A New Method for Fuzzy Group Decision

Making Based on α-Level Cut and Similarity

Jibin Lan1,2 , Liping He2 , and Zhongxing Wang2

1
School of Economics and Management, Southwest Jiaotong University,
610031 Chengdu, Sichuan, P.R. China
[email protected]
2
School of Mathematics and Information Science, Guangxi University,
530004 Nanning, Guangxi, P.R. China
[email protected]

Abstract. Let opinions of experts among group decision making be rep-

resented as L-R fuzzy numbers. The difference of two experts’ opinions
is reflected by two distances, which are called the left-hand side distance
and the right-hand side one. A method to calculate two types of distances
based on the same α-level is presented. Then the distances are employed
to construct a new similarity function to measure the similarity degrees
of both sides which represent the pessimistic and optimistic similarity de-
grees between the experts, respectively. The degree of importance of each
expert among group decision making is obtained by employing Saaty’s
analytic hierarchy process (AHP). The method of aggregating individual
fuzzy opinions into a group consensus opinion by combining similarity
degrees and the degree of importance of each expert is proposed. Finally
some properties of the proposed similarity measure are proved and some
numeric examples are shown to illustrate our method.

1 Introduction

Up to now, several methods have been proposed for drawing consensus from
opinions of experts. Methods in [5,9,10,15,17] considered the fuzzy preference
relation of each expert. Ishikawa et al.[8] and Xu et al.[19] proposed the indi-
vidual opinions represented by interval-values and got the consensus from cu-
mulative frequency distribution. Hsu [7] proposed a method called similarity
aggregation method (SAM) to aggregate individual opinions of experts based
on the similarity degree function. But it is not effective when experts’ opin-
ions which are represented as fuzzy numbers have no common intersection. Even
when they intersect, the difference between them still exists. For example, let the
opinions of expert E1 and expert E2 be represented as trapezoidal fuzzy num-
bers (1, 2, 4, 5) and (4, 5, 6, 7). Obviously, they intersect at the point P (4.5, 0.5).
Since the left- and right-hand membership functions are considered to reflect the
pessimistic and optimistic opinions of experts, respectively, for expert E1 , the
opinion at the point of intersection is optimistic while for expert E2 it is pes-
simistic. Thus, they aren’t thought to be identical. At the same time, there are a

L. Wang and Y. Jin (Eds.): FSKD 2005, LNAI 3614, pp. 503–513, 2005.


c Springer-Verlag Berlin Heidelberg 2005
504 J. Lan, L. He, and Z. Wang

number of approaches to determine the distance between fuzzy sets in the liter-
ature [2,11,12,13,14]. Most of them are based on the Hausdorff distance between
the α-level cuts. However these approaches can not keep producing output sets
with the same nature as the input sets [2,6]. Lee [3] proposed an iterative pro-
cedure for approximating the optimal consensus of experts’ opinions which are
represented by trapezoidal fuzzy numbers. Lee’s method based on the distance
between A
= (b1 , b2 , b3 , b4 ) is defined by:

= (a1 , a2 , a3 , a4 ) and B


4

B)
dp (A,
= ( (|ai − bi |)p )1/p . (1)
i=1

This distance satisfies the conditions of non-negativity, symmetry and trian-

gle inequality. However, Lee’s method may not be effective in some cases. For
example, let A
= (4, 5, 7, 8), B

= (2, 5, 7, 8) and C

= (5, 6, 8, 9) be three trape-
zoidal fuzzy numbers and assume p = 2. According to Lee’s method, we obtain

B)
d2 (A,
= 2, d2 (A,

C)

= 2. But in fact the two distances aren’t identical.
The main characteristics of the methods mentioned above are synthesized as
follows:
(1)most of them are limited to the use of trapezoidal fuzzy numbers;
(2) the shape of the membership function is not taken into consideration or
only a part of it is used, which leads to losing much information;
(3) the pessimistic opinion and the optimistic opinion which are two entirely
distinct conceptions aren’t considered, respectively.
In this paper, an aggregation based on the left- and right-hand side distances
between fuzzy numbers for the degrees of similarity among experts whose opin-
ions are represented by L-R fuzzy numbers is presented.

