interactive random fuzzy two-level programming through possibility-based probability model
TRANSCRIPT
Information Sciences 239 (2013) 191–200
Contents lists available at SciVerse ScienceDirect
Information Sciences
journal homepage: www.elsevier .com/locate / ins
Interactive random fuzzy two-level programming throughpossibility-based probability model
0020-0255/$ - see front matter � 2013 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/j.ins.2013.03.024
⇑ Corresponding author. Tel.: +81 82 424 7694; fax: +81 82 422 7195.E-mail address: [email protected] (M. Sakawa).
Masatoshi Sakawa ⇑, Takeshi MatsuiFaculty of Engineering, Hiroshima University, Higashi-Hiroshima 739-8527, Japan
a r t i c l e i n f o
Article history:Received 22 September 2011Received in revised form 7 March 2013Accepted 14 March 2013Available online 21 March 2013
Keywords:Two-level programmingRandom fuzzy programmingPossibilityProbability maximizationInteractive programming
a b s t r a c t
This paper considers interactive decision making methods for random fuzzy two-level lin-ear programming problems. Assuming that the decision makers concern about the proba-bilities that their own objective function values are smaller than or equal to certain targetvalues, fuzzy goals of the decision makers for the probabilities are introduced. Then, thepossibility-based probability model to maximize the degrees of possibility with respectto the attained probability is considered. Interactive fuzzy nonlinear programming toobtain a satisfactory solution for the decision maker at the upper level in considerationof the cooperative relation between decision makers is presented. An illustrative numericalexample demonstrates the feasibility and efficiency of the proposed method.
� 2013 Elsevier Inc. All rights reserved.
1. Introduction
Decision making problems in hierarchical managerial or public organizations are often formulated as two-level mathe-matical programming problems [21,33]. In the context of two-level programming, the decision maker at the upper level firstspecifies a strategy, and then the decision maker at the lower level specifies a strategy so as to optimize the objective withfull knowledge of the action of the decision maker at the upper level. In conventional multi-level mathematical program-ming models employing the solution concept of Stackelberg equilibrium, it is assumed that there is no communicationamong decision makers, or they do not make any binding agreement even if there exists such communication[2,14,33,34]. Compared with this, for decision making problems in such as decentralized large firms with divisional indepen-dence, it is quite natural to suppose that there exists communication and some cooperative relationship among the decisionmakers [21].
For two-level linear programming problems or multi-level ones such that decisions of decision makers in all levels aresequential and all of the decision makers essentially cooperate with each other, Lai [7] and Shih et al. [32] proposed fuzzyinteractive approaches. In their methods, the decision makers identify membership functions of the fuzzy goals for theirobjective functions, and in particular, the decision maker at the upper level also specifies those of the fuzzy goals for thedecision variables. The decision maker at the lower level solves a fuzzy programming problem with a constraint with respectto a satisfactory degree of the decision maker at the upper level. Unfortunately, there is a possibility that their method leadsa final solution to an undesirable one because of inconsistency between the fuzzy goals of the objective function and those ofthe decision variables. In order to overcome the problem in their methods, by eliminating the fuzzy goals for the decisionvariables, Sakawa et al. have proposed interactive fuzzy programming for two-level or multi-level linear programming prob-lems to obtain a satisfactory solution for decision makers [24,25]. Extensions to two-level linear fractional programming
192 M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200
problems [27], decentralized two-level linear programming problems [19,28], two-level linear fractional programmingproblems with fuzzy parameters [26], and two-level nonconvex programming problems with fuzzy parameters [20] wereprovided. The subsequent works on two-level or multi-level programming have been appearing [1,8,15,17,35] includingthe recent extension to fuzzy random two-level programming problems [18,23,29–31]. A recent survey paper of Sakawaand Nishizaki [22] is devoted to reviewing and classifying the numerous major papers in the area of so-called cooperativemulti-level programming.
It should be noted here that fuzzy random variables [6,16,37] are considered to be random variables whose realized val-ues are not real values but fuzzy numbers or fuzzy sets. Studies on linear programming problems with fuzzy random variablecoefficients, called fuzzy random linear programming problems, were initiated by Wang and Qiao [37] and mathematicalprogramming problems with fuzzy random variables have been developed [5,9,23,38].
On the other hand, from a viewpoint of ambiguity and randomness different from fuzzy random variables [6,16,37], byconsidering the experts’ ambiguous understanding of means and variances of random variables, a concept of random fuzzyvariables was proposed [10]. Mathematical programming problems with random fuzzy variables were formulated and havealso been developed [9,11,23,38–40].
