Proceedings of the 2012 International Conference on Industrial Engineering and Operations Management

Istanbul, Turkey, July 3-6, 2012

An α -Cut Approach for Fuzzy Product and Its Use in Computing

Solutions of Fully Fuzzy Linear Systems

R. Hassanzadeh and I. Mahdavi

Department of Industrial Engineering
Mazandaran University of Science and Technology
Babol, Iran

N. Mahdavi-Amiri
Faculty of Mathematical Sciences
Sharif University of Technology
Tehran, Iran

A. Tajdin
Department of Industrial Engineering
Mazandaran University of Science and Technology
Babol, Iran

Abstract

We propose an approach for computing the product of various fuzzy numbers using α-cuts. A regression model is used to
obtain the membership function of the product. Then, we make use of the approach to compute solutions of fully fuzzy linear
systems. We also show how to compute solutions of fully fuzzy linear systems with various fuzzy variables. Examples are
worked out to illustrate the approach.

Keywords
Fuzzy numbers; α-cut; regression model; fuzzy product; fully fuzzy linear system (FFLS)

1. Introduction
Over the years, a number of researchers have focused on computing various approximations of product of fuzzy numbers
(see, e.g., Ban 2008 and 2009, Mrówka 2005, Nasibov and Peker 2008, Yeh 2009, Ghanbari et al. 2010, Ezzati to appear ).
Here, we propose a new approach for computing fuzzy product of fuzzy numbers using α-cuts. The fuzzy numbers may be of
the same type or mixed. Linear systems arise from many areas of science and engineering. Moore (1979) pointed out that
exact numerical data might be unrealistic, but uncertain data can consider more aspects of a real world problem. A popular
way to describe uncertain data is using fuzzy data. Fuzzy linear systems under various assumption have been studied.
Buckley et al. (1991 and 2002) studied square fuzzy linear systems, Ax = b, where the components of A and b are fuzzy
numbers. Since an algebraic solution does not exist, they proposed three other types of solutions (based on α -cuts, extension
principal and interval arithmatic). Vorman et al. (2007) improved Buckely and Qu’s method (1991) by obtaining the
necessary parametric functions needed for the solution of the systems corresponding to the bounds of the considered α-level.
Muzzioli and Reynaerts (2006) provided a generalization of the vector solution proposed by Buckley and Qu (1991) to the
fuzzy system A1 x + b1 = A2 x + b2 , where the components of A1 , A2 , b1 and b2 are triangular fuzzy numbers and A1 and
A2 are square matrices. They applied their method to solve and analyze a particular fuzzy linear system having financial
applications (Muzzioli and Reynaerts 2007).

A general model for solving a fuzzy linear system whose coefficient matrix is crisp and the right-hand side column is an
arbitrary fuzzy vector was first proposed by Friedman et al. (1998). Friedman and his colleagues used the embedding method
and replaced the original fuzzy linear system by a crisp linear system. Following their work, Dehghan and Hashemi (2006)

2047
proposed some iterative methods. Ghanbari et al. (2010) have recently showed shortcomings of the approach proposed by
Friedman et al. (1998) and proposed anew approach for computing the solutions based on a least squares model.

As noticed by Rao and Chen (1998) and Muzzioli and Reynaerts (2006), although several investigations have been reported
in the literature for the solution of fuzzy systems (Buckley and Qu 1991, Zhao and Govind 1991, Kawaguchi and Da-Te
1993), very few methods are in fact practical. The main reason for this drawback lies in the structure at the problem. To be
more precise, as pointed out by Kreinovich et al. (1998), finding a solution of a system with interval coefficient is NP-hard.
Here, we discuss the case in which all parameters in a fuzzy linear system are fuzzy numbers, calling it a Fully Fuzzy Linear
~ ~ ~ ~
System (FFLS). We intend to propose an approach for solving A ⊗ ~ x = b , where A is a fuzzy matrix and ~ x and b are
fuzzy vectors. Rao and Chen (1998) gave a computational method to solve this problem, based on their proposed cuts. But a
computational method using Zadeh’s extension principle is not at hand (see, e.g., Buckley and Qu 1991). Since there is no
analytic formula for some arithmetic operators, Doubois and Prade (1980) proposed some approximations for arithmetic
operators lacking analytic formulas on a special and useful type of fuzzy numbers, that are used by other researchers
Wagenknecht et al. 2001, Sakawa 1973, Giachetti and Young 1997.

