On Multiple Objective Linear Programming Problems

with Fuzzy Rough Coefficients

‫ﻣﺸﺎآﻞ اﻟﺒﺮﻣﺠﺔ اﻟﺨﻄﻴﺔ ﻣﺘﻌﺪدة اﻷهﺪاف ذات اﻟﻤﻌﺎﻣﻼت اﻟﻔﺎزﻳﺔ اﻟﺨﺸﻨﺔ‬

M. L. Hussein M. E. A. El-Mikkay Y. A. S. El-Masry
Prof. of pure mathematics Prof. of computer science Ass. Lecture
Faculty of Education Faculty of Science Faculty of Science
Kafr El-Shiekh University Mansoura University Mansoura University
mlhussein@hotmail.com mikkawy@yahoo.com Yasser_elmasry@mans.edu.eg
: ‫اﻟﺨﻼﺻﺔ‬
‫ ﺗﻤﺖ‬.‫اﻟﺒﺤﺚ ﻳﺘﻨﺎول ﺑﺎﻟﺪراﺳﺔ ﻣﺸﺎآﻞ اﻟﺒﺮﻣﺠﺔ اﻟﺨﻄﻴﺔ ﻣﺘﻌﺪدة اﻷهﺪاف ذات اﻟﻤﻌﺎﻣﻼت اﻟﻔﺎزﻳﺔ اﻟﺨﺸﻨﺔ‬
‫ﺗﻠﻚ اﻟﺪراﺳﺔ ﺑﺪﻣﺞ اﻷﻋﺪاد اﻟﻔﺎزﻳﺔ اﻟﺨﺸﻨﺔ ﺑﺪوال اﻟﻬﺪف و آﺬاﻟﻚ دوال اﻟﻘﻴﻮد اﻟﺘﻲ ﺗﺸﻜﻞ اﻟﻬﻴﻜﻞ اﻟﻌﺎم‬
.‫ و ﺗﻮﺻﻠﻨﺎ إﻟﻰ ﺗﻌﺮﻳﻒ ﻟﺤﻞ اﻟﻤﺸﻜﻠﺔ ﺗﺤﺖ اﻻﻋﺘﺒﺎر ﻣﻦ ﺧﻼل إدﺧﺎل ﻣﻔﺎهﻴﻢ ﺣﻠﻮل اﻟﻜﻔﺎءة اﻟﻔﺎزﻳﺔ‬.‫ﻟﻠﻤﺸﻜﻠﺔ‬
.‫ و أﻋﻄﻴﻨﺎ ﻣﺜﺎﻻ ﻋﺪدﻳﺎ ﻟﺘﻮﺿﻴﺢ ﻓﻜﺮة اﻟﺒﺤﺚ‬.‫و ﺗﻢ اﻟﺤﺼﻮل ﻋﻠﻰ اﻟﺸﺮط اﻟﻀﺮوري و اﻟﻜﺎﻓﻲ ﻟﻠﺤﻞ‬

Abstract:
This paper deals with the multiple objective linear programming problems with fuzzy rough
coefficients. We consider the problem by incorporating fuzzy rough coefficients into a multiple objective
linear programming framework. A solution concept that is attractive from the stand points of feasibility
and rough efficiency is specified. A necessary and sufficient condition for such a solution is established.
A numerical example is also included for the sake of illustration.

