
road
acci
de
nt
sit
uation
sta
te
d
t
hat
the
ps
yc
holo
gical
su
f
fer
i
ngs
are
of
te
n
inten
se
,
la
sti
ng
an
d
ev
en
pe
rm
anen
t.
The
victi
m
s
m
ay
gen
erate
so
m
at
i
c
il
lnesses whic
h w
or
se
n
t
his
psy
cho
lo
gical
distress, c
reati
ng
a v
ic
io
us
cir
cl
e.
Ther
e
a
re
se
ve
ral
m
e
tho
ds
be
ing
us
ed
t
o
det
erm
ine
the
ho
t
sp
ots
for
r
oa
d
acci
den
ts
in
t
he
pr
e
vious
stud
ie
s.
Am
on
g
w
hich
are
a
sta
ti
sti
cal
package,
Crim
eSta
t
that
of
fe
rs
va
r
iou
s
al
gorit
hms
in
determ
ini
ng
the
sp
at
ia
l
patte
r
n
that
exist
within
the
data
[
3
-
6]
al
so
us
e
d
Ge
ogra
ph
ic
al
I
nf
or
m
at
ion
Syst
em
(G
IS
)
t
o
vis
ualiz
e
the
identifie
d
ho
ts
pots
to
fur
ther
stre
ng
t
hen
the
validit
y
of
the
res
ults.
I
n
this
stud
y,
t
he
f
ocu
s
was
on
the
acci
den
ts
patte
rn
al
on
g
the
North
-
S
ou
t
h
Ex
pr
ess
way
(
NS
E
).
T
he
N
SE
is
the
longest
co
ntr
olled
-
acce
s
s
express
w
ay
in
Ma
la
ysi
a
with
the
total
le
ng
t
h
of
a
bout
77
2
km
(4
80
m
i)
ru
nnin
g
from
B
uk
it
Kay
u
Hita
m
in
Ked
a
h
nea
r
t
he
Ma
la
ysi
an
-
T
hai
bord
e
r
(c
onnects
with
P
hetkasem
Road
(Route
4)
in
Thail
an
d)
to
Jo
hor
Ba
hru
at
the
so
ut
hern
portio
n
of
Pe
ninsula
r
Ma
la
ysi
a
and
to
Si
ngap
ore
.
The
ex
pr
e
ss
way
li
nk
s
m
any
m
ajo
r
ci
ti
es
and
to
wns
in w
est
e
rn
Pe
nin
s
ular
Ma
la
ysi
a,
act
ing
as
the
'
bac
kbone'
of
the
west
coa
st
of
the
p
eni
nsula
. I
t
pro
vid
es
a
fa
ste
r
al
te
r
native
t
o
th
e
old
Fede
ral
Ro
ute
1,
t
hus
reduci
ng
tra
velli
ng
tim
e
between
va
rio
u
s
town
s
& cit
ie
s.
2.
RESEA
R
CH MET
HO
D
This
stu
dy
fir
st
exam
ined
m
on
thly
rep
or
te
d
ro
a
d
acci
de
nts
involvi
ng
al
l
t
ypes
of
veh
ic
le
s
that
occurre
d
al
ong
North
-
S
ou
t
h
Ex
pr
ess
way
(
NS
E
)
sta
rtin
g
from
Su
ngai
P
et
ani
to
Bu
kit
Lan
j
an
from
20
11
t
o
2014.
T
he
dat
a
was
obta
ine
d
from
PLU
S
Ex
pr
ess
way
Be
rh
a
d
that
incl
ud
e
s
acci
den
t
locat
ion
s
(in
te
rm
s
of
kilom
et
er)
,
the
m
on
th,
date,
day,
tim
e,
veh
i
cl
e
inv
ol
ved,
c
auses
of
acci
de
nt,
colli
sio
n
ty
pe
an
d
in
j
ury
ty
pe.
Assum
ing
that
the
acci
de
nt
can
occ
ur
at
any
sp
ot
al
ong
the
express
w
ay
,
w
e
div
ide
d
t
he
le
ng
t
h,
a
lo
ng
S
ungai
Peta
ni to
B
ukit
Lanja
n,
i
nto
six
(6)
sect
io
ns
a
s d
esc
ribe
d
i
n Table
1.
