62 Int. J. Mathematical Modelling and Numerical Optimisation, Vol. 8, No.

1, 2017

Cost parametric analysis of linear programming

problems with fuzzy cost coefficients based on
ranking functions

Ali Ebrahimnejad
Department of Mathematics,
Qaemshahr Branch,
Islamic Azad University,
P.O. Box 163, 4765161964, Qaemshahr,
Iran
Fax: +98-1142155025
Email: [email protected]
Email: [email protected]

Abstract: Parametric analysis (PA) as a basic tool for studying perturbations in

optimisation problems is applied to study the effect of the continuous variations
of several parameters in the objective function coefficients on the optimum
solution. To the best of our knowledge, till now there is no study in the
literature to deal with the PA of the fuzzy linear programming problem in
which the cost coefficients are represented by LR flat fuzzy numbers. In this
paper, the parametric analysis for the same problem is investigated to determine
the optimal solution and the fuzzy optimal objective values as a function of
parameters when the fuzzy cost coefficients are perturbed along a new fuzzy
cost vector. Then, its main advantages over the existing fuzzy sensitivity
analysis problems are discussed. Finally, to show the application of fuzzy PA
problem, an existing real-life fuzzy transportation problem is solved and the
obtained results are explored.

Keywords: cost parametric analysis; fuzzy linear programming; fuzzy

numbers; ranking function.

Reference to this paper should be made as follows: Ebrahimnejad, A. (2017)

‘Cost parametric analysis of linear programming problems with fuzzy cost
coefficients based on ranking functions’, Int. J. Mathematical Modelling and
Numerical Optimisation, Vol. 8, No. 1, pp.62–91.

Biographical note: Ali Ebrahimnejad serves as an Associate Professor at the

Mathematics Department, Qaemshahr Branch, Islamic Azad University. He
obtained his PhD in Applied Mathematics in the Department of Mathematics at
the Islamic Azad University (Sciences and Research Branch), Tehran, Iran. He
received his BS from Mazandaran University, Iran and his MSc from Tehran
University, Iran. He is on the Editorial Board of the International Journal of
Fuzzy System Applications (IJFSA), Annals of Fuzzy Mathematics and
Informatics (AFMI), International Journal of Information and Decision
Sciences (IJIDS), Iranian Journal of Optimization (IJO), International Journal
of Strategic Decision Sciences (IJSDS) and International Journal of Enterprise
Information Systems (IJEIS). His research interests include operations research,
network OWS, data envelopment analysis and fuzzy optimisation.

Copyright © 2017 Inderscience Enterprises Ltd.

 A. Ebrahimnejad 63

1 Introduction

Linear programming (LP) is a well-known optimisation approach and can be used as a

decision-making aid in many different industries, and in financial and service
organisations. Although, in conventional LP problems, it is assumed that the parameters
of the problem are exactly known, there may be some situations in formulating real-life
problems when the parameters are not known in a precise way. A frequently used manner
to represent the imprecision in parameters is fuzzy numbers that enlarge the range of
applications of linear programming. Many different techniques have been used in the
literature to solve the LP problems in fuzzy environment. In what follows, a taxonomy
and review of some of these techniques for solving fuzzy linear programming (FLP)
problem are provided.
Theory of fuzzy mathematical programming has been introduced by Tanaka et al.
(1974) and the first formulation of FLP problem has been given by Zimmerman (1978).
Tanaka and Asai (1984) proposed a possibilistic LP formulation with crisp cost
coefficients and fuzzy decision variables. Verdegay (1984) presented the concept of a
fuzzy objective based on the fuzzification principle and used this concept to solve the
FLP problems. Herrera et al. (1993) examined the fuzzified version of the mathematical
problem assuming that the coefficients are given by fuzzy numbers and the relations in
the definition of the feasible set are also fuzzy. Maleki et al. (2000), based on ranking
functions, extended the primal simplex method for solving LP problems with fuzzy cost
coefficients. Ebrahimnejad et al. (2011) generalised their method for FLP problems with
bounded decision variables. Allahviranloo et al. (2008) proposed a new method for
solving FLP problems with inequality constraints where the FLP problem is first
converted into a crisp LP problem and then the obtained crisp LP problem is solved to
find the fuzzy optimal solution for the original FLP problem. Ezzati et al. (2015), based
on a new lexicographic ordering on triangular fuzzy numbers, proposed a novel algorithm
to solve fully FLP problem by converting it into a multi-objective linear programming
(MOLP) problem with three-objective functions. Ezzati et al. (2014), based on two new
fuzzy arithmetic operations and fuzzy ordering on symmetric trapezoidal fuzzy numbers,
proposed a new algorithm to find a lexicographic/preemptive fuzzy optimal solution of a
fuzzy lexicographic MOLP problem and a properly fuzzy efficient solution of a fuzzy
MOLP problem.
Payan and Noora (2014) presented a linearisation procedure to solve the multi-
objective linear fractional programming problem with fuzzy parameters. Razmi et al.
(2016) proposed a new intuitionistic fuzzy approach for solving multi-objective problems
an interactive procedure by considering the degrees of satisfaction and dissatisfaction of
objectives. Chinnadurai and Muthukumar (2016) proposed presented a new appraoch for
solving linear fractional programming problems with fuzzy coefficients and fuzzy
decision variables, as well as developing a method for computing them. Ren et al. (2016)
developed an interactive programming approach to obtain a balanced solution of the fully
fuzzy bilevel programming problem. Zhao et al. (2016) proposed three interactive
intuitionistic fuzzy methods for multi-level programming problems by defining a score
function to depict decision-makers’ satisfactory degree.
Ganesan and Veeramani (2006) formulated an FLP model with symmetric trapezoidal
fuzzy numbers. They extended the primal simplex algorithm to derive a fuzzy solution
64 Cost parametric analysis of linear programming problems

for the resulting FLP problem without converting it into a crisp one. Ebrahimnejad
(2011a) proved the finite convergence of their extended fuzzy algorithm and proposed a
revised algorithm with less computational complexity. In addition, a fully FLP problem
with all parameters and variables as triangular fuzzy numbers has been investigated by
Hosseinzadeh-Lotfi et al. (2009). Ebrahimnejad (2015a) proposed a new method based
on the bounded dual-simplex method to determine the fuzzy optimal solution of that kind
of FLP problems in which some or all variables are restricted to lie within fuzzy lower
and upper bounds. Hatami-Marbini and Tavana (2011) proposed a new method for
solving LP problems with fuzzy parameters. Their proposed method not only is
independent of restrictive conditions, but also considers the decision-makers’
participation in an interactive manner. Saati et al. (2012) proposed a twofold model
which consists of two new methods for solving FLP problems in which the variables and
the coefficients of the constraints are characterised by fuzzy numbers. Baykasoğlu and
Subulan (2015) introduced a new parametric method for solving fully FLP problems
based on α-cut representation of fuzzy intervals by incorporating the decision-maker’s
attitude towards risk. Selvaraj and Pandian (2016) developed a new fuzzy neural network
for solving fully FLP problems based on a fuzzy energy function. Najafi et al. (2016)
proposed a new efficient method for solving fully FLP with unrestricted fuzzy
coefficients and fuzzy variables based on crisp nonlinear programming techniques.
Das et al. (2016) introduced an efficient approach to solve fully FLP problems with
trapezoidal fuzzy numbers based on lexicographic ordering method. A survey on models
and methods for solving FLP problems can be found in Ebrahimnejad and Verdegay
(2016a). Also, R Package for solving FLP problems has been provided by Villacort et al.
(2016). It is worthwhile noting that although the FLP problem with all parameters as
fuzzy numbers is the general case of the FLP, considering different sources of fuzziness
for all FLP models is not often suitable. The FLP model investigated in this study belongs
to those categories in which the coefficients of the objective function are represented by
LR flat fuzzy numbers.
Although the flexibility in the values involved in the problem has been allowed in
FLP problems, even after formulating the problem under consideration in fuzzy
environment, it is quite possible that entered values were wrong or change due to time
variations; hence, it is required to use the sensitivity analysis for FLP problem. Although
several approaches have been proposed in the literature to obtain the solution of FLP
problems, there are only few papers investigating sensitivity analysis in FLP problems.
Based on this motivations, this paper investigates parametric analysis in FLP problems.
The rest of this paper was organised as follows. A comprehensive literature review on
fuzzy sensitivity analysis is presented in Section 2. The basic definitions, arithmetic
operations and ranking function of fuzzy numbers are reviewed in Section 3. Section 4
presents the mathematical formulation and solution approach of FTP problems with fuzzy
cost coefficients. Section 5 gives the fundamentals of proposed solution approach for
doing parametric analysis of fuzzy cost vector and illustrates it using an application
example. Advantages of the proposed methods are discussed in Section 6. To show the
application of proposed method, a real-life problem is solved by the proposed method in
Section 7. Finally, conclusions and future works are given in Section 8.
 A. Ebrahimnejad 65