1. Introduction Tamura and Miura [15], Vangeldere [17] and

Making decisions involving multiple
objectives is a daily task for a lot of researchers in
the more diverse fields. Hence multiple objective
decision making problems have defined a very
well studied topic in the general area of decision
making theory. In particular, multiobjective
decision making problems which can be modeled
as mathematical programming problems are also
well known.
The involvement of different kinds of
fuzziness in these problems is an interesting
matter. It has received a great deal of work since
the early 1980s. This is due to the fact that
decision makers have some lack of precision in
stating some of the parameters involved in the
model.
The multiple objective linear programming
problem arises when two or more non comparable
linear criterion functions are to be simultaneously
optimized over a polyhedral set.
Many researchers are concerned with this
subject, (see, for instance, Kuhn and Tucker [6],
1
Zadeh [18] ).
Tanaka and Asai [16] formulated
multiobjective linear programming problems with
fuzzy parameters. This has be done by following
the fuzzy decision or minimum operator proposed
by Bellman and Zadeh [2] together with triangular
membership functions for fuzzy parameters. They
considered two types of fuzzy multiobjective
linear programming problems. One is to determine
the non fuzzy solution and the other is to
determine the fuzzy solution. There are different
methods for comparison of fuzzy numbers
[1,3,4,9,13,14]. One of the most convenient
methods is the comparison by integration [1,4,9].
By using the concept of comparison of fuzzy
numbers, Maleki and Tata [8], introduced a new
method for solving the fuzzy number linear
programming problems.
The notion of rough sets has been introduced
by Pawlak [11], and subsequently the algebraic
approach to rough sets was studied by Iwinski [5].
The concept of fuzzy sets was investigated by
Zadeh [19]. A comparison between fuzzy sets and
rough sets was introduced by Pawlak [12], where
it is shown that these concepts are different.
Nanda and Majumdar [10], were introduced the
concept of fuzzy rough set.
The main objective of the current paper is to
investigate the problem of incorporating fuzzy
rough data into a multiple objective linear
programming framework. Inspired by the notions
of Luhandjula [7], we define a satisfying solution
for this problem via two concepts of α-fuzzy
rough feasibility and β-fuzzy rough efficiency. A
necessary and sufficient condition for a potential
action to be a satisfying one is established and
some techniques to single out such a solution are
described. A numerical example is given for the
sake of illustration.
The paper is organized as follows: In section
2, some preliminaries and definitions are given.
Section 3 is devoted to the statement of the fuzzy
rough multiple objective linear programming
problem ( P1 ) and the concepts of α-fuzzy rough
feasible, β-fuzzy rough efficient. Some techniques
for finding a satisfying solution of the problem
( P1 ) are described in section 4. An efficient
solution for this problem is described in section 5.
In section 6, we give a numerical example for the
sake of illustration. A conclusion and future work
are given in section 7.
2. Fuzzy Rough Set Theory
The notions of fuzzy rough set theory are
illustrated through the following definitions. For
more details see [10].
Definition 2.1 [19]
Let X be a set and L be a lattice. In particular L
could be the closed interval [0,1]. A fuzzy set A in
X is characterized by a membership function µA :
X → L, which associates with each point x ∈ X its
degree of membership µA(x) ∈ L.
Let A and B be fuzzy sets in X. Then,
(1) A = B ⇔ µ A ( x) = µ B ( x)∀x ∈ X ,
(2) A ⊂ B ⇔ µ A ( x) ≤ µ B ( x)∀x ∈ X ,
(3) C = A ∪ B ⇔ µ C ( x) = max[µ A ( x), µ B ( x)]∀x ∈ X ,
(4) D = A ∩ B ⇔ µ D ( x) = min[ µ A ( x), µ B ( x )]∀x ∈ X ,
2
Definition 2.2 [8]
The α-cut of a fuzzy set A is
Aα = {x ∈ X : µ A ( x) ≥ α } .
Definition 2.3 [8]
A fuzzy set A is said to be convex if for every
λ ∈ [0,1], x 1 , x 2 ∈ X then , µ A ((1 − λ ) x 1 + λ x 2 ) ≥ min( µ A ( x 1 ), µ A ( x 2 )).
Definition 2.4 [8]
A fuzzy set A is said to be normalized if there
exists x ∈ X such that µ A ( x) = 1.
Also, it can be easily verified that for any fuzzy
subset A of X;
(1) A is convex if and only if Aα is convex.
(2) If A is continuous, then A is convex if and
only if Aα is closed interval.
Definition 2.5 [8]
~
A fuzzy subset A of X which is continuous,
convex and normalized is said to be a fuzzy
number.
Definition 2.6 [7]
Let X denote the n product X 1 × X 2 × ... × X n
and let f be a function defined by The extension
principle states that f can be extend to n-tuples
(a 1 , a 2 ,..., a n ) where A is a fuzzy subset of X as
follows;
µ 1 2 n ( x1, x 2 ,..., x n ) = sup(min(µ 1 ( x1),
f (a ,a ,...,a ) a
µ 2 ( x2 ),..., µa n ( xn )))
a
Definition 2.7 [11]
Let U be a nonempty set and let β be a
complete sub algebra of the Boolean algebra P(U)
of subsets of U. The pair (U,β) is called a rough
universe.
Let ζ=( U,β) be a given fixed rough universe.
Let R be the relation defined as follows:
A=(AL ,AU )∈R if and only if AL ,AU ∈β, AL
⊂ AU elements of R are called rough sets and
elements of β are called exact sets. Identify the
element (X,X)∈R with the element X∈β and
hence an exact set is a rough set in the sense of the
above identification. But a rough set need not be
exact set; for example if U is any nonempty set,
then (Φ,β) is a rough set which is not exact.
Let A=(AL ,AU) and B=(BL ,BU ) be any two
rough sets, then;
(1) A = B ⇔ AL = BL , AU = BU ; Definition 2.12
(2) A ⊂ B ⇔ AL ⊂ BL , AU ⊂ BU ;
A fuzzy rough set A = ( AL , AU ) which is
(3) A ∪ B = ( AL ∪ B L , AU ∪ BU );
continuous, convex and normalized is said to be a
(4) A ∩ B = ( AL ∩ BL , AU ∩ BU ). fuzzy rough number.