In Section 2, basic definitions are given. Section 3 describes α -cuts. The design of fuzzy product operator on fuzzy numbers
is described in Section 4. Solutions of fully fuzzy linear systems are described in Section 5. We conclude in Section 6.

2. Definitions
To arrive at the product of various types of fuzzy numbers, we start with basic definitions of some well-known fuzzy
numbers.
~
Definition 1. An LR fuzzy number is represented by A = ( m, a, b) LR , with the membership function, µ a~ ( x ) , defined by

 m − x
L  a  x≤m
  
µ a~ ( x) = 
 R  x − m  x ≥ m,
  
  b 
where, L and R are non-increasing functions from R+ to [0,1] , L(0)=R(0)=1, m is the center, a is the left spread and b is
the right spread [18].
~ = (m, a, b)
Note that if L(x) = R(x)=1-x with 0<x<1, then x is a triangular fuzzy number and is represented by the triplet a
, with the membership function µ a~ ( x ) , defined by
 m−x
1 − ( a ) x≤m
µ a~ ( x) = 
x−m
 1− ( ) x ≥ m.
 b
Definition 2. A trapezoidal fuzzy number a~ is shown by a~ = (a1 , a 2 , a3 , a 4 ) , with the membership function as follows:
 0 x ≤ a1
 x − a1
 a1 ≤ x ≤ a2
 a − a1
 2
µ a~ ( x ) =  1 a2 ≤ x ≤ a3
 a4 − x
a − a a3 ≤ x ≤ a4
 4 3

 0 a4 ≤ x.
It is apparent that a triangular fuzzy number is a special trapezoidal fuzzy number with a2 = a3 .
−x 2
Definition 3. If L(x)=R(x)= e , with x ∈ ℜ , then x is a normal fuzzy number that is shown by (m, σ ) and the
corresponding membership function is defined to be:
x−m
−( )2
µ a~ ( x) = e σ , x∈ℜ ,

where, m is the mean and σ is the standard deviation.

2048
 ~ ~ ~
Definition 4. The α-cut and strong α-cut for a fuzzy set A are shown by Aα and Aα+ , respectively, and for α ∈ [0,1] are
defined to be:
~
{
Aα = x µ A~ ( x) ≥ α , x ∈ X , }
~
{
Aα+ = x µ A~ ( x) > α , x ∈ X , }
where, X is the universal set.
~ ~ ~
Upper and lower bounds for any α-cut Aα are shown by sup Aα and inf Aα , respectively. Here, we assume that the upper
~ ~ R and ~ ~
and lower bounds of α-cuts are finite values and for simplicity denote sup Aα by Aα
inf Aα by AαL .

Definition 5. Let A = (a~ij ) and B = bij be

~ ~ ~
( ) m×n
~ ~ ~
and n × p fuzzy matrices respectively. We define A ⊗ B = C to be
an m × p matrix, where,
n
~
c~ij = ∑ (a~ik ⊗ bkj ).
k =1
Definition 6. Consider the n×n linear system of equations:
(a~11 ⊗ ~
x1 ) ⊕ (a~12 ⊗ ~
x 2 ) ⊕ ⋅ ⋅ ⋅ ⊕ (a~1n ⊗ ~x n ) = b1 ,
~

(a~21 ⊗ ~
x1 ) ⊕ (a~22 ⊗ ~x 2 ) ⊕ ⋅ ⋅ ⋅ ⊕ (a~2 n ⊗ ~
x n ) = b2 ,
~

.
.

(a~n1 ⊗ ~
x1 ) ⊕ (a~n 2 ⊗ ~
x 2 ) ⊕ ⋅ ⋅ ⋅ ⊕ (a~nn ⊗ ~x n ) = bn .
~

In matrix form, the above equations are shown by
~ ~
A⊗~ x = b,
~
or simply A~
~ ~ ~
x = b , where the coefficient matrix A = a ( ) ~
ij , 1 ≤ i , j ≤ n is an n × n fuzzy matrix and x and b are n × 1
~
fuzzy vectors. This system is called a fully fuzzy linear system (FFLS).

