European Journal of Operational Research 180 (2007) 29–37

www.elsevier.com/locate/ejor

Continuous Optimization

Binary fuzzy goal programming

Ching-Ter Chang *

Chungyu Institute of Technology, Department of Information Management, Keelung, Taiwan, ROC

Received 12 August 2005; accepted 29 March 2006

Available online 15 May 2006

Abstract

Goal programming is an important technique for solving many decision/management problems. Fuzzy goal program-
ming involves applying the fuzzy set theory to goal programming, thus allowing the model to take into account the vague
aspirations of a decision-maker. Using preference-based membership functions, we can define the fuzzy problem through
natural language terms or vague phenomena. In fact, decision-making involves the achievement of fuzzy goals, some of
them are met and some not because these goals are subject to the function of environment/resource constraints. Thus, bin-
ary fuzzy goal programming is employed where the problem cannot be solved by conventional goal programming
approaches. This paper proposes a new idea of how to program the binary fuzzy goal programming model. The binary
fuzzy goal programming model can then be solved using the integer programming method. Finally, an illustrative example
is included to demonstrate the correctness and usefulness of the proposed model.
Ó 2006 Elsevier B.V. All rights reserved.

Keywords: Goal programming; Fuzzy programming; Binary programming

1. Introduction

Goal programming (GP) was originally proposed by Charnes and Cooper [1], and further development car-
ried out by Lee [2], Ignizio [3], Tamiz [4], and Romero [5], among others [6–8]. It has been applied to many
real-world problems in areas such as accounting, agriculture, economics, engineering, transportation, finance,
government, international context, and marketing [10,14]. GP is an important technique for decision-makers
(DMs) to consider simultaneously several objectives in finding a set of acceptable solutions. It can be said that
GP has been, and still is, the most widely used technique for solving multi-criteria and multi-objective deci-
sion-making problems. Attempting to stimulate OR practitioners in the right use of GP, Tamiz et al. [4]
and Romero [11] have proposed the overviews of the state-of-the-art in GP. The DMs for their goals set some
acceptable aspiration levels, gk (k = 1, 2, . . . , n), for these goals, and try to achieve a set of goals as closely as
possible [4]. The purpose of GP is to minimize the deviations between the achievement of goals, fk(X), and
these acceptable aspiration levels, gk (k = 1, 2, . . . , n). A mathematical formulation of GP is given below.

*
Tel.: +886920884730.
E-mail address: [email protected]

0377-2217/$ - see front matter Ó 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.ejor.2006.03.030
30 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37

(GP)
X
n
Minimize jfk ðX Þ  gk j
k¼1

Subject to X 2 F ; ðF is a feasible setÞ;

where fk(X) is the linear function of the kth goal, and gk is the aspiration level of the kth goal.
The majority of GP applications in the literature have been implemented using various methods such as
lexicographic GP (preemptive GP), weighted GP (Archimedean GP), and MINMAX GP (Chebyshev GP)
[5]. In order to solve GP, we let the function fk ðX Þ ¼ d þ 
k  d k þ gk , then GP can be formulated as the follow-
ing achievement function.
X
n
Minimize ðd þ 
k þ dk Þ
k¼1

Subject to fk ðX Þ  gk ¼ d þ 
k  dk ; k ¼ 1; 2; . . . ; n;
X 2 F; ðF is a feasible setÞ;
dþ 
k ; d k P 0; k ¼ 1; 2; . . . ; n;