2 Literature review

An overview of the papers which studied fuzzy sensitivity analysis in FLP problems and
discussed regarding solution methodologies can be summarised as follows: Sensitivity
analysis in FLP problems was first considered by Hamacher et al. (1978). They derived
the functional relationship between changes of the right-hand-side (RHS) vector and
those of objective function optimal value in an LP problem with one fuzzy objective
function and crisp and fuzzy constraints. Dutta et al. (1992) studied sensitivity analysis
for fuzzy linear fractional programming problem. Gupta and Bhatia (2001) considered the
measurement of sensitivity for multi-objective linear fractional programming problem in
fuzzy environment. Garcia-Aguado and Verdegay (1993) and Delgado et al. (1993) also
studied on sensitivity analysis of membership function in FLP problems. Precup and
Preitl (2006) presented a stability analysis method dedicated for some fuzzy control
systems. The concept of sensitivity analysis in linear programming problems with fuzzy
parameters has been investigated by Ebrahimnejad (2011b) using fuzzy primal and fuzzy
dual-simplex algorithms. Kheirfam and Hasani (2010) and Nasseri and Ebrahimnejad
(2011) studied the same sensitivity analysis for linear programming problems with
trapezoidal fuzzy variables. In a similar way, Kumar and Bhatia (2011) discussed some
other cases that are not considered by Kheirfam and Hasani (2010). In addition, Bhatia
and Kumar (2012) proposed a new method, named as Mehar’s method, to deal with the
sensitivity analysis of such FLP problems in which all the parameters are represented by
LR flat fuzzy numbers. Kheirfam and Verdegay (2012) analysed the strictly sensitivity
analysis for fuzzy quadratic programming when perturbations arise in both the right-hand
side of the constraints and the the cost coefficients. Farhadinia (2014) demonstrated that
study of the sensitivity analysis for the interval-valued trapezoidal fuzzy number LP
problems gives rise to the expected results as same as those obtained in FNLP problems.
However, to the best of our knowledge, till now there is no method in the literature to
deal with the parametric analysis in FLP problems. The parametric analysis is a general
case of sensitivity analysis in which one often finds a direction along which the objective
function or the constraints are perturbed, and then seeks to move along this direction. As
an extension of the proposed approach by Jia et al. (2013), in this paper, we seek the
optimal solutions to the FLP problem with fuzzy cost coefficients, by perturbing the
fuzzy cost vector along with another fuzzy vector.

3 Preliminaries

In this section, some basic definitions, arithmetic operations of LR flat fuzzy numbers
and an existing ranking approach for comparing LR flat fuzzy numbers are reviewed
(Dubois and Prade, 1980; Klir and Yuan, 1995; Bhatia and Kumar, 2012).

Definition 1: (Klir and Yuan, 1995): A fuzzy set A , defined on universal set X , is given
by a set of ordered pairs A = {( x, μ A ( x)) | x ∈ X } where μ A ( x) gives the membership
grade of the element x in the set A and is called membership function.
66 Cost parametric analysis of linear programming problems

Definition 2: (Klir and Yuan, 1995): A fuzzy set A , defined on universal set of real
numbers R , is said to be a fuzzy number if its membership function has the following
characteristics:

1 A is convex, i.e. ∀x, y ∈ R, ∀λ ∈ [0,1], μ A (λ x + (1 − λ ) y) ≥ min{μ A ( x), μ A ( y)} .

2 A is normal, i.e. ∃x ∈ R; μ A ( x ) = 1 .

3 μ A is piecewise continuous.

Definition 3: (Dubois and Prade, 1980): A function L :[0, ∞) → [0,1] (or

R :[0, ∞) → [0,1] ) is said to be reference function of fuzzy number if and only if L(0) = 1
(or R (0) = 1 ) and L or R is non-increasing on [0, ∞) .

Definition 4: (Bhatia and Kumar, 2012): A fuzzy number A = (m, n, α , β ) LR is said to be

an LR flat fuzzy number, if its membership function is given by
⎧ ⎛m−x⎞
⎪L ⎜ ⎟ , for x ≤ m, α > 0
⎪ ⎝ α ⎠
⎪
μ A ( x) = ⎨1, for m ≤ x ≤ n
⎪
⎪ R ⎛⎜ x − n ⎞⎟ , for x n, β > 0
⎪⎩ ⎝ β ⎠

Definition 5: A fuzzy number A = (m, n, α , β ) LR is said to be a non-negative LR flat

fuzzy number if m − α 0 and is said to be a non-positive LR flat fuzzy number if
n+β 0 .

Definition 6: Let A = (m, n, α , β ) LR be an LR flat fuzzy number and λ be a real

number in the interval [0,1] then the crisp set, A = {x ∈ R : μ  ( x) λ} = [m − α L−1 (λ ),
λ A

n + β R (λ )] is said to be λ -cut of A .
−1

Remark 1: If m = n and L( x) = R( x) = max{0,1 − x} , then an LR flat fuzzy number

A = (m, n, α , β ) is said to be a triangular fuzzy number and is denoted by A = (m, α , β ) .
LR

Remark 2: If m ≠ n and L( x) = R( x) = max{0,1 − x} , then an LR flat fuzzy number

A = (m, n, α , β ) LR is said to be a trapezoidal fuzzy number and is denoted by
A = (m, n, α , β ) . The membership function of a trapezoidal fuzzy number
A = (m, n, α , β ) is given by
 A. Ebrahimnejad 67

⎧ x − (m − α )
⎪ , for m − α ≤ x ≤ m
α
⎪
⎪1, for m ≤ x ≤ n
μ A ( x) = ⎨
⎪ (n + β ) − x , for n ≤ x ≤ n + β
⎪ β
⎪
⎩0, else

Definition 7 (Dubois and Prade, 1980): Let A1 = (m1 , n1 , α1 , β1 ) LR and

A2 = (m2 , n2 , α2 , β2 ) LR be two non-negative LR fuzzy numbers and
A3 = (m3 , n3 , α3 , β3 ) RL be a RL flat fuzzy number then,

1 A1 + A2 = (m1 + m2 , n1 + n2 , α1 + α 2 , β1 + β2 ) LR .

2 A1 − A3 = (m1 − n3 , n1 − m3 , α1 + β3 , β1 + α3 )LR .

3 k ≥ 0, k ∈ R; kA1 = (km1 , kn1 , kα1 , k β1 )LR .

4 k < 0, k ∈ R; kA1 = (kn1 , km1 , −k β1 , −kα1 )RL .

One important issue in decision-making under uncertainty is how to compare fuzzy

numbers in an efficient manner. One simple and efficient manner for comparing and
ranking fuzzy numbers is based on ranking functions in which a ranking function
ℜ : F ( R ) → R converts each fuzzy numbers into a real number. In this case, comparison
of fuzzy numbers is done as

A1 A2 if and onlyif ℜ( A1 ) ℜ( A2 )

A1 A2 if and only if ℜ( A1 ) > ℜ( A2 )

A1  A2 if and only if ℜ( A1 ) = ℜ( A2 )

Yager (1981) proposed a procedure for ordering fuzzy sets in which a linear ranking
index ℜ( A) is calculated for an LR flat fuzzy number A = (m, n, α , β ) LR according to
the following formula:

ℜ( A) = 12 ( ∫ [m − α L (λ )]d λ + ∫ [n + β R
1

0
−1
1

0
−1
(λ )]d λ ) (1)

Remark 3: For a trapezoidal fuzzy number A = (m, n, α , β ) , Yager’s ranking index is

1 1
reduced to ℜ( A) = (m + n) + ( β − α ) .
2 4
68 Cost parametric analysis of linear programming problems

3 Linear programming problem with fuzzy cost coefficients

In this section, we first review the steps of fuzzy primal simplex algorithm (Maleki et al.,
2000) for solving FLP problem with fuzzy cost coefficients and then explore it with the
help of an application example. The obtained results will be used for parametric analysis
in the next section.

3.1 The fuzzy primal simplex algorithm

A special kind of FLP problems in which all decision parameters except of decision
variables are represented by fuzzy numbers is called fuzzy numbers linear programming
(FNLP) problem. In this paper, we focus on a type of FNLP problems in which only the
cost coefficients are represented by fuzzy numbers. This kind of FLP problem is defined
as follows (Maleki et al., 2000):
max z  cx
s.t. Ax = b (2)
x≥0
In problem (2) , b ∈ Rm , cT ∈ ( F (R))n , A ∈Rm×n are given and x ∈ R n is to be
determined.

Definition 8: Any vector x ∈ R n which satisfies the constraints and non-negative

restrictions of (2) is said to be a feasible solution.

Definition 9: Let S be the set of all feasible solutions of (2) . Any solution x* ∈ S is
said to be a optimum solution to (2) if cx* cx for all x* ∈ S , where c = (c1 , c2 , … , cn )
and cx  c1 x1 + c2 x2 + + cn xn .

Definition 10 [Basic solution]: Consider the system of constraints Ax = b where

rank( A) = m . Partition A as [ B, N ] where B is an (m × m) invertible matrix and N is
m × (n − m) matrix. The solution x = ( xB , xN ) to the equations Ax = b , where
xB = B b and xN = 0 is called a basic solution of the system. If xB = B−1b ≥ 0 , then
−1

x = ( xB , xN ) = ( B−1b, 0) is called a feasible basic solution. Here, B and N are called

basic matrix and non-basic matrix, respectively. Also, x B and xN are called basic
variables and non-basic variables, respectively.
Suppose x is a basic feasible solution of (2) . Let y j and w be the solutions to
By = a j and wB  cB , respectively and define z j  cB B−1 . Maleki et al. (2000) proved
some important theorems for problem (2) and then proposed a new algorithm for solving
such problem. A summary of their method in tableau format is given here.
 A. Ebrahimnejad 69

Algorithm 1: [A fuzzy primal simplex algorithm for FLP problem: Maximisation

problem]
Initialisation Step
Suppose an initial fuzzy basic feasible solution with basis B is at hand. Form the initial
tableau similar to Table 1.

Table 1 The initial FLP primal simplex tableau

Basis xB xN R.H.S.

ℜ( z) 0 ℜ( z N − cN ) = ℜ(cBYN − cN ) ℜ( z ) = ℜ(cB B −1b)

z 0 z N − cN = cBYN − cN cB B −1b
xB I YN b = B −1b

Main Step

1 Calculate z j − c j for all nonbasic variables. Let ℜ( zk − ck ) = Max {ℜ( z j − c j )} ,

j∈T

where T is the index set of the current nonbasic variables.