3. Computing α-Cuts
One approach for multiplication of fuzzy numbers is to employ α-cuts. Here, using our recently developed approach for
addition of mixed fuzzy numbers in Tajdin (2010), we describe a methodology to multiply various fuzzy numbers using α-
cuts.
For an LR fuzzy number with L and R invertible functions, the α-cut is characterized by (see Fig. 1):
m−x m−x ~
α = L[ ] ⇒ = L−1 (α ) ⇒ AαL = x = m − aL−1 (α ),
a a
x−m x−m ~
α = R[ ] ⇒ = R −1 (α ) ⇒ AαR = x = m + bR −1 (α ).
b b
~ ~ ~
Thus, Aα = [ AαL , AαR ] is the corresponding α-cut.
µ A ( x )

1
α

0 x
Aα− Aα+
Figure 1. An example of an α-cut

Next, for specific L and R functions, the following cases are discussed.

2049
3.1. α-cuts for trapezoidal fuzzy numbers
Let a~ = (a , a , a , a ) be a trapezoidal fuzzy number. An α-cut for a~ , a~α is computed as:
1 2 3 4

x − a1
α= ⇒ a~αL = x = (a 2 − a1 )α + a1 ,
a 2 − a1
(1)
a −x
α= 4 ⇒ a~αR = x = a 4 − (a 4 − a 3 )α ,
a 4 − a3

~ = [a~ L , a~ R ] is the corresponding α-cut (see Fig. 2). The α-cuts for triangular fuzzy numbers are obtained by
where, aα α α

using the above equations considering a2 = a3 .

Figure 2. A trapezoidal fuzzy number with an α-cut

3.2. α-cuts for normal fuzzy numbers

~ = (m, σ ) is a normal fuzzy number, then
If a a~α , 0 < α ≤ 1, is computed as:

m− x 2
−( ) m−x ~ L = x = m − σ − ln α
α =e σ
⇒ − ln(α ) = ⇒ a α
σ (2)
x−m 2
−( ) x−m ~ R = x = m + σ − ln α ,
α =e σ
⇒ − ln(α ) = ⇒ a α
σ
where, aα α α[ ]
~ = a~ L , a~ R , with 0 < α ≤ 1, is the corresponding α-cut.

4. Fuzzy Product Operator

We now propose an approach for the product of mixed fuzzy numbers using α -cuts. Divide the interval [0,1] into n equal
subintervals using n + 1 equidistant points α 0 = 0 , α i = α i −1 + ∆α i , i = 1,..., n, with ∆α i = 1 , i = 1,..., n . For normal
n
α be equal to zero. Therefore, for this case we consider α ∈ (0,1] .
fuzzy numbers x ∈ ( −∞, + ∞) , it is improper to let
~ ~ ~ = [a~ − , a~ + ] and b~ = [b~ − , b~ + ] . The product of the
Assume two fuzzy numbers a and b with corresponding α-cuts a α α α α α α
α -cuts are:
~
a~α × bα = [ L , R ],
where,
~ ~ ~ ~
L = min (a~α− × bα− , a~α− × bα+ , a~α+ × bα− , a~α+ × bα+ ),
~ ~ ~ ~ (3)
R = max (a~ − × b − , a~ − × b + , a~ + × b − , a~ + × b + ).
α α α α α α α α

2050
Next, we show how to add up a trapezoidal fuzzy number together with a normal one. A numerical approach is given to
approximate the sum and its corresponding membership function.

Example 1: Consider the following two trapezoidal fuzzy numbers:

~
a~ = (1, 5, 6, 9) , b = (2, 3, 5, 8).
~
The diagram of their product, a~ × b , using equation (3), for 100 α i -cuts, i = 1,...,100, is shown in figure 3. As seen in the
diagram, the result is not exactly trapezoidal.

Figure 3. The product of fuzzy numbers using α-cuts Figure 4.The product of fuzzy numbers using α-cuts

Example 2: consider the following two normal fuzzy numbers:

