TELKOM
NIKA
, Vol. 11, No. 10, Octobe
r 2013, pp. 5
797 ~ 5
805
ISSN: 2302-4
046
5797
Re
cei
v
ed Fe
brua
ry 26, 20
13; Re
vised Ju
ly 3, 201
3; Accepted
Jul
y
16, 2013
Evaluation Studies on Client Satisfaction Degree of
Railway Statistic Information System
Hua
w
e
n
W
u
*
1
, Xingjun Sh
i
2
, Cheny
a
ng Duan
3
, Fuzh
a
ng Wa
ng
1
1
Institute of Electronic Com
put
ing T
e
chnol
og
y, China Aca
d
e
m
y
of Rai
l
w
a
y
Scienc
e
Beiji
ng 1
0
0
081
, Beijin
g, Chin
a
2
School of Eco
nom
y
& T
r
ade, Z
hejia
ng Ind
u
s
tr
y
& T
r
ade Vocatio
nal C
o
ll
e
g
e
W
enzho
u 32
50
03, Z
heji
a
n
g
, Chin
a
3
Compa
n
y
Fiv
e
of Cadet Brig
a
de, T
h
ird Milita
r
y
Med
i
cal U
n
i
v
ersit
y
Cho
ngq
in
g 40
0
038, Ch
on
gqi
n
g
, Chin
a
*Corres
p
o
ndi
n
g
author, e-ma
i
l
:
w
h
w
1
9
8
3
@
g
m
ail.com
A
b
st
r
a
ct
T
o
incre
a
se t
he acc
u
racy
and th
e effic
a
cy of c
lie
nt s
a
tisfaction
de
g
r
ee of r
a
ilw
ay
statistic
infor
m
ati
on sy
stem, w
e
giv
e
an AH
P-bas
e
d
co
mpr
e
h
ensi
v
e assess
men
t
about s
a
tisfa
c
tion d
egr
ee
of
infor
m
ati
on sys
tem, ho
pin
g
to
solve pr
obl
e
m
s
about ev
alu
a
ti
on difficu
lties o
f
multi-i
n
d
e
x, mu
lti-criteri
a
a
n
d
mu
lti-lev
e
l. Si
n
c
e the c
onve
n
t
i
on
al AHP-
bas
ed
meth
od
is
affected by s
u
bjectiv
e
factors
,
w
e
devel
op
an
enh
anc
ed AHP
meth
od to dec
rease l
i
mitatio
n
s
of convent
io
n
a
l metho
d
s. Our metho
d
, still
base
d
on ex
pe
rt
scorin
g
, perfor
m
c
l
uster
an
al
ysis of scor
i
n
g
data,
app
ly
t
he cl
usteri
ng
meth
od
of Eu
clid
Dista
nce
w
i
th
W
e
ight to eli
m
i
nate scores w
i
t
h
the lar
gest di
verge
n
ce
, an
d utili
z
e
th
e AHP
meth
od a
nd F
unctio
n
of W
e
i
ght
Averag
e to obt
ain w
e
ig
ht of evalu
a
tion i
n
d
e
x
,
w
h
ich
is useful to improv
e the accur
a
cy an
d efficacy and
can
e
n
h
a
n
c
e
e
ffe
cts o
f
th
e
mo
re p
i
vo
tal
e
v
al
ua
ti
o
n
i
n
de
x on re
su
l
t
s. Fi
n
a
lly, we
p
r
o
v
e
i
t
s ra
ti
o
n
a
l
i
t
y and
reliability in an
eval
uation of client satisfac
tion degr
ee of railway stat
istic information system
.
Ke
y
w
ords
: Ra
ilw
ay Statistical
Informati
on Sy
stem, AH
P, Users’
Satisfacti
on, Cluster Analysis
Copy
right
©
2013 Un
ive
r
sita
s Ah
mad
Dah
l
an
. All rig
h
t
s r
ese
rved
.
1. Introduc
tion
With a
d
van
c
ement of
rail
way info
rmati
onization, m
o
re
and
mo
re information
system
asso
ciated
with rail
way
statistics a
r
e
p
r
esent,
so h
o
w to
evalua
te these
syst
em an
d
clie
nt
satisfa
c
tion
d
egre
e
of the
m
are
hot for many rail
wa
y related
stu
d
ies. T
he
core of evaluatio
n of
client satisfa
c
tion de
gree
is determin
a
tion of
wei
g
hts of evalua
tion index. Here, we studi
ed
weig
hts of i
n
dex abo
ut rai
l
way stati
s
tic
informat
io
n system an
d a
nalyze
d
the
client sati
sfacti
on
degree.
A large amou
nts of method
s have bee
n use
d
by
rese
arche
r
s to investigate the
weig
ht of
index, includi
ng Analytic Hierarchy Proce
s
s(A
H
P), Fuzzy Synthetic Evaluation Method,
Data
Envelopme
n
t Analysis(DEA), Multi-
level
Extension Me
thod, BP Neu
t
ra
l Net
w
ork
etc. Wan
g
et al.