2 Let yk = B−1 ak . If yk 0 , then stop; the problem is unbounded. Otherwise,
determine the index r of variable xB leaving the basic as follows:
r

br ⎧b ⎫
= min ⎨ i yik > 0 ⎬ .
yrk 1 i m
⎩ yik ⎭
3 Update the tableau by pivoting at yrk . Update the basic and nonbasic variables
where xk enters the basis and xB leaves the basis, and go to (1).
r

3.2 Application example

Here, we solve an application example using Algorithm 1 and we apply this example to
deal with the parametric analysis on the FLP problem throughout this paper.

Example 1: A company produces two products P1 and P2 on two different machines M 1

and M 2 . The time required for manufacturing one unit of each products and the daily
capacity of the machines are given in Table 2.
70 Cost parametric analysis of linear programming problems

Table 2 The data of Example 1

Time per unit (min)

Machines P1 P2 Machine capacity (min/day)

M1 1 2 44

M2 3 1 42

Because of variations in price, profit for each product can be changed. At the same time,
the company wants to keep the profit somewhat close to $40 for P1 and close to $30 for
P2 . The company wants to determine the range of each product to be produced per day in
order to maximise its profit. It is assumed that all the amounts produced are consumed in
the market.
Since the profit from each product is uncertain, the problem will be a FLP problem.
We use trapezoidal fuzzy numbers for each uncertain value. The profit of product P1
which is close to 40 is model as (36, 42,1, 5) and the profit of product P2 which is close
to 30 is modeled as (28, 34, 5,1) . Assume x1 and x2 are the daily amount of products P1
and P2 to determine the optimal combination, respectively. In this case, the problem is
formulated as follows:
max z  (36, 42,1,5) x1 + (28,34,5,1) x2
s.t. x1 + 2 x2 ≤ 44
(3)
3x1 + x2 ≤ 42
x1 , x2 ≥ 0

We construct the standard form of the FLP problem (3) as follows, where x3 and x4 are
the slack variables:
max z  (36, 42,1,5) x1 + (28,34,5,1) x2
s.t. x1 + 2 x2 + x3 = 44
(4)
3x1 + x2 + x4 = 42
x1 , x2 , x3 , x4 ≥ 0

This FLP problem can be solved using the fuzzy primal simplex algorithm. Table 3
presents the initial primal simplex tableau.

Table 3 The initial fuzzy simplex tableau

Basis x1 x2 x3 x4 R.H.S.

ℜ( z) −40 −30 0 0 0
z (−42, −36, 5, 1) (−34, −28, 1, 5) (0, 0, 0, 0) (0, 0, 0, 0) (0, 0, 0, 0)
x3 1 2 1 0 44

x4 3 1 0 1 42
 A. Ebrahimnejad 71

In this table, x 1 and x 4 are entering variable and leaving variable, respectively. The
second simplex tableau presented in Table 4 is derived by pivoting on y21 = 3 .

Table 4 The first iteration simplex tableau

Basis x1 x2 x3 x4 R.H.S.

0 −50 0 40 560
ℜ( z)
3 3
(0,0,0,0) ⎛ 4 20 ⎞ (0,0,0,0) ⎛ 1 5⎞ (504, 588, 14, 70)
z ⎜ −22, −14, , ⎟ ⎜ 12,14, , ⎟
⎝ 3 3 ⎠ ⎝ 3 3⎠
0 5 1 −1 30
x3
3 3
1 1 0 1 14
x1
3 3

5
In this table, x 2 is an entering variable and x 3 is a leaving variable. Pivoting on y12 =
3
gives the third simplex tableau presented in Table 5.

Table 5 The optimal tableau

Basis x1 x2 x3 x4 R.H.S.

ℜ( z) 0 0 10 10 860
z (0,0,0,0) (0,0,0,0) ⎛ 42 66 20 4 ⎞ ⎛ 38 56 3 15 ⎞ (756, 984, 134, 94)
⎜ , , , ⎟ ⎜ , , , ⎟
⎝ 5 5 5 5⎠ ⎝ 5 5 5 5⎠
x2 0 1 3 −1 18
5 5
x1 1 0 −1 2 8
5 5

Table 5 is the optimal tableau because for all non-basic variables:

⎛ 42 46 20 4 ⎞
ℜ( z3 − c3 ) = ℜ ⎜ , , , ⎟ ≥ 0
⎝ 5 5 5 5⎠

⎛ 38 56 3 15 ⎞
ℜ( z4 − c4 ) = ℜ ⎜ , , , ⎟ ≥ 0
⎝ 5 5 5 5 ⎠
This shows that the company should produce eight units of product P1 and eighteen units
of product P2 for obtaining a profit close to $860. The membership function of the total
profit is as follows:
72 Cost parametric analysis of linear programming problems

⎧ x − 622
⎪ 134 for 622 ≤ x ≤ 756
⎪
μ z ( x) = ⎨1 for 756 ≤ x ≤ 984
⎪1078 − x
⎪ for 984 ≤ x ≤ 1078
⎩ 94

4 Shortcomings of the existing methods

There are several methods in the literatures to deal with the sensitivity analysis and
parametric analysis of FLP problems. In this section, the shortcomings of the existing
methods (Kheirfam and Hasani, 2010; Nasseri and Ebrahimnejad, 2011; Kumar and
Bhatia, 2011; Ebrahimnejad, 2011b; Bhatia and Kumar, 2012; Kheirfam and Verdegay,
2013a; Ebrahimnejad and Verdegay, 2014) are pointed out.
1 The existing methods (Kheirfam and Hasani, 2010; Kumar and Bhatia, 2011; Nasseri
and Ebrahimnejad, 2011) can deal with the sensitivity analysis of such FLP problems
in which the decision variables and RHS vectors are represented by fuzzy numbers
and rest of the parameters are represented by real numbers, i.e.
max(min) z ( x)  cx
s.t. Ax  b
x 0

At first, the existing methods (Kheirfam and Hasani, 2010; Kumar and Bhatia, 2011;
Nasseri and Ebrahimnejad, 2011) do not give a non-negative optimal solution after
accruing the changes in the data of the problem under consideration. For instance, the
values of x1 , x2 and x3 in the new optimal solution of Example 5 in Kheirfam and
⎛ 4 88 264 152 ⎞ ⎛ 97 27 207 389 ⎞
Hasani (2010) are as ⎜− , , , ⎟, ⎜− , , , ⎟ and
⎝ 7 7 7 7 ⎠ ⎝ 14 14 14 14 ⎠
⎛ 29 1 57 117 ⎞
⎜− , , , ⎟ , respectively. It is obvious that these values are not non-negative
⎝ 2 2 2 2 ⎠
fuzzy numbers as there are negative parts in these fuzzy values. It should be noted that in
the fuzzy problem under consideration, the constraint x 0 is not a soft constraint. Thus,
no violation would be acceptable. Secondly, the existing methods (Kheirfam and Hasani,
2010; Kumar and Bhatia, 2011; Nasseri and Ebrahimnejad, 2011) can only consider the
effect of the discrete variation of parameters on the optimal solution. They do not
perform a systematic study of how the optimal solution changes as many of the
parameters change simultaneously over some range.
2 The existing method (Ebrahimnejad, 2011b) can deal with the sensitivity analysis of
FLP problem where all parameters of the problem are fuzzy numbers, while the
decision variables are crisp, i.e.
 A. Ebrahimnejad 73

max(min) z ( x) = cx
s.t. Ax  b
x≥0

This approach described how to test specific changes in the model parameters on the
optimal solution. It cannot be used when one or more parameters continuously are
changed over some interval(s).
3 The existing method (Bhatia and Kumar, 2012) can be used to deal with the
sensitivity analysis of the FLP in which all the parameters are represented by LR
flat fuzzy numbers, i.e.
max(min) z ( x)  c ⊗ x
s.t. A⊗ x  b
x 0

where x = ( x j )n×1 , x j = (m j , n j , α j , β j ) . In the existing method, this problem is

converted to the following crisp problem:
max(min) ℜ( z ( x )) = ℜ(c ⊗ x)
s.t. ℜ( A ⊗ x) = ℜ(b ), .
n j − m j ≥ 0, m j − α j ≥ 0, α j ≥ 0, β j ≥ 0, J = 1, 2, … , n

At first, the existing method (Bhatia and Kumar, 2012) used an approximate formula for
the extended multiplication on LR flat fuzzy numbers. Secondly, by converting fuzzy LP
problem into crisp one, the number of decision variables and constraints are increased
significantly. It is worth noting that the number of constraints and variables has a direct
effect on the computational complexity of LP problems. But, the existing method (Bhatia
and Kumar, 2012) increases the number of decision variables and the number of
functional constraints which is the most important tool to determine computation time of
the simplex method. Thirdly, this kind of sensitivity analysis does not involve the
systematic study of how the optimal solution changes as one or more parameters change
continuously over some range.
4 The existing method (Kheirfam and Verdegay, 2013a; Ebrahimnejad and Verdegay,
2014) can deal with the sensitivity analysis of FLP problem in which the coefficients
of the objective function and the values of the right-hand side are represented by
symmetric trapezoidal fuzzy numbers while the elements of the coefficient matrix
are represented by real numbers., i.e.
max(min) z ( x)  cx
s.t. Ax  b
x 0

At first, the assumption that all trapezoidal fuzzy intervals are symmetric is too
restrictive. Most of fuzzy numbers happens to be bell-shaped ( LR fuzzy numbers), even
the non-symmetric trapezoidal fuzzy numbers. Secondly, decision-makers are interested
in crisp and not in fuzzy solutions. Thirdly, they claimed that their approach gives a non-
74 Cost parametric analysis of linear programming problems

negative fuzzy optimal solution, while there are negative parts in the fuzzy optimal
solution [in the optimal solution of Example 2 in Kheirfam and Verdegay (2013a), the
⎛ 1 3 7 7⎞
fuzzy value of x4 is ⎜ − , , , ⎟ and in the optimal solution of Example 7 in
⎝ 4 2 4 4⎠
⎛ 9 26 35 35 ⎞
Kheirfam and Verdegay (2013a), the fuzzy value of x2 is ⎜ − , , , ⎟ ]. Fourthly,
⎝ 11 11 11 11 ⎠
this kind of sensitivity analysis involves only changing one parameter at a time in the
original model to check its effect on the optimal solution. It cannot be applied to show the
effect of the continuous variations of several parameters in the objective function
coefficients on the optimum solution. In other words, the existing methods (Kheirfam and
Verdegay, 2013a; Ebrahimnejad and Verdegay, 2014) cannot be used for situations that
there are factors that cause the parameters to change together, i.e., for situations that the
two products competitive in some sense so that a larger-than-expected unit profit for one
implies a smaller-than-expected unit profit for the other.