Under these circumstances, in this paper, as a first attempt to tackle decision making problems in hierarchical organiza-tions under random fuzzy environments, assuming cooperative behavior of the decision makers, we first consider two-levellinear programming problems involving random fuzzy variables. To deal with the formulated random fuzzy two-level linearprogramming problems, we assume that the decision makers concern about the probabilities that their own objective func-tion values are smaller than or equal to certain target values. By considering the imprecise nature of the human judgments,we introduce the fuzzy goals of the decision makers for the probabilities. Then, assuming that the decision makers are willingto maximize the degrees of possibility with respect to the attained probability, we consider the possibility-based probabilitymodel for random fuzzy two-level programming problems. Interactive fuzzy nonlinear programming to obtain a satisfactorysolution for the decision maker at the upper level in consideration of the cooperative relation between decision makers ispresented. It is shown that all of the problems to be solved in the proposed interactive fuzzy nonlinear programming becomenonlinear programming problems and approximate optimal solutions can be obtained through the use of particle swarmoptimization for nonlinear programming (PSONLP) [12]. An illustrative numerical example is provided to demonstrate thefeasibility and efficiency of the proposed method.
2. Random fuzzy variables
In the framework of stochastic programming [4,36], it is implicitly assumed that the uncertain parameter which well rep-resents the stochastic factor of real systems can be definitely expressed as a single random variable. This means that the real-ized values of random parameters under the occurrence of some event are assumed to be definitely represented with realvalues. Depending on the situations, however, it is natural to consider that the possible realized values of these randomparameters are often only ambiguously known to the experts. In this case, it may be more appropriate to interpret the ex-perts’ ambiguous understanding of the realized values of random parameters as fuzzy numbers. From such a point of view, afuzzy random variable was first introduced by [6], and its mathematical basis was constructed by [16]. An overview of thedevelopments of fuzzy random variables was found in the recent article of [3].
From the expert’s experimental point of view, however, the experts may think of a collection of random variables to beappropriate to express stochastic factors rather than only a single random variables. In this case, reflecting the expert’s con-viction degree that each of random variables properly represents the stochastic factor, it would be quite reasonable to assignthe different degrees of possibility to each of random variables. For handling such an uncertain parameter, a random fuzzyvariable was defined by Liu [10] as a function from a possibility space to a collection of random variables, which is consideredto be an extended concept of fuzzy variable [13]. It should be noted here that the fuzzy variables can be viewed as anotherway of dealing with the imprecision which was originally represented by fuzzy sets. Although we can employ Liu’s defini-tion, for consistently discussing various concepts in relation to the fuzzy sets, we define the random fuzzy variables byextending not the fuzzy variables but the fuzzy sets.
Definition 1 (Random fuzzy variable). Let C be a collection of random variables. Then, a random fuzzy variable eC is definedby its membership function
leC : C! ½0;1�: ð1Þ
In Definition 1, the membership function leC assigns each random variable �c 2 C to a real number leC ð�cÞ. It should be noted
here that if C is defined as R, then (1) becomes equivalent to the membership function of an ordinary fuzzy set. In this sense,a random fuzzy variable can be regarded as an extended concept of fuzzy sets. On the other hand, if C is defined as a sin-
gleton C ¼ f�cg and leC ð�cÞ ¼ 1, then the corresponding random fuzzy variable eC can be viewed as an ordinary random
variable.
When taking account of the imprecise nature of the realized values of random variables, it would be appropriate to em-ploy the concept of fuzzy random variables. However, it should be emphasized here that if mean and/or variance of random
M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200 193
variables are specified by the expert as a set of real values or fuzzy sets, such uncertain parameters can be represented by notfuzzy random variables but random fuzzy variables.
As a simple example of random fuzzy variables, we consider a Gaussian random variable whose mean value is not def-initely specified as a constant. For example, when some random parameter �c is represented by the Gaussian random variableN(si, 102), where the expert identifies a set {s1, s2, s3} of possible mean values as (s1, s2, s3) = (90,100,110), if the membershipfunction leC is defined by8
leC ð�cÞ ¼0:5 if �c � Nð90;102Þ0:7 if �c � Nð100;102Þ0:3 if �c � Nð110;102Þ0 otherwise;
>>>><>>>>:
then eC is a random fuzzy variable. More generally, when the mean values are expressed as fuzzy sets or fuzzy numbers, thecorresponding random variable with the fuzzy mean is represented by a random fuzzy variable.3. Random fuzzy two-level programming
Consider the random fuzzy two-level linear programming problems formulated as
minimizefor DM1
z1ðx1; x2Þ ¼ eC 11x1 þ eC 12x2
minimizefor DM2
z2ðx1; x2Þ ¼ eC 21x1 þ eC 22x2
subject to A1x1 þ A2x2 6 bx1 P 0; x2 P 0;
9>>>>>>=>>>>>>;ð2Þ
where the two objective functions z1(x1, x2) and z2(x1, x2) are those of DM1 and DM2, respectively, and ‘‘minimizefor DM1
’’ and
‘‘minimizefor DM2
’’ mean that DM1 and DM2 are minimizers for their objective functions. Moreover, x1 is an n1 dimensional decision
variable column vector for the decision maker at the upper level (DM1), x2 is an n2 dimensional decision variable columnvector for the decision maker at the lower level (DM2), Aj, j = 1, 2 are m � nj coefficient matrices, and b is an m dimensionalcolumn vector.