[1] took AHP method to an
alyze facto
r
s, determi
n
ed the hierarchi
c
al st
ru
cture o
f
index and the
judgme
n
t mat
r
ix, gave the
singl
e ra
nki
n
g wei
ght an
d
overall
ran
k
in
g wei
ght of th
e eleme
n
ts in
all
layers,
an
d a
nalyze
d
fa
cto
r
s qu
alitatively and
qu
antit
atively. Shao
et al. [2]
set up
a fa
cto
r
set,
establi
s
h
ed t
he evalu
a
tion
set an
d weig
ht set
ba
se
d
on the
Delp
hi
Method, d
e
termin
ed
subj
ect
function
s of e
v
ery factor, constructe
d th
e singl
e facto
r
judgi
ng mat
r
ix and finally
cond
ucte
d the
fuzzy comp
rehen
sive eva
l
uation. Zhu
et al.
[3] intr
odu
ced a im
proved
DEA method, wh
ich
avoided t
he
shortcomin
gs
of the tra
d
itio
nal
D
EA met
hod th
at coul
d not
determi
ne the
seque
nce
of efficient DMU. Shao et
al. [4] developed the mu
lti
-
level extensi
b
le evaluatio
n model
with the
extensio
n me
thod a
s
th
e
core, in
whi
c
h the
risk
de
gree
s
of the
line an
d its subsy
s
tems a
r
e
determi
ned
a
nd a
ne
w m
e
thod fo
r
co
mpre
hen
sive
evaluation
i
s
p
u
t forward
.
Wan
g
et al.
[5
]
applie
d an i
m
prove
d
BP Neutral Network
Metho
d
to evaluate in
dex, in whi
c
h
comp
re
hen
sive
evaluation
b
a
se
d on
S-T
y
pe Fun
c
tion
wa
s p
e
rf
o
r
med to
solve
pro
b
lem
s
of
co
nventional
BP
Neutral Net
w
ork
with long training time, high sensiti
v
ity to
initial weight,
li
ability to converge to
local mini
mum.
Here, based
on expert scoring, we pe
rform clu
s
ter
analysi
s
of scori
ng data,
apply the
clu
s
terin
g
m
e
thod
of Eu
clid
Di
stance
with
Weig
h
t
to eliminat
e sco
r
e
s
wi
th the la
rge
s
t
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 579
7 –
5805
5798
diverge
n
ce, combi
ne the
AHP metho
d
and Fu
ncti
on of Weig
ht Average to
obtain weight
of
evaluation i
n
dex, whi
c
h i
s
useful to i
m
prove t
he a
c
curacy an
d efficacy an
d can
enha
nce effe
cts
of pivotal evaluation ind
e
x on re
sults.
2. Ev
aluation Sy
stem on Client Satisfac
tion De
gree of
Railw
ay
Statisti
c Informatio
n
Sy
stem
With a
d
van
c
ement of
rail
way info
rmati
onization, m
o
re
and
mo
re information
system
asso
ciated
with rail
way
statistics a
r
e
p
r
esent,
so h
o
w to
evalua
te these
syst
em an
d
clie
nt
satisfa
c
tion
d
egre
e
of the
m
are
hot for many rail
wa
y related
stu
d
ies. T
he
core of evaluatio
n of
client satisfa
c
tion de
gree
is determin
a
tion of
wei
g
hts of evalua
tion index. Here, we studi
ed
weig
hts of i
n
dex abo
ut rai
l
way stati
s
tic
informat
io
n system an
d a
nalyze
d
the
client sati
sfacti
on
degree.
2.1. Subhea
d
ings Con
s
ti
tution o
f
Ra
il
w
a
y
Statistic
Information
Sy
stem
Rail
way statistic
i
n
form
ation system
i
s
g
u
id
e
d
by
co
mplete
a
nd inta
ct in
stitution,
sci
entific
an
d rational
in
dex, advan
ced inve
sti
gat
ion te
chn
o
lo
gy, rapi
d a
n
d p
r
ompt
d
a
ta
treatment, a
c
curacy
and
reliability of st
atistics info
rmation, leg
a
l
mana
gem
en
t, standa
rd
a
n
d
seq
uential
inf
r
ast
r
u
c
ture
a
nd sup
e
rio
r
consulting se
rvice. Deci
sio
n
supp
ort a
n
d
an
alysi
s
i
s
at the
core, and
co
nstru
c
tion
of informatio
n reso
urce d
a
ta
base and
re
alizin
g re
sou
r
ce
sh
arin
g
and
automation
of statistic
wo
rk is ai
m to p
r
o
v
ide high q
u
a
lity and efficie
n
t statistic
se
rvice to railway
reform, d
e
velopment an
d manag
eme
n
t [6].
Rail
way stati
s
tic sy
stem contain
s
all pa
rts of
rail
way
-rel
a
ted a
c
tivities, inclu
d
in
g Ca
rgo
Tran
sp
ort
a
t
i
o
n
s
u
b
-
sy
st
e
m
,
P
a
s
s
en
g
e
r T
r
a
n
s
port
a
t
i
on S
t
at
ist
i
cs
s
u
b
-
sy
st
e
m
,
Lo
comot
i
ve
Statistics Sub
-
sy
stem, Lu
g
gage
an
d Pa
rcel Stat
isti
cs Sub-system, Labo
r
Statisti
cs Sub
-
sy
ste
m
,
Investment i
n
fixed Asse
ts Statistics
sub
-
sy
stem,
Tran
sp
ortatio
n
Equipm
ent
Statistics S
ub-
system, Ene
r
gy Saving Statistics Sub-system,
Enviro
nmental Protection
S
t
at
ist
i
cs S
u
b-
sy
st
e
m
,
Integrated In
quire a
bout T
r
an
spo
r
tation
Statis
tics Sub
-
sy
stem, as shown in Figu
re 1.