where d þ 
k ¼ maxð0; fk ðX Þ  g k Þ and d k ¼ maxð0; gk  fk ðX ÞÞ are, respectively, over- and under-achievement of
the kth goal; other variables are defined as in GP.
In a conventional GP formulation, goals are precisely defined. That is, the formulation assumes that the
DM is able to determine accurately goal values for their decision-making problems. In fact, many imprecise
aspiration levels may exist in decision-making problems such as ‘‘somewhat larger than’’, ‘‘substantially lesser
than’’, or ‘‘around’’ the vague goal gk due to DM’s ambiguous understanding of their nature. Thus, the DM
may find it is impossible to state precisely exact aspiration levels to the goals for their problems. In doing so, if
the imprecise aspiration level is introduced to each objective, then the problem is turned into fuzzy GP (FGP)
[12,13]. Since Narasinhan [13] applied the fuzzy set theory with preference-based membership function to GP,
many achievements have been reported in the literature such as preemptive FGP, weight additive model, sto-
chastic model, and real-life case studies [14–18]. The FGP can be expressed as follows:
(FGP)
A solution set X is obtained for FGP that let gk be the aspiration level of kth objective, fk(X)
(k = 1, 2, . . . , n). Then, the FGP can be expressed as the following form:
fk ðX Þ J gk ðor f k ðX Þ K gk Þ; k ¼ 1; 2; . . . ; n
Subject to X 2 F; ðF is a feasible setÞ;

where fk(X) J ([)gk indicates the kth fuzzy goal approximately greater than or equal to (approximately less
than or equal to) the aspiration level gk; other variables are defined as in GP.
With the MAXMIN approach of Zimmermann [12], the FGP can be expressed as follows:

Maximize k
Subject to k  lk ðfk ðX ÞÞ 6 0; k ¼ 1; 2; . . . ; n;
X 2 F; ðF is a feasible setÞ;

where k is an additional continuous variable; lk(fk(X)) is a membership function of the kth objective.
The FGP has the advantage of allowing for the vague aspirations of a DM, which can be qualified using
some natural language or vague phenomena. For representing the preference concept from DMs, lk(fk(X)) is
then characterized as follows:
 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37 31
8
>
> 1; if f k ðX Þ P gk ;
>
>
< ðf ðX Þ  l Þ
k k
lk ðfk ðX ÞÞ ¼ ; if lk 6 fk ðX Þ 6 gk for f x ðX Þ J gk ;
> g k  lk
>
>
>
:
0; if f k ðX Þ 6 lk :
8
>
> 1; if f k ðX Þ 6 gk ;
>
>
< ðu  f ðX ÞÞ
k k
lk ðfk ðX ÞÞ ¼ ; if gk 6 fk ðX Þ 6 uk for f k ðX Þ K gk ;
>
> uk  g k
>
>
:
0; if f k ðX Þ P uk :

where lk and uk are, respectively, lower and upper limits for the kth goal; fk(X) and gk are defined as in
GP.
The FGP problem has also been addressed using various methods such as probability distribution, penalty
function, fuzzy numbers, preemptive fuzzy goal programming, interpolated membership function, and the
weighted additive model [14,19]. Zimmermann [12] first proposed fuzzy programming for solving the multi-
objective linear programming problems. Narasimhan [13] presented the initial FGP model and solution
procedure. Hannan [20] introduced interpolated membership functions (i.e., piecewise linear membership
functions) into the FGP model, then the FGP model could be solved using the linear programming method.
However, the model of Hannan [20] is applicable only to FGP problems with concave membership functions.
To further improve the model of Hannan [20], Nakamura [21] proposed an extended FGP for solving quas-
siconcave membership functions. Later, from the model of Hannan [20], Yang et al. [22] further formulated
the FGP problem using fewer additional variables. The concept of different importance and priority use in
FGP was considered by Chen and Tsai [19]. However, in some cases the author believes that there may exist
a situation where a DM would like to make a decision concerning the FGP problem, that involves the
achievement of fuzzy goals, some of which are met and some not because these fuzzy goals are subject to
the function of environment/resource constraints, Rk(X) (k = 1, 2, . . . , n). That is, the fuzzy goals in the
decision-making problem could be either level achieved or not achieved completely. As an example, assume
a local company provides a FGP model to help a homebuyer get into a position to buy a home quickly. After
the model process, the homebuyer found a house, which has the desired 80% satisfaction level.
Unfortunately, it is near a railway and has noise pollution from the passing trains. This problem means that
the fuzzy goal is not completely achieved. This is a typical binary FGP (BFGP) problem. To the best knowl-
edge of the author, no work has been done on such problem. A mathematical formulation of the BFGP is
given below:
Rk(X): It is the function of environment/resource constraints for the kth goal (k = 1, 2, . . . , n).
bk: It is the binary decision variable for the kth goal (k = 1, 2, . . . , n); its behavior (i.e., 0 or 1) is bounded by
Rk(X);
(BFGP)
fk ðX Þbk J gk bk ðor f k ðX Þbk K gk bk Þ; k ¼ 1; 2; . . . ; n
Subject to X 2 F ðF is a feasible setÞ;
bk 2 Rk ðX Þ; k ¼ 1; 2; . . . ; n;
where all variables are defined as in FGP.
We realize that if bk = 1 then it will force BFGP to be FGP, otherwise, zero is achieved. It is interesting to
note that the BFGP has two characteristics: (i) all objectives of BFGP might not be fully achieved simulta-
neously, and (ii) some of the objectives of BFGP are achieved and others are not achieved completely due
to environment/resource limitations. According to the above two characteristics, however, the BFGP cannot
be solved by current GP approaches. In order to overcome the drawback, we introduce a lower (or an upper)
tolerance limit, lk (or uk), and a binary membership function (BMF), lbk ðfk ðX ÞÞðk ¼ 1; 2; . . . ; nÞ for the kth goal
to determine the achieved level of that goal as follows:
32 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37