Table
1. Lo
cat
ion
a
nd Secti
on
of Stu
dy
Ar
e
a
Sectio
n
Locatio
n
Distan
ce
N3
Su
n
g
ai Petani to
Jawi
5
4
.90
k
m
N4
Jawi to
Can
g
k
at Jering
5
6
.80
k
m
N5
Can
g
k
at Jering
to Ipo
h
5
5
.40
k
m
C1
Ipo
h
to Bid
o
r
6
4
.40
k
m
C2
Bid
o
r
to
T
an
ju
n
g
M
ali
m
5
9
.70
k
m
C3
Tanju
n
g
M
ali
m
to
Bu
k
it L
an
jan
6
0
.30
k
m
Descr
i
ptive
a
naly
sis
on
road
acci
de
nts
at
each
sect
ion
is
pr
e
sente
d
to
obse
rv
e
the
patte
r
n.
In
order
t
o
id
entify
the
hotspo
ts
,
tw
o
m
e
thods
f
ro
m
Cri
m
eSta
t,
Near
est
Neighb
or
Hierarc
hical
(
NNH
)
Cl
us
te
rin
g
A
na
ly
sis
and
S
patia
l
Te
m
po
ral
C
lusterin
g
A
nal
ysi
s
are
consi
de
red
a
nd
com
par
ed
.
The
se
m
et
hods
autom
at
ic
collect
the
s
urface locat
ion
of
the
targ
et
a
nd
s
olve
d
us
in
g
C
rim
e
Stat
,
a
s
patia
l
sta
ti
sti
cs
pr
ogra
m
fo
r
the an
al
ysi
s
of
crim
e incident locati
on
s
[7].
2.1.
Ne
ares
t Nei
ghbor
Ana
lysis (
NNA
)
Near
est
Nei
ghbor
A
naly
sis
(N
N
A)
pro
du
ce
s
a
cal
culat
ion
cal
le
d
the
Near
est
Neig
hbor
Inde
x
(NN
I
)
wh
ic
h
is
the
r
at
io
of
t
he
ob
s
erv
e
d
distance
div
ide
d
by
th
e
exp
e
ct
ed
distance.
If
t
he
in
dex
valu
e
sho
ws
le
ss
than
1,
it
m
ean
s
that
the
patte
rn
of
i
ncide
nt
exh
i
bits
cl
us
te
r
ing
a
nd
i
f
the
i
nd
e
x
is
great
er
than
1,
it
m
ea
ns
that
the
incide
nt
tr
end
is
ra
ndom
or
t
ow
a
r
d
di
sp
ersi
on.
I
n
N
NA,
the
Z
sc
or
e
value
s
m
e
asur
e
t
he
sta
ti
sti
cal
sign
ific
a
nce
w
hethe
r
the
p
at
t
ern
s
of i
ncide
nt
are
ra
ndom
ly
o
r
not ra
ndom
l
y dist
rib
uted.
NNI is calc
ulate
d
as
foll
ows:
N
e
i
g
h
b
o
r
N
e
a
r
e
s
t
A
v
g
E
x
p
e
c
t
e
d
N
e
i
g
h
b
o
r
N
e
a
r
e
s
t
A
v
g
I
n
d
e
x
N
e
i
g
h
b
o
r
N
e
a
r
e
s
t
W
he
re
:
A
c
c
i
d
e
n
t
s
of
N
u
m
b
e
r
D
i
s
t
a
n
c
e
N
e
i
g
h
b
o
r
N
e
a
r
e
s
t
A
v
e
r
a
g
e
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,
Vol
.
9
,
No.
1
,
Jan
ua
ry
201
8
:
146
–
151
148
P
o
i
n
t
s
of
N
u
m
b
e
r
A
r
e
a
2
1
N
e
i
g
h
b
o
r
N
e
a
r
e
s
t
A
v
e
r
a
g
e
E
x
p
e
c
t
e
d
2.2.
Ne
ares
t Nei
ghbor
Hie
rarchical
(
NNH) Clust
eri
ng
The
ne
xt
ste
p
is
to
identify
the
de
ns
it
y
of
r
oad
acci
dent
s
occurre
nce
or
the
hot
spo
ts
us
in
g
the
Near
est
Nei
ghbor
Hiera
rc
hical
(N
N
H)
Cl
ust
ering
.