Definition 2.8 [10] Definition 2.13

Let U be a set and β a Boolean sub algebra of Let
the Boolean algebra of all subsets of U. Let L be a X = ( X L , X U ) = (( X L1 , X U1 ), ( X L2 , X U2 ),..., ( X Ln , X Un ))
lattice. Let X be a rough set then, X=(XL,XU)∈β2
, and f be a function defined by
with XL ⊂ XU.
f : ( X L , X U ) → ( R, R) . The extension principle
A fuzzy rough set A=(AL ,AU) in X is
characterized by a pair of maps states that f can be extended to n-tuples ordered
µ A : X L → L, µ A : X U → L with the property pairs ((a 1L , aU1 ), (a L2 , aU2 ),..., (a Ln , aUn )) where
L U

that µ AL ( x) ≤ µ Au ( x) ∀ x ∈ X U . A = ( AL , AU ) is a fuzzy rough subset of

For any two fuzzy rough sets ( X L , X U ) as follows;
A = ( AL , AU ), B = ( B L , BU ) in X then;
(1) A = B ⇔ µ A ( x) = µ B ( x)∀x ∈ X L , µ A ( x) µ 1 1 2 2 n n
L L U 1 ),(a 2 ,a 2 ),...,(a n ,a n )) (( xL , xU ), ( xL , xU ),..., ( xL , xU )) =
f ((a1L ,aU L U L U
= µ B ( x)∀x ∈ X U ;
U (sup(min(µ 1 ( x1L ), µ 2 ( xL
2 ),..., µ ( x n ))), sup(min(µ
n L
aL 1
(2) A ⊂ B ⇔ µ A ( x) ≤ µ B ( x)∀x ∈ X L , µ A aL aL aU
L L U 1 ), µ ( x 2 ),..., µ ( x n )))
( x) ≤ µ B ( x)∀x ∈ X U ; ( xU 2 U aUn U
aU
U
(3) C = A ∪ B ⇔ µ CL ( x) = max{µ AL ( x), µ B L ( x)}∀x ∈ X L ,
µ C ( x) = max{µ A ( x), µ B ( x)}∀x ∈ X U ; 3. Problem Formulation
3.1. Statement of The Problem
U U U

(4) D = A ∩ B ⇔ µ D ( x) = min{µ A ( x), µ B ( x)}∀x ∈ X L ,

L L L
Consider the mathematical program
µ D ( x) = min{µ A ( x), µ B ( x)}∀x ∈ X U .
U U U ( P1 ) : max (c 1 x, c 2 x,..., c k x) ,
Ax ≤ b ,
Definition 2.9 x≥0
The α-cut of a fuzzy rough set A is
where c j ( j = 1,2,..., k ) are n-vectors, b is an m-
Aα = ( ALα , AUα ) where,
ALα = { x ∈ X L : µ AL ( x ) ≥ α }, AUα = {x ∈ X U : µ AU ( x ) ≥ α }.
vector and A an m × n matrix, all having fuzzy
rough components.
By using the concept of fuzzy rough sets,
Definition 2.10 problem ( P1 ) converts to the following two fuzzy
A fuzzy rough set A = ( AL , AU ) is said to be
problems. The first of the lower bound for the
convex if AL and AU are convex fuzzy sets. rough parameters and the second of the upper
i.e. bound for the rough parameters.
µ A ((1 − λ ) x1 + λx 2 ) ≥ min( µ A ( x1),
L L , ( PL ) : max(c L1 x, c L2 x,..., c Lk x) ,
µA 2 1 2
( x ))∀λ ∈ [0,1], x , x ∈ X
L
L AL x ≤ bL
µ A ((1 − λ ) x1 + λx 2 ) ≥ min( µ A ( x1), x ≥ 0.
U U
µ A ( x 2 ))∀λ ∈ [0,1], x1, x 2 ∈ X U
and
U
( PU ) : max(cU1 x, cU2 x,..., cUk x) ,
Definition 2.11
AU x ≤ bU
A fuzzy rough set A = ( AL , AU ) is said to be
x ≥ 0.
normalized if there exist an
Here, we must specify the appropriate
x ∈ X L such that µ AL ( x) = 1
solution concept for this problem. For this task,
we define the following solution concepts.