5 Parametric analysis of fuzzy cost coefficients

In this section, we obtain the optimal solution to a class of FLP problem (2) by
perturbing the fuzzy cost coefficients vector along a fuzzy vector.
Assume that B is an optimal basis of FLP (2) . Suppose that the fuzzy cost vector c
is perturbed along the fuzzy cost direction c ' , that is, c is replaced by c + λ c' where λ
is a parameter. The problem to be solved is of the following form:
max zλ ( x)  (c + λ c' ) x
s.t. Ax = b (5)
x≥0

The FLP (5) is called a fuzzy parametric cost problem (FPCP). We are interested in
finding the optimal points and corresponding fuzzy objective values as a function of
parameter λ .
Clearly, B is an optimal basis of FPCP (5) when λ = 0 . Decomposing A into
[ B, N ] , c into [cB , cN ] and c ' into [c'B , c'N ] , we get:

z − (cB + λ c'B ) xB − (cN + λ c'N ) xN  0

(6)
BxB + NxN = b

Updating the tableau and denoting c'B y j by z'j , we get

z − ∑[( z j − c j ) + λ ( z'j − c'j )] x j  cB b + λ c'B b

j∈T
(7)
xB + ∑ y j x j = b
j∈T

Let the canonical tableau with respect to basis B be Table 6.

 A. Ebrahimnejad 75

Table 6 The canonical tableau of fuzzy parametric cost problem

Basis xB xN R.H.S.

ℜ( z(λ )) 0 ℜ( z N − cN ) + ℜ( z'N − c'N ) ℜ( z (λ )) = ℜ(cBb + λcB' b )

z(λ) 0 ( z N − cN ) + λ ( z'N − c'N ) z (λ ) = cBb + λcB' b

xB I YN b = B −1b

5.1 Optimality interval of basis B

Here, we determine the range of parameter λ so that when the fuzzy cost coefficients are
distributed, the optimal solution remains invariant.
It is evident that the basis B is an optimal basis for all values of parameter λ
satisfying:

z j (λ ) − c j (λ ) = ( z j − c j ) + λ ( z'j − c'j ) 0, j ∈ T (8)

where T is the index set of the current non-basic variables. This implies that if
ℜ( z'j − c'j ) > 0 then

−ℜ( z j − c j )
λ≥ (9)
ℜ( z'j − c'j )

and if ℜ( z'j − c'j ) < 0 , then

−ℜ( z j − c j )
λ≤ . (10)
ℜ( z'j − c'j )

Hence, λ B ≤ λ ≤ λB where

⎧ ⎪⎧ −ℜ( z j − c j ) ⎪⎫
⎪⎪max ⎨ ℜ( z'j − c'j ) > 0⎬
λB = ⎨ j ∈T ' '
⎩⎪ ℜ( z j − c j ) ⎭⎪ (11)
⎪ ' '
⎪⎩ −∞ , if for all j ℜ( z j − c j ) ≤ 0
and
⎧ ⎪⎧ −ℜ( z j − c j ) ⎪⎫
⎪min ⎨ ' '
ℜ( z'j − c'j ) < 0 ⎬
⎪ j∈T
⎪⎩ ℜ( z j − c j ) ⎪⎭
⎪
λB = ⎨ (12)
⎪ ' '
⎪∞, if for all j ℜ( z j − c j ) ≥ 0
⎪⎩
76 Cost parametric analysis of linear programming problems

The values λ B and λB are called the lower and upper break point of basis B , respectively.
In addition, the closed interval [λ B , λB ] is called the optimality interval of basic B
.Therefore, in the optimality interval of basis, the optimal solution is the same. The only
change will occur in the fuzzy optimal objective value that will be equal to:

zλ ( x)  cB b + λ c'B b (13)

5.2 Solving the problem for λ > λB

Suppose that B is an optimal basis for the FPCP (5) with a non-empty optimality
−ℜ( zs − cs )
interval [λ B , λB ] and suppose λB is finite. Assume that λB = in (12) . When
ℜ( z's − cs' )
λ exceeds λB , the fuzzy relative coefficient of xs becomes negative, i.e.,
zs (λ ) − cs (λ )  ( zs − cs ) + λ ( z − c ) ≺ 0 . In this case, xs enters into the basic variable.
'
s
'
s

When this is done, two cases might occur.

Case 1: The FPCP (5) becomes unbounded, i.e., yis ≤ 0 for all i . In this case, this
problem is unbounded for all λ > λB .

Case 2: Assuming x s enters into the basic variables; the fuzzy primal simplex algorithm
is continued for finding the new optimal solution.
In this case, assume xr drops out of the basic variables and update the tableau by
pivoting on xrs . Let the new basis be B̂ . The same procedure can be repeated with this
new basis and the optimal solution of (5) determined as λ increases further. Some
important results are given using the following theorems.

Theorem 1: Bases B and B̂ are two alternative optimal bases for (5) when λ = λB .

−ℜ( zs − cs )
Proof: Note that λB = . This implies that ℜ( zs − cs ) + λℜ( z'j − c'j ) = 0 .
ℜ( z'j − c'j )
Therefore, the relative cost coefficients of any variable x j with respect to bases B and
B̂ are both equal when λ = λB as the entry in this row in the pivot column is zero. This
completes the proof.…

Theorem 2: The new basis B̂ is not an optimal basis for any λ < λB .

Proof: The relative cost coefficient of the dropping variable xr with respect to new basis
B̂ is
( zs − cs ) + λ ( z's − cs' )
zˆs (λ ) − cˆs (λ )  ( zˆs − cˆs ) + λ ( zˆ's − cˆs' ) 
yrs
 A. Ebrahimnejad 77

We note that the optimality interval of basis B is [λ B , λB ] . This implies that we have
( zs − cs ) + λ ( z's − cs' ) V 0 for any λ < λB . Now, noting that yrs > 0 , we conclude that
( zˆs − cˆs ) + λ ( zˆ's − cˆs' ) ≺ 0 . This implies that the new basis B̂ is not an optimal basis for
any λ < λB .…

Theorem 3: The lower breaking point of new basis B̂ is λ = λB .

Proof: Based on Theorem 1 , bases B and B̂ are two alternative optimal bases for (5)
when λ = λB and based on Theorem 2 , the new basis B̂ is not an optimal basis for any
λ < λB . These imply that the lower breaking point of new basis B̂ is equal to the upper
breaking point of basis B , i.e, λB = λ Bˆ .…

5.3 Solving the problem for λ < λ B

−ℜ( zt − ct )
Returning to the original basis B , suppose λ B is finite. Assume that λ B =
ℜ( zt' − ct' )
in (11) . In the case of λ < λ B , the fuzzy relative coefficient of xt becomes negative, i.e.,
zt (λ ) − ct (λ )  ( zt − ct ) + λ ( zt' − ct' ) ≺ 0 . In this case, xt enters into the basic variable. When
this done, two cases might occur. The FPCP (5) becomes unbounded for all λ < λ B , if
for all i , yit ≤ 0 . Otherwise, the fuzzy primal simplex algorithm is continued for finding
the new optimal solution. In fact, xr is dropped out of the basic variables and the tableau
is updated by pivoting on yrt . Let the new basis be B̂ . From results similar to the case of
λ > λB , it is evident that B and B̂ are two alternative optimal bases for (5) when
λ = λ B . The optimality interval of this new basis is another closed interval with its right
end point equals to the left end point of the optimality interval of basis B . The same
procedure can be repeated with this new basis and the optimal solution of (5) obtained as
λ decreases further.