In
N
NH,
a
thresho
ld
value
an
d
m
ini
m
u
m
nu
m
ber
of
acci
den
ts,
m
i
n
n
with
in
a
cl
us
te
r
nee
d
to
be
pr
e
dete
rm
ined.
Usi
ng
Cri
m
eSta
t
so
ftwar
e
,
1
kilom
et
er
has
bee
n
se
t
up
as
a
thre
sho
ld
distance,
a
nd
the
m
ini
m
um
nu
m
ber
per
cl
us
te
rs,
m
i
n
n
is
20
0.
At
the
sam
e
t
i
m
e,
NN
H
res
ul
ts
are
giv
e
n
with their
sever
it
y
va
lue,
w
hich
is
the
num
ber
of
acci
den
ts
locat
e
d
in
the
bounda
ry
of
a
cl
us
te
r
.
The
adv
a
ntage
to
this
te
ch
nique
is
that
it
ca
n
ide
ntify
sm
al
l
geogr
a
phic
al
env
i
ronm
ents
where
t
here
ar
e
con
ce
ntrate
d
i
ncide
nts.
T
his
can
be
us
ef
ul
f
or
sp
eci
fic
ta
r
geting,
ei
the
r
by
poli
ce
de
pl
oym
ent
or
c
om
m
un
ity
interve
ntio
n.
2.3.
Sp
at
i
al T
empor
al
Cluste
ri
ng
Analysis
(
ST
AC)
STA
C
is
one
of
the
wi
dely
use
d
m
et
ho
ds
in
detect
ing
ho
ts
po
ts
in
t
he
enti
re
stu
dy
area
ba
sed
on
it
s
sh
a
pe.
I
n
m
os
t
pr
e
vious
stu
die
s
it
is
su
ggest
e
d
to
us
e
r
ect
an
gu
la
r
if
the
ana
ly
sis
area
has
m
os
t
reg
ular
pa
tt
ern
,
wh
il
e
the
tria
ngula
r
patte
rn
is
m
os
t
su
it
ab
le
f
or
t
he
a
rea
w
hich
ge
ner
a
ll
y
has
an
irre
gu
la
r
patte
rn.
STAC
places
a
ci
rcle
on
e
ver
y
no
de,
counts
the
nu
m
ber
of
acci
de
nts
w
hich
fall
ing
within
eac
h
ci
rcle
and
ra
nks
the
ci
rcle
in
desce
nd
i
ng
orde
r.
T
he
X
a
nd
Y
co
ordinates
of
a
ny
no
de
with
at
le
ast
two
in
ci
de
nts
within
t
he
search
rad
i
us
a
re r
ec
orde
d, al
ong wit
h
the
num
ber
of
data points
f
ound
f
or
eac
h n
od
e
.
2.4.
C
ompari
so
n
of H
ot Sp
ot Tec
hnique
s
In
orde
r
to
obs
erv
e
t
he
acc
ur
a
cy
of
eac
h
m
eth
od,
Pr
e
dicti
on
Acc
ur
acy
Inde
x
(
PAI)
that
is
com
m
on
ly
us
e
d
in
cr
im
e
ho
t
s
pot
m
app
ing
is
a
da
pted.
A
cco
r
ding
to
[
7]
this
m
e
tho
d
was
de
vel
op
e
d
in
orde
r
to d
et
erm
ine
the
dif
fer
e
nces
betwee
n
eac
h
m
et
ho
d
i
n
ca
pturin
g
or
pr
e
dicti
ng
hotsp
ots
locat
ion.
Fi
ndin
g
10
0
per
c
ent
of
fu
t
ur
e
e
ven
ts
i
n
100
per
ce
nt
of
the
area
w
ou
l
d
gi
ve
a
PAI
value
of
1.
I
f
the
hit
rate
an
d
the
area
per
c
entage
fall
b
y an eq
ua
l
m
easur
e, the
value
w
ould b
e
co
m
pu
te
d
as 1
also.
