TELKOM
NIKA
, Vol.11, No
.1, Janua
ry 2013, pp. 287
~29
5
ISSN: 2302-4
046
287
Re
cei
v
ed Se
ptem
ber 29, 2012; Revi
se
d De
cem
ber
1, 2012; Acce
pted De
cem
b
er 8, 2012
Index Selection
Preference and
Weighting for Uncertain
Network Sentiment Emergency
Qianshe
ng Z
h
ang
1
, Fuch
un Liu
2
, Bail
in Xie
3
,
Yiro
ng Huang
*
4
1,3
School of Informatics, Guan
gdo
ng U
n
ivers
i
t
y
of F
o
reig
n Studi
es, Guangz
hou 5
1
0
420, C
h
in
a
2
F
a
cult
y
of Co
mputer, Guan
g
don
g Univ
ersit
y
of T
e
chnolo
g
y
, Guan
gzh
ou
510
00
6, Chin
a
*4
Sun Yat-sen
Busin
e
ss Scho
ol, Sun Yat-se
n Univ
ersit
y
, G
uan
gzh
ou 51
0
275, Ch
in
a
*Corres
p
o
ndi
n
g
author, e-ma
il: huan
g
y
r@m
a
il.s
y
su.e
du.cn
A
b
st
r
a
ct
T
o
mor
e
effectively co
pe w
i
th
the netw
o
rk p
ublic
s
enti
m
ent
emerge
ncy i
n
volve
d
many
i
n
terval-
valu
ed w
a
rn
in
g
ind
e
xes, w
e
present
a n
e
w
method
of
e
a
rly w
a
rning
in
dex selecti
on and
w
e
ight
ass
i
gn
me
n
t
for unc
ertain
n
e
tw
ork pub
lic
senti
m
e
n
t e
m
e
r
gency
dec
isio
n-makin
g
. By
usin
g a
n
e
w
in
terval fu
zz
y
A
H
P
,
the interv
al w
e
ight of
each
e
a
rly w
a
rni
ng
in
dex of
netw
o
rk pub
lic s
enti
m
ent e
m
er
ge
ncy
can
be o
b
tai
n
ed.
T
hen by
mea
n
s
of the w
e
i
ght
ed a
ggr
egati
o
n
valu
es of
a
ll th
e e
m
er
ge
ncy w
a
rni
ng i
n
d
e
xes,
w
e
can ra
nk th
e
severity of ever
y netw
o
rk publ
i
c
senti
m
ent e
m
erge
ncy an
d s
e
lect the
most
severe o
ne. F
i
nally, a
nu
mer
i
cal
exa
m
p
l
e is giv
en to illustr
a
te
the appl
icatio
n of
the prop
osed
meth
od
of w
a
rning in
d
e
x selecti
ng a
n
d
w
e
ightin
g to un
certain n
e
tw
ork publ
ic senti
m
e
n
t emerge
ncy decisi
on.
Key
w
ords
:
Network pu
blic sentim
ent em
ergen
cy
, Wa
rnin
g
index,
Wei
ght assignm
ent,
Prefere
n
ce re
lation
Copyrig
h
t
©
2013
Univer
sitas Ahmad
Dahlan. All rights res
e
rv
ed.
1. Introduc
tion
Network sent
iment is the
public
opini
ons of
some
event with some i
n
fluen
ce an
d
stren
g
th. Re
cently, Network
se
ntimen
t analysis
a
nd early warning be
com
e
very import
ant
resea
r
ch
i
s
sues.
As
i
s
well kn
own, the
un
co
ntrol
l
ed n
e
two
r
k
sentime
n
ts e
a
sily in
cu
r t
he
emergen
cy.
Simultaneo
usly, emergen
cy will affect
netwo
rk p
ubli
c
se
ntiment. So, in order
to
decrea
s
e th
e risk of eme
r
g
ency ma
nag
e
m
ent and d
e
cision [1, 2], there i
s
mu
ch n
eed to an
alyze
and control th
e netwo
rk p
u
b
lic sentiment
effectivel
y.
In the above a
r
ea
s, Zeng [3
, 4] and Zhan
g
[5] propo
se
d
the method
s of sel
e
ctin
g sentim
ent indexe
s
and
determi
ning t
heir
weig
hts
for
netwo
rk
senti
m
ent emerge
ncy. Peng [6] and Zha
ng [7
] discu
ssed th
e clo
s
e rel
a
tionship bet
we
en
netwo
rk p
ubli
c
se
ntiment a
nd emergen
cy. Also
some
authors [8, 9] have propo
sed ma
ny early
warning
de
ci
sion
or al
arm
severity p
r
io
rity or
de
ring method
s
for netwo
rk eme
r
gen
cy.
Ho
we
ver,
the mo
st exi
s
ting related
e
m
erg
e
n
c
y de
cisi
on m
e
tho
d
s
and
ala
r
m
seve
rity ra
n
k
ing
me
cha
n
i
cs
can
only de
a
l
with the e
m
erge
ncy u
n
d
e
r p
r
e
c
ise co
ndition a
nd
certain
enviro
n
ment. Altho
ugh
Lin [10] propo
sed
a metho
d
for net
work
sentiment
ea
rl
y warni
ng, it
exce
ssively d
epen
ded
on t
h
e
sele
cted fu
zzy reaso
n
ing rules, and the
weight of
ne
twork se
ntim
ent index is not con
s
ide
r
e
d
.
Thus, this p
r
opo
sed ap
proach is incon
v
enient in
so
me ca
se
s an
d it can not deal with network
sentime
n
t em
erge
ncy
with
interval lin
gui
stic te
rm
s.
In fac
t, due to
the inc
r
eas
i
ng complexity
o
f
the so
cio-eco
nomic e
n
viro
nment and th
e lack of
kno
w
led
ge abo
ut the problem
domain, mo
st of
the real-wo
r
ld
problem
s, like netwo
rk p
u
b
lic se
nt
imen
t analysis an
d unce
r
tain d
e
ci
sion
-ma
k
in
g,
ar
e involved
var
i
ety of fuzz
iness
,
lik
e f
u
zz
y val
ue a
nd interval
va
lue. Esp
e
ciall
y
, in the pro
c
ess
of un
certai
n
netwo
rk em
erge
ncy
de
ci
sion
ma
king,
a de
ci
sion
make
r m
a
y
provide
hi
s/h
e
r
prefe
r
en
ce
s
over the alte
rnate eme
r
g
e
n
cie
s
with in
t
e
rval nu
mbe
r
s [11, 12] o
r
i
n
terval ling
u
i
s
tic
values [13] ra
ther than real
numbe
rs.