5.4 Example
Example 2: Consider Example 1a Suppose that management of the company now have
the option of making trade-offs in the profitability of the two activities, whereby the
objective function coefficient of x1 can be increased (decreased) by simultaneously
decreasing (increasing) the objective function coefficient of x2. In this case, the aim is to find the
optimal solution and the fuzzy optimal objective values of the class problem whose fuzzy
objective function (c + λ c' )  ((36, 42,1,5) + λ (17, 21,1,5),(28,34,5,1) + λ (−11, −7,5,1))
is a function of λ ; that is, the profits of products P1 and P2 are perturbed along the
fuzzy profit vector c'  ( (17, 21,1, 5), (−11, −7, 5,1) ) . In fact, we should solve the
following fuzzy cost parametric problem:
78 Cost parametric analysis of linear programming problems

max z  ( (36, 42,1,5) + λ (17, 21,1,5) ) x1 + ( (28,34,5,1) + λ (−11, −7,5,1) ) x2

s.t. x1 + 2 x2 ≤ 44
(14)
3 x1 + x2 ≤ 42
x1 , x2 ≥ 0

The optimal solution for λ = 0 is given in Table 5. In order to find the range over which
this tableau is optimal, we first find ( z3' − c3' ) and ( z'4 − c4' ) as follows:

⎛ 3 ⎞
⎜ 5 ⎟ ⎛ −54 −38 20 4 ⎞
( z − c )  c y3 − c  ( (17, 21,1, 5), ( −11, −7, 5,1) ) ⎜ ⎟ − (0, 0, 0, 0)  ⎜
' ' ' '
3 3 B 3
, , , ⎟
−
⎜ ⎟ 1 ⎝ 5 5 5 5⎠
⎜ ⎟
⎝ 5 ⎠

⎛ −1 ⎞
⎜ 5 ⎟ ⎛ 41 51 3 15 ⎞
( z 4 − c4 )  cB y4 − c4  ( (17, 21,1, 5), ( −11, −7, 5,1) ) ⎜ ⎟ − (0, 0, 0, 0)  ⎜ , , , ⎟
' ' ' '

⎜ 1 ⎟ ⎝ 5 5 5 5⎠
⎜ ⎟
⎝ 5 ⎠

We note that ℜ( z3' − c3' ) < 0 and ℜ( z'4 − c4' ) > 0 . Hence, based on Eqs. (11) and (12) we
have λ B = −1 and λB = 1 , respectively. This implies that the optimality interval of basis
B = [a2 , a1 ] is [λ B , λB ] = [−1,1] . The fuzzy optimal objective function value in this
interval based on (13) is given as follows:

zλ ( x)  cB b + λ cB' b  (756,984,134,94)
⎛ 18 ⎞
+ λ ( (−11, −7, 5,1), (17, 21,1,5) ) ⎜ ⎟
⎝8⎠
 (756,984,134,94) + λ ( −62, 42,98,58)
In this case, the optimal solution in this optimality interval is given as Table 7.

Table 7 The optimal tableau in the range of [–1,1]

Basis x1 x2 x3 x4 R.H.S.

ℜ( z(λ )) 0 0 10 − 10λ 10 + 10λ 860 − 20λ

0 0 ⎛ 42 66 20 4 ⎞ ⎛ 38 56 3 15 ⎞ (765, 984, 134, 94) +

⎜ , , , ⎟ + ⎜ , , , ⎟ +
⎝ 5 5 5 5⎠ ⎝ 5 5 5 5 ⎠ ( − 62, 42, 98, 58)λ
z(λ) ⎛ −54 −38 20 4 ⎞ ⎛ 41 53 3 15 ⎞
⎜ , , , λ ⎜ , , , ⎟
⎝ 5 5 5 5 ⎟⎠ ⎝ 5 5 5 5 ⎠
λ
1 0 3 −1 18
x2
5 5
0 1 −1 2 8
x1
5 5
 A. Ebrahimnejad 79

Now, we should solve the problem for λ > λB = 1 and λ < λ B = −1 . For λ > λB = 1 , the
variable x3 does not satisfy the optimality condition. In fact, at λB = 1 , the value of
ℜ( z3 (λ ) − c3 (λ )) = 10 − 10λ is zero, and x3 is introduced into the basis leading to
Table 8.

Table 8 The updated tableau for λ > 1

Basis x1 x2 x3 x4 R.H.S.

0 −50 50 0 40 20 560 + 280λ

ℜ( z (λ )) + λ + λ
3 3 3 3
0 ⎛ 4 20 ⎞ 0 ⎛ 52 78 29 49 ⎞ (360, 732, 158,
⎜ −22, − 14, 3 , 3 ⎟ + ⎜ 5 , 5 , 15 , 15 ⎟ + 214)+(166, 366,
⎝ ⎠ ⎝ ⎠ 122, 178) λ
z (λ )
⎛ 38 54 4 20 ⎞ ⎛ 69 121 19 49 ⎞
⎜ 3 , 3 , 3 , 3 ⎟λ ⎜ 15 , 15 , 15 , 15 ⎟ λ
⎝ ⎠ ⎝ ⎠
0 5 1 −1 30
x3
3 3
1 1 0 1 14
x1
3 3

In order to obtain the optimality interval of basis Bˆ = [a3 , a1 ] , we should have

ℜ( z2 (λ ) − c2 (λ )) ≥ 0 and ℜ( z4 (λ ) − c4 (λ )) ≥ 0 :

⎛⎛ 14 20 ⎞ ⎛ 38 54 4 20 ⎞ ⎞ 50 50
ℜ( z2 (λ ) − c2 (λ )) = ℜ ⎜ ⎜ −22, −14, , ⎟ + λ ⎜ , , , ⎟ ⎟ = − + λ ≥ 0 ⇒ λ ≥ 1
⎝⎝ 3 3 ⎠ ⎝ 3 3 3 3 ⎠⎠ 3 3

⎛ ⎛ 52 78 29 49 ⎞ ⎛ 69 121 29 49 ⎞ ⎞ 40 20
ℜ( z4 (λ ) − c4 (λ )) = ℜ ⎜ ⎜ , , , ⎟ + λ ⎜ , , , ⎟⎟ = + λ ≥ 0 ⇒ λ ≥ −2
⎝ ⎝ 5 5 15 15 ⎠ ⎝ 15 15 15 15 ⎠ ⎠ 3 3

These imply that the basis Bˆ = [a3 , a1 ] is optimal for any λ ≥ 1 . In fact, since
⎛ 38 54 4 20 ⎞ 50
ℜ ( z'2 − c2' ) = ℜ ⎜ , , , ⎟ = >0 and
⎝ 3 3 3 2 ⎠ 3
⎛ 69 121 29 49 ⎞ 20
ℜ ( z '4 − c4' ) = ℜ ⎜ , , , ⎟= > 0 , based on Eqs. (11) and (12) we conclude
⎝ 15 15 15 15 ⎠ 3
that

⎧ 50 40 ⎫
⎪− − ⎪
λ Bˆ = max ⎨ 3 , 3 ⎬ = 1
⎪ 50 20 ⎪
⎩ 3 3 ⎭
80 Cost parametric analysis of linear programming problems

and λBˆ = ∞ , respectively.

As we see the left end point of new basis Bˆ = [a3 , a1 ] is equal to the right end point of the
basis B = [a2 , a1 ] . The fuzzy optimal objective function value in this new optimality
interval is equal to zλ ( x)  (360, 732,158, 214) + λ (166, 366,122,178) .
Now, we solve the problem for λ < λ B = −1 . For λ < λ B = −1 , the variable x4 does
not satisfy the optimality condition with regard to Table 7. In fact, at λ = λ B = −1 , the
value of ℜ( z4 (λ ) − c4 (λ )) = 10 + 10λ is zero, and x4 is introduced into the basis leading
to Table 9.

Table 9 The updated tableau for λ < − 1

Basis x1 x2 x3 x4 R.H.S.

ℜ( z(λ )) −25 − 25λ 0 15 − 15λ 0 660 − 220 λ

z(λ) 15 3 0 ⎛ 61 94 43 23 ⎞ 0 (532, 832, 194,

(−28,−19, , )+ ⎜ , , , ⎟+ 106)+ (−274,−122,
2 2 ⎝ 5 5 10 10 ⎠
158, 70) λ
⎛ −53 −41 15 3 ⎞ ⎛ −67 −23 43 23 ⎞
⎜ , , , ⎟ ⎜ , , , ⎟
⎝ 15 15 2 2 ⎠ ⎝ 10 10 10 10 ⎠
λ λ
x2 1 1 1 0 22
2 2
x4 5 0 −1 1 20
2 2

In order to obtain the optimality interval of basis B = [a2 , a4 ] , we should have

ℜ( z1 (λ ) − c1 (λ )) ≥ 0 and ℜ( z3 (λ ) − c3 (λ )) ≥ 0 :

⎛⎛ 15 3 ⎞ ⎛ −53 −41 15 3 ⎞ ⎞
ℜ( z1 (λ ) − c1 (λ )) = ℜ ⎜ ⎜ −28, −19, , ⎟ + λ ⎜ , , , ⎟ ⎟ = −25 − 25λ ≥ 0 ⇒ λ ≤ −1
⎝⎝ 2 2⎠ ⎝ 2 2 2 2 ⎠⎠
⎛ ⎛ 61 94 43 23 ⎞ ⎛ −67 −23 43 23 ⎞ ⎞
ℜ( z3 (λ ) − c3 (λ )) = ℜ ⎜ ⎜ , , , ⎟ + λ ⎜ , , , ⎟ ⎟ = 15 − 15λ ≥ 0 ⇒ λ ≤ 3
⎝ ⎝ 5 5 10 10 ⎠ ⎝ 10 10 10 10 ⎠ ⎠

These imply that the basis Bˆ = [a2 , a4 ] is optimal for any λ ≤ − 1 . In fact, since

⎛ −53 −41 15 13 ⎞
ℜ ( z1' − c1' ) = ℜ ⎜ , , , ⎟ = −25 ≤ 0
⎝ 2 3 2 2⎠

and
⎛ −67 −23 43 23 ⎞
ℜ ( z3' − c3' ) = ℜ ⎜ , , , ⎟ = −5 ≤ 0 ,
⎝ 10 10 10 10 ⎠
based on Eqs. (11) and (12) we conclude that λ Bˆ = −∞ and
 A. Ebrahimnejad 81

⎧ 25 −15 ⎫
λBˆ = min ⎨ , ⎬ = −1
⎩ −25 −5 ⎭

As we see the right end point of new basis Bˆ = [a2 , a4 ] is equal to the left end point of
the basis B = [a2 , a1 ] . The fuzzy optimal objective function value in the new optimality
interval ( −∞, −1] is equal to zλ ( x)  (532, 832,194,106) + λ (−274, −122,158, 70) .
Figures 1–3 show the fuzzy objective function values in the intrvals ( −∞, −1] , [−1,1]
and [1, + ∞ ) , respectively.

