Lin and Li Fuzzy Ranking Method

An
lntematimal
Journal
computers &
mathematics
mIlppl-
PERGAMON
Computers and Mathematics with Applications 45 (2003) 709-713

www.elsevier.com/locate/camwa
Inconsistent Property
of Lee and Li
Fuzzy Ranking Method
H.-C.
TANG
Department of Industrial Engineering and Management
National Kaohsiung University of Applied Sciences
Kaohsiung, Taiwan 807, R.O.C.
[email protected]
January
(Received
2002; accepted February
2002)
Abstract-This
paper analyzes the Lee and Li fuzzy ranking method for the normalized trapezoidal fuzzy numbers (NTFNs). Every NTFN has associated with it another NTFN, which possesses
an inconsistent property. An inconsistent property is that the results of comparison of these two
NTFNs are conflicting when two different comparison indices are used. Based on this property,
Lee and Li fuzzy ranking method suffers from lack of discrimination and occasionally conflicts with
intuition. @ 2003 Elsevier Science Ltd. All rights reserved.
Keywords--Fuzzy
ranking method, Fuzzy number.
INTRODUCTION
A common problem in operations
which is best.
a set of alternatives
to determine
Fuzzy ranking method is a good way of evaluating these alternatives.
research involves comparing
The method
of fuzzy ranking seems to have been proposed first by Jain

methods
have been developed in an attempt
to simplify the comparison

represent
the FNs.
adopts possibilistic
the approaches
occasionally
probability
to order the fuzzy numbers
This simplified
representation
in the literature
with intuition.
probability
distributions
(FNs)
indices.
[2-lo].
In order
indices to
into two classes.
the method
One
A major drawback of all

and
of Lee et al. [7] based on the

According
they proposed two groups of comparison
example to show the results are conflicting
to the uniform and

indices. However,
when two different
are used. The main result of this paper is that a theorem is provided which further
extends this property to any normalized trapezoidal

In the following sections,
Then,
a pathological
a large variety of
is that they suffer from lack of discrimination
In particular,
distributions,
Lee and Li gave a pathological
other.
may be categorized
measure considers the mean and variance of the FNs.
proportional
method.
Since then,
of FNs, several scholars propose one or two comparison
indices and the other adopts probabilistic
proposed
conflict
[l].
a theorem
property
FNs (NTFNs).
we first provide a concise review of the Lee and Li fuzzy ranking

is presented
for the NTFNs.
With the theorem,
which the two groups of comparison
Finally, we give some concluding
there always exists
indices are inconsistent
with each
remarks.
0898-1221/03/$
- see front matter @ 2003 Elsevier Science Ltd. All rights reserved.
PII: SO898-1221(03)0003C-0
Typeset by 414-W
710
H.-CH.
LEE AND LI FUZZY
TANG
RANKING
METHOD
In this section, we present the mean and variance formulae of the Lee and Li fuzzy ranking
method for the NTFNs.
For the sake of the completeness,
we recall the Lee and Li ranking
method. For brevity, we adopt the same notations as Lee and Li. Two kinds of distributions
of fuzzy events are distinguished.
In the case of the uniform distribution,
the mean of a fuzzy
event A with membership function PA is defined as
M
and the variance
.f~~/l.d~)~~
k4(Ic)dx
,q =
11
_f,
(1)
of A is defined as
0A
G;
In the case of the proportional

by the integrals
X2PA
(x) dx
= _,A
.f/, PA cx)dx
distribution,
M,"
(A).
the mean and variance
(2)
of A are, respectively,
given
(3)
P
and
s,X2&(X)dx
s,&(x)dx
-Mi
(4)
The NTFN, denoted symbolically by writing A = (a, b, c, d), means that the membership function
PA(x) is 1 at b to c, and becomes zero at the two end points, a and d. More precisely, the
membership function of the NTFN is piecewise linear function and is defined as follows:
a
I b_ax-
if a 5 x 5 b,
b-a
if b 2 x 5 c,
if c 5 x < d,
otherwise.
From the definition
of the NTFN,
-a-b+c+d
A PA(x)dx=
JA
JA
x2,.LA(x) dx =
x/LA (x) dx =
A&(x)dx
we can deduce the following results:
-a2 - b2 -t c2 + d2 - ab + cd
>
-a3 - b3 + c3 + d3 - ab(a + b) + cd(c + d)