As we
kno
w
,
the unexpe
cted em
erg
e
n
cy gen
er
all
y
involves many publi
c
sentiment
factors, incl
u
d
ing the imp
o
rtan
ce of to
pic, t
he tide
of sentime
n
t, the attention
degree of to
pic,
and th
e p
opu
larity of topi
c,
as
well
as the
sp
readi
ng
sp
eed
of to
p
i
c. Also, the
values of a
b
o
v
e
emergen
cy in
fluence fact
ors a
r
e
ea
s
ily e
x
presse
d by i
n
terval n
u
mb
ers.
In thi
s
p
aper we aim
to
prop
ose an e
ffective method for dete
r
m
i
ning the ala
r
m
severity pri
o
rity orde
ring
of the uncert
a
in
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 287 – 2
9
5
288
netwo
rk senti
m
ent em
erg
e
n
cy, whic
h can
effici
ently con
d
u
c
t
the
emergen
cy u
r
gent
de
cisi
o
n
in
the uncertain
environ
ment.
We fi
rst
pro
p
o
se
some
ba
sic op
erato
r
s between
int
e
rval n
u
mb
ers in
sectio
n
2. And i
n
se
ction
3, we
introd
uce th
e ala
r
m
seve
rity prio
rity orderin
g p
r
o
c
e
s
s of
network
publi
c
sentim
ent
emergen
cy
with inte
rval
lingui
stic val
ue. In
sectio
n 4, on
e nu
mero
us
exa
m
ple i
s
give
n to
demon
strate the effectiven
ess
of the propo
sed n
e
twork
se
ntiment
emerg
e
n
c
y deci
s
io
n ma
king
approa
ch by usin
g the pro
posed interva
l
fuzzy fu
sio
n
operator a
nd
interv
al preference matrix.
Re
cently, ma
ny re
se
arche
r
s stu
d
ied
th
e ap
pro
a
che
s
of
index
a
nalysi
s
a
nd
weig
ht
assignm
ent. For exampl
e, Kahrama
n
e
t
al. [14]
employed fuzzy AHP in sup
p
lie
r sele
ction a
n
d
servi
c
e
qualit
y evaluation,
but the inte
rv
al fuzzy
ind
e
x sel
e
ction
an
d weight a
s
si
gnment
we
re
not
con
s
id
ere
d
. In fact, most of the existing fu
zzy index analysi
s
methods
ca
nnot effectively
determi
ne th
e weig
hts of interval fuzzy indexes.
Th
us, in this pa
per we try to propo
se a n
e
w
effective app
roa
c
h fo
r e
a
r
ly wa
rnin
g i
ndex se
le
ction an
d wei
ght assig
n
m
ent of net
work
sentime
n
t e
m
erg
e
n
c
y wi
th interval li
ngui
stic term
s, and th
en
deal
with th
e network p
ublic
sentime
n
t e
m
erg
e
n
c
y d
e
ci
sion
probl
em involved
interval ev
aluation val
u
e in u
n
cert
ain
environ
ment.
To the
end
, it will grea
tly fac
ilitate the net
wo
rk
publi
c
sentim
ent eme
r
ge
n
cy
manag
eme
n
t adopting th
e corre
s
p
o
n
d
ing de
cisi
o
n
strategy to
cope
with the most severe
alternate
net
work
publi
c
sentiment em
e
r
gen
cy a
c
cord
ing to the
severity ran
k
in
g re
sult of all
the
possibl
e network p
ubli
c
se
ntiment emergen
ce
s.
2. Preliminaries
Interval valued fuzzy set (IvFS) is a usef
ul ge
ne
ralizatio
n of the ordi
nary fuzzy set,
whi
c
h h
a
s be
en p
r
oved to
be m
o
re
suitable way for dealin
g
with
vaguen
ess
and u
n
certai
nty.
Particul
arly, the informatio
n entropy [15
], similari
ty measure a
nd di
stan
ce mea
s
ure [16] of IvFSs
play very imp
o
rtant role
s i
n
the a
pplication a
r
ea
s of
pattern
re
cog
n
ition, medi
cal diag
no
sis,
and
deci
s
io
n-m
a
ki
ng [12,17,18].
Defini
tion 1
. An interval fuzz
y s
e
t
A
in the universe
}
,
,
,
{
2
1
n
x
x
x
X
is define
d
as
,
{(
i
x
A
}
/
]
1
,
0
[
)]
(
),
(
[
)
(
X
x
x
A
x
A
x
A
i
i
i
i
, where
)
(
),
(
i
i
x
A
x
A
are
calle
d me
m
bership
de
gree
and no
n-m
e
mbershi
p
deg
ree of elem
en
t
i
x
to s
e
t
A
, res
p
ec
tively.
For sim
p
licit
y, in
this paper
we cal
l
]
,
[
~
a
a
a
to be an
interval number, if
1
0
a
a
.
Defini
tion 2
. Let
]
,
[
~
a
a
a
,
]
,
[
~
b
b
b
be two interval nu
mbers, so
me
basi
c
ope
ra
tions
betwe
en the
m
are defin
ed
as follows.
b
a
~
~
]
,
[
a
a
+
]
,
[
b
b
=
]
,
[
b
a
b
a
(1)
b
a
~
~
]
,
[
a
a
]
,
[
b
b
=
]
,
[
b
a
b
a
a
w
~
]
,
[
]
,
[
wa
wa
a
a
w
, if
0
w
.
b
a
~
/
~
]
,
[
a
a
]
,
/[
b
b
=
]
/
,
/
[
b
a
b
a
,
if
0
,
b
a
.
Defini
tion 3
.