Figure 1 The fuzzy objective function value as a function of λ in the intrval ( −∞, −1] (see
online version for colours)

Figure 2 The fuzzy objective function value as a function of λ in the intrval [−1,1] (see online
version for colours)
82 Cost parametric analysis of linear programming problems

Figure 3 The fuzzy objective function value as a function of λ in the interval [1, + ∞ ) (see
online version for colours)

6 Advantages of the proposed method

The fuzzy parametric analysis can provide a very useful extension of fuzzy sensitivity
analysis (Kheirfam and Hasani, 2010; Nasseri and Ebrahimnejad, 2011; Kumar and
Bhatia, 2011; Ebrahimnejad, 2011b; Bhatia and Kumar, 2012; Kheirfam and Verdegay,
2013a; Ebrahimnejad and Verdegay, 2014), e.g., to check the effect of correlated
parameters that change together due to exogenous factors such as the state of the
economy. In this section, it is considered that using the proposed method we can
overcome the shortcomings described in Section 4.
1 The proposed method finds a crisp solution that is acceptable for decision-makers in
real-world problems.
2 Contrary to the existing fuzzy sensitivity analysis problem (Kheirfam and Verdegay,
2013a; Ebrahimnejad and Verdegay, 2014), the fuzzy coefficients of objective
function are not limited to the symmetric trapezoidal fuzzy numbers in the proposed
fuzzy sensitivity analysis problem. The bell-shaped membership function used in the
proposed approach is used most often for representing fuzzy numbers in real-life
applications.
3 The existing fuzzy sensitivity analysis problems (Kheirfam and Hasani, 2010;
Kheirfam and Verdegay, 2013a) involve changing one parameter at a time in the
original model to check its effect on the optimal solution. By contrast, the proposed
fuzzy parametric analysis problem involves the systematic study of how the optimal
solution changes as many of the parameters change simultaneously over some range.
4 The main advantage of the proposed fuzzy parametric analysis problem over the
existing fuzzy sensitivity analysis problems (Kheirfam and Hasani, 2010; Nasseri
and Ebrahimnejad, 2011; Kumar and Bhatia, 2011; Ebrahimnejad, 2011b; Bhatia and
Kumar, 2012; Kheirfam and Verdegay, 2013a; Ebrahimnejad and Verdegay, 2014) is
 A. Ebrahimnejad 83

the investigation of trade-offs in parameter values. For example, if the c j values

represent the unit profits of the respective activities, it may be possible to increase
some of the c j values at the expense of decreasing others by an appropriate shifting
of personnel and equipment among activities.
5 Contrary to the proposed method by Bhatia and Kumar (2012), the number of
decision variables and the number of functional constraints as the most important
tools to determine computation time of the simplex method are not increaed in the
proposed method.

7 A real-life fuzzy transportation problem

The purpose of transportation problem as a special kind of network flow problems

(Ebrahimnejad, 2014, 2015b, 2016; Ebrahimnejad and Verdegay, 2016b) is to determine
a feasible ‘shipping pattern’ from origins to destinations that minimises the total
transportation cost. In this section, a real-life fuzzy transportation problem (adopted from
Kaur and Kumar, 2012) is used in order to show the applicability of the fuzzy parametric
analysis problem.
The owner of a regional coal company situated in Jharia (Dhanbad, India) wants to
know how to minimise the cost of shipping coals from four collieries to six washeries.
The data are given in Table 10.

Table 10 Summary of the fuzzy transportation problem

Washery

Colliery D1 D2 D3 D4 D5 D6 Supply
S1 (27,35,7,8) (11,12,2,2) (15,18,5,2) (20,22,5,2) (15,18,5,2) (12,15,3,3) 124

S2 (25,35,5,6) (11,12,2,4) (11,12,2,2) (15,21,5,2) (15,18,5,5) (12,15,3,3) 120

S3 (10,12,1,4) (70, 74,5,2) (25,35,5,6) (15,22,3,2) (12,14,2,4) (20,26,5,2) 150

S4 (11,12,2,2) (15, 21,5,3) (25,35,5,6) (15,18,5,6) (10,12,2,2) (20,25,5,3) 170

Demand 112 90 84 92 106 80

Since the owner is certain about the values of availability of the coal at different collieries
and the values of demand of the coal at different washeries, these parameters are
represented by real numbers in Table 10. However, since the rates of diesel vary
frequently, he \ she generally cannot estimate the the transportation cost precisely. In this
kind of situation, the usual way is to obtain the fuzzy data based upon past experience or
expert advice. The objective is to minimise the total transportation costs. Figure 4 is a
graphical representation of the regional coal company’s problem.
84 Cost parametric analysis of linear programming problems

Figure 4 Graphical representation of fuzzy transportation problem (see online version for
colours)

To formulate this problem as an FLP problem, we begin by defining a variable for each
decision that the owner of the company must make. Because the owner of the company
must determine how much coal is sent from each colliery to each washery, we define (for
i = 1, 2, 3, 4 and j = 1, 2, 3, 4, 5, 6 ):

xij = amount of coal(ton) produced at collieryi and sent to washery j .

In terms of these variables, the total cost of supplying the coal demands to washeries may
be written as
(20, 27, 35, 41) x11 + (9,11,12,14) x12 + (10,15,18, 20) x13 + (15, 20, 22, 24) x14 + (10,15,18, 20) x15
+ (9,12,15,18) x16 (cost of shipping coal from colliery1) + (20, 25, 35, 41) x21 + (9,11,12,16) x22
+ (9,11,12,14) x23 + (10,15, 21, 23) x24 + (10,15,18, 23) x25 + (9,12,15,18) x26
(cost of shipping coal from colliery 2) + (9,10,12,16) x31 + (65, 70, 74, 76) x32
+ (20, 25, 35, 41) x33 + (12,15, 22, 24) x34 + (10,12,14,18) x35 + (15, 20, 26, 28) x36
(cost of shipping coal from colliery 3) + (9,11,12,14) x41 + (10,15, 21, 24) x42 +
(20, 25, 35, 41) x43 + (10,15,18, 20) x44 + (8,10,12,14) x45 + ((15, 20, 25, 28) x46
(cost of shipping coal from colliery14).
 A. Ebrahimnejad 85

The company faces two types of constraints. First, the total coal supplied by each colliery
cannot exceed the colliery’s capacity. For example, the total amount of coal sent from
colliery 1 to the six washeries exceed 124 ton. Each variable with first subscript 1
represents a shipment of coal from colliery 1, so we may express this restriction by the
constraint x11 + x12 + x13 + x14 + x15 + x16 ≤ 124 .
In a similar fashion, we can find constraints that reflect colliery 2’s, colliery 3’s and
colliery 4’s capacities. In this case, this problem contains the following four supply
constraints:
x11 + x12 + x13 + x14 + x15 + x16 ≤ 124 (colliery1 supply constraint)
x21 + x22 + x23 + x24 + x25 + x26 ≤ 120 (colliery 2 supply constraint)
x31 + x32 + x33 + x34 + x35 + x36 ≤ 150 (colliery 3 supply constraint)
x41 + x42 + x43 + x44 + x45 + x46 ≤ 170 (colliery 4 supply constraint)

Second, we need constraints that ensure that washery will receive sufficient coal to meet
its peak demand. For example, washery 1 must receive at least 112 ton. Each variable
with second subscript 1 represents a shipment of coal to washery 1, so we obtain the
constraint x11 + x21 + x31 + x41 ≥ 112 .
Similarly, we obtain a constraint for each of washeries 2, 3, 4, 5 and 6. A constraint
that ensures a location receives its demand is a demand constraint. The company must
satisfy the following six demand constraints:
x11 + x21 + x31 + x41 ≥ 112 (Washery1 demand constraint)
x12 + x22 + x32 + x42 ≥ 90 (Washery 2 demand constraint)
x13 + x23 + x33 + x43 ≥ 84 (Washery 3 demand constraint)
x14 + x24 + x34 + x44 ≥ 92 (Washery 4 demand constraint)
x15 + x25 + x35 + x45 ≥ 106 (Washery 5 demand constraint)
x16 + x26 + x36 + x46 ≥ 80 (Washery 6 demand constraint)

Because all the xij s must be non-negative, we add the sign restrictions
xij ≥ 0, (i = 1, 2,3, 4 j = 1, 2,3, 4,5,6) .
Combining the objective function, supply constraints, demand constraints and sign
restrictions yields the following FLP formulation of regional coal company’s problem:
min z  (20, 27, 35, 41) x11 + (9,11,12,14) x12 + (10,15,18, 20) x13
+ (15, 20, 22, 24) x14 + (10,15,18, 20) x15 + (9,12,15,18) x16
+ (20, 25, 35, 41) x21 + (9,11,12,16) x22 + (9,11,12,14) x23
+ (10,15, 21, 23) x24 + (10,15,18, 23) x25 + (9,12,15,18) x26
+ (9,10,12,16) x31 + (65, 70, 74, 76) x32 + (20, 25,35, 41) x33
+ (12,15, 22, 24) x34 + (10,12,14,18) x35 + (15, 20, 26, 28) x36
+ (9,11,12,14) x41 + (10,15, 21, 24) x42 + (20, 25,35, 41) x43
 86 Cost parametric analysis of linear programming problems

s.t.
x11 + x12 + x13 + x14 + x15 + x16 ≤ 124 (Supply constraints)
x21 + x22 + x23 + x24 + x25 + x26 ≤ 120
x31 + x32 + x33 + x34 + x35 + x36 ≤ 150
x41 + x42 + x43 + x44 + x45 + x46 ≤ 170
x11 + x21 + x31 + x41 ≥ 112 (Demandconstraints)
(15)
x12 + x22 + x32 + x42 ≥ 90
x13 + x23 + x33 + x43 ≥ 84
x14 + x24 + x34 + x44 ≥ 92
x15 + x25 + x35 + x45 ≥ 106
x16 + x26 + x36 + x46 ≥ 80
xij ≥ 0, i = 1, 2,3, 4 j = 1, 2,3, 4,5, 6.