Figure 1. Structure
s
of Railway
st
at
ist
i
c inf
o
rmat
io
n sy
st
em.
Rail
way st
atistic
system
is a
n
imp
o
rt
ant
pa
rt of
Rail
way intel
ligent tra
n
sp
ortation
manag
eme
n
t, and Rail
way
statistic info
rmationization
is based on
publi
c
platform con
s
tru
c
tio
n
,
whi
c
h con
s
ist
of communi
cation network platform,
Rai
l
way statisti
c
informatio
n sharin
g platform,
informatio
n secu
rity assu
rance platform
and p
ubli
c
pl
atform of ne
w Railway st
atistic info
rm
ation
etc.
2.2. Ev
aluation Index
of Client S
a
ti
sfac
ti
on
De
gree o
f
Rai
l
w
a
y
Statis
tic Informa
t
io
n
Sy
stem
The
evaluati
on in
dex of
rail
way
stat
istic i
n
form
ation system
reflect
s
situa
t
ions of
transpo
rtation
,
prod
uctio
n
,
and m
anip
u
l
a
tion in
th
e railway, in
clud
ing ma
ny aspect
s
, such
as
passe
nge
r transportatio
n
, carg
o tran
spo
r
tation,
pass
enger
c
a
rs
, freight cars
, loc
o
motive,
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Evaluatio
n Studie
s
on Clie
nt Satisfactio
n
Deg
r
ee of
Rail
wa
y Statistic … (Hu
a
wen Wu
)
5799
disp
atchi
ng, tran
spo
r
tation
se
cu
rity and
ope
ration
in
come
et
c.
A
c
cor
d
ing
t
o
t
h
ese
f
a
ct
o
r
s,
we
con
s
tru
c
ted
Evaluation In
dex of Cli
ent
Satisfactio
n
Deg
r
e
e
. Th
ese fa
cto
r
sets a
r
e
sho
w
n in
Table 1.
Table 1. Evaluation Indi
ce
s of
Client Sa
tisfaction
Deg
r
ee
2.3. Scoring Standa
rd
Usi
ng
scores
perfo
rmed
by
expe
rts to
e
s
tab
lish
a pairwise comp
ari
s
on
matrix U=[uij],
in
whi
c
h the imp
o
rtan
ce of the
compa
r
ative
values of
ui to uj is rep
r
e
s
ented by uij, and i and j is
the
index de
scrib
ed above. Scoring
st
and
ard is sh
own in table 2.
Table 2. Sco
r
ing Standa
rd
Score Meaning
1 Equal
importance
3 Moderate
impo
rt
ance
5 Strong
importa
nce
7 Ver
y
Imp
o
rtance
9 Extreme
I
m
porta
nce
2,4,6,8
Compromise
Value
Inverse uij=1/uji
3. Fuzzy
Comprehensiv
e Ev
aluation Based o
n
Cl
uster
Analy
s
is
With the
ad
vancem
ent i
n
networkin
g and
multi
m
edia te
ch
n
o
logie
s
e
n
a
b
les th
e
distrib
u
tion
a
nd sha
r
ing
of multimedi
a co
ntent
wi
dely. In the
meantime,
pira
cy be
co
mes
increa
singly
ramp
ant as
the cu
stome
r
s ca
n eas
ily
duplicate a
nd redi
strib
u
te the received
multimedia
content to
a
larg
e a
udie
n
ce. In
su
ri
ng
the copyri
g
h
ted
multim
e
d
ia conte
n
t is
approp
riately use
d
ha
s be
come incre
a
si
ngly critical.
3.1. Normalizatio
n of Sc
ores
Since the ei
genvalu
e
of xi(xi
∈
X) is not in the clo
s
e interval [
0
,1], it is essential to
norm
a
lize ra
w data [7]. as sho
w
n in eq
uation
s
1-4:
Average Valu
e
x
j
:
1
1
n
x
x
j
ij
n
i
; (1)
S
y
stematic layer
Element la
y
e
r
Index la
yer
Evaluation Index
S
y
stem of
Client Satisfactio
n
Degree
S
y
stem Cont
rolling Element
1
U
Functionality
11
r
Realiability
12
r
Pr
acticability
13
r
Efficiency
14
r
Staff Controlling
Element
2
U
Proficiency
21
r
O
ccupational Ca
pability
22
r
Cooperation Cap
ability
23
r
Consciousness
24
r
Environmental Equipment Elements
3
U
Equipment
31
r
Net
w
ork Band
w
i
dth and
Environment
32
r
Working Environment
33
r
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 579
7 –
5805
5800
Standard dev
iation
S
j
:
1
21
/
2
((
)
)
1
n
Sx
x
j
ji
j
n
i
; (2)
Standard value of raw d
a
ta:
()
'
x
x
j
ij
x
ij
S
j
; (3)
Extreme value stand
ardi
za
tion:
''
mi
n
(
)
''
''
ma
x
(
)
m
i
n
(
)
xx
ij
i
j
x
ij
x
x
ij
ij
; (4)
Whe
n
''
mi
n
(
)
x
x
ij
i
j
, x=
0; when
''
max
(
)
x
x
ij
i
j
, x=1. So sta
nda
rdi
z
a
t
ion value
s
of
ra
w d
a
ta is
s
i
tuated at [0,1].