Let
}
,
,
,
{
2
1
n
c
c
c
C
be the
early warni
n
g index set of netwo
rk sentime
n
t
emergen
cy, sup
p
o
s
e (
n
n
ij
g
)
is the pair-wi
se
compa
r
i
s
on
interval fuzzy prefe
r
en
ce relation
matrix con
s
tructed
by the
kn
owl
edge
of expert
s
,
where
ij
g
represe
n
ts the
interv
al preferen
c
e
degree of in
d
e
x
i
c
over ind
e
x
j
c
,
ij
ji
g
g
/
1
. The inte
rval weig
hts of
indexe
s
can
be given
by
following formula
n
i
n
ij
n
j
n
ij
n
j
i
g
g
w
1
/
1
1
/
1
1
)
(
)
(
,
for
n
j
i
,
,
2
,
1
,
. (2)
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Index Sele
cti
on Preferen
ce and Weighti
ng fo
r Un
ce
rt
ain Net
w
ork
… (Qia
nsh
e
n
g
Zhang
)
289
Defini
tion 4
. Let
]
,
[
~
a
a
a
,
]
,
[
~
b
b
b
be any two interval numbe
rs,
the prefere
n
c
e
degree of
a
~
ove
r
b
~
is defined a
s
}
0
},
0
,
)
(
)
(
max{
1
max{
)
~
~
(
b
b
a
a
a
b
b
a
P
, (3)
Notably,
1
)
~
~
(
)
~
~
(
a
b
P
b
a
P
, a
nd
5
.
0
)
~
~
(
a
a
P
.
Defini
tion 5
. Let
}
1
]
,
[
~
{
n
i
r
r
r
i
i
i
be an
interval
nu
mber sequ
e
n
ce,
by pai
r-wi
se
comp
ari
s
o
n
betwe
en i
n
te
rval nu
mbe
r
s we
con
s
tru
c
t
the preference relation
matrix
n
n
ij
P
)
(
,
whe
r
e
)
~
~
(
j
i
ij
r
r
P
p
. The non-d
o
min
ance degree of the
i
th interval numbe
r is d
e
fined a
s
}
1
{
min
*
ji
i
j
i
p
NDD
, (4)
whe
r
e
}
0
,
max{
*
ij
ji
ji
p
p
p
represents the de
gree to whi
c
h
i
r
~
strictly domin
ated by
j
r
~
.
Defini
tion 6.
The full
ran
k
ing
of all th
e
n
interval
nu
mbers a
r
e
d
e
termin
ed by
the a
s
cendi
ng
orde
r of
i
NDD
, and
the set of non-do
minate
d
alternate e
m
e
r
gen
cy is defi
ned a
s
}
{
max
/
{
j
j
i
i
ND
NDD
NDD
e
e
. (5)
3.
Early
Warning Index Sel
ection
and
Weigh
t
Assi
gnment
for
Net
w
o
r
k
Pu
blic Sentime
n
t
Emergenc
y
w
i
th
Interv
al Values
As
we
kn
o
w
, ma
ny ki
nds of in
de
xes p
r
o
babl
y incu
r
net
work publi
c
se
ntiment
emergen
cy.
Espe
cially in
the un
ce
rtain
eme
r
ge
nc
y
deci
s
io
n e
n
vironm
ent, the
accurate valu
e of
early wa
rnin
g
index inform
ation is difficult to meas
u
r
e. However, by interval linguisti
c
value,
we
can
conveni
e
n
tly com
pare
the p
r
eferen
ce de
gr
ee
bet
wee
n
two e
m
erge
ncy
inde
xes a
n
d
get t
h
e
interval fuzzy
preferen
ce relation on
ea
rly wa
rning i
ndex set. Th
rough the i
n
te
rval fuzzy AHP
analysi
s
meth
od, we can weight all the e
a
rly wa
rn
in
g indexe
s
of network sentim
ent emergen
cy.
Gene
rally, b
y
emergen
cy
man
agem
e
n
t expe
rt
qu
estion
naire
survey a
nd
statistical
analysi
s
fro
m
network p
ublic
se
ntime
n
t emergen
cy manage
me
nt we
can
e
a
sily get
so
me
importa
nt indexes which p
o
ssibly cau
s
e
the net
work
publi
c
sentim
ent emergen
cy. Also, thro
ugh
emergen
cy supervi
so
rs an
d sea
r
ch engi
nes, we c
an
obtain mu
ch i
n
formatio
n of network pu
bl
ic
sentime
n
t em
erge
ncy
wa
rn
ing ind
e
xes i
n
clu
d
ing
su
bj
ective an
d ob
jective ind
e
xe
s. For the
sa
ke
of dealing wi
th early warning and e
m
erge
ncy de
ci
sion ma
kin
g
, we firstly choo
se the finite
comp
re
hen
si
ve and hi
era
r
chi
c
al in
dexe
s
from
all the
possibl
e alte
rnate in
dexe
s
based o
n
th
e
well-esta
blish
ed pri
n
cipl
e
that each i
ndex
sh
ould
posse
ss i
n
depe
nden
cy
,
sen
s
itivity,
and
rep
r
e
s
entatio
n, as
well a
s
guida
nce qu
a
lity. Ther
efore, we n
eed to
sele
ct the
rel
a
tive importa
nt
early warni
n
g index with
highe
r scores. Gen
e
rall
y, after index early warn
ing analy
s
is
and
sele
ction, the
r
e are
still multi-level wa
rning index
e
s
that should b
e
taken into
accou
n
t. Usu
a
lly,
every network publi
c
senti
m
ent eme
r
ge
ncy mainly
compri
se
s the
following first-gra
de ind
e
xes:
netwo
rk pu
blic sentime
n
t emer
gen
cy
power i
nde
x, network
s
entiment inte
nsity index
and
emergen
cy coping
cap
a
cit
y
index.
Additionally, each first-g
r
a
de ea
rly warn
ing
ind
e
x also ha
s ma
ny seco
nd-grade
warnin
g
indexe
s
. In
gene
ral, network p
ubli
c
se
ntiment
emergen
cy powe
r
index briefly con
s
ist
s
of the
followin
g
se
cond-grade in
dexes
，
i
n
cl
ud
ing time duration, ext
ent of diffusion
, environme
n
t
disruption d
e
g
ree, severit
y
of econom
ic loss.
And network se
ntiment inten
s
ity index briefly
con
s
i
s
ts of th
e followi
ng
se
con
d
-g
ra
de in
dexes
，
i
n
cl
ud
ing sentime
n
t attention d
e
g
r
ee,
spreadi
n
g
spe
ed of
net
work
se
ntiment, emotion
tenden
cy,
b
ehavior tend
ency, auth
e
n
t
icity of network
publi
c
sentim
ent. The
gov
ernm
ent em
e
r
gen
cy
copi
n
g
capa
city bri
e
fly con
s
i
s
ts
of the follo
wi
ng
se
con
d
-g
ra
de
indexe
s
, in
cludi
ng resp
onse spee
d, informatio
n
tran
spa
r
en
cy, emerge
ncy
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 287 – 2
9
5
290
evacu
a
tion capa
city, emerge
ncy re
source allo
ca
tion capa
city, governmen
t respo
n
si
bili
ty,
peopl
e-cent
ra
lized d
egree, etc.