,
12
-a-2b+2c+d
3
_,2 - 3b2 + 3c2 + d2 - 2ab + 2cd

xp; (x) dx =
,
12
sA
x&(x)
.I
dx =
-a3 - 4b3 + 4c3 + d3 - ab(2a + 36) + cd(3c + 2d)

30
711
From equations
M
(l)-(4),
11
0
Gz
it further
G;
(5)
= -a3 - b3 + c3 + d3 - ab(a + b) + cd(c + d) _ M2
7L(
6(-a-b+c+d)
A
(6)
11
= -a2 - 3b2 + 3c2 + d2 - 2ab + 2cd
(7)
4(-a-2b+2c+d)
A
0
are as follows:
3(-a-b+c+d)
A
for the NTFNs
= -a2 - b2 + c2 + d2 - ab + cd
0
M
follows that the formulae
= -a3 - 4b3 + 4c3 + d3 - ab(2a + 3b) + cd(3c + 2d) _ M2

P
lO( -a - 2b + 2c + d)
INCONSISTENT
j
(8)
PROPERTY
The following example illustrates the inconsistent property of the Lee and Li ranking method.
Consider two NTFNs A = (4,5,6,11) and B = (2,7,8,9).
After using formulae (5)-(g) to obtain
M,(A) = 6.75, G;(A) = 2.3542, M&$
= 6.3889, G;(A;) = 1.5265, M,(B) = 6.25, GE(B) =
2.3542, M,(B)
= 6.6111, G;(B)
= 1.5265.
S ince n/l,(A)
> M,(B)
and G:(A)
= GE(B),
A is
larger than fi if A and fi have the uniform distribution,

whereas for proportional
distribution
B
is larger than A. In fact, this inconsistent property is not innocent. A rigorous statement of this
property is the following.
THEOREM 1. For any NTFIV 2, there exists another
relationships:
(a) MU(A) > M,(B)
(b) M,(A) = h/l,(B)
(c) M,(A) < MU(h)
Moreover,
we have Gt (A) = Gz (B) and Gz (A) = Gz (fi)
in each of these relationships
M
Subtracting
11
0
11
(
cr, a + Q + /?, a + Q t- p + y), where a, p, y 2 0. Let

as (a-q,a-q+Y,a-q+@+Y,a-q+cr+b+y).
the means of fuzzy events A and A.
and B is uniform, we use (5) to obtain
= 2a2+3P2+Y2+3a(Y+6ap+3ay+6a/3+3cuy+3py
3(a + 2P + Y)
one of the following
and M,(A) < M,(B),

and MP@) = MP(@,
and M*(A) > M,(B).
PROOF. Assume that a NTFN A is (a, a +

17= (y-CY)/~, anddefineanother
NTFNB
First, we discusses the relationship between
When, the distribution
of fuzzy events A
M
hJTFN fi satisfying
= a2 + 3P2 + 2y2+3(a
- q)cY+6(a - v)P+3(a
>
(9)
- q)y+3crP+3wy+6Py
3(a + 2P + Y)
(10)
(10) from (9), we find

Mu ()
-Mu
(B)
(r2--+3~~+6qP+3r/~+3ap-3&
3(0 + 2P + Y)
= (a + r)(~
- Y) + 3rl(~ f Y) + 3P(a - Y) + 6P77