3.2. Cluste
r Analy
s
is
Here, cl
uste
r analysi
s
i
s
based o
n
weighted E
u
cli
d
dista
n
ce, i
n
whi
c
h
the
relative
distan
ce
but
not ab
sol
u
te
distan
ce
is
m
o
re
ac
cu
ratel
y
res
pon
se
to data
di
strib
u
tion, when
we
have no any
domain
kno
w
l
edge a
bout th
e data obje
c
t
s
[8].
1/
2
2
(,
)
1
p
dx
x
w
x
x
ij
ki
k
j
k
k
(5)
(Where
(1
,
2
,
,
)
wk
q
k
rep
r
ese
n
ts weight
values of variable
s
.)
Input: k--num
ber of clu
s
te
rs, dataset U containin
g
n d
a
ta.
Output: numb
e
rs of
clu
s
ters whi
c
h h
a
ve minimum vari
ance.
Steps:
(1) S
e
le
ct
k
s
a
mple
s f
r
om
n
sampl
e
s as initial cluster;
(2)Ba
s
e
d
on
the averag
e value of overall
samp
les, obtain
Euclid di
stan
ce of al
l
sampl
e
s,
an
d
clust
e
r all sa
mples f
r
om m
i
nimum w
e
igh
t
E
u
clid dist
a
n
ce.
(3)Cal
cul
a
te the avera
ge value of all sa
mples.
(4)Retu
r
e to (2) and
(3) u
n
til there is no
cha
nge of an
y cluste
r.
3.3. Weight
Determinatio
n
This pa
pe
r ap
plies AHP me
thod to establ
ish
a matrix[9
,10], as sho
w
n in equatio
n
s
6-1
2
:
11
12
1
21
22
2
12
aa
a
n
aa
a
n
A
aa
a
nn
nn
(6)
(Where
1/
(
)
aa
i
j
ij
ji
,
1(
)
ai
j
ij
and value is dete
r
mined a
c
cord
ing to Table 2
.
)
Sum-Pro
d
u
c
t Method in AHP:
(1)
Normali
z
a
t
ion of every line of all variable
s
in matri
x
()
*
Aa
ij
nn
:
(,
1
,
2
,
3
,
)
1
a
ij
ai
j
n
n
ij
a
ij
i
(7)
(2)Su
m
of every ro
w of all variable
s
in
matrix
()
*
Aa
ij
nn
after normaliz
a
tion:
(,
1
,
2
,
3
,
,
)
1
n
Wa
i
j
n
i
ij
j
(8)
(3)
Normali
z
a
t
ion of normal
i
zed
[,
,
,
,
]
12
3
WW
W
W
W
n
:
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Evaluatio
n Studie
s
on Clie
nt Satisfactio
n
Deg
r
ee of
Rail
wa
y Statistic … (Hu
a
wen Wu
)
5801
(1
,
2
,
3
,
,
)
1
W
i
Wi
n
n
W
i
j
(9)
(Where
[,
,
,
,
]
12
3
T
WW
W
W
W
n
are n
eede
d weig
ht
vectors.)
(4) Maxim
u
m cha
r
a
c
teri
ze
d
roots fro
m
Matrix Theory:
()
1
1
ma
x
11
n
aW
ij
j
AW
nn
j
i
nW
n
W
ii
ii
(10
)
(5) Con
s
i
s
ten
c
y
Test:
ma
x
1
n
CI
n
(11
)
CI
CR
R
I
(12
)
(Where CI is Con
s
i
s
ten
c
y Index, CR is
Con
s
i
s
ten
c
y Ratio, and RI
is Average
Ran
dom
Con
s
i
s
ten
c
y Index.)
4. Example
4.1. Scoring b
y
Experts
For
example,
we i
n
vited 6
e
x
perts to eval
uat
e railway statistic
info
rm
ation system,
whi
c
h
includes Systematic
Controlling Layer containing
Functionality, Reli
ability, Practi
cability
and
Efficiency, Staff Controlli
n
g
Layer
co
ntaining Pr
ofici
ency, O
c
cup
a
tional
capa
b
ility, Coopera
t
ion
cap
ability an
d Con
sci
ou
sness, a
nd E
n
vironm
ental
Equipm
ent
Layer
contai
ning
equi
pm
ent,
netwo
rk
ban
d
w
idth a
nd en
vironme
n
t an
d wo
rki
ng En
vironme
n
t. As mentio
ned
above, pai
rwi
s
e
comp
ari
s
o
n
i
s
u
s
ed
to
score[11,12].Fi
r
st, Elem
e
n
t layers were
scored,
whi
c
h i
s
sho
w
n
in
Table 3.