~
a~ = (40,4) , b = (50,9) .
~
~ × b , using (3), for 100 α -cuts, i = 1,...,100 , is shown in figure 4. As seen in the diagram,
The diagram of their product, a i
the result appears to be normal.
In general, the membership function for 𝑎� × 𝑏� can be computed as follows:
~
Consider two nonnegative fuzzy numbers a~ = (a1 , a2 , a3 , a4 ) and b = (b1 , b2 , b3 , b4 ) . The end points of the
α-cuts for them are:

a~αL = x = (a 2 − a1 )α + a1 , ~
bαL = x = (b2 − b1 )α + b1 ,
~ R ~ R
aα = x = a 4 − (a 4 − a3 )α bα = x = b4 − (b4 − b3 )α .
Thus, we have
~
c~αL = a~αL × bαL = [(a 2 − a1 )α + a1 ][(a 2 − a1 )α + a1 ] = Aα 2 + Bα + C , (4)
where, A = (a 2 − a1 )(b2 − b1 ) , B = b1 (a 2 − a1 ) + a1 (b2 − b1 ) and C = a1b1 .
Now, using (4), we have
~ B B2 B2 4 Ax − 4 AC + B 2 − B
c~αL = a~αL × bαL = x = A(α 2 + α + − )+C ⇒ α = . (5)
A 4 A2 4 A2 2A
Also, for the right side of the α-cuts, we have
~
c~αR = a~αR × bαR = A′α 2 + B ′α + C ′, (6)
where, A′ = (a 4 − a3 )(b4 − b3 ) , B ′ = b4 (a3 − a 4 ) + a 4 (b3 − b4 ) and C ′ = a 4 b4 .
Using (6), we have
~ B′ B′2 B′2 4 A′x − 4 A′C ′ + B ′ 2 − B ′
c~αR = a~αR × bαR = x = A′(α 2 + α + − ) + C ′ ⇒ α = . (7)
A′ 4 A′ 2 4 A′ 2 2 A′
Therefore, using (5) and (7), the membership function for the product of the two nonnegative fuzzy numbers is:

2051
 0 x ≤ a1b1

 4 Ax − 4 AC + B − B
2

 a1b1 < x < a 2 b2

2A

µ c~ ( x) = 1 a 2 b2 ≤ x ≤ a 3 b3

 4 A′x − 4 A′C ′ + B ′ − B ′
2
a 3 b3 < x < a 4 b4
 2 A′

0 x ≥ a 4 b4 ,
where, for Example 1, the membership function is:
0 x≤2

 16 x + 49 − 9 2 < x < 15
 8

µ c~ ( x) = 1 15 ≤ x ≤ 30

 36 x + 7 − 51 30 < x < 72
 18
0 x ≥ 72.

~ = (m , σ ) and b~ = (m , σ ) . The end points of the
Next, consider two nonnegative normal fuzzy numbers a 1 1 2 2
α-cuts for them are:
~
 a~αL = m1 − σ 1 − ln α bαL = m2 − σ 2 − ln α
~ R , ~ R , 0 < α ≤ 1.
aα = m1 + σ 1 − ln α bα = m2 + σ 2 − ln α
Then,
~
c~αL = a~αL × bαL = x = (m1 − σ 1 − ln α )(m2 − σ 2 − ln α ) = A( − ln α ) 2 + B − ln α + C , (8)
where, A = σ 1σ 2 , B = −(m1σ 2 + m2σ 1 ) and C = m1 m2 .
Using (8), we have
4 Ax − 4 AC + B 2 + B 2
−( )
α =e 2A
. (9)
Also, for the right end of the α-cuts, we obtain:
4 Ax − 4 AC + B 2 − B 2
−( )
α =e 2A
. (10)
Using (9) and (10), the membership function for the product of the two nonnegative normal fuzzy numbers is:

 −( 4 Ax − 4 AC + B 2 + B 2
)
e 2A
x ≤ m1 m2
µ c~ ( x) = 
4 Ax − 4 AC + B − B 2
 −(
2
)
e 2A
x > m1 m2 ,
where, for Example 2, the membership function is:
 −( 144 x + 25600 −560 2
)
e 72
x ≤ 2000
µ c~ ( x) = 
144 x + 25600 + 560 2
 −( )
e
72
x > 2000.
We now show how to multiply a trapezoidal fuzzy number with a normal one. We present a numerical approach to construct
a model for the product and its corresponding membership function.
~
Let a~ = (a1 , a2 , a3 , a4 ) and b = (m, σ ) be the trapezoidal and normal fuzzy numbers, respectively. Given α i ∈ (0,1],
the αi-product of these fuzzy numbers using equations (1) and (2) is obtained to be:
~
a~α × bα = [ L , R ],
where,