3
3.2. α-Fuzzy Rough Feasible, β-Fuzzy Rough Definition 3.2.3
Efficient and satisfying solution for (p1). x 0 ∈ X is an (α,β) satisfying solution for (p1) if
and only if x 0 is β-fuzzy rough efficient for the
Definition 3.2.1 [α-Fuzzy Rough Feasible] program
( P3 ) : max (c 1 x, c 2 x,..., c k x) ,
Consider α = (α 1 , α 2 ,..., α m ), α i ∈ [0,1].
x∈ Xα
x ∈ X = {x ∈ R n : x ≥ 0} is said to be α-fuzzy
where X α denotes the set of all α-fuzzy rough
rough feasible for ( P1 ) if feasible actions for ( P1 ) .
µ F~i ( FiL ) ≥ α i , µ F~i ( FiU ) ≥ α i .
L U

where, 4.Characterization of An (α,β)-Satisfying

~ ~ ~ Solution for (p1).
FiL = { AiL x ≤ biL , x ≥ 0},
~ ~ ~
FiU = { AiU x ≤ biU , x ≥ 0}, i = 1,2,..., m. Consider the mathematical program
Using the extension principle we have
µ F~ = sup min(µa~i1 (ai1L ), µa~i 2 (ai 2 L ),..., µa~in ( P4 ) : max ((c 1 ) β x, (c 2 ) β x,..., (c k ) β x) ,
iL L L L

(ainL ), µb~ (biL )) x∈ Xα

where (c i ) β = ((c1 ) β (c 2 ) β ...(c n ) β ) and (c j ) β
iL i i i i

µ F~ = sup min(µa~i1 (ai1U ), µa~i 2

iU U U denotes the β-cut of the fuzzy rough number c ji .
(ai 2U ),..., µa~in (ainU ), µb~ (biU )) By the convexity assumption,
U iU
~
(c j ) β , j = 1,2,..., n , i = 1,2,..., k are rough real
i

Consider now the following mathematical intervals that will be denoted as [C isj (λ ), C im
j (θ )]
program; and given by
[(1 − λ ) c j u + λ c j L , (1 − θ ) c j L + θ c j U ],0 ≤ λ , θ ≤ 1
is is im im

( P2 ) : max (c 1 x, c 2 x,..., c k x) ,
.
x ∈ D ⊂ X = {x ∈ R n : x ≥ 0}
Let now φ β (λ ,θ ) be the set of all k × n rough
Definition 3.2.2 [β-Fuzzy Rough Efficient] real matrices C (λ , θ ) = (cij (λ ,θ )) with,
cij (λ ,θ ) ∈ [(1 − λ ) c is
j u + λ c j L,
is

x 0 ∈ D is β-fuzzy rough efficient for ( P2 ) ; if (1 − θ ) c im

j L + θ c j U ] , 0 ≤ λ , θ ≤ 1.
im
there is no
It is clear that ( P4 ) may be written as:
x ∈ D, i ∈ (1,2,..., k ) : µ C~ (C L ) ≥ β , µ C~ (CU ) ≥ β
L U
max (cx : x ∈ X α , c ∈ φ β (λ , θ )) .
where,
k ) ∈ R kn : c1 x ≥ c1 x 0 ,..., c i −1x ≥ c i −1x 0 , c i x > c i x 0 ,
C L = {(c1L ,..., c L ( P4 ) is then a family of pair multiple objective
L L L L L L
c iL+1x ≥ c iL+1x 0 ,..., c L
k x ≥ c k x 0 },
L
linear programs.
1 k kn 1
CU = {(cU ,..., cU ) ∈ R : cU x ≥ cU 1 x 0 ,..., c i −1x ≥ c i −1x 0 , c i x > c i x 0 ,
U U U U
cUi +1x ≥ c i +1x 0 ,..., c k x ≥ c k x 0 }, Definition 4.1
U U U
~ ~ are the set of all lower and upper fuzzy
C L , CU
x 0 is rough efficient for ( P4 ) if and only if there
~ is no c(λ , θ ) ∈ φ β (λ , θ ), x ∈ X α : cx ≥ cx 0 with at
coefficient for the fuzzy rough coefficient C to
problem ( P1 ) . least strict one inequality holds. In other words,
On account of the extension principle, x 0 is rough efficient for ( P4 ) if and only if x 0 is
efficient for
µ c~ (c L ) = sup min ( µ C (c 1L ), µ C (c L2 ),..., µ C (c Lk )) ≥ β ,
1 2 k

∀ c(λ , θ ) ∈ φ β (λ ,θ ),
L
max
L L L
cx
µ c~ (cU ) = sup min ( µ C (cU1 ), µ C (cU2 ),..., µ C (cUk )) ≥ β
1 2 k

x∈ Xα
U U U U

.
Now we are going to give the following result.