3(a + 2P + Y)
= (a + r)( - Y + 377)+ 3P(a - Y + 377) - 9Pv + 6P77

3(a + 2P + Y)
= (a + Y + 3P)(a - Y + 3V) - 3PV
3(a + 2p + y)
Substituting
77= (y - a)/2
Mu (>
-Mu
into the above equation

(fi)
yields
(cx + 3P + y)(-1/2cu
+ l/2?)
- 3/2P(y
- a)
3(Q + 2P + Y)
(11)
= (Y - a) (ct. + 3P + Y - 3P) = (Y - a) (Y + Q)
6(cy + 2P + y)
6(~ + 2P + Y)
(9) Bu!Qids
uoynq!w!p
um3!un
ayl30 am3 aqq UI 8 pue y 30 sam+v~
aq$ az&&
MO;
aM
0 > L J! (@%! < (y)%l

= (IyA!
PuQ (@A! = (y)Jv (z) D < L JI
Pue (@A! > (y)vv
(E) 0 = L 31 (@%
(@Iv > (y)m PuQ (a)?w
< (y)Vv (1) a*eq ah (~1) puv (11) uovlap!suo3
ow! ~upm
CL+ dr: + D)Z

(0 - Lc)d =
(ZT)
CL+ dE + o)z
WZ+(dV+a+L)(kZ-D-L)
CL + dE + m)z
449 - (0 - L)dp
+ (L + D)&
- (0 + L)(x, - L) =
(L + dE + D)P
Ld8 + d;\78 - Lb
- dazr
- o&
- zLz + $z--
= (y)
VI
(8)
%I
Pus
CL+ d!?+ O>P
Lm + LDp+ dv + LP - O)P+ d(h -
)ZI + D(h - o)f7 + zLc + ,dg + $J
3NVL
= 0 &l
Al
'HD-'H
ZIL
713
Reconsider the NTFN A = (4,5,6,11) in the above example, using the notation of Theorem 1.
we have a = 4, (Y= 1, p = 1, and y = 5, this implies that 77= (y - o)/2 = 2, so fi = (2,7,8,9).
Since y > cr, we get A&(A) > M,(B), G:(j)
= G:(B), A&(A) < M,(B), and G;(A) = G:(B)
by utilizing Theorem 1. Therefore, we can obtain the same result as that of example.
It is straightforward
to show the inconsistent
property of the Lee and Li ranking method for
the normalized triangular FNs.
4. CONCLUSION
The comparison of FNs is a very important topic in decision-making
problem. In the literature,
several articles have addressed this issue and various kinds of techniques have been devised. Most
of them suffer from some pathological examples, which cannot compare all the FNs. This paper
focuses on the inconsistent
property of the Lee and Li fuzzy ranking method for the NTFNs.
With the Theorem 1, every NTFN A = (a, a + a, a + a + /3,a + cr + /? + y), a, p, y 2 0, has
associatedwithitanotherNTFNB=(a-rl,a-77+y,(I-77+P+y,a-rl+cu+P+y)with
q = (y - o)/2 satisfying one of the following relationships:
and G:(a)
= G;(B)
(2) A&(A) = i&(a),
(1) A&(A) > 11/I,(B), G:(A)
GE(A) = G:(B),
= G:(B),
MP(A) = IMP(B), and G:(A)
= G:(B)
if y = (Y,
(3) A&(A) < A&(B),
G:(A)
&,(A)
= G;(B)
if y < cr.
= G:(B),
n/i,(A) < M,(B),