2 L-R Fuzzy Numbers

A fuzzy number is a fuzzy set in the real line and is completely defined by its
membership function as follows:

µ(x) : X → [0, 1] (2)

= {(x, µ(x))|x ∈ X} is called a fuzzy set.
where X is the subset of real line, R
The membership degree of x to the fuzzy set R
is expressed by the mem-
bership function µ(x). For computational purpose, this definition is generally
restricted to those fuzzy numbers which are both normal and convex:
(1) Normality: sup µ(x) = 1, where x ∈ X
x∈X
(2) Convexity: µ{λx1 + (1 − λ)x2 } ≤min{µ(x1 ), µ(x2 )} , where x ∈ R
and
λ ∈ [0, 1].
The definition and the two requirements imply that the α-level cut R
α = {x :


µR
(x)>α} of such a fuzzy subset R is a closed interval [al , ar ] for any α ∈ (0, 1],
α α

and the highest membership values are clustered around a given interval (or
point). And the 1-level cut R
1 = {x : µ
(x) = 1} is not an empty set.
R
 A New Method for Fuzzy Group Decision Making 505

A general type of normal and convex fuzzy number is L-R fuzzy number and
is defined by:
⎧ m−x
⎨ L( δ ), x < m, δ > 0
µ(x) = 1, m≤x≤n (3)
⎩ R( x−n ), x > n, γ > 0
γ

where δ, γ are the left-hand and the right-hand spread, respectively; L(x) is a
monotone increasing function and satisfies the condition L(0) = 1; R(x) which
is a monotone decreasing function and also satisfies the condition R(0) = 1 is
not necessarily symmetric to L(x). If L(x) and R(x) are linear functions, we can
call the fuzzy number trapezoidal fuzzy number. A trapezoidal fuzzy number
is determined by four parameters a, m, n, b; where R
0 = [a, b]. We denote a
trapezoidal fuzzy number as (a, m, n, b), and obtain a triangular fuzzy number
while m = n.
In this paper, we assume that the pessimistic opinion of an expert’s is re-
flected by the left-hand membership function while the optimistic opinion is
reflected by the right-hand one, and the most possible opinion is reflected by
the support set which is the closed interval [m, n]. Obviously, the closed inter-
val [m, n] is obtained when the pessimistic and optimistic opinions are decided.
Since the pessimistic opinion and the optimistic opinion are two entirely distinct
conceptions, the pessimistic opinion should be compared with the pessimistic
one. So should the optimistic one. Therefore, the similarity degrees between pes-
simistic opinions and the similarity degrees between optimistic ones are merely
considered in this paper, when we consider the similarity degrees of experts.

3 Distance and Similarity Between L-R Fuzzy Numbers

In order to obtain a rational distance, in this paper the left- and right-hand
membership functions of L-R fuzzy number are assumed to be strictly monotonic
functions. We have the theorem as follows:

Theorem 1. The inverse function of g(x) is also strictly monotonic if and only
if g(x) is a strictly monotonic function.

Distance is an important concept in science and engineering. In the following,

two types of distances’ definitions between two L-R fuzzy numbers based on the
same α-level are given.

Definition 1. Let R
i and R

j ,representing the subjective estimate of the ration
to an alternative under a given criterion of expert Ei and Ej , be two L-R fuzzy
numbers with membership functions µR
i (x) and µR
j (x). The left-hand side dis-
tance which reflects the difference of their pessimistic opinions at the same α-
level cut, is defined by:
−1 −1
ρLR

(α) =| µLR
(α) − µLR
(α) | (4)
i ,Rj i j
506 J. Lan, L. He, and Z. Wang

The right-hand side distance which reflects the difference of their optimistic opin-
ions at the same α-level cut, is defined by:
−1 −1
ρR


R
(α) =| µR

(α) − µR

(α) | (5)
i ,Rj R i R j

where α ∈ [0, 1]; µLR
and µR

i are the left- and right-hand membership functions
R
i

i while µ and µ are the left- and right-hand ones of R
of R L R
j , respectively.
R
j