Observing that the real data with uncertainty are often distributed normally, from the practical point of view, we assumethat each of eCljk; k ¼ 1;2; . . . ;nj of eC lj, l = 1, 2, j = 1, 2 is the Gaussian random variable with fuzzy mean value eMljk which isrepresented by an L–R fuzzy number characterized by the membership function
leMljkðsÞ ¼
L mljk�saljk
� �if mljk P s;
R s�mljk
bljk
� �if mljk < s;
8><>: ð3Þ
where the shape functions L and R are nonincreasing continuous functions from [0,1) to [0,1], mljk is the mean value, and aljk
and bljk are positive numbers which represent left and right spreads. Fig. 1 illustrates an example of the membership functionleMljk
ðsÞ.
Let C be a collection of all possible Gaussian random variables N(s,r2), where s 2 (�1,1) and r2 2 (0,1). Then, eCljk isexpressed as a random fuzzy variable with the membership function
Fig. 1. An example of the membership function leMljkðsÞ.
194 M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200
leC ljk
ð�cljkÞ ¼ leMljkðsljkÞ j �cljk � N sljk;r2
ljk
� �� �; 8�cljk 2 C: ð4Þ
Through the Zadeh’s extension principle, in view of (4), the membership function of a random fuzzy variable correspondingto each of objective functions zl(x1,x2), l = 1, 2 is given as
leC lxð�ulÞ ¼ sup
�cl
min16k6nj ; j¼1;2
leC ljk
ð�cljkÞ �ul ¼X2
j¼1
Xnj
k¼1
�cljkxjk
�����( )
¼ supsl
min16k6nj ; j¼1;2
leMljkðsljkÞ �ul � N
X2
j¼1
Xnj
k¼1
sljkxjk;VlðxÞ �����
!( ); ð5Þ
where �cl ¼ ð�cl11; . . . ; �cl1n1 ; �cl21; . . . ; �cl2n2 Þ, sl ¼ ðsl11; . . . ; sl1n1 ; s121; . . . ; sl2n2 Þ, and
VlðxÞ ¼X2
j¼1
Xnj
k¼1
r2ljkx2
jk:
Assuming that the decision makers (DMs) concern about the probabilities that their own objective function values eC lx aresmaller than or equal to certain target values fl, l = 1, 2, we introduce the probabilities Pðx j eC lðxÞx 6 flÞwhich are expressedas fuzzy sets ePl with the membership functions
lePlðplÞ ¼ sup
�ul
leCeC lxð�ulÞ j pl ¼ Pðx j ulðxÞ 6 flÞ
� �; ð6Þ
where fl, l = 1, 2 are target values specified by the DMs as constants.Considering the imprecise nature of the DMs’ judgments for the probabilities ePl with respect to the random fuzzy objec-
tive function values eC lx, l = 1, 2, we introduce the fuzzy goals eGl; l ¼ 1;2 such as ‘‘ePl should be greater than or equal to acertain value’’. Such fuzzy goals eGl; l ¼ 1;2 can be quantified by eliciting corresponding membership functions
leGlðpÞ ¼
0 if p 6 p0l
glðpÞ if p0l 6 p 6 p1
l ; l ¼ 1;21 if p1
l 6 p;
8><>: ð7Þ
where gl(p), l = 1, 2 are nondecreasing functions. Fig. 2 illustrates a possible shape of the membership function for the fuzzygoal eGl.
Recalling that the membership function is regarded as a possibility distribution, the degree of possibility that the prob-ability ePl attains the fuzzy goal eGl is expressed as
PePlðeGlÞ ¼ sup
pl
minflePlðplÞ; leGl
ðplÞg; l ¼ 1;2: ð8Þ
Fig. 3 illustrates the degree of possibility PePlðeGlÞ.
Now, assuming that the DMs are willing to maximize the degrees of possibility with respect to the attained probability,we consider the possibility-based probability model for random fuzzy two-level programming problems formulated as
maximizefor DM1
ZP;P1 ðx1; x2Þ ¼ PeP1
ðG1Þ
maximizefor DM2
ZP;P2 ðx1; x2Þ ¼ PeP2
ðG2Þ
subject to A1x1 þ A2x2 6 bx1 P 0; x2 P 0
9>>>>>=>>>>>;ð9Þ
Fig. 2. An example of a membership function leGlðyÞ of a fuzzy goal eGl .
Fig. 3. The degree of possibility PeP lðeGlÞ.
M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200 195
or equivalently
maximizefor DM1
h1
maximizefor DM2
h2
subject to PeP1ðfG1ÞP h1
PeP2ðfG2ÞP h2
A1x1 þ A2x2 6 bx1 P 0; x2 P 0:
9>>>>>>>>>>>>=>>>>>>>>>>>>;ð10Þ
From (8), the constraints PePlð eGlÞP hl; l ¼ 1;2 in (10) is equivalently replaced by the condition that there exists a p such that
lePlðplÞP hl and leGl
ðplÞP hl, namely
sup�sl
min16k6nj ;j¼1;2
leMljkðsljkÞ pl ¼ Pðxj�ulðxÞ 6 flÞ; �ul � N
X2
j¼1
Xnj
k¼1
sljkxjk;VlðxÞ !�����
( )P hl; ð11Þ
and pl P lHeGl
ðhlÞ; l ¼ 1;2, where lHeGl
ðhlÞ are pseudo inverse functions defined as lHeGl
ðhlÞ ¼ inffpljleGlðplÞP hlg; l ¼ 1;2. This
implies that there exists a vector ðpl; sl; �ulÞ; l ¼ 1;2 such that
min16k6nj ;j¼1;2
leMljkðsljkÞP hl; �ul � N
X2
j¼1
Xnj
k¼1
sljkxjk;VlðxÞ !