Notably, in
compl
e
x un
certain d
e
ci
si
on
environm
ent the ab
o
v
e-mentio
ned
netwo
rk
publi
c
se
ntiment emergen
cy early warning ind
e
xe
s are difficult
to measure
by preci
s
e
real
numbe
rs, inst
ead, they are
easily
a
s
sessed by eme
r
ge
ncy man
age
rs an
d rel
a
ted
field expert
s
i
n
terms
of interval linguisti
c
values, li
ke
strong
e
m
erge
ncy po
we
r, seriou
s e
c
o
n
o
m
ic lo
ss,
sev
e
re
environ
ment
disruption,
enormou
s
diffusion
de
gree,
wicke
d
behavio
r li
ability, unreli
able
sentime
n
t re
port, low
re
spo
n
se sp
ee
d, wea
k
em
erge
ncy eva
c
uatio
n ca
pa
city, incompl
e
te
emergen
cy re
scue facility, and so on.
More
over, th
e evalu
a
tion
value of
every al
ternate
n
e
twork publi
c
se
ntiment
e
m
erg
e
n
c
y
with multiple
early warni
n
g
indexe
s
are
easily ex
p
r
e
s
sed
by the in
terval fuzzy lingui
stic term
s
like extre
m
el
y strong, ve
ry str
ong,
stro
ng, mediu
m
, wea
k
, very weak, extre
m
e
l
y wea
k
rath
er
than by usin
g
accurate rea
l
numbe
rs. In
orde
r to
sim
p
lify the treatment of judg
ment expre
ssion
of network p
ublic sentim
ent warni
ng
index, a
uni
f
i
ed
set of i
n
terval ling
u
istic varia
b
le
s is
pred
etermi
ne
d in Table 1.
Table 1. Ling
uistic T
e
rm
s for Evaluating
Networ
k Sent
iment Emerg
ency with Interval Index
Linguistic terms
Interval numbers
Extremel
y Stron
g
(ES) / E
x
treme
l
y
High(EH
) /
Extremel
y Big (E
B)
[0.9, 1.0]
Very
very
stro
ng
(VVS) / Very
v
e
ry
high
(VVH) /
Very
very
Big (V
VB)
[0.8, 0.9]
Ver
y
St
rong (VS
)
/ Ver
y
high (VH
)
/ Ver
y
Big (
VB)
[0.7, 0.8]
Strong (S
) /
High (H)
/
Big (B)
[0.6, 0.7]
Medium (M)
[0.4, 0.6]
Weak (W) /
Lo
w
(L
) /
Tin
y
(T)
[0.3, 0.4]
Very
Weak (VW
)
/ Very
Lo
w
(V
L
)
/ Very tin
y
(VT)
[0.2, 0.3]
Very
very Weak(VVW) / Very very
L
o
w
(V
VL
) /
Very
very tin
y
(V VT)
[0.1, 0.2]
Extremel
y Weak
(EW) / Ext
r
emel
y Lo
w
(EL)
/ E
x
t
r
emel
y
tin
y
(ET
)
[0.0, 0.1]
Based
on
the
above
analy
s
is an
d the
p
r
eviou
s
fo
rmu
l
ae, we aim
to develo
p
a
n
interval
fuzzy A
H
P
approa
ch to
determi
ne t
he rational
weig
ht of warnin
g ind
e
x and the
n
make
emergen
cy d
e
ci
sion fo
r ne
twork pu
blic
sentime
n
t
em
erge
ncy invol
v
ed interval li
ngui
stic valu
es
in uncertain e
n
vironm
ent.
Step 1
. By
statisti
cal
que
stionn
aire
and th
e
score
s
assi
gned
by e
m
erg
e
n
cy
manag
eme
n
t experts, we
first con
s
tru
c
t all the in
terval fuzzy p
r
eferen
ce rel
a
tions ove
r
e
a
ch
warning index level. Let
n
n
ij
b
B
)
~
(
~
)
1
(
)
1
(
,
m
m
k
ij
k
b
B
)
~
(
~
)
2
(
)
2
(
rep
r
e
s
ent
int
e
rv
al
f
u
zzy
prefe
r
en
ce m
a
trixes of the
first-grade e
m
erg
e
n
c
y early warni
ng in
dexes, and th
e se
cond
-g
ra
de
indexe
s
of first-g
r
ad
e ind
e
x
k
c
, res
p
ec
tively, where
ij
b
~
take some interval values listed in Table
2. The
n
, by i
n
terval fu
zzy
AHP meth
od
and fo
rmul
a
(2)
we
can first assign
the
weig
ht vecto
r
to
each first
-
grade ind
e
x le
vel and to
each seco
nd
-grade in
dex
level. Ultimately, by using
multiplicatio
n
of the wei
ght
s of all th
e warnin
g in
d
e
xe
s of top
-
level
and its sub-l
e
vel, we o
b
ta
in
the overall weight of ea
ch
wa
rning
ind
e
x re
g
a
rdi
n
g
the net
work publi
c
senti
m
ent eme
r
g
e
n
cy
deci
s
io
n goal.
Table 2. Ling
uistic te
rms f
o
r Co
mpa
r
ing
t
he Importan
c
e Degree of Early Warnin
g Indexes
Intensit
y
of impor
tance
Defini
tion of grad
e
Interval degree
9
~
Extremel
y strong
importance
[8,10]
7
~
Ver
y
stro
ng impo
rtance
[6, 8]
5
~
Strong importa
nce
[5,6]
3
~
Moderate impo
rt
ance
[2,4]
2
~
Fair importance
[1,3]
1.5
Just Equal importance
[1,2]
1
~
Equal importance
[1,1]
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Index Sele
cti
on Preferen
ce and Weighti
ng fo
r Un
ce
rt
ain Net
w
ork
… (Qia
nsh
e
n
g
Zhang
)
291
Step 2
. By u
s
ing the ab
ove-evalu
a
ted
weig
ht
vector and formula
(1), we can
compute
the interval
weighted
arith
m
etic a
g
g
r
eg
ation valu
e
i
e
~
of
ea
ch potenti
a
l
net
work pu
blic se
ntimen
t
emergen
cy
i
e
.