4
Theorem 4.1 where,
x 0 is an (α,β)-satisfying solution for ( P1 ) if and
Tst = {(c1 ,..., c k ) ∈ R nk | c1 x 2 ≥ c1 x 0 ,..., c s −1 x 2 ≥ c s −1 x 0 , c s x 2 > c s x 0 ,
only if x 0 is an rough efficient solution for ( P4 ) .
c s +1 x 2 ≥ c s +1 x 0 ,..., c k x 2 ≥ c k x 0 }, t = L,U
Proof.
By using the extension principle then, we have
Suppose that x 0 is an (α,β) -satisfying solution
for ( P1 ) then by definition, x 0 is α-fuzzy rough Sup min ( µ c 1 (c 1L ), µ c 2 (c L2 ),..., µ c k (c Lk )) ≥ β ,
feasible and β- fuzzy rough efficient for ( P3 ) . L L L

and
Assume now that x 0 is not efficient for ( P4 ) . (3)
Then there is Sup min ( µ c 1 (cU1 ), µ c 2 (cU2 ),..., µ c k (cUk )) ≥ β .
x1 ∈ X α , (c1, c 2 ,..., c k ) with ci ∈ (c i ) β ;
U U U

and For this supremum to exist there is ( p 1 , p 2 ,..., p k )

i 1 i 0
c x ≥ c x ∀i ∈ {1,2,..., k} satisfying the following constraints;
p1x 2 ≥ p1x 2 , p 2 x 2 ≥ p 2 x 0 ,..., p s −1x 2 ≥ p s −1x 0 , p s x 2 > p s x 0 ,
i0 ∈ {1,2,..., k} such that s +1 2
x ≥ p s +1x 0 ,..., p k x 2 ≥ p k x 0 .
c i0 x1 > c i0 x 0 (1) (4)
As c i ∈ (c i ) β , i = 1,2,..., k we have also Suppose now that for all ( p 1 , p 2 ,..., p k ) satisfying
the system (4), we have min
min ( µ c 1 (c 1 ), µ c 2 (c 2 ),..., µ c k (c k )) ≥ β and ( µ c 1 ( p ), µ c 2 ( p ),..., µ c k ( p )) < β , and
1 2 k
L L L L L L

min ( µ c 1 (c ), µ c 2 (c ),..., µ c k (c )) ≥ β
1 2 k min
( µ c 1 ( p ), µ c 2 ( p ),..., µ c k ( p k )) < β , then,
1 2
U U U

(2) U U U

Sup min
By (1),(2) we have ( µ c 1 ( p 1 ), µ c 2 ( p 2 ),..., µ c s −1 ( p s −1 ), µ c s ( p s ), µ c s +1 ( p s +1 ),..., µ c k ( p k )) < β ,
L L L L L L

Sup min
sup 1 2 min(µ 1 (c1), µ 2 (c 2 ),..., µ k (c k )) = µ ~ L (c1x1 ≥ c1x 0 , ( µ c 1 ( p 1 ), µ c 2 ( p 2 ),..., µ c s −1 ( p s −1 ), µ c s ( p s ), µ c s +1 ( p s +1 ),..., µ c k ( p k )) < β ,
(c ,c ,...,c k )∈T L cL cL cL c
i0 U U U U U U

i −1 i −1 i i i +1 i +1
..., c 0 x1 ≥ c 0 x 0 , c 0 x1 > c 0 x 0 , c 0 x1 ≥ c 0 x 0 ,..., c k x1 ≥ c k x 0 ) ≥ β ,
sup 1 2 min(µ 1 (c1), µ 2 (c 2 ),..., µ k (c k )) = µ ~U (c1x1 ≥ c1x 0 , Contradicting (3). There is then ( p 1 , p 2 ,..., p k )
(c ,c ,...,c k )∈T U c
U
c
U cU c
i0
i −1 i −1 i i i +1 i +1
satisfying (4) such that
..., c 0 x1 ≥ c 0 x 0 , c 0 x1 > c 0 x 0 , c 0 x1 ≥ c 0 x 0 ,..., c k x1 ≥ c k x 0 ) ≥ β .
min ( µ c 1 ( P ),..., µ c K ( P )) ≥ β ,
1 K
L L

where, min ( µ c 1 ( P ),..., µ c K ( P K )) ≥ β .