;
pl ¼ Pðxj�ulðxÞ 6 flÞ;pl P lHeGl
;
which can be equivalently transformed into the condition that there exists a vector ðsl; �ulÞ such that
leMljkðsljkÞP hl; �ul � N
X2
j¼1
Xnj
k¼1
sljkxjk;VlðxÞ !
; Pðxj�ulðxÞ 6 flÞP lHeGl
; l ¼ 1;2; j ¼ 1;2; k ¼ 1; . . . ;nj: ð12Þ
In view of (3), it follows that
leMljkðsljkÞP hl () sljk 2 ½mljk � LHðhlÞaljk;mljk þ RHðhlÞbljk�;
where Lw(hl) and Rw(hl) are pseudo inverse functions defined as Lw(hl) = sup{tjL(t) P hl} and Rw(hl) = sup{tjL(t) P hl}. Hence,(12) is rewritten as the equivalent condition that there exists a �ul such that
Pðxj�ulðxÞ 6 flÞP lHeGl
; �ul � NX2
j¼1
Xnj
k¼1
fmljk � LHðhlÞaljkgxjk;VlðxÞ !
: ð13Þ
Since Prðxj�ulðxÞ 6 flÞ is transformed into
P x�ul �
P2j¼1
Pnj
k¼1fmljk � LHðhlÞaljkgxjkffiffiffiffiffiffiffiffiffiffiffiVlðxÞ
p 6
fl �P2
j¼1
Pnj
k¼1fmljk � LHðhlÞaljkgxjkffiffiffiffiffiffiffiffiffiffiffiVlðxÞ
p����� !
;
in consideration of
�ul �P2
j¼1
Pnj
k¼1fmljk � LHðhlÞaljkgxjkffiffiffiffiffiffiffiffiffiffiffiVlðxÞ
p � Nð0;1Þ:
196 M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200
(13) is equivalently transformed as
Ufl �
P2j¼1
Pnj
k¼1fmljk � LHðhlÞaljkgxjkffiffiffiffiffiffiffiffiffiffiffiVlðxÞ
p !P lHeGl
ðhlÞ; ð14Þ
where U is a probability distribution function of the standard Gaussian random variable N (0,1).From the monotone increasingness of U, (14) is rewritten as
X2
j¼1
Xnj
k¼1
mljk � LHðhlÞaljk�
xjk þU�1 lHeGl
ðhlÞ � ffiffiffiffiffiffiffiffiffiffiffi
VlðxÞp
6 fl; ð15Þ
where U�1 is the inverse function of U.From (11)–(15), it holds that
PePlðeGlÞP hl ()
X2
j¼1
Xnj
k¼1
fmljk � LHðhlÞaljkgxjk þU�1 lHeGl
ðhlÞ � ffiffiffiffiffiffiffiffiffiffiffi
VlðxÞp
6 fl: ð16Þ
Consequently, (9) is equivalently transformed into
maximizefor DM1
h1;
maximizefor DM2
h2;
subject toX2
j¼1
Xnj
k¼1
fm1jk � LHðh1Þa1jkgx1jk þU�1 lHeG1
ðh1Þ � ffiffiffiffiffiffiffiffiffiffiffiffi
V1ðxÞp
6 f1;
X2
j¼1
Xnj
k¼1
fm2jk � LHðh2Þa2jkgx2jk þU�1 lHeG2
ðh2Þ � ffiffiffiffiffiffiffiffiffiffiffiffi
V2ðxÞp
6 f2;
A1x1 þ A2x2 6 b;
x1 P 0; x2 P 0:
9>>>>>>>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>>>>>>>;
ð17Þ
In order to obtain an initial candidate for an overall satisfactory solution to (9) or (17), it would be useful for DM1 to find asolution which maximize the smaller degree of satisfaction between the two DMs by solving the maximin problem
maximize min ZP;P1 ðx1; x2Þ; ZP;P
2 ðx1; x2Þn o
;
subject to A1x1 þ A2x2 6 b;x1 P 0; x2 P 0;
9>>=>>; ð18Þ
or equivalently
maximize vsubject to ZP;P
1 ðx1; x2ÞP v ;ZP;P
2 ðx2; x2ÞP v ;A1x1 þ A2x2 6 b;x1 P 0; x2 P 0:
9>>>>>>=>>>>>>;ð19Þ
Similar to the equivalent transformation (16), in view of ZP;Pl ðx1; x2Þ ¼ PePl
ðGlÞ, it holds that
ZP;Pl ðx1; x2ÞP v ()
X2
j¼1
Xnj
k¼1
fmljk � LHðvÞaljkgxjk þU�1 lHeGl
ðvÞ � ffiffiffiffiffiffiffiffiffiffiffi
VlðxÞp
6 fl; ð20Þ
where U�1 is the inverse function of the probability distribution function U of the standard Gaussian random variable N(0,1),and lHeGl
ðvÞ ¼ supfpl j leGlðplÞP vg and Lw(v) = sup{tjL(t) P v} are pseudo inverse functions.