Table 3. Sco
r
es of Element
Layers
Expert
Pairw
i
se Com
par
ison Among 3 Element La
yers
U1-U2
U1-U3
U2-U3
1 3
3
2
2 4
3
2
3 4
4
2
4 5
2
3
5 2
3
2
6 4
3
2
Average
3.7
3.0
2.2
4.2. Data
Clu
s
tering
Firstly, we
made
a
clu
s
tering
analy
s
is
of
Syste
m
atic
Co
ntro
lling Elem
ent
s
U1
to
eliminate
the
value
with th
e la
rge
s
t d
e
viation a
pplyin
g
Eu
clid
Di
stance T
heo
ry. Since p
r
o
c
e
s
se
s
of Fuzzy Cl
u
s
terin
g
Meth
o
d
are compli
cated an
d
the comp
utation power
h
uge, so we relie
d on
SPSS17.0 to treat matrix
through a
computer
aided way according to
steps mentioned
above[13].
Figure 2 sh
o
w
ed
results o
f
data treatm
ent and
we can kn
ow that
all 6 sampl
e
s are in
the clu
s
teri
ng
analysi
s
. Fig
u
re 3
sho
w
e
d
step
s of
treatment an
d we can find t
hat relatio
n
sh
ip
betwe
en clu
s
ters vari
ed, a
nd co
efficient
was la
rg
e
r
d
u
ring p
r
o
c
ee
ding of clu
s
tering, indi
cati
ng
their
co
rrel
a
tion b
e
co
me l
e
ss an
d deviat
i
on be
co
me
l
a
rge. In
Figu
re 3, the l
e
ft h
a
lf part
sh
owed
the expe
rts
were
clu
s
tere
d in eve
r
y st
ep, and
we
coul
d find th
e expe
rt 2 a
nd 5
we
re firstly
clu
s
tere
d. Th
e right half pa
rt sho
w
ed the
firstly cluste
red step n
u
mb
er.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 579
7 –
5805
5802
Treatme
nt Collectiona,b
Example
Efficient
Defficient
Total
N
percent
N
percent
N
percent
6
100.0
0
.0
6
100.0
a. Euclidean Dist
ance had been u
s
ed
b. average link(b
e
tween g
r
oups)
Figure 2. Dat
a
Treatm
ent Colle
ction
Rank
Cluster Gr
oup
Coefficient
First Cluster
Next R
ank
Cluster 1
Cluster 2
Cluster 1
Cluster 2
1
2
5
1.000
0
0
4
2
1
4
1.000
0
0
3
3
1
3
1.584
2
0
4
4
1
2
2.183
3
1
5
5
1
6
3.742
4
0
0
Figure 3. Steps of Fu
zzy T
r
eatme
nt
From ab
ove analysi
s
, we
coul
d kn
ow t
hat score
s
b
y
the expert 6 sho
w
e
d
the large
s
t
deviation, wh
ich m
a
y be t
he results of
subj
ective
fa
ctors, so sco
r
es from
him
we
re elimi
n
a
t
ed
from ove
r
all
data. With
a
simila
r meth
o
d
, re
sults
fro
m
the expe
rt
4 we
re
also e
liminated.
Du
ring
the clu
s
terin
g
analysi
s
of Staff Controllin
g Elem
ent U2 and Enviro
nment Equip
m
ent U3, sco
r
es
from the exp
e
rt 4 and
5 showed the la
rge
s
t devia
tio
n
and elimi
n
ated re
sp
ecti
vely. After these
treatment
s,
n
e
w scori
ng
ta
bles we
re obt
ained,
a
nd T
able 4
sh
owe
d
Element
La
yer sco
r
ing
a
n
d
Index Layer scori
ng respe
c
tively.
Table 4. Sco
r
es of Index L
a
yers
expe
rts U1
r11-r
1
2
r11-r
1
3
r11-r
1
4
r12-r
1
3
r12-r
1
4
r13-r
1
4
1 6
4
3
1
1/2
1/4
2 5
5
4
2
1/3
1/3
3 6
5
2
1
1/3
1/3
4 7
4
3
1
1/2
1/4
5 6
5
4
2
1/3
1/3
Average
6.0
4.6 3.2 1.4
0.4
0.3
expe
rts U2
r21-r
2
2
r21-r
2
2
r21-r
2
2
r21-r
2
2
r21-r
2
2
r21-r
2
2
1
1/4
2
2
5 5 1
2
1/3
3
1
4 6 1
3
1/3
2
2
5 4 2
4
1/4
3
2
4 5 1
5
1/3
3
1
4 5 2
Average
0.3
2.6 1.6 4.4
5
1.4
expe
rts
U3
r31-r
3
2
r31-r
3
2
r31-r
3
2
1 2
2
2/5
2 2
3
1/2
3 3
4
3/4
4 3
4
1/2
5 2
2
1/2
Average
2.4
3
0.5
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Evaluatio
n Studie
s
on Clie
nt Satisfactio
n
Deg
r
ee of
Rail
wa
y Statistic … (Hu
a
wen Wu
)
5803
4.3. Compre
hensiv
e Fuzz
y
Analy
s
is
After analy
s
is with A
H
P-ba
sed
metho
d
as
de
scribe
d above, weig
h
t
s
of U1, U2
and U3
were obtain
e
d
. RI is determined by large amou
nt
s of experimen
t data, which
were sho
w
n
in
Table 5.