8
> 1; if f k ðX Þ P gk and bk ¼ 1;
>
< ðf ðX Þ  l Þb
k k k
lbk ðfk ðX ÞÞ ¼ ; if lk 6 fk ðX Þ 6 gk and bk ¼ 1 for f k ðX Þ J gk ; ð1Þ
>
> g k  lk
:
0; if f k ðX Þ 6 lk or bk ¼ 0:
8
> 1; if f k ðX Þ 6 gk and bk ¼ 1;
>
< ðu  f ðX ÞÞb
k k k
lbk ðfk ðX ÞÞ ¼ ; if gk 6 fk ðX Þ 6 uk and bk ¼ 1 for f k ðX Þ K gk ; ð2Þ
>
> uk  g k
:
0; if f k ðX Þ P uk or bk ¼ 0:
where lk and uk are, respectively, lower and upper limits for the kth goal; bk is a binary variable; fk(X) and gk
are defined as in FGP.
The lbk ðfk ðX ÞÞ is a standard version of membership function if the binary variable bk = 1. In real cases, the
behavior of bk (i.e., 0 or 1) is subject to the function of environment/resource constraints, Rk(X), such as stor-
age space available, total budgets, manpower and so on. The standard version of membership function has
been used in many approaches such as max–min, two-phase, weight max–min, and multi-step approaches
[12,14]. However, the BMF cannot be solved by current GP approaches. The contributions of this paper lie
in two areas: (i) linearization strategies are derived to convert BFGP into a standard version of FGP, which
can be easily solved by common integer programming packages, and (ii) a new idea for BFGP with the BMF is
proposed to formulate decision problems involving the achievement of fuzzy goals, some of which are met and
some not.
The rest of the paper is organized as follows. In Section 2, linearization strategies are derived for linearizing
the BFGP. Then, the linearized BFGP is solved using the integer programming approach. To demonstrate the
correctness and usefulness of the proposed model, an illustrative example is included in Section 3. Finally, con-
clusions are presented in Section 4.