Consequently, (19) is equivalently transformed as
M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200 197
maximize v ;
subject toX2
j¼1
Xnj
k¼1
fm1jk � LHðvÞa1jkgxjk þU�1 lHeG1
ðvÞ � ffiffiffiffiffiffiffiffiffiffiffiffi
V1ðxÞp
6 f1;
X2
j¼1
Xnj
k¼1
fm2jk � LHðvÞa2jkgxjk þU�1 lHeG2
ðvÞ � ffiffiffiffiffiffiffiffiffiffiffiffi
V2ðxÞp
6 f2;
A1x1 þ A2x2 6 b;x1 P 0; x2 P 0:
9>>>>>>>>>>>>=>>>>>>>>>>>>;ð21Þ
Although (21) is nonconvex, an approximate solution can be obtained through particle swarm optimization for nonlinearprogramming (PSONLP) [12].
If DM1 is satisfied with the degrees ZP;Pl x�1; x
�2
� ; l ¼ 1;2, the corresponding optimal solution x�1; x
�2
� to (21) is regarded as
the satisfactory solution. However, if DM1 is not satisfied, by introducing the constraint that ZP;P1 ðx1; x2Þ is larger than or
equal to the minimal satisfactory level d 2 (0,1) specified by DM1, we consider the maximization problem formulated as
maximize ZP;P2 ðx1; x2Þ;
subject to ZP;P1 ðx1; x2ÞP d;
A1x1 þ A2x2 6 b;x1 P 0; x2 P 0;
9>>>>=>>>>; ð22Þ
or equivalently
maximize h;
subject to ZP;P2 ðx1; x2ÞP h;
ZP;P1 ðx1; x2ÞP d;
A1x1 þ A2x2 6 b;x1 P 0; x2 P 0:
9>>>>>>=>>>>>>;ð23Þ
Similar to the equivalent transformation (16), it follows that
ZP;P2 ðx1; x2ÞP h ()
X2
j¼1
Xnj
k¼1
fm2jk � LHðhÞa2jkgxjk þU�1 lHeG2
ðhÞ � ffiffiffiffiffiffiffiffiffiffiffiffi
V2ðxÞp
6 f2;
ZP;P1 ðx1; x2ÞP d ()
X2
j¼1
Xnj
k¼1
fm1jk � LHðdÞa1jkgxjk þU�1 lHeG1
ðdÞ � ffiffiffiffiffiffiffiffiffiffiffiffi
V1ðxÞp
6 f1:
9>>>>>=>>>>>;ð24Þ
Consequently, (23) is rewritten as
maximize h;
subject toX2
j¼1
Xnj
k¼1
fm2jk � LHðhÞa2jkgxjk þU�1 lHeG2
ðhÞ � ffiffiffiffiffiffiffiffiffiffiffiffi
V2ðxÞp
6 f2;
X2
j¼1
Xnj
k¼1
fm1jk � LHðdÞa1jkgxjk þU�1 lHeG1
ðdÞ � ffiffiffiffiffiffiffiffiffiffiffiffi
V1ðxÞp
6 f1;
A1x1 þ A2x2 6 b;x1 P 0; x2 P 0:
9>>>>>>>>>>>>=>>>>>>>>>>>>;ð25Þ
In view of the nonconvexity of (25), quite similar to (21), particle swarm optimization for nonlinear programming (PSONLP)[12] can be applied for solving this problem.
Realizing that the objective functions of DM1 and DM2 often conflict with each other, it should be noted here that thelarger the minimal satisfactory level d for the DM1’s objective function value is specified, the smaller the DM2’s objectivefunction value becomes, which may lead to the unbalanced satisfactory degrees of DM1 and DM2 due to the large differencebetween the objective function values of both DMs. In order to derive the satisfactory solution which has well-balancedobjective function values of DM1 and DM2, by introducing the ratio D between the satisfactory degrees of both DMs ex-pressed as
D ¼ ZP;P2 ðx1; x2Þ
ZP;P1 ðx1; x2Þ
; ð26Þ
we assume that DM1 specifies the lower bound Dmin and the upper bound Dmax of D, which are used for evaluating theappropriateness of the ratio D. To be more specific, if it holds that
Table 1Values
a1
a2
a3
a4
198 M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200
D 2 ½Dmin;Dmax�;
then DM1 regards the corresponding solution as a promising candidate for the satisfactory solution with well-balancedmembership function values.