1
U U

i −1 i −1 i i
(5)
Tit = {(c1,..., c k ) ∈ R nk | (c1x1 ≥ c1x 0 ,..., c 0 x1 ≥ c 0 x 0 , c 0 x1 > c 0 x 0 ,
0 By (5), µ c i ( p i ) ≥ β , i = 1,2,..., k
i +1 i +1
c 0 x1 ≥ c 0 x 0 ,..., c k x1 ≥ c k x 0}, t = L, U .
i.e. p i ∈ (c i ) β , i = 1,2,..., k
0
This contradicts the β-rough efficiency of x for
(6)
( P3 ) and the if part of the theorem is established.
(4) and (6) contradict the efficiency of x 0 for (p4)
To show the only if part, suppose x 0 is efficient and we have done.
for ( P4 ) and not (α,β) -satisfying solution for
( P1 ) . Then there is x 2 ∈ X α ; S ∈ {1,2,..., k} such 5. A Solution for (p4)
that
The following notations will facilitate further
µ c~L (c L ) ≥ β , µ c~U (cU ) ≥ β
discussions.
where M β (λ ,θ ) denotes the subset of φ β (λ ,θ )
composed by matrices C (λ , θ ) having elements of
µ ~c t (c1 x 2 ≥ c1 x 0 ,..., c s −1 x 2 ≥ c s −1 x 0 , c s x 2 > c s x 0 ,
each column at the upper bound or at the lower
c s +1 x 2 ≥ c s +1 x 0 ,..., c k x 2 ≥ c k x 0 ) ≥ β , t = L, U bound.
i.e. i.e. if C (λ , θ ) ∈ M β (λ , θ ) then either
sup ( c1 ,c 2 ,...,c k )∈T t min(µ c~ t (c 1 ),..., µ c~ t (c k )) ≥ β .
s

5
 ⎛ c sj1 (λ ) ⎞ ⎛ c mj1 (θ ) ⎞ ( P6 ) max ( c 1 x, c 2 x )
⎜ ⎟ ⎜ ⎟
⎜ c sj 2 (λ ) ⎟ ⎜ c mj 2 (θ ) ⎟ such that
C sj (λ ) = ⎜ ⎟ or C mj (θ ) = ⎜ ⎟ x1 + x 2 ≥ 250,
⎜# ⎟ ⎜# ⎟
⎜⎜ sn ⎟⎟ ⎜⎜ mn ⎟⎟ x1 ≤ 200,
⎝ c j (λ ) ⎠ ⎝ c j (θ ) ⎠ 2 x 2 ≤ 200,
where c sij (λ ) and c mi j (θ ) are the left and the right 2 x1 + 1.5 x 2 ≤ 480,
~i
end point of (C j (λ , θ )) β respectively. 3x1 + 4 x 2 ≤ 900; x1 , x 2 ≥ 0.

Lemma 5.1 where c 1 = (c11 , c12 ), c 2 = (c 21 , c 22 ) and

A necessary and sufficient condition for x 0 cij = (cijL , cijU ) are fuzzy rough numbers
to be efficient for the multiple objective program
characterized by the membership functions shown
max C x , x ∈ X
in Figure (1a-d).
Is that there is λ > 0 such that x 0 solves the This problem can be converted into the following
mathematical program max λ C x , x ∈ X . two fuzzy problems. The first problem is,
A fascinating point is that the following program
yields an (α , β ) -satisfying solution for ( P1 ) : 1
( P6 L ) max ( c L x, c L x )
2

( P5 ) : max q 0 x , x ∈ X α such that

0
where q is a solution of the system x1 + x 2 ≥ 250,
V i C i (λ ,θ ) − q = 0 ∀i such that x1 ≤ 200,
C i (λ , θ ) ∈ M β (λ , θ ) , (7) 2 x 2 ≤ 200,
2 x1 + 1.5 x 2 ≤ 480,
Proposition 5.1 3x1 + 4 x 2 ≤ 900; x1 , x 2 ≥ 0.
If x 0 is optimal for ( P5 ) then x 0 is efficient for
( P4 ) . 1 2
where, c L = (c11L , c12 L ), c L = (c 21L , c 22 L ) and cijL
Proof. are fuzzy numbers characterized by the
As q 0 is a solution of (7), ∀i such that membership functions shown in Figure (2a-d).
C i (λ , θ ) ∈ M β (λ , θ ) , there is V i ∈ R k ,V i > 0
The second problem is,
such that V i C i (λ ,θ ) = q 0 .
i.e. ∀i such that C i (λ , θ ) ∈ M β (λ , θ ) , x 0 solves 1
( P6U ) : max ( cU x, cU x )
2