> M,(B),
and G;(A)
Therefore, the two groups of indices of the Lee and Li fuzzy ranking
with each other for the NTFNs.
method
if y > (Y,
are inconsistent
REFERENCES
1. R. Jain, Decision-making in the presence of fuzzy variables, IEEE 7kansactions on Systems Man and
Cynernetics 6, 698-703 (1976).
2. J.F. Baldwin and N.C.F. Guild, Comparison of fuzzy sets on the same decision space, fizzy Sets and
Systems 2, 213-231 (1979).
3. G. Bortolan and R. Degani, A review of some methods for ranking fuzzy subsets, FUZZYSets and Systems
15, 1-19 (1985).
4. C.H. Cheng, A new approach for ranking fuzzy numbers by distance method, Fuzzy Sets and Systems 95,
307-317 (1998).
5. P. Grzegorzewski, Metrics and orders in space of fuzzy numbers, fizzy Sets and Systems 97, 83-94 (1998).
6. K. Kim and K.S. Park, Ranking fuzzy numbers with index of optimism, Wzzy Sets and Systems 35, 143-150
(1990).
7. E.S. Lee and R.J. Li, Comparison of fuzzy numbers based on the probability measure of fuzzy events.
Computers Math. Applic. 15 (lo), 887-896 (1988).
8. T.S. Liou and M.J.J. Wang, Ranking fuzzy numbers with integral value, fifty Sets and Systems 50, 247-255
(1992).
9. M. Modarres and S. Sadi-Nezhad, Ranking fuzzy numbers by preference ratio, fitzy Sets and Systems 118,
429-436 (2001).
10. P.A. Raj and D.N. Kumar, Ranking alternatives with fuzzy weights using maximizing set and minimizing
set, Fuzzy Sets and Systems 105, 365-375 (1999).

Lin and Li Fuzzy Ranking Method

Uploaded by

Copyright:

Available Formats

Lin and Li Fuzzy Ranking Method

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lin and Li Fuzzy Ranking Method

Uploaded by

Copyright:

Available Formats

An

Computers and Mathematics with Applications 45 (2003) 709-713

2002; accepted February

ranking method, Fuzzy number.

Fuzzy ranking method is a good way of evaluating these alternatives.

research involves comparing

of fuzzy ranking seems to have been proposed first by Jain

have been developed in an attempt

to simplify the comparison

to order the fuzzy numbers

into two classes.

A major drawback of all

of Lee et al. [7] based on the

they proposed two groups of comparison

example to show the results are conflicting

to the uniform and

extends this property to any normalized trapezoidal

is that they suffer from lack of discrimination

Lee and Li gave a pathological

measure considers the mean and variance of the FNs.

of FNs, several scholars propose one or two comparison

indices and the other adopts probabilistic

we first provide a concise review of the Lee and Li fuzzy ranking

for the NTFNs.

With the theorem,

which the two groups of comparison

Finally, we give some concluding

there always exists

indices are inconsistent

LEE AND LI FUZZY

and the variance

In the case of the proportional

the mean and variance

From the definition

we can deduce the following results:

-a3 - b3 + c3 + d3 - ab(a + b) + cd(c + d)

_,2 - 3b2 + 3c2 + d2 - 2ab + 2cd

-a3 - 4b3 + 4c3 + d3 - ab(2a + 36) + cd(3c + 2d)

= -a3 - b3 + c3 + d3 - ab(a + b) + cd(c + d) _ M2

= -a2 - 3b2 + 3c2 + d2 - 2ab + 2cd

for the NTFNs

follows that the formulae

= -a3 - 4b3 + 4c3 + d3 - ab(2a + 3b) + cd(3c + 2d) _ M2

larger than fi if A and fi have the uniform distribution,

we have Gt (A) = Gz (B) and Gz (A) = Gz (fi)

in each of these relationships

cr, a + Q + /?, a + Q t- p + y), where a, p, y 2 0. Let

one of the following

and M,(A) < M,(B),

PROOF. Assume that a NTFN A is (a, a +

(10) from (9), we find

- Y) + 3rl(~ f Y) + 3P(a - Y) + 6P77

= (a + r)( - Y + 377)+ 3P(a - Y + 377) - 9Pv + 6P77

into the above equation

ayl30 am3 aqq UI 8 pue y 30 sam+v~

0 > L J! (@%! < (y)%l

CL+ dr: + D)Z

CL+ d!?+ O>P

Lm + LDp+ dv + LP - O)P+ d(h -