R j

According to the theorem 1, the inverse function has only one value for each
α, and both side distances satisfy the conditions of non-negativity, symmetry
and triangle inequality. It has to be noted that either of the side distances is
−1 −1 −1 −1
equal to zero if and only if µLR
(α) = µLR
(α), or µR

i
R
(α) = µR

j
R
(α). It
i j
means that the two experts have the same degrees of membership at the same
point with the same opinion that is pessimistic or optimistic.
Now the side similarities between Ei and Ej at the same α-level are consid-
ered. It is well known that an exponential operation is highly useful in dealing
with a similarity relation, Shannon entropy and in cluster analysis. Therefore,
we choose
f (x) = e−x . (6)

The left- and right-hand side similarities between the two experts at the same
α-level are defined by:

−ρL

(α) −ρR
j (α)

i ,R
fRL

(α) = e Ri ,Rj
, fRR
(α) = e

,R
R
(7)
i ,Rj i j

where fRL
,R
(α) is named the left-hand side similarity while fRR
j (α) is named

i ,R
i j
the right-hand one.
Obviously, if either of side distances between them is zero then the corre-
sponding similarity is equal to one. Since the side similarities depend on the side
distances and the α-level, the curvilinear integral is taken into account. That is

 −ρL

(α) 

i , R
S L (R
j ) = e Ri ,Rj
ds/ ds
ρL (α) ρL (α)

i ,R
R
j
i ,R
R
j
 −ρR  (8)

i , R

j ) =
j (α)

i ,R
S R (R e R
ds/ ds
ρR (α) ρR (α)

i ,R
R
j
i ,R
R
j

i , R
where α ∈ [0, 1]. S L (R
j ) is named left-hand side similarity measure function
R


while S (Ri , Rj ) is named right-hand side similarity measure function.
We get S L (R
i , R

j ) = 1 and S R (R
i , R

j ) = 1 when expert Ei and expert Ej
have the same opinions, that is R
i = R

j . In other words, the opinions of expert
Ei and expert Ej are identical, and then the agreement degree between them is
equal to one.
 A New Method for Fuzzy Group Decision Making 507

Since the left- and right-hand side similarity measure functions have been
defined, now the side average agreement degrees of expert Ei (i = 1, 2, · · · , n) are
given by

1 
n
1 
n
A(EiL ) =
i , R
S L (R
j ), A(EiR ) =
i , R
S R (R
j ) (9)
n−1 n−1
j=1,j=i j=1,j=i

where A(EiL ) is named the left-hand side average agreement degree while A(EiR )
is named the right-hand one.
Now we compute the relative left- and right-hand side agreement degrees of
expert Ei (i = 1, 2, · · · , n) as follows:


n 
n
SALi = A(EiL )/ A(EjL ), SAR R
i = A(Ei )/ A(EjR ) (10)
j=1 j=1

where SALi is named the relative left-hand side agreement degree while SAR i is
named the right-hand one.
According to the above definitions, an expert’s opinion is determined by his
pessimistic and optimistic opinions. The relative agreement degree of expert’s
opinion depends on the side relative agreement degrees. We take the average
value of SALi and SARi as the relative agreement degree of expert’s opinion by:

SAi = (SALi + SAR

i )/2 . (11)

In practice, the group decision making is heavily influenced by the degrees

of importance of participants. Sometimes there are important experts in the
decision group, such as the executive manager of a company, or some experts
who are more experienced than others. The final decision is influenced by the
different importance of each expert. Therefore, a good method to aggregate
multi-experts’ opinions must consider the degree of importance of each expert
in the aggregating procedure. We employ Saaty’s [16] analytic hierarchy process
(AHP) to deal with the weight of each expert. The construction of the square
reciprocal matrix (aij )n×n is performed by comparing expert Ei with expert
Ej , with respect to the degree of importance. The other values are assigned
as follows: aij = 1/aji ; aii = 1. To solve the reciprocal matrix, the maximum
eigenvalue is cardinal ratio scale for the experts compared. The eigenvector is
then normalized and the weight γi (i = 1, 2, · · · , n) of each expert is obtained.
As is stated above, we get the relative agreement degree and the degree
of importance of each expert. Now the consensus degree coefficients of expert
Ei (i = 1, 2, · · · , n) can be defined by:

ωi = βSAi + (1 − β)γi (12)

where β ∈ [0, 1].