Now can we construct a procedure of interactive fuzzy programming for the possibility-based probability model in orderto derive an overall satisfactory solution.
3.1. Interactive fuzzy programming in the possibility-based probability model
Step 1: Ask each DM to specify the membership function leGland the target values fl, l = 1, 2.
Step 2: For the current target values fl, l = 1, 2, solve the maximin problem (18).Step 3: DM1 is supplied with the objective function values ZP;P
1 x�1; x�2
� and ZP;P
2 x�1; x�2
� for the optimal solution x�1; x
�2
� obtained in step 2. If DM1 is satisfied with the current membership function values, then stop. If DM1 is not satisfiedand prefers to update fl, l = 1, 2, ask DM1 to update fl, and return to step 2. Otherwise, ask DM1 to specify the minimalsatisfactory level d and the permissible range [Dmin,Dmax] of D.
Step 4: For the current minimal satisfactory d, solve (22).Step 5: DM1 is supplied with the current values of ZP;P
1 x�1; x�2
� , ZP;P
2 x�1; x�2
� and D. If D 2 [Dmin, Dmax] and DM1 is satisfied
with the current objective function values, then stop. Otherwise, ask DM1 to update the minimal satisfactory leveld, and return to step 4.
4. Numerical example
To demonstrate the feasibility and efficiency of the proposed method, consider the following two-level linear program-ming problem involving random fuzzy variable coefficients:
minimizefor DM1
z1ðx1; x2Þ ¼ eC 11x1 þ eC 12x2;
minimizefor DM2
z2ðx1; x2Þ ¼ eC 21x1 þ eC 22x2;
subject to a11x1 þ a12x2 6 b1;
a21x1 þ a22x2 6 b2;
a31x1 þ a32x2 6 b3;
a41x1 þ a42x2 6 b4;
x1 ¼ ðx11; x12; x13ÞT P 0;
x2 ¼ ðx21; x22; x23ÞT P 0:
9>>>>>>>>>>>>>>>>>=>>>>>>>>>>>>>>>>>;
ð27Þ
Table 1 shows values of coefficients of constraints ai, i = 1, 2, 3, 4 and bi, i = 1, 2, 3, 4 and Table 2 shows values of parametersof random fuzzy variables mljk, aljk and r2
ljk; l ¼ 1;2; j ¼ 1;2; k ¼ 1; . . . ;6.Through the use of this numerical example, it is now appropriate to illustrate the proposed interactive fuzzy nonlinear
programming.The parameter values of particle swarm optimization for nonlinear programming (PSONLP) are set as swarm size N = 50,
maximal search generation number Tmax = 3000, c1 = 2.0, c2 = 2.0, w0 = 1.2 and wTmax ¼ 0:1.Although the membership function does not always need to be linear, for the sake of simplicity, we adopt the linear mem-
bership function defined as
llðzlðx1; x2ÞÞ ¼
1; if zlðx1; x2Þ < z1l ;
zlðx1 ;x2Þ�z0l
z1l�z0
l; if z1
l 6 zlðx1; x2Þ 6 z0l ;
0; if zlðx1; x2Þ > z0l :
8>><>>:
Here, the parameter z1l is determined as z1l ¼ zmin
l ¼ zl xlo1 ; x
lo2
� , and the parameter z0
l is specified as z0l ¼ zm
l ¼ zl xjo1 ; x
jo2
� �; l – j,
where xlo1 ; x
lo2
� is a feasible solution minimizing zl(x1,x2). Through PSONLP, the parameter values characterizing the linear
membership functions are determined as z11 ¼ �169:23; z0
1 ¼ �168:40; z12 ¼ �117:20, and z0
2 ¼ �112:08.
of coefficients in constraints.
al11 al12 al13 al21 al22 al23 b
25 15 17 0 0 0 12000 0 0 7 9 10 4004 3 5 3 2 2 1505 5 4 3 4 3 220
Table 2Values of mljk, aljk and r2
ljk .
�~cl11�~cl12
�~cl13�~cl21
�~cl22�~cl23
m1jk �4.30 �3.50 �5.00 �1.30 �1.50 �1.80m2jk �2.90 �2.50 �2.30 �1.50 �1.20 �1.00a1jk 1.20 0.80 0.90 0.40 0.60 0.30a2jk 0.70 0.50 1.10 0.40 0.70 0.60r2
1jk0.30 0.90 1.00 0.50 0.70 1.10
r21jk
0.70 0.60 1.20 0.70 1.00 0.90
Table 3Interaction process.