2052
 ~ ~ ~ ~
L = min (a~α− × bα− , a~α− × bα+ , a~α+ × bα− , a~α+ × bα+ ),
(11)
~ ~ ~ ~
R = max (a~ − × b − , a~ − × b + , a~ + × b − , a~ + × b + ).
α α α α α α α α

According to (11), using α i = α i −1 + ∆α i , α 0 = 0, and ∆α i =

1 ~
, i = 1,..., n, 2n points are gained for c , n points for the
n
c~α+i and n points for the c~α−i . The membership function of the product is obtained using the resulting points via α-cuts and a
regression method for fitting a membership function to the product. We propose an exponential membership function for
approximating the product as follows (later, we will see that this choice would indeed provide a good model for the
approximating product of trapezoidal and normal fuzzy numbers). Let xi = c~αRi and y i = µ (c~αRi ) , and for the n points
x −λ
−( )2
β
( xi , y i ) , consider the fitting model to be y = e . As defined, the unknown parameters λ and β appear
nonlinearly. To linearize the model, we note that for any x i > λ , we must have
xi − λ
ln y i = −( )2 . (12)
β
Since 0 < y i ≤ 1 , then ln( y i ) ≤ 0, and thus we can write
xi − λ ,
− ln y i = ( ) (13)
β
and hence
β − ln y i + λ = xi . (14)
Therefore, the problem of least squares for the minimization of error is:
n
min E = ∑ ( β − ln y i + λ − xi ) 2 , (15)
i =1

where, xi = c~αRi is given by (11). To solve (15), we need to have:

∂E
∂β
[
= ∑ (−2 − ln y i )( x i − β − ln y i − λ ) = 0 , ] (16)
i

∂E
∂λ
[
= ∑ − 2( xi − β − ln y i − λ ) = 0, ]
i
and thereby, the following so-called normal equations are to be solved:

β ∑ ln y i − λ ∑ − ln y i = −∑ xi − ln y i , (17)
i i i

− β ∑ − ln y i − nλ = −∑ xi . (18)
i i

From (17) and (18), λ and β are explicitly determined to be:

− ∑ x i − ln y i − ∑ − ln y i
i i
− ∑ xi −n n∑ ( x i × − ln y i ) − ∑ − ln y i × ∑ x i , (19)
β= i
⇒β = i i i

∑ ln y i − ∑ − ln y i − n∑ ln y i − ∑ − ln y i × ∑ − ln y i
i i i i i

− ∑ − ln y i −n
i

2053
 ∑ ln y i −∑ ( − ln y i × x i )
∑ ln y ( − ∑ x ) − ∑ ( )  
i i
− ∑ − ln y i − ∑ xi i i − ln y i × x i ×  ∑ − ln y i  (20)
 i 
λ= ⇒λ =
i i i i i
.
∑ ln y i − ∑ − ln y i − n∑ ln y i − ∑ − ln y i × ∑ − ln y i
i i i i i

− ∑ − ln y i −n
i
λ ′− x
~ L and y = µ (c~ L ) , and consider the model y = e −(
Now, similarly let xi = c
β′
)2
. Using the above approach, λ′ and
αi i αi
β ′ are found to be:
λ ′ − xi
ln yi = −( ), (21)
β′
n∑ xi − ln y i − ∑ xi × ∑ − ln y i
β′ = i i i , (22)
n∑ ln y i + ∑ − ln y i × ∑ − ln y i
i i i

∑ ln y × ∑ x + ∑ (x × − ln y )× ∑ − ln y
i i i i i
λ′ = i i i i . (23)
n∑ ln y + ∑ − ln y × ∑ − ln y i i i
i i i
Thus, the membership function is determined to be:
 −( λβ′−′ x ) 2
e , x < λ′

µ c~ ( x) =  1, λ′ ≤ x ≤ λ (24)
 −( x −λ ) ) 2
e β , x > λ,

with λ , β , λ ′, and β ′ as defined by (19), (20), (22), and (23), respectively.
Next, we provide a numerical illustration of our proposed model for the product of trapezoidal and normal fuzzy numbers.