If β = 1, the degree of importance of expert is not considered in the aggre-
gation process. If β = 0, only the degree of importance of expert is reflected in
508 J. Lan, L. He, and Z. Wang

the consensus. The membership function of aggregating the consensus opinion

R
can be defined by:

µR
(z) = sup

n
min µωi xi (xi ), (i = 1, 2, · · · , n) . (13)
z= i=1 ωi x i

We summarize the criterion which is discussed above and propose an algo-

rithm to combine all experts’ opinion into the consensus opinion of group decision
making.

Algorithm 3.1
step 1: For the criterion and an alternative under group decision making
environment, each expert Ei (i = 1, 2, · · · , n) proposes his opinion as a L-R
fuzzy number denoted by R
i . Suppose the left- and right hand-side membership


functions of Ri are strictly monotonic.
step 2: Calculate the left- and right hand-side distances between R
i and R

j .
k


step 3: Calculate the side similarity degrees S (Ri , Rj ), (k = L, R) of the
opinions between each pair of experts.
step 4: Calculate the side average agreement degrees A(EiL ) and A(EiR ).
step 5: Calculate the side relative agreement degrees SALi and SAR i of expert
Ei (i = 1, 2, · · · , n).
step 6: Calculate the relative agreement degree SAi of expert Ei (i = 1,
2, · · · , ).
step 7: Define the degree of importance of expert Ei (i = 1, 2, · · · , n) by
employing Saaty’s AHP.
step 8: Calculate the membership function of the group consensus opinion
by equation(13).
The aggregation method preserves some important properties. These prop-
erties are as follows:
Corollary 1. If R
j for all i and j, then R

i = R
=R
i . In other words, if all
estimates are identical, the combined result should be the common estimate.

n n
Proof. If all R
i is equal, then R

= ωi  R
i = R
i [βSAi + (1 − β)γi ] =
i=1 i=1

i [β  SAi + (1 − β)  γi ] = R
n n
R
i [β + (1 − β)] = R

i .
i=1 i=1
Agreement preservation is a consistency requirement. 

Corollary 2. The result of the aggregation method would not depend on the
order with which individual opinions are combined. That is, if {(1), (2), · · · , (n)}

= f (R
is a permutation of {1, 2, · · · , n}, then R
1 , R

2 , · · · , R

n ) = f (R

(1) , R

(2) , · · · ,
R
(n) ). The result is also a consistency requirement.

Corollary 3. Let the uncertainty measure H(R
i ) of individual estimate R

i be
defined as the area under its membership function µR
i (x),
 ∞

i ) =
H(R µR
i (x)dx (14)
−∞
 A New Method for Fuzzy Group Decision Making 509

The uncertainty measure H is defined to fulfil the following equation.


n

=
H(R)
i ) .
ωi × H(R (15)
i=1

Corollary 4. If an expert’s estimate is far from the consensus, then his estimate
is less important.

Corollary 5. The common intersection of supports of all experts’ estimates is

n
the aggregation result, namely
i ⊆ R.
R

i=1

n

i be R
Proof. Let α − cut of R
α = [aα , bα ]; let
i be [aα , bα ], and then
R
i i i
i=1

α =  ωi  R

α = [  ωi aα ,  ωi bα ].
n n n
R i i i
i=1 i=1 i=1

n 
n
Since ω i aα
i ≤ max{aα
i } ≤ aα and i ≥ min{bi } ≥ b , we have
ωi bα α α
i=1 i i=1 i
n

i ⊆ R.
R


i=1

n
Corollary 6. If
i = φ, a consensus also can be derived.
R
i=1

4 Numerical Example
Example 1. Consider a group decision problem evaluated by three experts. The
experts’ opinions are represented as trapezoidal fuzzy numbers as follows:

1 = (1, 2, 3, 4), R
R
2 = (1.5, 2.5, 3.5, 5), R

3 = (2, 2.5, 4, 6).

We employ our method to deal with this problem and consider two cases:
1. Do not consider the degree of importance of expert; i.e. β = 0
2. Consider the degree of importance of each expert; i.e. 0 < β < 1 .

is calculated in full details as follows:
The result of R
Case 1: Do not consider the importance degree of each expert.

i and R
step 2: Calculate the left- and right hand-side distances between R
j .