Step 3
. Acc
o
rding to fo
rm
ulae (
3
)
-(5
), we r
a
n
k
all the possibl
e ne
twork pu
blic
sentiment
emergen
cie
s
.
That i
s
, if
k
i
NDD
NDD
, then the
altern
ate n
e
twork
publi
c
sentime
n
t
emergen
cy
i
e
is severe tha
n
the emergen
cy
k
e
, then we must deal
with emergen
cy
i
e
earli
e
r
than
k
e
. if
k
i
e
e
~
~
,
then the severity of emergen
cy
i
e
i
s
s
a
m
e
a
s
k
e
, we can si
multaneo
usl
y
deal with the
two network publi
c
se
ntiment emergen
cies with the
same eme
r
ge
ncy sol
u
tion.
By the above emerg
e
n
c
y deci
s
io
n app
roach,
the net
work p
ublic
sentiment em
erge
ncy
manag
eme
n
t can
cope
wi
th the eme
r
g
ency m
o
re
ef
fi
ciently acco
rding
to the
severity of every
potential net
work p
ubli
c
sentiment em
erge
ncy.
F
r
om th
e
s
e
le
c
t
ed
e
a
r
ly
w
a
rn
in
g
in
de
xe
s
an
d
the severity ranki
ng of all the alternat
e netwo
rk p
u
b
lic sentimen
t emerge
ncie
s, we ca
n al
so
desi
gn the
d
e
ci
sion m
e
ch
anism
and
a
dopt the
co
rresp
ondi
ng e
m
erg
e
n
c
y re
spon
se o
r
d
e
cision
strategy to de
cre
a
se the po
ssi
ble lo
sses
of network pu
blic sentiment
emerg
e
n
c
y.
4. Applicatio
n Example
In un
ce
rtain
setting, the
n
e
twork senti
m
ent
em
erge
ncy m
ana
ge
ment exp
e
rts usually
use
the lin
gu
istic valu
e to
evaluate th
e
importa
nc
e
of the ind
e
x
and to
rate t
he alte
rnativ
es
involved vari
ous
wa
rnin
g i
ndexe
s
. Mo
st of the
existin
g
eme
r
ge
ncy
deci
s
io
n met
hod
s have
o
n
ly
pre
c
ise value
s
for th
e pe
rforma
nce ratin
g
s a
nd the
i
n
dex wei
ghting
.
Therefo
r
e, i
n
ord
e
r to
sel
e
ct
the mo
st severe
one
from
a nu
mbe
r
of
po
ssibl
e n
e
twork
se
ntime
n
t eme
r
gen
ci
es
with diffe
rent
interval indexes, we
will develop an
interval fuzzy A
H
P to determi
ne the priority of different early
warning
inde
xes, an
d the
n
choo
se th
e
most
se
ve
re
netwo
rk
sent
iment eme
r
g
ency fo
r n
e
twork
publi
c
se
ntiment emergen
cy managem
e
n
t.
Example 1.
Suppo
se the
netwo
rk se
ntiment
eme
r
gen
cy mana
gement
depa
rtments
acq
u
ire mu
ch information
of uncertai
n
early
wa
rning indexe
s
for some p
o
tential network
sentime
n
t
em
erge
nci
e
s by
employing su
pervisor co
ntrol platform
s
or sea
r
ch en
gine
s, and th
ey
need to
eval
uate the
sev
e
rity of all th
e po
ssi
ble
city emerg
e
n
c
ie
s, then m
a
ke
final eme
r
ge
ncy
deci
s
io
n ma
ki
ng. No
w a
ssume the
r
e ex
ist mult
iple al
ternate n
e
two
r
k
se
ntiment
emergen
cie
s
E
={
1
e
,
2
e
,
3
e
,
4
e
}, which
may po
sse
s
ses
many
un
certain
ea
rly
warning
ind
e
x
es. By the
aid of
statistical q
u
e
s
tionn
aire
fro
m
eme
r
ge
ncy
de
cisi
o
n
exp
e
rts
and
thro
ugh
our e
s
ta
blish
ed p
r
in
ci
ple
of early warn
ing ind
e
x sel
e
ction, h
e
re
we
ch
o
o
se three fi
rst-gra
de warni
ng i
ndexe
s
in
clu
d
ing
netwo
rk
publ
ic sentiment
emergen
cy
power in
de
x (
1
c
), govern
m
ent eme
r
g
ency
copi
ng
cap
a
cit
y
(
2
c
) an
d netwo
rk
se
ntiment inten
s
ity index (
3
c
). More
over, in first-g
r
ad
e wa
rning in
dex
level
1
c
we
sel
e
ct the
follo
wing
se
cond
-g
rade
ind
e
xes:
seve
rity of e
c
on
omic lo
ss
)
(
11
c
, extent
of diffusio
n
)
(
12
c
and time
du
ra
tion
)
(
13
c
. And i
n
first-g
r
ad
e
warnin
g in
dex
level
2
c
we also
sele
ct
t
h
e
f
o
ll
owin
g
se
con
d
-g
rad
e
in
de
x
e
s in
clu
d
ing
re
spo
n
s
e
sp
eed
)
(
21
c
, netwo
rk se
ntiment
informatio
n transparen
cy
)
(
22
c
. Also, in th
e n
e
twork
se
ntiment inten
s
it
y index
3
c
we
choo
se th
e
followin
g
su
b
-
indexe
s
: sen
t
iment attention deg
ree
)
(
31
c
, emotional liabil
i
ty of network sentime
n
t
)
(
32
c
, and authenti
c
ity of network se
ntiment
)
(
33
c
.
More
over, by
emerg
e
n
c
y experts
com
parin
g ea
ch
pair of warni
ng indexe
s
,
we can
easily get the
interval fuzzy preferen
ce
matrix ov
er e
a
ch
wa
rning i
ndex level as sho
w
n in Ta
ble
s
3. Also, the
evaluated
val
ues of all
the
alter
nate
net
work sentime
n
t eme
r
ge
nci
e
s
with
re
sp
e
c
t to
the unce
r
tain
warning ind
e
x
es are given
by relat
ed expertise as sh
own in the followin
g
Table
4.
Our next task is to dete
r
mine the se
verity r
anki
n
g of all the potential net
work sentim
ent
emergen
cie
s
involved int
e
rval lin
guisti
c
terms.