Example 3: Consider the following two fuzzy numbers, one being normal and the other being trapezoidal:
~
a~ = (5,9) , b = (1,2,4,6) .
x −λ
−( )2
, where λ and β are the mean and standard deviation,
β
For the normal fuzzy number, we consider y = e
~
~ × b , using equation (11), is shown in figure 5, respectively. As appeared, the
respectively. The diagrams of their product, a
result is neither trapezoidal nor normal.
Using the points obtained from the αi-cuts considering n = 100 , the amounts of (λ , β ) for the right side and (λ ' , β ' ) for
the left side are obtained by solving the corresponding least squares problems of the regression model.
To do this, we used the function CURVEFITTING in MATLAB. The obtained result is:

 −( λβ'−' x ) 2
e , x < λ'

µ c~ ( x) =  1, λ'< x < λ
 −( x −βλ ) )2
e , x > λ,


2054
  10− x
 −( 9.62 )2
e , x ≤ 10

µ c~ ( x) = 1 , 10 ≤ x ≤ 20,
 x − 20 2
 −( ) , x ≥ 20.

e 10.24
The obtained membership function using the least squares method and the product of fuzzy numbers using α-cuts is shown in
figure 6.

Figure 5. The product of fuzzy numbers using α-cuts Figure 6. The obtained membership function using the
least squares method and the product of fuzzy numbers
using α-cuts

5. Solving Fully Fuzzy Linear Systems

To conceptualize the solution method, consider the following 2 × 2 system,
~
a~x  e~ 
~ b
~~  =  ~ ,
~
y  f 
c d 
~, b, ..., ~f are known fuzzy numbers of arbitrary types and ~
where, a x, ~y are unknown fuzzy numbers (without loss of
generality, all fuzzy numbers are considered to be nonnegative). For any α-cuts, a corresponding system is configured as
follows:
[
 a~α− , a~α+ ] [b~
α
−~
, bα+ ] [~x
α ,~ ] [
xα+   e~α− , e~α+
−
] .
 −  =  ~− ~+
[
 c~α , c~ +
α ] [d~− ~+
α , dα ] [~y
−
α , ~
y +
α  ] [
  f α , f α ]
Using the fuzzy addition and product, we have
 [~− ~+ ] [
~− ~+ ~−
( aα , aα ⊗ xα , xα ) ⊕ ( bα , ] [ ~
bα+ ⊗ ~ ] [
yα− , ~
yα+ ) = ] [e~ α
−
, e~α+]
 ~− ~+
 [ ] [
~− ~+ ~−
( cα , cα ⊗ xα , xα ) ⊕ ( d α , ] [ ~
d α+ ⊗ ~ ] [
yα− , ~
yα+ ) = ] [~f − ~+
α , fα . ]
Then, using (3), this system is turned into a 2 × 2 system with real numbers as follows:
~ − ~− ~ + ~+
a~α− ~
xα− + bα− ~
yα = eα a~α+ ~
xα+ + bα+ ~
yα = eα
~ − ~ − ~ − ~ − ~ − ~ + ~ + ~ + ~ + ~ +
cα xα + d α yα = f α ,cα xα + d α yα = f α .
After solving these two systems for the α-cuts, the four unknowns ~
xα− , ~
xα+ , ~yα− and ~yα+ are obtained. Next, using the
x and ~y are obtained.
regression the membership functions for ~

6. Conclusions
We proposed a practical approach for computing the product of fuzzy numbers using α-cuts. To obtain the corresponding
membership function for the resulting fuzzy product, we constructed a regression model by using a least squares model for
the regression. The proposed model, while being practically simple, has the flexibility to construct the product operator on
2055
various types of fuzzy numbers having the same type or being mixed. We made use of the approach to compute the solutions
of fully fuzzy linear systems. The approach can also be used for computing solutions of fully fuzzy linear systems with
various fuzzy variables.