The left hand-side distances are as follows:

ρLR

(α) = ρLR

(α) = ρLR

(α) = 0;
1 ,R1 2 ,R2 3 ,R3
−1 −1
ρLR

(α) = ρLR

(α) =| µLR
(α) − µLR
(α) |= 12 ;
1 ,R2 2 ,R1 1 2
L −1 L −1
ρLR

(α) = ρLR

(α) =| µR
(α) − µR
(α) |= 1 − α2 ;
1 ,R3 3 ,R1 1 3
−1 −1
ρLR

(α) = ρLR

(α) =| µLR
(α) − µLR
(α) |= 1
2 − α2 .
2 ,R3 3 ,R2 2 3
510 J. Lan, L. He, and Z. Wang

The right hand-side distances are as follows:

ρR


R
(α) = ρR


(α) = ρR


(α) = 0;
1 ,R1 R2 ,R2 R 3 ,R3
−1 −1
ρR


R
(α) = ρR


(α) =| µR
1 (α) − µR
2 (α) |= 1 − α2 ;
1 ,R2 R2 ,R1 R R
R −1 R −1
ρR


R
(α) = ρR


(α) =| µR
(α) − µR
(α) |= 2 − α;
1 ,R3 R3 ,R1 1 3
−1 −1
ρR


R
(α) = ρR


(α) =| µR
2 (α) − µR
3 (α) |= 1 − α2 .
2 ,R3 R3 ,R2 R R

step 3: Calculate the side similarity degrees S k (R
i , R

j ), (k = L, R) of the
opinions between each pair of experts.

1 , R
S L (R
1 ) = S L (R

2 , R

2 ) = S L (R

3 , R
3 ) = 1;
R

1 ) = S (R
R
2 , R

2 ) = S (R
R
3 , R

3 ) = 1;
S (R1 , R
L

L


1 , R

2 ) = S R (R
2 , R
1 ) = 0.4773;
S (R1 , R2 ) = S (R2 , R1 ) = 0.6065, S R (R
L

L

R

R

1 ) = 0.2325;
S (R1 , R3 ) = S (R3 , R1 ) = 0.4773, S (R1 , R3 ) = S (R3 , R

2 , R
S L (R
3 ) = S L (R

3 , R

2 ) = 0.7869, S R (R
2 , R

3 ) = S R (R
3 , R
2 ) = 0.4773.
step 4: Calculate the side average agreement degrees A(EiL ) and A(EiR ).
A(E1L ) = 0.5419,A(E2L) = 0.6967,A(E3L ) = 0.6321;
A(E1R ) = 0.3549,A(E2R ) = 0.4773,A(E3R ) = 0.3549.
step 5: Calculate the side relative agreement degrees SALi and SAR
i of expert
Ei (i = 1, 2, · · · , n).
SAL1 = 0.2897,SAL2 = 0.3724,SAL3 = 0.3379;
SAR R R
1 = 0.2990,SA2 = 0.4020,SA3 = 0.2990.

step 6: Calculate the relative agreement degree SAi of expert Ei (i = 1,

2, · · · , n).
SA1 = 0.2943, SA2 = 0.3873, SA3 = 0.3184.
Because we do not consider the degree of importance of each expert in this
case (β = 0), the consensus degree coefficients of the experts E1 , E2 and E3 are
ω1 = SA1 = 0.2943,ω2 = SA2 = 0.3873,ω3 = SA3 = 0.3184.
The “overall” fuzzy number of combing experts’ opinions is
R
= ω1  R
1 + ω2  R

2 + ω3  R

3 = (1.5121, 2.3528, 3.5121, 5.0241).
Case 2: Consider the degree of importance of experts.
Suppose that the degrees of importance of each expert are γ1 = 0.42, γ2 =
0.25,and γ3 = 0.33 by employing Saaty’s AHP. We take β = 0.6; the aggregation
coefficients of the experts E1 , E2 and E3 can be computed as
ω1 = 0.6SA1 + 0.4γ1 = 0.3446, ω2 = 0.6SA2 + 0.4γ2 = 0.3324,
ω3 = 0.6SA3 + 0.4γ3 = 0.3230.
The “overall” fuzzy number of combining experts’ opinions is
R
= ω1  R
1 + ω2  R

2 + ω3  R

3 = (1.4892, 2.3277, 3.4892, 4.9785).

is smaller than the one using Hus’
In this example, the width of the result R
method, and the uncertainty of the aggregation result for Hus’ method in case
 A New Method for Fuzzy Group Decision Making 511

= 2.341) and in case 2 (H(R)

1 (H(R)
= 2.33)is larger than the one using our

= 2.3356) and in case 2 (H(R)
method in case 1 (H(R)
= 2.3253).