Ultimately, we
make
final u
r
gent
de
cisi
o
n
to
sele
ct
t
he mo
st
sev
e
re on
e we mu
st
d
eal wit
h
f
i
rst of all out of all the alternat
e netwo
rk pu
blic
sentime
n
t em
erge
nci
e
s.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 287 – 2
9
5
292
Table 3. Interval Fuzzy Preferen
ce
Rel
a
tion ov
er Network Senti
m
ent Emerg
e
ncy Index Le
vels
C1
C2
C3
C11
C12
C13
C21
C22
C31
C32
C33
C1
[1,1]
[2,4]
[4,6]
C2
[1/4,1/2]
[1,1]
[1/8,1/6]
C3
[1/6,1/4]
[6,8]
[1,1]
C11
[1,1]
[4,6]
[6,8]
C12
[1/6,1/4]
[1,1]
[1/4,1/2]
C13
[1/8,1/6]
[2,4]
[1,1]
C21
[1,1]
[2,4]
C22
[1/4,1/2]
[1,1]
C31
[1,1]
[4,6]
[2,4]
C32
[1/6,1/4]
[1,1]
[1/8.1/6]
C33
[1/4,1/2]
[6,8]
[1,1]
Table 4. Net
w
ork Sentime
n
t Emergen
cy Deci
si
on System with Int
e
rval Ling
uisti
c
Terms
Emergenc
y
C11
C12
C13
C21
C22
C31
C32
C33
1
e
VT
B
V S
V VH
H
M
B
S
2
e
EB VB
VW
H
V
VL
S
T
W
3
e
VB T
M
VH
V
VH
W
VT
VVW
4
e
T
B
V S
V VH
H
V VW
M
VS
First, from th
e pai
r-wise
compa
r
ison p
r
eferen
ce
mat
r
ix of the first-gra
de i
ndex
es
and
se
con
d
-g
ra
de
indexes in
Table 3, by employ
ing int
e
rval fuzzy AHP and form
ula (2)
we can
comp
ute the weig
ht vector and prio
rity of each ea
rly warning in
de
x level as listed in Table 5.
Table 5. The
Priority Weig
hts in the Ne
t
w
ork Sentime
n
t Emergen
cy Warnin
g Index Levels
Weight of top ind
e
x
Weight of sub-index
Overall
w
e
ight o
f
sub-index
Index 1
[0.4366,0.87
01]
Sub-index 11
[0.576,0.941
2]
[0.2515,0.81
89]
Sub-index 12
[0.0692,0.12
95]
[0.0302,0.11
27]
Sub-index
13
[0.1258, 0.22
63]
[0.0549, 0.19
69]
Index 2
[0.0688, 0.13
18]
Sub-index 21
[0.5224,1.04
48]
[0.0359, 0.13
77 ]
Sub-index 22
[0.1847,0.36
94]
[0.0127,
0.04
87]
Index 3
[0.2183, 0.38
01]
Sub-index 31
[0.4151, 0.84
34]
[0.0906, 0.32
06]
Sub-index 32
[0.0571, 0.10
14]
[0.0125, 0.03
85]
Sub-index
33
[0.2376, 0.46
42]
[0.0519, 0.17
64]
From
Table
5
,
we
see th
at
the wei
ghts
22
w
=[0.0127, 0.04
87],
32
w
=[0.012
5
,
0.0385] a
r
e
very small, so the two wa
rning sub-i
nde
xes
32
22
,
c
c
can be o
m
itted. Now
we only sel
e
ct the six
warning i
nde
x {
33
31
21
13
12
11
,
,
,
,
,
c
c
c
c
c
c
}, whi
c
h a
r
e viewed
as
the six criteri
a
}
,
,
,
{
6
2
1
r
r
r
of
the netwo
rk p
ublic
sentime
n
t emerg
e
n
c
y deci
s
ion p
r
o
b
lem.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Index Sele
cti
on Preferen
ce and Weighti
ng fo
r Un
ce
rt
ain Net
w
ork
… (Qia
nsh
e
n
g
Zhang
)
293
From ling
u
isti
c term Table
1, we transl
a
te
Table 4 regarding the
sele
cted six
warning
indexe
s
into the followi
ng interval fuzzy deci
s
io
n matrix
6
4
)
~
(
ij
r
D
,
,
]
8
.
0
,
7
.
0
[
]
2
.
0
,
1
.
0
[
]
9
.
0
,
8
.
0
[
]
8
.
0
,
7
.
0
[
]
7
.
0
,
6
.
0
[
]
4
.
0
,
3
.
0
[
]
2
.
0
,
1
.
0
[
]
4
.
0
,
3
.
0
[
]
8
.
0
,
7
.
0
[
]
6
.
0
,
4
.
0
[
]
4
.
0
,
3
.
0
[
]
8
.
0
,
7
.
0
[
]
4
.
0
,
3
.
0
[
]
7
.
0
,
6
.
0
[
]
7
.
0
,
6
.
0
[
]
3
.
0
,
2
.
0
[
]
8
.
0
,
7
.
0
[
]
0
.
1
,
9
.
0
[
]
7
.
0
,
6
.
0
[
]
6
.
0
,
4
.
0
[
]
9
.
0
,
8
.
0
[
]
8
.
0
,
7
.
0
[
]
7
.
0
,
6
.
0
[
]
3
.
0
,
2
.
0
[
)
~
(
6
4
ij
r
D
ij
r
~
is the interva
l
membershi
p
degree
of netwo
rk
senti
m
ent emerge
ncy
i
e
with respec
t to j-t
h
index co
nsi
d
e
r
ed.
From T
able 5
we kno
w
tha
t
the weight vector
of all the sele
cted
six
warning
sub
-
indexe
s
is
W
=([0.2
515, 0.
8189], [0.03
02, 0.1127],
[0.0549,
0.1969], [0.03
59, 0.1377],
[0.0906, 0.3206],
[0.0519,0.17
6
4
]).
And by formu
l
a (1
)
we
cal
c
ulate the i
n
terv
al weig
hted
arithmeti
c
ag
greg
ation val
ue
i
e
~
of
each network public
sentim
ent emergen
cy
i
e
as bel
ow.
j
j
j
r
w
e
1
6
1
1
~
~
~
=[0.202
9, 0.9218];
j
j
j
r
w
e
2
6
1
2
~
~
= [0.3499, 1.