Interaction 1st 2nd 3rd
d̂ – 0.80 0.90
x11 0.40 15.61 17.99x12 17.20 22.38 20.46x13 8.91 0.30 0.78x21 0.68 0.88 0.18x22 9.13 3.65 0.76x23 2.78 2.59 2.72
ZP;P1 ðxÞ 0.74 0.80 0.90
ZP;P2 ðxÞ 0.74 0.77 0.68
D 1.0 0.96 0.76Iime (sec) 12.14 11.83 9.44
M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200 199
The target values fl; p0l ; p1
l ; l ¼ 1;2 are set as f1 ¼ �160:00; f 2 ¼ �115:00; p01 ¼ 0:20; p0
2 ¼ 0:10; p11 ¼ 0:65; p1
2 ¼ 0:60.The obtained result is shown at the column labeled ‘‘1st’’ in Table 3. DM1 is not satisfied with this solution, but he doesnot desire to update fl, l = 1, 2. Thus, DM1 determines the minimal satisfactory level d̂ ¼ 0:80 to improve ZP;P
1 ðx1; x2Þ at theexpense of ZP;P
2 ðx1; x2Þ. Furthermore, DM1 specifies the upper bound Dmax = 0.90 and the lower bound Dmin = 0.70 for the ra-tio of objective functions D ¼ ZP;P
2 ðx1; x2Þ=ZP;P1 ðx1; x2Þ.
For the updated value of d̂, (27) is solved by PSONLP. The obtained result is shown at the column labeled ‘‘2nd’’ in Table 3.DM1 considers that ZP;P
1 ðx1; x2Þ is improved but D is greater than Dmax. Hence, DM1 is not satisfied with this solution andupdates the minimal satisfactory level d̂ from 0.80 to 0.90. (27) is solved for the updated value of d̂, and the obtained resultis shown at the column labeled ‘‘3rd’’ in Table 3. Since D exists in the interval [Dmin, Dmax] and DM1 is satisfied with thebalance between ZP;P
1 ðx1; x2Þ and ZP;P2 ðx1; x2Þ, the interactive algorithm is terminated.
In the proposed interactive fuzzy nonlinear programming, through a series of update procedures of the minimal satisfac-tory level d̂ and the target values fl, l = 1, 2, it can be possible to obtain a satisfactory solution, where the satisfactory degree ofDM1 is guaranteed to be greater than or equal to the minimal satisfactory level d̂ and is well balanced with that of DM2.
Finally, it is appropriate to point out here that our numerical experiments were performed on a personal computer (PC)with Intel Core 2 Duo 6300 (1.86 GHz) and, as shown in Table 3, the computational times of PSONLP were 12.14, 11.83 and9.44 s, respectively.
5. Conclusion
In this paper, interactive decision making methods for random fuzzy two-level linear programming problems have beenconsidered. Through the introduction of the possibility-based probability model, the original random fuzzy two-level pro-gramming problem was reduced to a deterministic one. In order to obtain a satisfactory solution for the decision makerat the upper level in consideration of the cooperative relation between decision makers, interactive fuzzy nonlinear pro-gramming for random fuzzy two-level linear programming problems was proposed. It was shown that all of the problemsto be solved in the proposed interactive fuzzy nonlinear programming can be solved through particle swarm optimization fornonlinear programming (PSONLP). An illustrative numerical example demonstrated the feasibility and efficiency of the pro-posed method. Extensions to other stochastic programming models will be considered elsewhere. Also extensions to randomfuzzy two-level linear programming problems with two decision makers under noncooperative environments will be re-ported in the near future.
References
[1] M.A. Abo-Sinna, I.A. Baky, Interactive balance space approach for solving multi-level multi-objective programming problems, Information Sciences 177(2007) 3397–3410.
[2] W.F. Bialas, M.H. Karwan, Two-level linear programming, Management Science 30 (1984) 1004–1020.
200 M. Sakawa, T. Matsui / Information Sciences 239 (2013) 191–200
[3] M.A. Gil, M. Lopez-Diaz, D.A. Ralescu, Overview on the development of fuzzy random variables, Fuzzy Sets and Systems 157 (2006) 2546–2557.[4] P. Kall, J. Mayer, Stochastic Linear Programming: Models, Theory, and Computation, second ed., Springer, New York, 2011.[5] H. Katagiri, M. Sakawa, Interactive multi-objective fuzzy random programming through the level-set based probability model, Information Sciences
181 (2011) 1641–1650.[6] H. Kwakernaak, Fuzzy random variables – I. Definitions and theorems, Information Sciences 15 (1978) 1–29.[7] Y.J. Lai, Hierarchical optimization: a satisfactory solution, Fuzzy Sets and Systems 77 (1996) 321–325.[8] E.S. Lee, Fuzzy multiple level programming, Applied Mathematics and Computation 120 (2001) 79–90.[9] J. Li, J. Xu, Multi-objective portfolio selection model with fuzzy random returns and a compromise approach-based genetic algorithm, Information
Sciences 220 (2013) 507–521.[10] B. Liu, Random fuzzy dependent-chance programming and its hybrid intelligent algorithm, Information Sciences 141 (2002) 259–271.[11] Y. Liu, W. Tang, X. Li, Random fuzzy shock models and bivariate random fuzzy exponential distribution, Applied Mathematical Modelling 35 (2011)