Th
us
, th
e
g
reater the nu
m
ber
o
f
acci
den
ts i
n
a
ho
ts
pot
area
that
is
sm
a
ll
er
in
siz
e
to
the
whole
stu
dy
ar
ea,
the
higher
the
PAI
valu
e
.
The
P
A
I
is
ea
sy
to
cal
culat
e, cons
iderin
g
t
he
nu
m
ber
o
f
acci
de
nts that
fall
into
the
area
deter
m
ined
as
ho
ts
pots agai
ns
t t
he
siz
e o
f
the hots
po
t a
nd
the size
of the
stud
y a
rea.
The
h
ig
he
r
the
P
AI the m
or
e acc
urat
e the m
et
hod
is.
Pr
e
dicti
on
Acc
ur
acy
Ind
e
x
is
cal
culat
ed
as
f
ollows:
100
)
A
a
(
100
)
N
n
(
I
n
d
e
x
(
P
A
I
)
A
c
c
u
r
a
c
y
P
r
e
d
i
c
t
i
o
n
,
Wh
e
re
:
n
is t
he
total
num
ber
of accide
nts
fou
nd in
t
hat part
ic
ular
cl
us
te
r,
N
is t
he
total
num
ber
of accide
nts i
n
t
he
st
ud
y a
rea,
a
e
qua
ls t
o
total
a
rea
of hotsp
ots a
nd
A
e
qua
ls t
o
the
en
ti
re
area in
the st
ud
y regi
on.
3.
RESU
LT
S
A
ND AN
ALYSIS
Figure 1
s
hows
, f
r
om
2
01
1
to
2014 secti
on
s
C3 (
Ta
njung
Ma
lim
to
Buk
it
Lan
j
a
n)
r
ec
orded
30.5% o
f
total
acci
den
ts,
the
highest
nu
m
ber
of
r
oad
a
cci
den
t
com
par
ed
to
t
he
ot
he
r
sect
ion
s
.
T
his
sect
ion
rem
ain
s
th
e
record
of
ha
vi
ng
the
highest
cases
of
ro
a
d
acci
den
t
fro
m
20
11
-
2014.
Wh
il
e,
the
lo
west
nu
m
ber
of
r
oa
d
acci
den
t
is
sect
ion
N3
wh
ic
h
is
from
Su
ng
ai
Peta
ni
to
Jawi
wh
ic
h
co
ntri
bute
d
only
6.7%
of
the
t
otal
num
ber
of
a
cci
de
nt
in
al
l
sect
ion
s.
The
se
ve
rity
rati
ng
of
the
va
st
m
ajo
rity
of
r
oa
d
acci
de
nt
acro
s
s
N
ort
h
Sout
h
Ex
pr
ess
way is
descr
i
bed in
Figure
2.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c
En
g
&
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Determi
ning
H
otspots
of R
oad
Acci
den
ts
Us
ing
Spatial
An
alysis
(
S.S
ar
if
ah R
adia
h Sh
ar
i
ff
)
149
Figure
1.
Ef
fec
ts of sel
ect
ing
diff
e
re
nt sw
it
c
hing
unde
r dynam
ic
co
nd
it
io
n
Figure
2
.
Ef
fec
ts of sel
ect
ing
diff
e
re
nt
switc
hing
unde
r dynam
ic
co
nd
it
io
n
3.1.
R
andom
ness P
attern
I
dent
ific
at
i
on
As
m
entioned
earli
er,
1
kilo
m
et
er
hav
e
be
en
set
up
as
a
thres
hold
distance
wh
il
e
the
m
ini
m
u
m
nu
m
ber
per
cl
ust
ers
is
200,
usi
ng
Crim
eSta
t
so
ft
war
e
.
I
n
Ta
ble
2,
it
sho
ws
that
the
r
oad
a
cci
den
t
cases
a
cro
s
s
North
South
E
xpress
way
(
N
SE)
a
re
ha
ving
a
cl
us
te
red
distrib
ution
be
cause
the
Nea
r
est
Neig
hbor
I
nd
e
x
is
0.005
06
w
hich
is less t
han
1.
The p
-
value
al
so
s
hows
th
at
it
less than
, th
us we r
e
j
ect
n
ull
h
yp
oth
esi
s
of
the
occurre
nce
of
ro
a
d
a
cci
den
ts
acr
oss
NS
E
was
ra
ndom
.