3595];
j
j
j
r
w
e
3
6
1
3
~
~
=
[
0.2646, 1.092];
j
j
j
r
w
e
4
6
1
4
~
~
=[0.206
1, 0.8931].
Also, acco
rdi
ng to form
ula
(3)
we
com
p
ute
the interv
al prefe
r
en
ce
degree
s bet
ween a
n
y
two fuzz
y numbers
as
follows
:
)
~
~
(
2
1
e
e
P
0.3309,
)
~
~
(
3
1
e
e
P
0.4
25,
)
~
~
(
4
1
e
e
P
0.5091,
)
~
~
(
3
2
e
e
P
0.596,
)
~
~
(
4
2
e
e
P
0.6798,
)
~
~
(
4
3
e
e
P
0.585.
And we then
con
s
tru
c
t the
followin
g
interval possibility matrix
P
, where
)
~
~
(
j
i
ij
e
e
P
p
,
4
4
)
(
ij
p
P
=
5000
.
0
415
.
0
3202
.
0
4909
.
0
5850
.
0
500
.
0
4040
.
0
5750
.
0
6798
.
0
596
.
0
5000
.
0
6691
.
0
5091
.
0
425
.
0
3309
.
0
5000
.
0
.
By using formulae (4
), (5) we can calculat
e the non
-domi
nan
ce d
egre
e
of each netwo
rk
sentime
n
t em
erge
ncy
i
e
~
as
follows
.
1
NDD
=
}
4
1
1
{
min
*
1
1
j
p
j
j
=0.661
8,
2
NDD
=
}
4
1
1
{
min
*
2
2
j
p
j
j
=1,
3
NDD
=
}
4
1
1
{
min
*
3
3
j
p
j
j
=0.808,
4
NDD
=
}
4
1
1
{
min
*
4
4
j
p
j
j
=0.6
404.
Evaluation Warning : The document was created with Spire.PDF for Python.
ISSN: 23
02-4
046
TELKOM
NIKA
Vol. 11, No
. 1, Janua
ry 2013 : 287 – 2
9
5
294
Since
}
{
max
2
j
j
NDD
NDD
,
}
{
2
e
e
ND
, that is
to
s
a
y,
2
e
is
the mos
t
severe
emergen
cy.
And du
e to t
hat
2
NDD
>
3
NDD
>
1
NDD
>
4
NDD
, we
obtain th
at th
e severity ra
nkin
g
of all the alternate network publi
c
se
ntiment emergen
cies is a
s
4
e
1
e
3
e
2
e
.
Thus, the net
work sentime
n
t emerge
ncy
2
e
is the optimal deci
s
io
n alternative. That is
to s
a
y,
2
e
is the most severe net
wo
rk
sentime
n
t emerg
e
n
c
y in all the possible n
e
two
r
k
sentime
n
t em
erge
nci
e
s, th
e netwo
rk
se
ntiment
eme
r
gen
cy manag
ement de
cisi
on-m
a
ker mu
st
firstly deal
wi
th this n
e
two
r
k
se
ntiment
emergen
cy, next to co
pe
with the
se
con
dary
sev
e
re
emergen
cy
3
e
, then
1
e
, and so
4
e
. The rel
a
te
d network se
ntiment eme
r
gen
cy mana
g
e
ment
will rai
s
e
the
corre
s
p
ondin
g
ea
rly wa
rni
ng an
d take u
r
gent
de
cisio
n
me
cha
n
ism
to co
ordi
nate
all
kinds of emergency facilit
ies among different muni
cipal zones an
d district
s to av
oid or decrease
the ri
sk lo
ss
of the
unexp
e
cted
net
wo
rk
sent
im
ent
significa
nt em
erge
ncy
befo
r
e im
pleme
n
ting
some e
m
erge
ncy re
spo
n
se
.
5. Conclusio
n
In this pap
er,
we em
ploy a
n
interval fuzzy
AHP meth
od to assig
n
the ratio
nal weights of
sele
cted
ea
rl
y warning
in
dexes for
ne
twork p
ubl
i
c
sentime
n
t e
m
erg
e
n
c
y. And the
n
by u
s
ing
interval weig
hted ag
gre
g
a
tion op
erato
r
of all the
warning i
nde
x value we
can
ran
k
all
the
severity
of n
e
twork pu
blic se
ntiment
e
m
erg
e
n
c
ie
s a
nd m
a
ke em
erge
ncy
de
ci
sion
to
sele
ct
the
most severe
netwo
rk p
ubli
c
se
ntiment
emergen
cy
, whi
c
h can he
lp the related
network p
u
b
lic
sentime
n
t em
erge
ncy ma
n
ageme
n
t dep
artment take
the co
rrespon
ding eme
r
ge
n
c
y strategy a
n
d
mech
ani
sm i
n
accord with
the obtained
severity
ran
k
ing re
sult of the network p
ublic
sentime
n
t
emergen
cy with interval linguisti
c
value
s
.
Ackn
o
w
l
e
dg
ement
This
wo
rk i
s
sup
porte
d by
the Hum
anit
i
es a
nd So
ci
al Scien
c
e
s
Youth Fou
n
d
a
tion o
f
Ministry of Education of Ch
in
a (No. 12YJCZH281
), the Guang
zhou Soci
al Scien
c
e Plan
ning
Comm
on Co
nstru
c
tion Project “Th
e
st
udy of early
warning ind
e
x
selection a
nd urg
ent de
cisi
on
mech
ani
sm f
o
r
city signif
i
cant em
erge
ncy
in u
n
certain enviro
n
m
ent”
(No.
2012
GJ31), t
h
e
Nation
al Natural S
c
ien
c
e
Found
ation
(No
s
. 612
022
71,
61273
11
8
, 610700
61,
6126
301
4),
the
Guan
gdo
ng
Natural S
c
ie
nce
Fou
ndati
on (No. S2
0
1204
0007
184
,
9451
009
00
1002
686
), the
Youth Proj
ect of Hu
manit
i
es
and
Soci
al Scie
nc
es
of Minist
ry o
f
E
ducatio
n of
Chi
na (No.
10YJC79
010
4), the Fun
d
a
m
ental Rese
arch Fun
d
s f
o
r the Central
Universitie
s
in Chin
a, and
the
Guan
gdo
ng
Province Hig
h
-level Tal
ent
s Proje
c
t.
Referen
ces
[1]
Ashley
WL, Morrison JL.