2. Linearization strategies

BFGP is a typical fractional mixed binary problem, which cannot be solved by the conventional GP
approach [23]. Hence, we use linearization strategies to transform the BFGP into the standard version of
FGP. The standard version of FGP is then solved using common integer programming packages. The linear-
ization procedures are as follows.
In fuzzy programming approaches, the highest possible value of BMF is 1. By referring to [9], as the defined
BMF in (1) and (2), the flexible membership function goals with the aspired level 1 can be presented as
ðfk ðX Þ  lk Þbk
þ d þ
k  d k ¼ 1; for f k ðX Þ J gk ; ð3Þ
g k  lk
ðuk  fk ðX ÞÞbk
þ d þ
k  d k ¼ 1; for f k ðX Þ K gk : ð4Þ
uk  g k
The kth membership goal in (3) can be expressed as follows:
1
Lk fk ðX Þbk  Lk lk bk þ d  þ
k  d k ¼ 1; where Lk ¼ : ð5Þ
g k  lk
(5) can be represented as: Lk fk ðX Þbk  Lk bk þ d  þ 
k  d k ¼ 1, where Lk ¼ Lk lk .
The kth membership goal in (4) can also be expressed as follows:
1
I k uk bk  I k fk ðX Þbk þ d  þ
k  d k ¼ 1; where I k ¼ : ð6Þ
uk  g k
(6) can be represented as: I k bk  I k fk ðX Þbk þ d  þ 
k  d k ¼ 1, where I k ¼ I k uk .
In order to obtain the global optimal solutions, the quadratic terms fk(X)bk in (5) and (6) should be line-
arized [23]. Let Wk = fk(X)bk, where Wk is an additional continuous variable. (5) can be represented as follows:
Lk W k  Lk bk þ d  þ 
k  d k ¼ 1, where Lk ¼ Lk lk . The quadratic mixed binary term fk(X)bk can be linearized using
the following program.
 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37 33

Program P1
Minimize W k
Subject to 0 6 W k 6 bk M ð7Þ
ðbk  1ÞM þ fk ðX Þ 6 W k 6 ð1  bk ÞM þ fk ðX Þ; ð8Þ
where M is a big positive value; bk is a binary variable.
Proposition 1. Quadratic mixed binary term fk(X)bk and P1 are equivalent in the sense that they have the same
optimal solutions.

Proof

(i) If bk = 0 then 0 6 Wk 6 0 (from (7)), M + fk(X) 6 Wk 6 M + fk(X) (from (8)). This forces Wk = 0 in
the objective function of P1.
(ii) If bk = 1 then 0 6 Wk 6 M (from (7)), fk(X) 6 Wk 6 fk(X) (from (8)). This forces Wk = fk(X) in the
objective function of P1.

Wk is bounded by (7) and (8), thus reduce to


0; if bk ¼ 0;
Wk ¼
fk ðX Þ; if bk ¼ 1:
It is obvious that P1 is equivalent to fk(X)bk. This completes the proof of Proposition 1. h
According to the above linearization process, the BFGP problem can be formulated as the following pro-
posed model (i.e., a standard achievement function of GP).
(Proposed model)
Minimize d  k ; k ¼ 1; 2; . . . ; n
Subject to Lk fk ðX Þbk  Lk bk þ d  þ
k  d k ¼ 1; k ¼ 1; 2; . . . ; n for f k ðX Þ J gk ; ð9Þ
  þ
I k bk  I k fk ðX Þbk þ d k  d k ¼ 1; k ¼ 1; 2; . . . ; n for f k ðX Þ K gk ; ð10Þ
X 2 F ; ðF is a feasible setÞ;
bk 2 Rk ðX Þ; k ¼ 1; 2; . . . ; n;
1
where Lk ¼ gk l k
, Lk ¼ Lk lk , I k ¼ uk g
1
k
and I k ¼ I k uk ; The quadratic mixed binary term fk(X)bk can be linearized
using P1; other variables are defined as in BFGP.
In GP, the deviational variables, d  þ
k and d k are included in the objective function for minimizing them. In

the proposed model, only d k is required to be minimized to achieve the highest aspired level 1 of the fuzzy
goal. Notably, any over-deviation from a fuzzy goal indicates full achievement of the membership value
[25]. For instance, let us consider (9) for fk(X) J gk: if bk = 0 then d  þ
k ¼ 1 and d k ¼ 0 (form (9)). This forces

under-deviational variable d k ¼ 1 in the objective function, which indicates the zero achievement of the mem-
bership value. (i.e., lbk ðfk ðX ÞÞ ¼ 0Þ. The same idea can be applied to (10) for fk(X) [ gk. From the proposed
model, we conclude that when solving a decision problem with n objectives, some of them will be achieved and
some will not. To determine the achieved levels of these objectives, n BMFs should be introduced. In order to
linearize these BMFs (i.e., quadratic mixed binary fractional programming problem), 2n auxiliary constraints
and n additional continuous variables will be introduced (see Chang [23] for more details).