On solving the fuzzy transportation problem (15) using Algorithm 1, the obtained
solution and minimum total fuzzy transportation cost are:
x11 = 90, x12 = 0, x13 = 0, x14 = 0, x15 = 0, x16 = 34, x21 = 0, x22 = 0, x23 = 84, x24 = 0, x25 = 0,
x26 = 36, x31 = 112, x32 = 0, x33 = 0, x34 = 38, x35 = 0, x36 = 0, x41 = 0, x42 = 0, x43 = 0,
x44 = 54, x45 = 106, x46 = 10, and z  (5198, 6514, 7812, 9244).

Thus, basis B1 = [a1,5 , a1,10 , a2,7 , a2,10 , a3,5 , a3,8 , a4,8 , a4,9 , a4,10 ] is the optimal basis of the
fuzzy transportation problem (15) where ai , j = ei − e j and ei and e j are unit vectors in
R 10 , with 1s in the i th and j th positions, respectively.
Suppose that we now want to apply fuzzy parametric linear programming analysis to
this problem. Assume that the fuzzy transportation cost is perturbed along the fuzzy cost
vector direction c ' given in Table 11.

Table 11 Direction of perturbed fuzzy transportation cost

Washery

Colliery D1 D2 D3 D4 D5 D6

S1 (−0.4, (1.8,7.2,0.9,0.9) (0.8,3.2,0.4,0.4) (1.4,5.6,0.7,0.7) (0.8,3.2,0.4,0.4) (1,4,0.5,0.5)

−1.6,0.2,0.2)
S2 (−1.2, (1.4,5.6,0.7,0.7) (0.4,1.6,0.2,0.2) (1,4,0.5,0.5) (−0.8, (0.6,2.4,0.3,0.3)
−0.3,0.15,0.15) −0.2,0.1,0.1)
S3 (1.2,4.8,0.6,0.6) (0, 0,0,0) (1.4,5.6,0.7,0.7) (2,8,1,1) (1.6,6.4,0.8,0.8) (1.6,6.4,0.8,0.8)

S4 (0.6,2.4,0.3,0.3) (1.8,7.2,0.9,0.9) (1.2,4.8,0.6,0.6) (1.4,5.6,0.7,0.7) (1,4,0.5,0.5) (1.2,4.8,0.6,0.6)

Now as explained in the proposed algorithm, the optimality interval of basic B1 is

[ λ B , λ B ] = [ −5, 5.75] . The fuzzy optimal objective function value in this optimality
1 1

interval is equal to zλ ( x)  (6474, 7802,1326,1442) + λ (653.2, 2525, 326.6, 326.6) .

 A. Ebrahimnejad 87

As described in Section 5, we can determine when and how the optimal solution
changes (if it does) as increases from 5.75 to any specified positive number or it
decreases from to any specified negative number. The optimal solutions and
corresponding fuzzy objective function values as the functions of parameter are shown in
Table 12 for four iterations.

Table 12 The optimal solution for four iterations

Optimality
interval Optimal solution Fuzzy optimal value

[−5,5.75] x 12 = 80, x 16 = 44, x 23 = 84, (6474,7802,1326,1442)

+ λ (653.2,2525,326.6,326.6)
x 26 = 36, x 31 = 112, x 34 = 38,
x 42 = 10, x 44 = 54, x 45 = 106
[5.75, 7.875] x 12 = 44, x 16 = 80, x 23 = 84, (6798,8380,1542,1802)
+ λ (602.8,2519.4,319.4,319.4)
x 25 = 36, x 31 = 112, x 34 = 38,
x 42 = 46, x 44 = 54, x 45 = 70

[7.875,8] x 13 = 44, x 16 = 80, x 23 = 40, (7370,9304,1938,1978)

+ λ (541.2,2404.8,256.6,256.6)
x 25 = 80, x 31 = 112, x 34 = 38,
x 42 = 90, x 44 = 54, x 45 = 26
[8,8.625] x 13 = 70, x 16 = 54, x 23 = 14, (6752,9760,2198,2134)
+ λ (504.8,2337.2.4,278.4,278.4)
x 25 = 106, x 31 = 112, x 34 = 12,
x 36 = 26, x 42 = 90, x 44 = 80

8 Concluding remarks

The main purpose of the fuzzy cost parametric analysis is to obtain the optimal solution
and the fuzzy optimal objective values as a function of parameters when the fuzzy cost
coefficients are perturbed along a fuzzy cost vector. To the best of our knowledge, till
now there is no method in the literature to deal with the parametric analysis of such FLP
problems without converting them to the crisp linear programming problems. In this
paper, we proposed a new method based on ranking function to handle the fuzzy cost
parametric analysis of that kind of FLP problems in which the cost coefficients of
objective function are represented by fuzzy numbers, and the remaining parameters are
represented by real numbers.
We see the following four possible directions for future researches:
1 In this paper, we studied the parametric analysis of fuzzy vector in FLP problem with
fuzzy cost coefficients. The proposed method can generalise for a situation in which
the RHS vector is perturbed along another RHS vector.
88 Cost parametric analysis of linear programming problems

2 The parametric analysis of FLP problems, in which both fuzzy cost coefficients and
RHS constraints simultaneously are perturbed, is left to the next research work.

3 It is worth noting that for the fuzzy number A , the properties A − A  0 and are not
satisfied. Recently, Kheirfam and Verdegay (2013b) defined new arithmetic
operations of symmetric trapezoidal fuzzy numbers satisfying the above-mentioned
properties. Then, they proposed a new approach for solving fully FLP problems
where all parameters and variables are symmetric trapezoidal fuzzy numbers.
However, as mentioned in Section 4, the assumption that all trapezoidal fuzzy
intervals are symmetric ones is too restrictive. In future, it may be tried to think some
new arithmetic operations of flat fuzzy numbers so that the properties and are
satisfied and then fully FLP problems (all parameters and variables are flat fuzzy
numbers) and the corresponding fuzzy sensitivity and parametric analysis problem
can be solved without converting it into crisp linear programming problems.
4 The proposed approach cannot be used for situations in which the coefficients of
decision variable in the objective function are changed using separate lambda values,
simultaneously. In the future, we will attempt to develop new methods to overcome
this defect.

References
Allahviranloo, T., Lotfi, F.H., Kiasary, M.K., Kiani, N.A. and Alizadeh, L. (2008) ‘Solving full
fuzzy linear programming problem by the ranking function decision variables,’ Applied
Mathematical Sciences, Vol. 29, No. 1, pp.19–32.
Baykasoğlu, A. and Subulan, K. (2015) ‘An analysis of fully fuzzy linear programming with fuzzy
decision variables through logistics network design problem’, Knowledge-Based Systems,
Vol. 90, pp.165–184.
Bhatia, N. and Kumar, A. (2012) ‘Mehar’s method for solving fuzzy sensitivity analysis problems
with LR flat fuzzy numbers’, Applied Mathematical Modelling, Vol. 36, No. 9, pp.4087–4095.
Chinnadurai, V. and Muthukumar, S. (2016) ‘Solving the linear fractional programming problem in
a fuzzy environment: Numerical approach’, Applied Mathematical Modelling, Vol. 40,
Nos. 11–12, pp.6148–6164.
Das, S.K., Mandal, T., and Edalatpanah, S.A. (2016) ‘A mathematical model for solving fully fuzzy
linear programming problem with trapezoidal fuzzy numbers’, Applied Intelligence. doi:
10.1007/s10489-016-0779-x.
Delgado, M., Herrera, F., Verdegay, J.L. and Vila, M.A. (1993) ‘Solution of a fully fuzzy multi-
item economic order quantity problem by using fuzzy ranking functions’, Fuzzy Sets and
Systems, Vol. 53, No. 3, pp.289–297.
Dubois, D. and Prade, H. (1980) Fuzzy Sets and Systems: Theory and Applications, Academic
Press, New York.
Dutta, D., Rao, J.R. and Tiwari, R.N. (1992) ‘Sensitivity analysis in fuzzy linear fractional
programming problem’, Fuzzy Sets and Systems, Vol. 48, No. 2, pp.211–216.
Ebrahimnejad, A. (2011a) ‘Some new results in linear programming with trapezoidal fuzzy
numbers: finite convergence of the Ganesan and Veeramani’s method a fuzzy revised simplex
method’, Applied Mathematical Modelling, Vol. 35, No. 9, pp.4526–4540.
Ebrahimnejad, A. (2011b) ‘Sensitivity analysis in fuzzy number linear programming problems’,
Mathematical and Computer Modelling, Vol. 53, No. 9, pp.1878–1888.
 A. Ebrahimnejad 89