References
Ban, A.I., Approximation of fuzzy numbers by trapezoidal fuzzy numbers preserving the expected interval, Fuzzy Sets and
Systems 159 (2008)1327–1344.
Ban, A.I., Trapezoidal and triangular approximations of fuzzy numbers—inadvertences and corrections, Fuzzy Sets and
Systems 160 (2009) 3048–3058.
Buckley, J.J., Eslami, E., and Feuring, T., Fuzzy Mathematics in Economics and Engineering, Studies in Fuzziness and Soft
Computing, Physica-Verlag, 2002.
Buckley, J.J., and Qu, Y., Solving systems of linear fuzzy equations, Fuzzy Sets and Systems 43 (1991) 33–43.
Dehghan, M., and Hashemi, B., Iterative solution of fuzzy linear systems, Applied Mathematics and Computation 175(1),
(2006) 645-674.
Dehghan, and M., Hashemi, B., Solution of the fully fuzzy linear systems using the decomposition procedure, Applied
Mathematics and Computation 182 (2006) 1568–1580.
Dehghan, M., Hashemi, B., and Ghatee, M., Computational methods for solving fully fuzzy linear systems, Applied
Mathematics and Computation 179 (2006) 328–343.
Dubois, D., and Prade, H., Fuzzy Sets and Systems: Theory and Applications, Academic Press, New York, 1980.
Ezzati, R., Khezerloo, S., Mahdavi-Amiri, N., and Valizadeh, Z., New models and algorithms for solutions of single-signed
fully fuzzy LR linear systems, Iranian Journal of Fuzzy Systems, To appear.
Friedman, M., Ming, M., and Kandel, A., Fuzzy linear systems, Fuzzy Sets and Systems 96 (1998) 201–209.
Ghanbari, R., Mahdavi-Amiri, N., and Yousefpour, R., Exact and approximate solutions of fuzzy LR linear systems: new
algorithms using a least squares model and the ABS approach, Iranian Journal of Fuzzy Systems 7(2), (2010) 1-18.
Giachetti, R.E., and Young, R.E., Analysis of the error in the standard approximation used for multiplication of triangular and
trapezoidal fuzzy numbers and the development of a new approximation, Fuzzy Sets and Systems 91 (1997) 1–13.
Giachetti, R.E., and Young, R.E., A parametric representation of fuzzy numbers and their arithmetic operators, Fuzzy Sets
and Systems 91 (1997) 185–202.
Kawaguchi, M.F., and Da-Te, T., A calculation method for solving fuzzy arithmetic equations with triangular norms, in
proceedings of the 2nd IEEE International Conference on Fuzzy Systems, San Francisco, CA, USA, 1993, March 28–
April 1, pp. 470-476.
Kreinovich, V., Lakeyev, A.V., Rohn, J., and Kahl, P.T., Computational Complexity and Feasibility of Data Processing and
Interval Computations, Applied Optimization, vol. 10, Springer, 1998.
Moore, R.E., Methods and Applications of Interval Analysis, SIAM, Philadelphia, 1979.
Mrówka, P. E., Trapezoidal approximations of fuzzy numbers, Fuzzy Sets and Systems 153 (2005) 115–135.
Muzzioli, S., and Reynaerts, H., Fuzzy linear systems of form A1 x + b1 = A2 x + b2 , Fuzzy Sets and Systems 157 (2006)
939–951.
Muzzioli, S., and Reynaerts, H., The solution of fuzzy linear systems by nonlinear programming: a financial application, Eur.
J. Oper. Res. 177 (2007) 1218–1231.
Nasibov, E.N., and Peker, S., On the nearest parametric approximation of a fuzzy number, Fuzzy Sets and Systems 159 (2008)
1365–1375.
Rao, S.S., and Chen, L., Numerical solution of fuzzy linear equations in engineering analysis, International Journal for
Numerical Methods in Engineering 42 (1998) 829–846.
Sakawa, M., Fuzzy Sets and Interactive Multiobjective Optimization, Plenum Press, New York and London, 1973.
Tajdin, A., Mahdavi, I., Mahdavi-Amiri, N., and Sadeghpour-Gildeh, B., Computing a fuzzy shortest path in a network with
mixed fuzzy arc lengths using α-cuts, Computers and Mathematics with Applications 60 (2010) 989–1002.
Vroman, A., Deschrijver, G., and Kerre, E., Solving systems of linear fuzzy equations by parametric functions-an improved
algorithm, Fuzzy Sets and Systems 158 (2007) 1515–1534.
Wagenknecht, M., Hampel, R., and Schneider, V., Computational aspects of fuzzy arithmetics based on Archimedean t-
norms, Fuzzy Sets and Systems 123 (2001) 49–62.
Yeh, C.-T., Weighted trapezoidal and triangular approximations of fuzzy numbers, Fuzzy Sets and Systems 160 (2009) 3059–
3079.
Zhao, R., and Govind, R., Solutions of algebraic equations involving generalized fuzzy numbers, Information Science 56
(1991) 199–243.

2056