Table 5. Valu
e of RI
n
1 2
3 4
5 6
RI
0.00 0.00
0.58 0.90
1.12 1.24
n
7
8
9 10
11 7
RI
1.32 1.41
1.45 1.49
1.51 1.32
All weight dat
a after scori
n
g:
(1)
Weig
ht of Elements Layer U1,
U
2 an
d U3
is
T
A
=
[
0
.
604
,
0
.
2
68
,
0
.
1
28]
,
λ
=3
.
0
8
ma
x
;
Cons
is
tenc
y Tes
t:
CI
=
0
.
0
4
,
CR
=
0
.069
.
(2 )Weig
h
t of Index Layer in Elements Layer
U1, inclu
d
ing F
u
n
c
tionality, Re
liability,
Practi
cability
and Efficiency, were
T
W
=
[
0
.
5
89,
0.112,
0
.
08
2,
0.217]
U1
res
p
ec
tively,
λ
=4
.
0
8
u1
m
a
x
;
Cons
is
tenc
y Tes
t:
CI
=
0
.
0
2
7
u1
,
CR
=
0
.030
u1
(3) Weight
of
Index
Layer in Ele
m
ent
s Laye
r
U2, i
n
clu
d
ing
Prof
icien
c
y, O
c
cu
pational
cap
ability, Coope
ration
capability and
Con
sci
ou
sn
ess, we
re
T
W
=
[0.
209
,
0
.5
60,
0.
12
6,
0.
105
]
U2
r
e
spec
tively,
λ
=4
.
1
4
u2m
ax
; Cons
is
tenc
y
Tes
t:
CI
=
0
.
047
u2
,
CR
=
0
.
052
u2
(4)
Weig
ht of Index Lay
er in Elem
e
n
ts
Laye
r
U3, inclu
d
ing
Equipme
n
t, Network
Bandwi
d
th a
nd Enviro
n
m
ent and
Wo
rkin
g En
vironme
n
t were
T
W
=
[
0
.5
91,
0.178
,
0
.2
31]
U3
r
e
spec
tively,
λ
=3
.
0
8
u3
ma
x
; Cons
is
tenc
y
Tes
t:
CI
=
0
.
0
4
u3
,
CR
=
0
.
0
6
9
u3
.
After eliminating scores by
exper
ts
with the larg
est de
viation:
(1) eig
h
t of Elements La
yer U1,U2 and U3 is
T
A
=
[0.609,
0.259,
0.132]
,
λ
=3
.
0
6
max
;
Cons
is
tenc
y Tes
t:
CI
=
0
.03
,
CR
=
0
.
0
5
2
。
(2)
Wei
ght o
f
Index Laye
r
in Elem
ent
s Laye
r
U1, inclu
d
ing
F
u
nction
ality,
Reliability,
Practi
cability
and Effici
ency,
were
T
W
=
[0.
5
6
9
,
0
.1
03
,
0
.0
89
,
0
.
2
3
9
]
U1
respec
tively,
λ
=4
.
0
7
u1m
a
x
;
Cons
is
tenc
y Tes
t:
CI
=
0
.
0
23
u1
,
CR
=
0
.
0
2
6
u1
(3) Weight
of
Index
Layer in Ele
m
ent
s Laye
r
U2, i
n
clu
d
ing
Prof
icien
c
y, O
c
cu
pational
cap
ability, Coope
ration
capabilit
y an
d Con
s
ciou
sness, we
re
T
W
=
[0.20
9
,
0
.5
67,
0
.
118,
0.10
6]
U2
r
e
spec
tively,
λ
=4
.
0
8
u2m
a
x
; Cons
is
tenc
y
Tes
t:
CI
=
0
.
027
u2
,
C
R
=
0
.
030
u2
(4)
Wei
ght
of Index La
yer in Elem
ents
L
a
yer
U3, in
cludin
g
Equipm
ent
, Netwo
r
k
Bandwi
d
th a
nd Environm
ent and Working Environ
m
ent we
re
T
W
=
[
0
.
574,
0.
184,
0.
242]
U3
resp
ectiv
e
ly
,
λ
=3
.
0
1
u3
m
a
x
, Cons
is
tenc
y
Tes
t:
C
I
=
0
.005
u3
,
C
R
=
0
.
0086
u3
.
4.4. Consis
tenc
y
Test of Weigh
t
CI wa
s i
ndex
of con
s
i
s
ten
cy, and CI
=0 i
ndica
ted full
con
s
i
s
ten
c
y. Whe
n
ap
proa
chin
g to
0, CI
sh
owed
more
sati
sfie
d con
s
iste
ncy
,
and, i
n
cont
rast,
la
rge
r
CI indicated
l
o
wer co
nsi
s
ten
c
y.