max ( V i C i (λ ,θ ) x : x ∈ X α ). such that

By lemma 5.1., x 0 is efficient for max x1 + x 2 ≥ 250,
( C i (λ ,θ ) x : x ∈ X α ) ∀C i (λ , θ ) ∈ φ β (λ , θ ) . x1 ≤ 200,
i.e. x 0 is efficient for ( P4 ) as desired. 2 x 2 ≤ 200,
2 x1 + 1.5 x 2 ≤ 480,
Corollary 5.1 3x1 + 4 x 2 ≤ 900; x1 , x 2 ≥ 0.
If x 0 is optimal for ( P5 ) then x 0 is an (α , β ) -
satisfying for ( P1 ) . 1 2
where, cU = (c11U , c12U ), cU = (c 21U , c 22U ) and cijU
This statement follows directly from Proposition
are fuzzy numbers characterized by the
5.1. and theorem 4.1.
membership functions shown in Figure (3a-d).
6. A Numerical Example In ( P6 ), let β = 0 then we obtain
In this section we are going to give a
numerical example.
Consider the following multiple objective
linear fuzzy rough program

6
(c11 ) β = [0(1 − λ ) + 1λ ,2(1 − θ ) + 3θ ] = [λ , θ + 2],
As well as the system
(c12 ) β = [0(1 − λ ) + 1λ ,1(1 − θ ) + 2θ ] = [λ ,θ + 1],
⎛ 2 1⎞ ⎛V12 ⎞ ⎛ 4.4 ⎞
(c 21 ) β = [−2(1 − λ ) + 0λ ,0(1 − θ ) + 2θ ] = [2λ − 2,2θ ], ⎜⎜ ⎟⎟ ⎜⎜ 2 ⎟⎟ = ⎜⎜ ⎟⎟
⎝ 0 1 ⎠ ⎝ V 2 ⎠ ⎝ 2 .9 ⎠
(c 22 ) β = [−3(1 − λ ) + 1λ ,1(1 − θ ) + 2θ ] = [4λ − 3,θ + 1]. has positive solutions given by
and 0 ≤ λ ,θ ≤ 1
By putting λ = 1 , θ = 0 then we get ⎛V11 ⎞ ⎛ 1.5 ⎞ ⎛V12 ⎞ ⎛ 0.75 ⎞
⎜ 1 ⎟ = ⎜⎜ ⎟⎟ , ⎜ 2 ⎟ = ⎜⎜ ⎟⎟ .
(c11 ) β = [1,2], (c12 ) β = {1} = (c 22 ) β , (c 21 ) β = {0}. ⎜V ⎟ ⎜ ⎟
⎝ 2 ⎠ ⎝ 2 .9 ⎠ ⎝ V 2 ⎠ ⎝ 2 .9 ⎠
The subset M β of φ β is composed of the matrices respectively.
⎛ 1 1⎞ ⎛ 2 1⎞ As constraints are crisp, the set X of points of R 2
⎜⎜ ⎟⎟ , ⎜⎜ ⎟⎟ . satisfying ( P6 ) is nothing but the set of 1-fuzzy
⎝ 0 1 ⎠ ⎝ 0 1 ⎠
rough actions. By the corollary, the program max
⎛ 4.4 ⎞
For q 0 = ⎜⎜ ⎟⎟ the system (q 0 x | x ∈ X α ) yields a (1,0)-satisfying solution
⎝ 2.9 ⎠ for ( P6 ) .
⎛ 1 1⎞ ⎛V11 ⎞ ⎛ 4.4 ⎞
⎜⎜ ⎟⎟ ⎜⎜ 1 ⎟⎟ = ⎜⎜ ⎟⎟
⎝ 0 1 ⎠ ⎝ V 2 ⎠ ⎝ 2 .9 ⎠ Solving this linear program, we obtain the
solution ( x1 , x 2 ) = (150,100) .