3. An illustrative example

The education system in Taiwan is facing a major problem, increasing number of teachers are becoming
unemployed due to the falling birth rate. Colleges and universities that mushroomed in the higher education
market over the past decade may soon face closure. In this context, a university of science and technology in
Taiwan plans to promote their programs using E-learning approaches via Internet to increase the graduation
34 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37

rate of students enrolled. In order to achieve this objective, a new E-learning system should be established in
the university. The E-learning system contains E-learning web servers (ELWS), learning contents, graphic user
interface, teaching tools, audio/video tools and distributed databases. ELWS should be allocated at different
cities in Taiwan for direct services. However, the university cannot build ELWS everywhere because of the
constraint of capital availability. Under this constraint, the university is establishing an investment plan to
build at least four ELWS, which is the goal in the E-learning system. Although a maximum of six ELWS
can be established but it is not required. To achieve reliability of the E-learning system, ELWS should be
established in three major cities, namely Taipei, Taichung, and Kaohsiung in Taiwan to connect each other
as a basic three-hub-nodes ring trunking network. The topology of the ring trunking network is shown in
Fig. 1. Other nodes are to be precisely connected to one of the three-hub-nodes. Problem parameters are given
as: six ELWS locations, capital requirements, and the number of students in each city are listed in Table 1.
There are two goals of the E-learning system as follows.
(g1) The system must serve at least 9000 students in six cities as shown in Table 1. That is, some cities have
ELWS and others will have no ELWS for student services.
(g2) The university must implement at least four ELWS for the E-learning system to achieve the greater
satisfaction of students.
In view of resource limitations of the university, some constraints are added to the E-learning system as
follows:

(1) Total manpower must not exceed 10 persons for the routine operations of the E-learning system.
(2) The total available investment budget must not exceed one million.
(3) A basic ring trunking network should be implemented.

Based on the proposed model, the above-mentioned problem can be formulated as the following FGP
program.

Fig. 1. Network topology of E-learning system.

Table 1
ELWS parameters
Decision Decision variable Capital requirements The number of students The number of manpower
ELWS in Taipei city x1 0.1 million 3 thousand 3
ELWS in Hsinchu city x2 0.2 million 1.5 thousand 2
ELWS in Taichung city x3 0.2 million 2 thousand 3
ELWS in Kaohsiung city x4 0.2 million 2.5 thousand 3
ELWS in Tainan city x5 0.2 million 1 thousand 2
ELWS in Hualien city x6 0.3 million 0.5 thousand 1
 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37 35

Maximize ðg1 Þ 3x1 þ 1:5x2 þ 2x3 þ 2:5x4 þ x5 þ 0:5x6 J 9;

ðg2 Þ x1 þ x2 þ x3 þ x4 þ x5 þ x6 J 4;
constraints :
3x1 þ 2x2 þ 3x3 þ 3x4 þ 2x5 þ x6 6 10 ðManpower constraintÞ; ð11Þ
0:1x1 þ 0:2x2 þ 0:2x3 þ 0:2x4 þ 0:2x5 þ 0:3x6 6 1 ðCapital constraintÞ; ð12Þ
x1 þ x3 þ x4 ¼ 3 ðBasic ring trunking network constraintÞ; ð13Þ
where xi (i = 1, 2, . . . , 6) are binary variables.
Assume that the tolerance limits of the above two fuzzy objective goals are (1, 1), respectively. The mem-
bership functions of the problem are obtained as follows:
3x1 þ 1:5x2 þ 2x3 þ 2:5x4 þ x5 þ 0:5x6  1
l1 ðf1 ðX ÞÞ ¼ ;
8
x1 þ x2 þ x3 þ x4 þ x5 þ x6  1
l2 ðf2 ðX ÞÞ ¼ :
3
Then, this problem can be formulated as the following program:

Minimize d  1 þ d2


Subject to
(i) The following constraints for g1.
1
8
ð3x1 þ 1:5x2 þ 2x3 þ 2:5x4 þ x5 þ 0:5x6 Þ  18 þ d  þ
1  d 1 ¼ 1, use to replace
 þ 1
L1 f1 ðX Þ  L1 l1 þ d 1  d 1 ¼ 1, where L1 ¼ g1 l 1
:
(ii) The following constraints for g2.
1
3
ðx1 þ x2 þ x3 þ x4 þ x5 þ x6 Þ  13 þ d  þ
2  d 2 ¼ 1, use to replace
 þ 1
L2 f2 ðX Þ  L2 l2 þ d 2  d 2 ¼ 1, where L2 ¼ g2 l 2
:
(11)–(13)

This problem is solved using LINDO [24] to obtain the optimal solutions as (x1, x2, x3, x4, x5,
x6)=(1, 0, 1, 1, 0, 1). We realize that g1 = 8 has a negative value (0.125) under the aspiration level and
g2 = 4 has reached the aspiration level exactly (i.e., l1(f1(X)) = 0.875 and l2(f2(X)) = 1).
In order to demonstrate the binary behavior of the model, the following scenario is then added to g2: Taipei
city is the center of government and culture in Taiwan. Hsinchu Science Park in Hsinchu city is the cradle of
the semiconductor and information industries in Taiwan, even in the world. The university needs to promote
their programs for students in both cities to increase their prestige. According to this issue, g2 has not been
completely achieved, if Taipei or Hsinchu city has no ELWS. To represent this scenario, g2 should be bound
by an extra binary variable b2 and the behavior of b2 is decided by whether Taipei and Hsinchu cities are
included in the E-learning system. This problem can be expressed as the following BFGP program.
Maximize ðg1 Þ 3x1 þ 1:5x2 þ 2x3 þ 2:5x4 þ x5 þ 0:5x6 J 9;
ðg2 Þ ðx1 þ x2 þ x3 þ x4 þ x5 þ x6 Þb2 J 4b2 ;
constraints : 3x1 þ 2x2 þ 3x3 þ 3x4 þ 2x5 þ x6 6 10 ðManpower constraintÞ; ð14Þ
0:1x1 þ 0:2x2 þ 0:2x3 þ 0:2x4 þ 0:2x5 þ 0:3x6 6 1 ðCapital constraintÞ; ð15Þ
x1 þ x3 þ x4 ¼ 3 ðBasic ring trunking network constraintÞ; ð16Þ
b2 6 x 1 x 2 ðTaipei or Hsinchu city has no ELWSÞ; ð17Þ
where xi (i = 1, 2, . . . , 6) are binary variables.
The above problem can be easily formulated according to the proposed model. The quadratic mixed binary
terms xib2 in (g2) and x1x2 in (17) can be linearized using Proposition 1. For example, we let Wk = xib2 where
Wk satisfies the following inequalities: (i) 0 6 Wk 6 b2M; (ii) (b2  1)M + xi 6 Wk 6 (1  b2)M + xi. We solve
36 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37

this problem to obtain the optimal solution as (x1, x2, x3, x4, x5, x6, b2)=(1, 0, 1, 1, 0, 1, 0). We realize that the first
goal g1 = 8 has a negative value (0.125) under the aspiration level (i.e., lb1 ðf1 ðX ÞÞ ¼ 0:875Þ. The second goal
g2 = 0 is achieved to a level of zero (i.e., lb2 ðf2 ðX ÞÞ ¼ 0) because it is bound by the binary variable b2 6 x1x2 in
(17) where x1 = 1 and x2 = 0. That is, Hsinchu city has no ELWS, which forces g2 = 0 to be not completely
achieved. In order to add Hsinchu city to the E-learning system, the university increases the number of per-
sonnel from 10 to 11 persons. Under this condition, we resolve the problem to obtain the optimal solutions as
(x1, x2, x3, x4, x5, x6, b2)=(1, 1, 1, 1, 0, 0, 1). From the solution we can see that both goals g1 = 9 and g2 = 4 have
precisely reached the aspiration levels (i.e., lb1 ðf1 ðX ÞÞ ¼ 1 and lb2 ðf2 ðX ÞÞ ¼ 1Þ.