Comp
ari
s
o
n
of data from this exampl
e wa
s sh
owed
here:
Weig
ht of Element Layer:
CIBefore
=0.0
4
＞＞
；
CIafter=0.
03
0
Weig
ht of Systemati
c
Cont
rolling Elem
e
n
tsU1: CIU1B
e
fore
=0.02
7
＞＞
CIU1
after=0.023
0;
Weig
ht of Systemati
c
Cont
rolling Elem
e
n
tsU2: CIU2B
e
fore
=0.04
7
＞＞
CIU2
after=0.027
0;
Weig
ht of Systemati
c
Cont
rolling Elem
e
n
tsU3: CIU3B
e
fore
=0.04
＞＞
CIU3
after=0.005
0;
Here, CI Bef
o
re i
ndi
cated
weig
ht of al
l data b
e
fore
clu
s
teri
ng
a
nalysi
s
an
d
CIafter
indicated
wei
ght of d
a
ta
after cl
uste
ri
ng a
nalysi
s
f
o
llowin
g
eli
m
inating
data
with the
larg
est
deviation. F
r
o
m
the
s
e
re
sul
t
s we
coul
d
know that
elim
inating
data
with the
large
s
t deviatio
n
can
improve con
s
isten
c
y.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 10, Octobe
r 2013 : 579
7 –
5805
5804
Whe
n
CR<0.1, weight of matrix is ad
mitted,
and satisfactio
n
co
nsi
s
ten
c
y is accepte
d
.
Comp
ari
s
o
n
of data from this exampl
e wa
s sh
owed
here:
Weig
ht of Element Layer:
CRafter=0.05
2
＜＜
CRbefo
r
e =0.06
9
0.1
Weig
ht of Systematic
Controlling
ElementsU1:
CRu1after =0.02
6
＜
CRu1 befo
r
e
=0.
0
3
0
＜
0.1
Weig
ht of Systematic
Controlling
ElementsU2:
CRu2after =0.03
0
＜
CRu2 befo
r
e
=0.
0
5
2
＜
0.1
Weig
ht of S
y
stematic
Contro
llin
g El
ementsU3:
CRu3after =0.0086
＜
CRu
3
befo
r
e
=0.
0
6
9
＜
0.1
After test with
CI and
CR,
we
can find t
hat
eliminatin
g scores
with
large
s
t devia
tion and
cal
c
ulatin
g the weig
ht woul
d impr
ove sat
i
sfactio
n
co
nsisten
c
y.
4.5. Consis
tenc
y
Test of Weigh
t
Integration
o
f
above
wei
ght value
s
after cl
uste
ri
ng, applyin
g
weig
hted a
v
erage
comp
re
hen
si
ve function, we coul
d obtai
n weig
hts of all indices, a
s
sho
w
ed in T
able 6.
Table 6. Inde
x of Railway Statistic Information Syste
m
U1
Functionality 0.347
Reliability
0.063
Pr
acticability
0.054
Efficiency 0.146
U2
Proficiency 0.054
O
ccupational Ca
pability
0.147
Cooperation Cap
ability
0.031
Consciousness 0.027
U3
Equipment 0.076
Net
w
ork Band
w
i
dth and Environ
m
ent
0.024
Working Environment
0.032
5. Conclusio
n
This
study
ap
plies Clu
s
te
ri
ng Analy
s
is to AHP M
e
tho
d
, analy
z
e
s
the
client
sati
sfactio
n
degree of rail
way statisti
c i
n
formatio
n sy
stem, and a
s
se
sses the
weight of all in
dice
s, re
sulti
n
g
into some m
eanin
g
ful dat
a.
Usi
ng
Clu
s
te
ring An
alysi
s
-ba
s
ed
AHP
method
i
s
o
b
jective
a
nd sci
entific
to evaluate
indices of rail
way
statistic i
n
formatio
n
system,
wh
ich i
s
u
s
eful
to im
prove
availab
l
e metho
d
s a
nd
decrea
s
e
problem
s of di
rectly
sco
r
in
g by experts. In addition, this method
increa
se
s the
accuracy a
n
d
validity of data evaluation,
redu
cin
g
the difficulty and
raisi
ng the eff
i
cien
cy. Similar
method
s ca
n also b
e
used
in other field
s
and have go
od appli
c
atio
n value.
Ackn
o
w
l
e
dg
ements
This work
wa
s
sup
porte
d
by scien
c
e
a
nd te
chn
o
log
y
re
sea
r
ch to
pic
of China
Rail
way
bure
au (
20
07X
008
-G,
2
008X
015
-H
)
.
Referen
ces
[1]
T
homas L Saa
t
y
.
Makin
g
an
d
Valid
ating
Co
mple
x D
e
cisi
o
n
s
w
i
t
h
the AH
P/ANP.
Journa
l of Systems
Scienc
e an
d Systems En
gin
e
e
rin
g
. 200
5; 29
(1): 1-36.
[2]
Lin S
h
a
o
, Bol
i
an
g L
i
n. Env
i
ronme
n
t Impa
ct
Assessment of Desert
Rail
w
a
y
B
a
sed on Fuz
z
y
Compre
he
nsiv
e Evalu
a
tio
n
.