µC 11

0 C11s U 1 C11s L C11mL 2 C11mU 3 R

Fig 1-a

7
 µC 12

1
0 C12s U C12s L C12mCU m
12L 2 R
Fig 1-b

µC 21

-2 C 21s U C 21s L 0 C 21m m

C 21 2 R
L U

Fig 1-c

8
 µC 22

-3 C 22s U 0 C 22s L 1 C 22
m m
C 22 2
L U
R

Fig 1-d

µC 11 L

R
1 2
Fig 2-a

9
µC
12 L

0 R
1
Fig 2-b

µC 21L

0 R
Fig 2-c

µC 22 L

0 1 R
Fig 2-d

10
 µC
11U

0 1 2 3 R
Fig 3-a

µC
12U

2
0 1 R
Fig 3-b

µC 21U

-2 0 2 R

Fig 3-c

11
 µC
-3
Fig 3-d
7. Conclusion
There are a large number of real world problem
that can be cast in form of multiple objective
fuzzy rough programming problems. A significant
outcome of this paper is the investigation of the
problem by incorporating fuzzy rough numbers
into a multiple objective linear programming
framework. Inspired by the techniques of
Luhandjula in [7], we characterized the solution
for this problem via the concepts of α-fuzzy rough
feasibility and β-fuzzy rough efficiency. A
numerical example is given for the sake of
development theory.
The extension of the current work to real life
problems and parametric studies is now under
active consideration. This will be the subject of
the next paper.
References
[1] F. F. Baldwin, N. C. F. Guild, Comparison of
fuzzy sets on the same decision space, Fuzzy
Sets and Systems Vol. 2, pp. 213-231, 1979.
[2] R. Bellman and L. A. Zadeh, Decision making
in a fuzzy environment. Management Sci.
Vol. 17, pp. 141-164, 1970.
[3] D. Dubois, H. Prade, Ranking fuzzy numbers
in the setting of possibility theory, Inform.
Sci. Vol. 30, pp. 183-224, 1983.
[4] P. Fortemps and M. Roubens, Ranking and
defuzzification methods based on area
compensation, Fuzzy Sets and Systems Vol.
82,pp. 319-330, 1996.
12
22U
0 1 2 R
[5] T. B. Iwinski, Algebraic approach to rough
sets, Bull. Polish Acad. Sci. Math. Vol. 35,
pp. 673-683, 1987.
[6] H. W. Kuhn and A. W. Tucker, Nonlinear
programming, Proc. 2nd Berkeley Sympos. On
Mathematics, statistics and Probability,
University of California Press, Berkeley,
pp.481-492, 1951.
[7] M.K.Luhandjula, On multiple objective
programming with possibilistic coefficients,
Fuzzy sets and systems Vol. 21, pp. 135-145,
1987.
[8] H. R. Maleki, M. Tata and M. Mashinchi,
Linear programming with fuzzy variables,
Fuzzy Sets and Systems Vol. 109, pp. 21-33,
2000.
[9] K. Nakamura, Preference relations on a set of
fuzzy utilities, Fuzzy Sets and Systems Vol.
20,pp. 147-162, 1986.
[10] S. Nanda and S. Majumdar, Fuzzy rough sets,
Fuzzy Sets and Systems Vol. 45,pp. 157-160,
1992.
[11] Z. Pawlak, Rough sets, Internat J. Inform.
Comput. Sci. Vol. 11(5), pp. 341-356, 1982.
[12] Z. Pawlak, Rough sets and fuzzy sets, Fuzzy
Sets and Systems Vol. 17,pp. 99-102, 1985.
[13] J. Ramik and J. Rimanek, Inequality relation
between fuzzy numbers and its use in fuzzy
optimization, Fuzzy Sets and Systems Vol.
16,pp. 123-138, 1985.
[14] M. Roubens, Comparison of flat fuzzy
numbers, in : N. Badler, A. Kandel (Eds.),
Proc. NAFIPS Vol. 86, pp. 462-476, 1986.
[15] K. Tamura and S. Miura, On linear vector
maximization problems, J. Oper. Res. Sci.
Japan Vol. 20, pp. 139-149, 1977.
[16] H. Tanaka and K. Asai, Fuzzy linear
programming based on fuzzy functions, Bull.
Univ. Osaka Perfect. Ser. A Vol. 29, pp.113-
125, 1980.
[17] J. Vangeldere, Pareto weak minimality and
various connected properties, in W. Oettli and
F. Steffans (Eds.), Proc. Third Sympos. On
Operations Research, Band 31, University of
Mannheim, Mannheim pp. 663-676, 1978.
[18] L. A. Zadeh, Optimality and nonscalar-valued
performance criteria, IEEE Trans. Automatic
control, A C-8 ,pp. 59-61, 1963. [19] L. A.
Zadeh, Fuzzy sets, Information and Control
Vol. 8, pp. 338-353, 1965.

13