Anticipat
or
y
m
a
n
a
geme
n
t:
T
ools
for better dec
ision m
a
kin
g
.
The futurist.
199
7; 31: 47-5
0
.
[2]
Morrison J
L
, W
ilson I. Anal
yzi
ng e
n
viro
n
m
ents and
de
velo
pin
g
scen
a
rios for u
n
ce
rtain times. In:
Plan
nin
g
an
d mana
geme
n
t for a chan
gi
ng
envir
onme
n
t, San F
r
ancisc
o
, 199
7.
[3]
Z
eng R
X
,
Xu
XL. T
he
earl
y
w
a
rn
in
g s
y
stem, ind
e
x
and
mecha
n
ism
o
f
net
w
o
rk
pu
bl
ic sentim
en
t
emerg
enc
y.
Journal of Informatics
. 2009; 1
1
: 52-55.
[4]
Z
eng R
X
. T
he
constructio
n
of earl
y
w
a
rni
ng i
nde
x of n
e
t
w
or
k publ
ic sentim
ent.
Informatic
s T
heory a
n
d
Practice.
201
0;
33: 77-80.
[5]
Z
hang
W
P
. Stud
y o
n
th
e i
n
d
e
x sel
e
ctin
g a
n
d
w
e
ig
ht eva
l
u
a
tion f
o
r em
ergenc
y
ear
l
y
w
a
rnin
g.
Jour
nal
of Chin
ese Pe
opl
e'
s Public S
e
curity Un
ivers
i
ty.
2008; 6: 80
-89.
[6]
Peng
Z
H
. Disc
u
ssio
n
o
n
po
p
u
lati
on
emer
ge
nc
y a
n
d
net
w
o
rk pu
blic
se
nti
m
ent.
Jour
na
l of
Sha
n
g
hai
Publ
ic Security
Colle
ge
. 2
008;
1: 46-50.
Evaluation Warning : The document was created with Spire.PDF for Python.
TELKOM
NIKA
ISSN:
2302-4
046
Index Sele
cti
on Preferen
ce and Weighti
ng fo
r Un
ce
rt
ain Net
w
ork
… (Qia
nsh
e
n
g
Zhang
)
295
[7]
Z
hang
YW
, Qi YJ. T
he inter
a
ction
mech
an
ism bet
w
e
e
n
net
w
o
rk
pu
blic
sentime
n
t a
n
d
a
bnorm
a
l
emerg
enc
y.
Journal of Informatics.
2010; 9:
1-6.
[8]
W
u
SZ
, Li S
H
. T
he stud
y
of mech
anism
for net
w
o
rk
pub
lic s
entime
n
t earl
y
w
a
r
n
i
ng, Jo
urna
l of
Chin
ese Pe
op
l
e
'
s
Public Sec
u
rit
y
Un
iversit
y
, 2008; 3: 38-4
2
.
[9]
Z
hang
LL. T
h
e inv
e
stig
atio
n
on t
he m
o
d
e
l
of em
erge
nc
y net
w
o
rk p
ubl
i
c
sentim
ent e
a
rl
y
w
a
rni
ng.
Journ
a
l of Li
brary and Infor
m
ation Serv
ice.
201
0; 11: 135-
138.
[10]
Lin C, Li BC
, W
ang J.
T
h
e method of
net
w
o
rk p
ubl
ic
sentim
ent ea
rl
y
w
a
rni
ng b
a
sed on fuzz
y
reaso
n
in
g.
Jou
r
nal of Infor
m
at
ion a
nd En
gin
e
e
rin
g
Univ
ersit
y
.
2011; 12: 72
-76.
[11]
Z
eng W
,
Lu
o C
.
Compre
hens
i
v
e dec
isio
n mo
del
of interv
al-
numb
e
r.
Jour
n
a
l of T
h
e
o
ry a
n
d
Practice
of
Syste
m
Eng
i
nee
ri
ng
. 199
7; 17
(11): 48-5
0
.
[12]
Z
hang
JJ, W
u
DS. T
he method
of
gre
y
r
e
l
a
ted a
nal
ys
is to
multip
le
attrib
ute d
e
cisi
on
pr
obl
ems
w
i
t
h
i
n
te
rval
n
u
m
bers.
Mathematic
al an
d Co
mp
ut
er Mode
lli
ng.
2
005; 42: 9
91-9
98.
[13]
Rodri
g
u
e
z RM,
Martinez
L, He
rrera F
.
He
sitant fuzz
y li
n
guistic term s
e
ts for decis
ion
makin
g
, IEEE
T
r
ansactions o
n
F
u
zz
y
S
y
ste
m
s, 2012; 20(
1
)
: 109-11
9.
[14]
Kahram
an G, Cebeci U, U
l
uka
n
Z
.
Multi-cr
iteria sup
p
li
er selectio
n u
s
ing
fuzz
y
A
H
P.
Lo
gistics
Information Ma
nag
e
m
ent.
20
0
3
; 16(6): 38
2-3
94.
[15]
Burill
o P, Busti
n
ce H. E
n
tr
op
y on
i
n
tuiti
onis
t
ic
fuzz
y
s
e
ts and on interv
a
l
-valu
ed
f
u
zz
y sets.
Fu
z
z
y
Sets and Syste
m
s
. 19
96; 78: 305-
316.
[16]
Grzegorze
w
s
ki
P. Distanc
es
bet
w
e
en
intuiti
onistic fuzz
y s
e
ts and/
or inte
rval-val
ue
d fuz
z
y
s
e
ts bas
e
d
on Hausdorff
metric.
F
u
zz
y
Sets and Syste
m
s.
200
4; 148
: 319-32
8.
[17]
Patricia M, Vic
t
or H, Danni
el
a R, F
e
vrier V, Oscar C. Ge
netic
Optimizat
i
on of N
eura
l
Net
w
orks fo
r
Person R
e
cog
n
itio
n Base
d o
n
the Iris.
TEL
K
OMNIKA
. 2012; 10(2): 3
09-
320.
[18]
Nasuti
on H, Ja
malu
ddi
n H, S
y
eriff JM. Energy Ana
l
ysis for Air Cond
itio
nin
g
S
y
stem Usi
n
g F
u
zz
y
Lo
gic
Contro
ller.
TELKOMNIKA
. 2011; 9(1):13
9
-1
50.
Evaluation Warning : The document was created with Spire.PDF for Python.