Example 2. Consider a group decision making problem with three experts. The
datum of the experts’ opinions are given as follows:
µR
1 (x)=1 − 4(5 − 4x)2 , 98 ≤ x ≤ 11
8 ; µR
2 (x)=1 − 4( 52 − 4x)2 , 48 ≤ x ≤ 68 ;
µR
3 (x)=1 − (3 − 4x)2 , 12 ≤ x ≤ 1.
We employ our method to deal with this problem and assume β = 0

i and R
step 2: Calculate the left- and right hand-side distances between R
j .
The left hand-side distances are as follows:
ρLR

(α) = ρLR

(α) = ρLR

(α) = 0;
1 ,R1 2 ,R2 3 ,R3
−1 −1
ρLR

(α) = ρLR

(α) =| µLR
(α) − µLR
(α) |= 58 ;
1 ,R2 2 ,R1 1 2
√
L −1 L −1
ρLR

(α) = ρLR

(α) =| µR
(α) − µR
(α) |= 1
2 + 1−α
8 ;
1 ,R3 3 ,R1 1 3
√
−1 −1
ρLR

(α) = ρLR

(α) =| µLR
(α) − µLR
(α) |= 1
8 − 1−α
8 .
2 ,R3 3 ,R2 2 3

The right hand-side distances are as follows:

ρR


R
(α) = ρR


(α) = ρR


(α) = 0;
1 ,R1 R 2 ,R2 R 3 ,R3
−1 −1
ρR


R
(α) = ρR


(α) =| µR

1 (α) − µR
2 (α) |= 58 ;
1 ,R2 R 2 ,R1 R R
√
−1 −1
ρR


R
(α) = ρR


(α) =| µR

1 (α) − µR
3 (α) |= 1
2 − 1−α
8 ;
1 ,R3 R 3 ,R1 R R
√
−1 −1
ρR


R
(α) = ρR


(α) =| µR

2 (α) − µR
3 (α) |= 1
8 + 1−α
8 .
2 ,R3 R 3 ,R2 R R

i , R
step 3: Calculate the side similarity degrees S k (R
j ), (k = L, R) of the
opinions between each pair of experts.

1 , R
S L (R
1 ) = S L (R

2 , R

2 ) = S L (R

3 , R

3 ) = 1;
R

R

R

3 ) = 1;
S (R1 , R1 ) = S (R2 , R2 ) = S (R3 , R

1 , R
S L (R
2 ) = S L (R

2 , R

1 ) = 0.5353, S R (R
1 , R

2 ) = S R (R

2 , R

1 ) = 0.5587;
L

3 ) = S (R
L
3 , R

1 ) = 0.9588, S (RR
1 , R

3 ) = S (R
R
3 , R

1 ) = 0.5353;
S (R1 , R
L

L

R

R

2 ) = 0.8130.
S (R2 , R3 ) = S (R3 , R2 ) = 0.6590, S (R2 , R3 ) = S (R3 , R
step 4: Calculate the side average agreement degrees A(EiL ) and A(EiR ).
A(E1L ) = 0.5470,A(E2L) = 0.7471,A(E3L ) = 0.7588;
A(E1R ) = 0.5971,A(E2R ) = 0.6741,A(E3R ) = 0.7360.
step 5: Calculate the side relative agreement degrees SALi and SAR
i of expert
Ei (i = 1, 2, · · · , n).
SAL1 = 0.2665,SAL2 = 0.3639,SAL3 = 0.3696;
SAR R R
1 = 0.2975,SA2 = 0.3358,SA3 = 0.3667.

step 6: Calculate the relative agreement degree SAi of expert Ei (i = 1,

2, · · · , n).
512 J. Lan, L. He, and Z. Wang

SA1 = 0.2820, SA2 = 0.3499, SA3 = 0.3681.