This
go
es
to
the
co
nc
lusio
n
that
the
ro
a
d
acci
den
ts e
xhibit
a cluste
rin
g patt
er
n.
Table
2.
Neare
st Neig
hbor
Index Re
su
lt
Test Statistic
(
Z)
:
-
2
3
7
.67
8
8
P
V
alu
e On
e
Ta
il :
0
.00
0
1
Mean Ne
arest
Nei
g
h
b
o
r
Distan
ce (
m
)
Exp
ected Near
est
Neig
h
b
o
r
Distan
ce (
m
)
Near
est
Neig
h
b
o
r
Ind
ex
3
.84
3
2
7
5
8
.807
0
.00
5
0
6
3.2.
H
ot
s
pots
Det
ermi
n
ati
on
Accor
ding
to
[
8],
ch
oice
of
t
hr
es
hold
li
m
i
ts
is
i
m
po
rtant,
wh
ic
h
basical
ly
def
ine
s
the
s
cop
e
of
the
cl
us
te
rin
g
anal
ysi
s.
Hen
ce
,
usi
ng
Crim
eSta
t
so
ftwa
re
in
cond
ucting
t
his
Near
est
Neighb
or
Hierarc
hical
Cl
us
te
rin
g
acr
os
s
North
Sou
th
Ex
press
way,
1
kilom
et
er
hav
e
bee
n
set
up
as
a
th
res
hold
distance
w
it
h
the
m
ini
m
u
m
nu
m
ber
pe
r
cl
ust
er
s
is
20
0.
T
able
3
s
how
s
that,
there
a
re
12
hot
spots
with
t
he
highest
num
ber
of
ro
a
d
acci
den
t a
re
349
case
s.
Du
e
to
it
s
natu
re
of
detect
ing
the
ho
ts
pots
ba
sed
on
patte
r
n
or
s
hap
e
,
there
is
a
need
to
de
te
rm
ine
the
search
ra
dius
and
the
m
ini
mu
m
nu
m
ber
of
acci
de
nts
pe
r
cl
us
te
r
.
Af
te
r
doin
g
t
he
se
nsi
ti
vity
analy
si
s,
th
e
search
ra
diu
s
a
ll
ow
ed
in
t
his
stud
y
ha
ve
bee
n
set
as
0.5
kil
om
et
er
with
n
m
i
n=
10
0
ba
sed
on
tria
ng
ular
pa
tt
ern
for
sca
n
ty
pe
. A
s
s
how
n
i
n
T
able
3,
th
ere
a
r
e
14 h
ots
pots
f
ound w
it
h
the
h
ig
hest n
um
ber
o
f
r
oad
acci
de
nts
ar
e
220
ca
ses.
The
num
ber
of
ho
t
spots
distr
ibu
ti
on
withi
n
the
six
sect
ion
s
is
su
m
m
arized
in
Table
4.
Bo
th
m
et
hods
sh
ow t
hat the
m
os
t ho
tsp
ots
are fo
und
i
n
se
ct
ion
C
3
(
Tan
jong Mal
im
to
Buk
it
La
njan
).
Table
3.
H
ot s
po
ts
Identifie
d base
d on N
NH an
d
S
TAC
NNH
STAC
Clu
ster
Mean X
(lon
g
itu
d
e)
Mean Y
(latitud
e)
Frequ
en
cy
(
No
.
Of
Acciden
t
Clu
ster
Mean X
(lon
g
itu
d
e)
Mean Y
(latitud
e)
Frequ
en
cy
(
No
.