4. Conclusions

Some BFGP problems still exist in the real world. However, they cannot be solved by conventional GP
approaches. In order to solve these problems, an extension approach of GP is developed in this paper. Lin-
earization strategies are derived to convert BFGP into a standard version of FGP, which can then be solved
by common integer programming packages. In addition, the proposed models can be easily applied to deci-
sion/management problems, which involve the achievement of fuzzy goals, some of which are met and some
not. As to the results, the author suggests the need for further research on the mixed binary non-linear GP for
solving more complex decision-making problems.

Acknowledgements

The author is grateful to the anonymous referees and professor Boaz Golany for their constructive sugges-
tions, which led to an improvement in this paper.

References

[1] A. Charnes, W.W. Cooper, Management Model and Industrial Application of Linear Programming, vol. 1, Wiley, New York, 1961.
[2] S.M. Lee, Goal Programming for Decision Analysis, Auerbach, Philadelphia, PA, 1972.
[3] J.P. Ignizio, Introduction to Linear Goal Programming, Sage, Beverly, Hills, CA, 1985.
[4] M. Tamiz, D. Jones, C. Romero, Goal programming for decision making: An overview of the current state-of-the-art, European
Journal of Operational Research 111 (1998) 567–581.
[5] C. Romero, Extended lexicographic goal programming: A unifying approach, Omega 29 (2001) 63–71.
[6] R.B. Flavell, A new goal programming, Omega 4 (1976) 731–732.
[7] B. Vitoriano, C. Romero, Extended interval goal programming, Journal of the Operational Research Society 50 (1999) 1280–1283.
[8] C.-T. Chang, A modified goal programming model for piecewise linear functions, European Journal of Operational Research 139
(2002) 62–67.
[9] R.H. Mohamed, The relationship between goal programming and fuzzy programming, Fuzzy Sets and Systems 89 (1997) 215–
222.
[10] M.J. Schniederjans, Goal Programming: Methodology and Applications, Kluwer, Boston, MA, 1995.
[11] C. Romero, A general structure of achievement function for a goal programming model, European Journal of Operational Research
153 (2004) 675–686.
[12] H.J. Zimmermann, Fuzzy programming and linear programming with several objective functions, Fuzzy Sets and Systems 1 (1978)
45–55.
[13] R. Narasimhan, Goal programming in a fuzzy environment, Decision Sciences 11 (1980) 325–338.
[14] Y.J. Lai, C.L. Hwang, Fuzzy Multiple Objective Decision Making-Methods and Applications, Springer-Verlag, 1994.
[15] G.A. Chdam, Fuzzy programming (FGP) approach to a stochastic transportation problem under budgetary cobstraint, Fuzzy Sets
and Systems 66 (1994) 293–299.
[16] A.K. Dhingra, A.A. Rao, H. Miura, Multiobjective decision making in a fuzzy environment with applications to helicopter design,
AIAA Journal 28 (1990) 703–710.
[17] H. Min, A model-based decision support system for locating banks, Information and Management 17 (1989) 207–215.
[18] S.B. Sinha, K.A. Rao, B.K. Mangaraj, Fuzzy goal programming in multi-criteria decision systems: A case study in agricultural
planning, Socio-Economic Planning Science 22 (1988) 93–101.
[19] L.H. Chen, F.C. Tsai, Fuzzy goal programming with different importance and priorities, European Journal of Operational Research
133 (2001) 548–556.
[20] E.L. Hannan, On fuzzy goal programming, Decision Sciences 12 (1981) 539–546.
[21] K. Nakamura, Some extension of fuzzy linear programming, Fuzzy Sets and Systems 14 (1984) 211–229.
 C.-T. Chang / European Journal of Operational Research 180 (2007) 29–37 37

[22] Y. Yang, J.P. Ignizio, H.J. Kim, Fuzzy programming with nonlinear membership functions: Piecewise linear approximation, Fuzzy
Sets and Systems 11 (1991) 39–53.
[23] C.-T. Chang, An efficient linearization approach for mixed integer problems, European Journal of Operational Research 123 (2000)
652–659.
[24] L. Schrage, LINGO Release 6.0, LINDO System, Inc., 1999.
[25] R.G. Dyson, Maximin programming, fuzzy linear programming and multicriteria decision making, Journal of the Operational
Research Society 31 (1980) 263–267.