Because we do not consider the degree of importance of each expert in this
case (β = 0), the consensus degree coefficients of the experts E1 , E2 and E3 are
ω1 = SA1 = 0.2820, ω2 = SA2 = 0.3499, ω3 = SA3 = 0.3681.
The membership function of the group consensus opinion is
µR
(x) = 1 − 34.1911(x − 0.8473)2
In this example, the opinions of experts are not expressed by trapezoidal
fuzzy numbers, so Lee’s method can’t deal with it. Meanwhile, Hsu’s method
can’t be also effective because R1 has obviously no common intersection with
R2 and R3 .

5 Conclusion

In this paper, the problem aggregating individual opinions into group consensus
under group decision environment is addressed. A simple similarity measure to
deal with the L-R fuzzy numbers has been employed. The distance and similarity
function is proposed. The degree of importance of each expert is taken into
account further. Meanwhile, the membership function and the pessimistic and
optimistic opinions of experts’ are considered in the method. This aggregation
method preserves some important properties which other aggregation methods
processed.

References
1. A.Bardossy, L.Duckstein and Bogardi: Combination of fuzzy numbers representing
expert opinions. Fuzzy Sets and Systems. 57(1993)173-181
2. P.Diamond, P.Kloeden: Metric Spaces of Fuzzy Sets: Theory and Application.
World Scientific, Singapore. 1994
3. H.S.Lee: Optimal consensus of fuzzy opinions under group decision making envi-
ronment. Fuzzy Sets and Systems. 132(2002)303-315
4. Jiulun Fan and Weixin Xie: Distance measure and induced fuzzy entropy. Fuzzy
Sets and Systems. 104(1999)305-314
5. M.Fedrizzi and J.Kacprzyk: On measuring consensus in the setting of fuzzy prefer-
ence relations, in: J.Kacprayk and M.Roubens, Eds. Non-conventional preference
Relations in Decision Making (Springer, Berlin, 1988). 129-141
6. R.Goetschel, W.Voxman: Topological Properties of Fuzzy Sets. Fuzzy Sets and
Systems. 10(1983) 87-99
7. H.M.Hsu, C.T.Chen: Aggregation of fuzzy opinions under group decision making.
Fuzzy Sets and Systems. 79(1996)279-285
8. A.Ishikawa, M.Ambiguous, T.Shiga, G,Tomizawa, R. Tactic and H.Mileage: The
max-min Delpi method and fuzzy Delphi method via fuzzy intergration. Fuzzy sets
and Systems. 55 (1993)241-253
9. J.Kacprzyk and M.Federation: A soft measure of consensus in the setting of par-
tial(fuzzy) preferences. Eur.J.OR. 34(1988)315-325
 A New Method for Fuzzy Group Decision Making 513

10. J.Kacprzyk, M.Federation and H.Norm: Group decision making and consensus un-
der fuzzy preferences and fuzzy majority. Fuzzy Sets and Systems. 49(1992)21-31
11. O.Kaleva, S.Siekkala: On fuzzy metric spaces. Fuzzy Sets and Systems.
12(3)(1987)301-317
12. G.J.Klir, B.Yuan:Fuzzy Sets and Fuzzy Logic: Theory and Applications, Prentice
Hall. Englewood Cliffs, NJ. 1995
13. L.T.Koczy, K.Hirota: Ordering and closeness of fuzzy sets. Fuzzy Sets and Systems.
90(1997)103-111
14. V.B.Kuz’min: A parametric approach to the description of linguistic variables and
hedges. Fuzzy Sets and Systems. 6(1981)27-41
15. H.Nurmi: Approaches to collective decision making with fuzzy preference relations.
Fuzzy Sets and Systems. 6(1981)249-259
16. T.L. Saaty: Modeling unstructured decision problems-the theory of analytical hi-
erarchies. Math. Comput. Simulation. 20(1978)147-158
17. T.Tanino: On group decision making under fuzzy preferences, in: J.Kacprzyk,
M.Fedrizzi Eds, Multiperson Decision Making Using Fuzzy Sets and Prossibility
Theory. Kilowatt Academic Publishers, Dordrecht. (1990)172-185
18. Jon Williams and Nigel Steele: Difference, distance and similarity as a basis for
fuzzy decision support based on prototypical decision classes. Fuzzy Sets and Sys-
tems. 131(2002)35-46
19. R.N.Xu and X.Y.Zhai:Extensions of the analytic hierarchy process in fuzzy envi-
ronment. Fuzzy Sets and Systems. 52(1992)251-257