Of
Acciden
t)
1
1
0
0
.9703
4
.69
3
6
6
349
1
1
0
0
.672
4
.92
6
9
220
2
1
0
1
.57
3
.23
0
1
349
2
1
0
1
.5668
3
.23
3
6
4
218
3
1
0
0
.6718
4
.92
7
5
5
328
3
1
0
0
.4891
5
.19
1
5
5
161
4
1
0
1
.023
4
.69
2
0
6
298
4
1
0
1
.542
3
.43
6
3
4
148
5
1
0
1
.5579
3
.37
0
6
2
298
5
1
0
0
.983
4
.68
8
4
8
133
6
1
0
1
.545
3
.43
1
0
2
295
6
1
0
1
.5603
3
.37
5
0
8
130
7
1
0
1
.5202
3
.66
1
7
5
241
7
1
0
1
.5549
3
.36
4
2
6
129
8
1
0
1
.5841
3
.19
3
234
8
1
0
1
.1708
4
.45
5
6
3
127
9
1
0
1
.2082
4
.40
7
8
9
223
9
1
0
1
.5557
3
.38
8
8
6
114
10
1
0
1
.5468
3
.30
6
5
4
222
10
1
0
1
.5766
3
.22
3
1
1
112
11
1
0
0
.8173
4
.77
8
6
5
213
11
1
0
1
.0037
4
.68
7
8
2
104
12
1
0
1
.4904
3
.69
1
5
9
210
12
1
0
1
.553
3
.35
1
7
6
102
13
1
0
1
.026
4
.68
6
1
1
101
14
1
0
1
.5657
3
.24
5
7
9
101
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,
Vol
.
9
,
No.
1
,
Jan
ua
ry
201
8
:
146
–
151
150
Table
4.
N
um
ber
of
H
ot S
po
t
by Secti
ons
us
i
ng NN
H
Sectio
n
Nu
m
b
e
r
o
f
Hot Sp
o
ts
Usin
g
NNH
Usin
g
ST
AC
N3
(
Su
n
g
ai Petani
to
Jawi)
0
1
N4
(
Jawi to
C
an
g
k
at Jering
)
1
1
N5
(
Can
g
k
at Jerin
g
to Ipo
h
)
3
3
C1
(
Ipo
h
to Bid
o
r)
1
1
C2
(
Bid
o
r
to
T
an
ju
n
g
M
ali
m
)
1
0
C3
(
Tanju
n
g
M
ali
m
to
Bu
k
it L
an
jan
)
6
8
3.
3
.
Predi
cti
on Acc
urac
y
The
pe
rfo
rm
a
nce
of
the
tw
o
m
et
ho
ds
in
identify
ing
t
he
hotsp
ots
is
m
easur
ed
us
i
ng
P
A
I
an
d
su
m
m
arized
in
Table
5.
Since
the
PA
I
val
ue
s
fo
r
ST
AC
ar
e
hig
he
r,
it
can
be
con
cl
ud
e
d
that
STA
C
perform
s
bette
r
com
par
e
d
to
NNH fo
r
a
ll
inj
ury
ty
pes a
nd
fatal
it
ie
s.
STA
C al
s
o
m
a
nag
e
d
to i
de
ntify t
wo
m
or
e
s
pots f
or
po
s
sible
ho
ts
pots c
om
par
ed
t
o NNH.
Table
5.
C
om
par
iso
n
Be
twee
n NNH a
nd ST
AC
Area
(sq
m
e
tre
)
Frequ
en
cy
Hit Rate
PAI
All I
n
ju
r
y
T
y
p
e
NNH
1
.39
3260
2
2
.29
1
6
3
.69
STAC
0
.29
1900
1
2
.99
4
6
1
.57
Fatalitie
s
NNH
0
.80
198
1
.35
1
7
.28
STAC
0
.51
77
0
.53
1
0
5
4
.5
8
4.
CONCL
US
I
O
N
Ov
e
rall
,
the
pa
tt
ern
of
r
oa
d
acci
den
t
on
North
S
ou
t
h
Ex
pr
ess
way
(
NS
E
)
res
ulted
in
cl
us
te
ring
patte
rn
which m
eans
that
the acci
den
t was grou
ped
i
n
on
e
l
ocati
on.
F
or
bo
th
m
et
ho
ds
u
se
d
in d
et
erm
ining
th
e
ho
ts
pots,
the
m
os
t
ho
tspo
ts
are
fou
nd
in
C
3
sect
ion
wh
i
c
h
was
f
r
om
Ta
njung
Ma
li
m
t
o
Bu
kit
Lanj
a
n.
This
is
al
so
par
al
le
l
w
it
h
descr
ipti
ve
sta
ti
sti
cs
resu
lt
on
the
hi
gh
e
st
ro
a
d
acci
den
ts
occurre
nce.