Journ
a
l of the C
h
in
a Rai
l
w
a
y Society
. 200
9; 3
1
(10): 84-
89.
[3]
Ying
xi Z
hu,
Xu
emei
Xia
o
, W
e
i
Li
etc. R
e
sear
ch
o
n
P
e
rform
ance
Eva
l
uati
o
n
of Com
p
il
atio
n
of Ra
il
w
a
y
Emerge
nc
y
P
l
an Bas
ed o
n
Improve
d
DEA
Method.
Jo
urn
a
l of the ch
ina
railw
ay Soci
ety
. 2011; 3
3
(4):
1-6.
[4]
Ke Pa
n, Ho
ng
de W
a
n
g
, Ji
an
yu
n S
h
i. Ap
pli
c
ati
on
of Multi-
level
E
x
tens
ibl
e
Meth
od to
U
r
ban S
u
b
w
a
y
Operatio
n Safe
t
y
Eval
uatio
n.
Journ
a
l of the c
h
in
a railw
ay So
ciety
. 2010; 3
3
(
5):14-1
9
.
[5]
Yanh
ui W
a
ng,
Qian S
un, Ji
ng
Li
etc. Res
ear
ch o
n
Safet
y
E
v
alu
a
tion
of R
a
il F
r
ei
ght V
ehic
l
e Op
eratin
g
Status & Its Ap
pliaction
. Jo
urn
a
l of the Chi
na
Railw
ay Soc
i
et
y
. 2011; 33(
4): 38-4
5
.
[6]
Qingfu Li
u. Rai
l
w
a
y St
atistic information layout.
Chin
ese R
a
ilw
ay
. 200
6; 1
8
(8): 12-1
5
.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Evaluatio
n Studie
s
on Clie
nt Satisfactio
n
Deg
r
ee of
Rail
wa
y Statistic … (Hu
a
wen Wu
)
5805
[7]
Xi
amia
o Li, Yo
ng Sh
e, Xi
ao
li
Z
hu, et al. Appl
icatio
n of the F
u
zz
y
Cluster
in
g Anal
ys
is in th
e Plan
ni
ng o
f
Expr
ess F
r
eig
h
t
T
r
ansportatio
n
.
Chin
a Rai
l
w
a
y Scienc
e
. 20
05; 26(4): 1
24-
129.
[8]
Z
hong
lin
Z
h
a
n
g
, Z
h
i
y
u
Ca
o, Yua
n
tao
Li.
Rese
arch B
a
s
ed
on E
u
cli
d
Distanc
e
w
i
t
h
W
e
ights
of
K_
me
an
s Al
gori
t
h
m
.
Journa
l of Z
heng
z
h
o
u
Univers
i
ty (Eng
ine
e
rin
g
Scie
n
c
e)
. 2010; 3
1
(1
): 89-92.
[9]
Bang
ya
n Ma
o, Junmen
g W
u
, Hong
pin
g
Y
u
, et
al. Stud
y o
n
T
r
ack Alignm
ent in H
u
ang
gu
oshu
–
Beip
anj
ia
ng S
e
ction
of S
h
ang
hai-K
unm
in
g Pass
en
ger
De
dicate
d
L
i
ne.
Jo
urna
l
of R
a
ilw
ay
Engi
neer
in
g Socitey
. 201
1; 1
57(1
0
):19-2
4
.
[10]
T
homas L S
a
at
y
.
Decis
i
on
Makin
g
-T
he Anal
ytic
Hi
erarc
h
y a
n
d
Net
w
o
r
k Process
e
s
(AHP/ANP).
Journ
a
l of Systems Sci
enc
e a
nd Syste
m
s En
gin
eeri
n
g
. 20
0
4
; 57(1): 38
5-4
04.
[11]
Maomin
g F
u
. Appl
icatio
n of
F
u
zz
y
C
o
mp
rehe
ns
ive A
n
a
l
y
s
is to Infor
m
ation S
e
curit
y
Eva
l
u
a
tion.
Journ
a
l of Sout
hw
est Jiaoton
g
University
. 20
10; 45(3): 4
40-
444.
[12]
Aliev
RA, Guir
i
m
ov BG, F
a
zl
olla
hi
Bij
an, A
l
i
e
v R
R
. Evol
uti
onar
y
Alg
o
rith
m-base
d
L
ear
nin
g
of
F
u
zz
y
Neur
al N
e
t
w
or
ks, Part2:
Rec
u
rrent F
u
zz
y
N
eura
l
Netw
orks
. F
u
zz
y
Sets a
nd Syste
m
s
. 2
009; 1
60(
17):
255
3-25
66.
[13]
Ding
l
e
i
W
a
n
g
,
Jian
yo
ng
L
i
, W
ensh
e
n
g
Xu.
T
he M
e
tho
d
s of
F
a
ctor W
e
ig
ht’
s
Determ
ine
i
n
the Pr
ocess
of Cluster.
T
e
l
K
omnik
a
Indo
n
e
sia
n
Journ
a
l o
f
Electrical Eng
i
ne
erin
g
. 201
2; 10(5): 113
7-1
141.
Evaluation Warning : The document was created with Spire.PDF for Python.