Ebrahimnejad, A. (2014) ‘A simplified new approach for solving fuzzy transportation problems
with generalized trapezoidal fuzzy numbers’, Applied Soft Computing, Vol. 19, pp.171–176.
Ebrahimnejad, A. (2015a) ‘A duality approach for solving bounded linear programming problems
with fuzzy variables based on ranking functions and its application in bounded transportation
problems’, International Journal of Systems Science, Vol. 46, No. 11, pp.2048–2060.
Ebrahimnejad, A. (2015b) ‘An improved approach for solving fuzzy transportation problem with
triangular fuzzy numbers’, Journal of Intelligent and Fuzzy Systems, Vol. 29, No. 2,
pp.963–974.
Ebrahimnejad, A. (2016) ‘New method for solving fuzzy transportation problems with LR flat
fuzzy numbers’, Information Sciences, Vol. 37, pp.108–124.
Ebrahimnejad, A., Nasseri, S.H. and Mansourzadeh, S.M. (2011) ‘Bounded primal simplex
algorithm for bounded linear programming with fuzzy cost coefficients’, International Journal
of Operations Research and Information Systems, Vol. 2, No. 1, pp.96–120.
Ebrahimnejad, A. and Verdegay, J.L. (2014) ‘A novel approach for sensitivity analysis in linear
programs with trapezoidal fuzzy numbers’, Journal of Intelligent and Fuzzy Systems, Vol. 27,
No. 1, pp.173–185.
Ebrahimnejad, A. and Verdegay, J.L. (2016a) ‘A survey on models and methods for solving fuzzy
linear programming problems’, in Fuzzy Logic in Its 50th Year, published in Studies in
Fuzziness and Soft Computing, Vol. 341, Springer International Publishing, Switzerland,
pp.327–368.
Ebrahimnejad, A. and Verdegay, J.L. (2016b) ‘An efficient computational approach for solving
type-2 intuitionistic fuzzy numbers based transportation problems’, International Journal of
Computational Intelligence Systems, Vol. 9, No. 6, pp.1154–1173.
Ezzati, R., Khorram, E. and Enayati, R. (2014) ‘A particular simplex algorithm to solve fuzzy
lexicographic multi-objective linear programming problems and their sensitivity analysis on
the priority of the fuzzy objective functions’, Journal of Intelligent and Fuzzy Systems, Vol.
26, No. 5, pp.2333–2358.
Ezzati, R., Khorram, E. and Enayati, R. (2015) ‘A new algorithm to solve fully fuzzy linear
programming problems using the MOLP problem’, Applied Mathematical Modelling, Vol. 29,
No. 12, pp.3183–3193.
Farhadinia, B. (2014) ‘Sensitivity analysis in interval-valued trapezoidal fuzzy number linear
programming problems’, Applied Mathematical Modelling, Vol. 38, No. 1, pp.50–62.
Ganesan, K. and Veeramani, P. (2006) ‘Fuzzy linear programming with trapezoidal fuzzy
numbers’, Annals of Operations Research, Vol. 1, pp.305–315.
Garcia-Aguado, C. and Verdegay, J.L. (1993) ‘On the sensitivity of membership functions for
fuzzy linear programming problems’, Fuzzy Sets and Systems, Vol. 56, No. 1, pp.47–49.
Gupta, P. and Bhatia, D. (2001) ‘Sensitivity analysis in fuzzy multi-objective linear fractional
programming problem’, Fuzzy Sets and Systems, Vol. 122, No. 2, pp.229–236.
Hamacher, H., Liberling, H. and Zimmermann, H.J. (1978) ‘Sensitivity analysis in fuzzy linear
programming’, Fuzzy Sets and Systems, Vol. 1, No. 4, pp.269–281.
Hatami-Marbini, A. and Tavana, M. (2011) ‘An extension of the linear programming method with
fuzzy parameters’, International Journal of Mathematics in Operational Research, Vol. 3,
No. 1, pp.44–55.
Herrera, F., Kovacs, M. and Verdegay, J.L. (1993) ‘Optimality for fuzzified mathematical
programming problems: a parametric approach’, Fuzzy Sets and Systems, Vol. 54, No. 3,
pp.279–284.
Hosseinzadeh-Lotfi, F., Allahviranloo, T., Jondabeh, M.A. and Alizadeh, L. (2009) ‘Solving a full
fuzzy linear programming using lexicography method and fuzzy approximate solution’,
Applied Mathematical Modelling, Vol. 33, No. 7, pp.3151–3156.
90 Cost parametric analysis of linear programming problems

Jia, Y.L., Yang, Y. and Zhong, Y.h. (2013) ‘Parametric study of fuzzy number linear
programming’, Proc.9th Int. Conf. IEEEComput. Intell. Secur. (CIS), pp.339–343.
Kaur, A. and Kumar, A. (2012) ‘A new approach for solving fuzzy transportation problems using
generalized trapezoidal fuzzy numbers’, Applied Soft Computing, Vol. 12, No. 3,
pp.1201–1213.
Kheirfam, B. and Hasani, F. (2010) ‘Sensitivity analysis for fuzzy linear programming problems
with fuzzy variables’, Advanced Modeling and Optimization, Vol. 12, No. 2, pp.257–272.
Kheirfam, B. and Verdegay, J.L. (2012) ‘Strict sensitivity analysis in fuzzy quadratic
programming’, Fuzzy Sets and Systems, Vol. 198, No. 1, pp.99–111.
Kheirfam, B. and Verdegay, J.L. (2013a) ‘The dual simplex method and sensitivity analysis for
fuzzy linear programming with symmetric trapezoidal numbers’, Fuzzy Optimization and
Decision Making, Vol. 12, No. 1, pp.171–189.
Kheirfam, B. and Verdegay, J.L. (2013b) ‘Optimization and reoptimization in fuzzy linear
programming problems’, Proc.8th Conf. Eur. Soc. Fuzzy Logic Technol., Atlantis Press, Italy,
pp.527–533.
Klir, G.J. and Yuan, B. (1995) 11Fuzzy Sets and Fuzzy Logic: Theory and Applications, Prentice
Hall P T R, New Jersey.
Kumar, A. and Bhatia, N. (2011) ‘A new method for solving fuzzy sensitivity analysis problems’,
International Journal of Applied Science and Engineering, Vol. 9, No. 1, pp.49–64.
Maleki, H.R., Tata, M. and Mashinchi, M. (2000) ‘Linear programming with fuzzy variables’,
Fuzzy Sets and Systems, Vol. 109, No. 1, pp.8499–8507.
Najafi, H.S., Edalatpanah, S.A. and Dutta, H. (2016) ‘A nonlinear model for fully fuzzy linear
programming with fully unrestricted variables and parameters’, Alexandria Engineering
Journal, Vol. 55, No. 3, pp.2589–2595.
Nasseri, S.H. and Ebrahimnejad, A. (2011) ‘Sensitivity analysis on linear programming problems
with trapezoidal fuzzy variables’, International Journal of Operations Research and
Information Systems, Vol. 2, No. 2, pp.22–39.
Payan, A. and Noora, A.A. (2014) ‘A linear modelling to solve multi-objective linear fractional
programming problem with fuzzy parameters’, International Journal of Mathematical
Modelling and Numerical Optimisation, Vol. 5, No. 3, pp.210–228.
Precup, R.E. and Preitl, S. (2006) ‘Stability and sensitivity analysis of fuzzy control systems,
mechatronics applications’, Acta Polytechnica Hungarica, Vol. 3, No. 1, pp.61–76.
Razmi, J., Jafariana, E. and Amin, S.H. (2016) ‘An intuitionistic fuzzy goal programming approach
for finding pareto-optimal solutions to multi-objective programming problems’, Expert
Systems with Applications, Vol. 65, pp.181–193.
Ren, A., Wang, Y., and Xue, X. (2016) ‘Interactive programming approach for solving the fully
fuzzy bilevel linear programming problem’, Knowledge-Based Systems, Vol. 99, pp.103–111.
Saati, S., Hatami-Marbini, A., Tavana, M. and Hajiahkondi, E. (2012) ‘A two-fold linear
programming model with fuzzy data’, International Journal of Fuzzy System Applications,
Vol. 2, No. 3, pp.1–12.
Selvaraj, G. and Pandian, P. (2016) ‘A fuzzy neural network for fully fuzzy linear programming
problems’, International Journal of Mathematical Modelling and Numerical Optimisation,
Vol. 7, Nos. 3/4, pp.231–243.
Tanaka, H. and Asai, K. (1984) ‘Fuzzy solution in fuzzy linear programming problems’, IEEE
Transactions on Systems, Man and Cybernetics, Vol. 14, No. 2, pp.325–328.
Tanaka, H., Okuda, T. and Asai, K. (1974) ‘On fuzzy mathematical programming’, Journal of
Cybernetics, Vol. 3, No. 4, pp.37–46.
Verdegay, J.L. (1984) ‘A dual approach to solve the fuzzy linear programming problem’, Fuzzy
Sets and Systems, Vol. 14, No. 2, pp.131–141.
 A. Ebrahimnejad 91

Villacort, P.J., Rabelo, C.A., Pelta, D.A. and Verdegay, J.L. (2016) ‘Fuzzy LP: An R package for
solving fuzzy linear programming problems’, in Granular, Soft and Fuzzy Approaches for
Intelligent Systems, published in Studies in Fuzziness and Soft Computing, Vol. 344, Springer
International Publishing, Switzerland, pp.209–230.
Yager, R.R. (1981) ‘A procedure for ordering fuzzy subsets of the unit interval’, Information
Sciences, Vol. 24, No. 2, pp.143–161.
Zhao, X., Zheng, Y. and Wan, Z. (2016) ‘Interactive intuitionistic fuzzy methods for multilevel
programming problems’, Expert Systems with Applications, Vol. 72, pp.258–268. doi:
10.1016/j.eswa.2016.10.063.
Zimmerman, H.J. (1978) ‘Fuzzy programming and linear programming with several objective
functions’, Fuzzy Sets and Systems, Vol. 1, No. 1, pp.45–55.