It
is
al
so
ob
se
rv
e
d
that
the
ty
pes
of
r
oad
al
ong
t
he
sect
ion
is
3
r
oad
la
ne.
Ba
s
ed
on
t
he
res
ul
ts
ob
ta
ine
d,
it
is
i
m
po
rtant
for
th
e
con
ce
r
ned
bodi
es
to
ta
ke
a
re
m
edial
act
ion
at
sel
ect
ed
hot
spots.
T
his
in
cl
ud
es
a
ddin
g
m
or
e
cauti
on
s
ign
a
ge
or im
ple
m
ents sp
eci
fic inte
rv
e
ntions to
preve
nt roa
d
acci
de
nt
.
Both
ap
plied
m
et
ho
ds
ca
n
be
us
ed
a
re
pra
ct
ic
al
in
determ
ining
the
h
ot
sp
ot
base
d
on
choosi
ng
a
n
accurate
pa
ram
et
er
value
su
c
h
as
m
uch
s
m
al
l
er
thres
ho
l
d
va
lue.
Furthe
r
an
al
ysi
s
on
m
or
e
su
it
able
par
a
m
et
er
values
to be
use
d
s
houl
d be
done
for bett
er
re
su
lt
s [9,
10]
.
It
is
adv
isa
ble
to
f
ur
the
r
r
esea
rch
t
o
ad
d
m
ore
of
hot
spot
t
echn
i
qu
e
m
et
hods
s
uch
a
s
Ke
r
nel
De
ns
it
y
Estim
at
ion
(KDE),
K
-
Me
an
s
Cl
us
te
rin
g,
G
et
is
-
Ord
Stat
ist
ic
s
and
M
or
a
n’s
I
ndex
.
Be
sid
es
that,
the
st
udy
can
be
div
ide
d
i
nto
an
oth
er
sub
gr
oup
s
uc
h
as
c
ol
li
sion
ty
pe,
in
jury
ty
pe
or
ve
hicle
s
in
vo
l
ved.
Com
par
ed
to
oth
e
r
stud
ie
s
wh
ic
h
consi
der
t
he
ne
twork
desig
n
su
c
h
as
r
ounda
bout
an
d
inte
rs
ect
ion
s,
t
his
stud
y
ca
nnot
ta
ke
any
desig
n
ty
pe
i
n
consi
der
at
io
n
s
ince
the
highwa
y
is
a
strai
gh
t
ro
a
d
netw
ork.
Howe
ver,
the
nu
m
ber
of
la
ne
s
or
ro
a
d
ty
pes
co
ul
d
be
use
d
to
unde
rstan
d
the
r
el
at
ion
sh
i
p
bet
ween
t
he
hot
sp
ots
a
nd
ro
a
ds
.
Sp
ee
d
sta
tus
of
the
par
ti
cula
r
r
oa
d segm
ents can b
e u
se
d w
hen an
al
yz
ing
the
hot
spots
.
ACKN
OWLE
DGE
MENT
We
w
ou
l
d
li
ke
to
acknow
le
dg
e
Re
sea
rch
Ma
nag
em
ent
In
sti
tute
(RMI)
of
U
niv
e
rsiti
Teknolo
gi
MARA,
Ma
la
ysi
a,
for
f
undi
ng
t
his
resea
rch
t
hro
ugh
t
he
G
ra
nt
N
o:
600
-
IRMI/
D
ANA
5/3
/LES
TAR
I
(00
39
/
2016).
REFERE
NCE
S
[1]
Sham
suddin,
S.,
Anill
,
M.
,
Othm
an
C
.
,
As
m
ah,
H
.
S.
,
&
T
armiz
i,
I
.
Spa
ti
a
l
and
T
e
m
pora
l
Pat
te
rn
o
f
Road
Acc
id
ents
and
Casua
lt
i
es
i
n
Peninsula
r
Ma
lay
s
ia.
Jur
nal
T
e
knol
ogi
,
2015
.
7
6(14),
57
–
65
.
[2]
Kare
em,
A.
R
e
vie
w
of
Glob
al
Mena
ce
of
R
oad
Acc
ide
nts
with
Spe
cial
R
efe
re
n
ce
to
Ma
lay
s
ia
-
A
Soc
ial
Perspec
ti
v
e.