2408–2418.[12] T. Matsui, M. Sakawa, K. Kato, T. Uno, K. Tamada, Particle swarm optimization for interactive fuzzy multi-objective nonlinear programming, Scientiae
Mathematicae Japonicae 68 (2008) 103–115.[13] S. Nahmias, Fuzzy variables, Fuzzy Sets and Systems 1 (1978) 97–110.[14] I. Nishizaki, M. Sakawa, Computational methods through genetic algorithms for obtaining Stackelberg solutions to two-level mixed zero-one
programming problems, Cybernetics and Systems: An International Journal 31 (2000) 203–221.[15] S. Pramanik, T.K. Roy, Fuzzy goal programming approach to multilevel programming problems, European Journal of Operational Research 176 (2007)
1151–1166.[16] M.L. Puri, D.A. Ralescu, Fuzzy random variables, Journal of Mathematical Analysis and Applications 114 (1986) 409–422.[17] E. Roghanian, S.J. Sadjadi, M.B. Aryanezhad, A probabilistic bi-level linear multi-objective programming problem to supply chain planning, Applied
Mathematics and Computation 188 (2007) 786–800.[18] M. Sakawa, H. Katagiri, T. Matsui, Interactive fuzzy random two-level linear programming through fractile criterion optimization, Mathematical and
Computer Modelling 54 (2011) 3153–3163.[19] M. Sakawa, I. Nishizaki, Interactive fuzzy programming for decentralized two-level linear programming problems, Fuzzy Sets and Systems 125 (2002)
301–315.[20] M. Sakawa, I. Nishizaki, Interactive fuzzy programming for two-level nonconvex programming problems with fuzzy parameters through genetic
algorithms, Fuzzy Sets and Systems 127 (2002) 185–197.[21] M. Sakawa, I. Nishizaki, Cooperative and Noncooperative Multi-Level Programming, Springer, New York, 2009.[22] M. Sakawa, I. Nishizaki, Interactive fuzzy programming for multi-level programming problems: a review, International Journal of Multicriteria
Decision Making 2 (2012) 241–266.[23] M. Sakawa, I. Nishizaki, H. Katagiri, Fuzzy Stochastic Multiobjective Programming, Springer, New York, 2011.[24] M. Sakawa, I. Nishizaki, Y. Uemura, Interactive fuzzy programming for multi-level linear programming problems, Computers & Mathematics with
Applications 36 (1998) 71–86.[25] M. Sakawa, I. Nishizaki, Y. Uemura, Interactive fuzzy programming for multi-level linear programming problems with fuzzy parameters, Fuzzy Sets
and Systems 109 (2000) 3–19.[26] M. Sakawa, I. Nishizaki, Y. Uemura, Interactive fuzzy programming for two-level linear fractional programming problems with fuzzy parameters, Fuzzy
Sets and Systems 115 (2000) 93–103.[27] M. Sakawa, I. Nishizaki, Y. Uemura, Interactive fuzzy programming for two-level linear and linear fractional production and assignment problems: a
case study, European Journal of Operational Research 135 (2001) 142–157.[28] M. Sakawa, I. Nishizaki, Y. Uemura, A decentralized two-level transportation problem in a housing material manufacturer – interactive fuzzy
programming approach, European Journal of Operational Research 141 (2002) 167–185.[29] M. Sakawa, T. Matsui, Interactive fuzzy random cooperative two-level linear programming through level sets based probability maximization, Expert
Systems with Applications 40 (2013) 1400–1406.[30] M. Sakawa, T. Matsui, Interactive fuzzy programming for fuzzy random two-level linear programming problems through probability maximization
with possibility, Expert Systems with Applications 40 (2013) 2487–2492.[31] M. Sakawa, T. Matsui, Interactive fuzzy random two-level linear programming based on level sets and fractile criterion optimization, Information
Sciences 238 (2013) 163–175.[32] H.S. Shih, Y.J. Lai, E.S. Lee, Fuzzy approach for multi-level programming problems, Computers and Operations Research 23 (1996) 73–91.[33] K. Shimizu, Y. Ishizuka, J.F. Bard, Nondifferentiable and Two-Level Mathematical Programming, Kluwer Academic Publishers, Boston, 1997.[34] M. Simaan, J.B. Cruz Jr., On the Stackelberg strategy in nonzero-sum games, Journal of Optimization Theory and Applications 11 (1973) 533–555.[35] S. Sinha, Fuzzy programming approach to multi-level programming problems, Fuzzy Sets and Systems 136 (2003) 189–202.[36] I.M. Stancu-Minasian, Stochastic Programming with Multiple Objective Functions, D. Reidel Publishing Company, Dordrecht, 1984.[37] G.-Y. Wang, Z. Qiao, Linear programming with fuzzy random variable coefficients, Fuzzy Sets and Systems 57 (1993) 295–311.[38] S. Wang, J. Watada, A hybrid modified PSO approach to VaR-based facility location problems with variable capacity in fuzzy random uncertainty,
Information Sciences 192 (2012) 3–18.[39] M. Wen, R. Kang, Some optimal models for facility location–allocation problem with random fuzzy demands, Applied Soft Computing 11 (2011) 1202–
1207.[40] Y. Xu, J. Hu, Random fuzzy demand newsboy problem, Physics Procedia 25 (2012) 924–931.