The
Malay
sian
Jour
nal
of
Me
di
cal S
ci
en
ce
s
,
2003.
1
0(2).
[3]
Prasanna
kum
ar
,
V
.
,
V
ij
i
th,
H.
,
C
har
utha,
R.
,
&
Gee
tha,
N.
Spat
io
-
T
emporal
Clu
steri
ng
of
Road
Acc
ide
n
ts :
GIS
Based
Ana
l
y
s
is
and
As
sess
m
ent
.
Pr
oce
dia
-
So
cial
and
Be
ha
vi
or
al
Sc
ie
nc
es
,
2
01
1.
21
,
317
–
325
.
[4]
T
ruong,
L
.
,
&
Som
ena
hal
li,
S.
Us
ing
GIS
to
ide
nti
f
y
ped
estrian
-
vehicle
cr
ash
hotspots
and
u
nsafe
bu
s
stops.
Journal
of
Publi
c
T
rans
portati
on
,
201
1.
14(1), 99
–
1
14.
Evaluation Warning : The document was created with Spire.PDF for Python.
Ind
on
esi
a
n
J
E
le
c
En
g
&
Co
m
p
Sci
IS
S
N:
25
02
-
4752
Determi
ning
H
otspots
of R
oad
Acci
den
ts
Us
ing
Spatial
An
alysis
(
S.S
ar
if
ah R
adia
h Sh
ar
i
ff
)
151
[5]
Xie,
K
.
,
Ozba
y
,
K.,
&
Y
ang,
H.
Spati
al
an
aly
sis
of
highwa
y
inci
dent
dur
at
ions
i
n
the
con
te
xt
of
Hurric
ane
Sand
y
.
Acc
ide
n
t
Ana
l
y
s
is
&
Prev
ent
ion
,
2015.
74
,
77
–
86
.
[6]
Karpa
gavalli.
S
,
&
B
alam
uruga
n,
V.
Onlin
e
Mar
i
ne
Sat
el
l
it
e
S
y
s
t
em
Us
ing
GIS
.
I
ndonesian
Journ
al
of
Elec
tric
al
Engi
ne
ering
and
Computer
Sc
ie
n
ce
,
2017.
8(2).
[7]
Chai
ne
y
,
S.
,
T
o
m
pson,
L.
,
&
U
hli
ng,
S.
Th
e
Ut
il
ity
of
Ho
tspot
Mapping
for
Pre
dic
ti
ng
Spat
ia
l
P
at
t
ern
s
of
Cr
ime.
Sec
urit
y
Journal
,
2008
.
21(2)
.
[8]
Le
vin
e,
N.
Crim
eSt
at
III:
A
Spat
ial
Statis
ti
cs
Pro
gram
for
the
Anal
ysis
of
Crime
Inc
ide
n
t
Locati
o
ns
(ve
rsion
3.
3).
2010,
Ned
Le
v
in
e
&
As
socia
t
es: H
ouston,
TX/
Nati
on
al
Inst
it
ut
e of
Jus
ti
ce:
W
ash
ingt
on,
DC.
[9]
MM
Othm
an
&
Mus
iri
n,
I.
A
novel
appr
o
ac
h
to
determ
ine
t
ra
nsm
ission
re
liabil
i
t
y
m
arg
in
using
par
ametr
i
c
bootstra
p
te
chn
i
que.
I
n
te
rnat
ion
al
Journal
of
Ele
ct
rical P
ow
er
&
Ene
rgy
S
yste
ms
,
2011.
33
(10), 1
666
-
1674
.
[10]
Aw
al
in,
L.
J
.
,
H.
Mokhlis,
M.
K.
Rahmat
,
Shi
lpa,
S,
Alba
tsh,
F.
&
Ism
a
il,
B
.
Fa
ult
Dist
ance
Id
e
nti
ficat
ion
Us
in
g
Im
peda
nce
and
Matc
hing
Appro
ac
hes
on
Distrib
uti
on
Network.
I
ndonesian
Journal
o
f
Elec
tri
cal
Engi
ne
ering
an
d
Computer
Scien
ce
,
2017.
8(3).
Evaluation Warning : The document was created with Spire.PDF for Python.