Professional Documents
Culture Documents
8 Efficiency Analysisand Rankingof DMUswith Fuzzy Data
8 Efficiency Analysisand Rankingof DMUswith Fuzzy Data
net/publication/227226428
CITATIONS READS
214 526
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Saber Saati on 13 August 2016.
A. MEMARIANI memar@modares.ac.ir
Department of Industrial Engineering, Tarbiat Modarres University, Tehran, Iran
G. R. JAHANSHAHLOO
Department of Mathematics, Teacher Training University, Tehran, Iran
Abstract. In this paper, a fuzzy version of CCR model (Charnes, Cooper and Rhodes (1978)) with asymmetrical
triangular fuzzy number is presented and a procedure is suggested for its solution. The basic idea is to transform
the fuzzy CCR model into a crisp linear programming problem by applying an alternative -cut approach.
Thereby, the problem is converted to an interval programming. In this method, instead of comparing the equality
(or inequality) of two intervals, a variable is defined in the interval, not only satisfies the set of constraints, but
also maximizes the efficiency value. We also propose a ranking method for fuzzy DMUs using presented fuzzy
DEA approach. To demonstrate the concept, numerical examples are solved and solutions are compared with Guo
and Tanaka (2001).
Keywords: data envelopment analysis, membership function, possibilistic linear programming, ranking
1. Introduction
which is not often very practicable. A fuzzy K-means clustering approach as a means
of identifying unusual and/or extreme efficiency behavior is proposed in Seaver and
Triantis (1992). In Cooper, Park and Yu (1999), the imprecise DEA method has been
developed. This method permits mixture of imprecisely data within specified bounds
and exactly known data. Kao and Liu (2000) formulated a pair of parametric
programs to derive bounds on the membership functions of the efficiency measures
in fuzzy BCC model. More recently, in Guo and Tanaka (2001) a fuzzy DEA model
is proposed. They considered the data as symmetrical triangular fuzzy vectors. After
using -cuts of constraints and comparison of intervals, they used a pair of linear
programming problems to evaluate the efficiency of the DMU under consideration.
Furthermore, with using the relationship between DEA and regression analysis (RA),
they proposed an extension of the fuzzy DEA model.
Our objective here is to explore the application of fuzzy measures and fuzzy
mathematical programs in the DEA models, where the observed data set provides
vague and imprecise knowledge about the generating process. To do so, a fuzzy issue
of CCR model is suggested and an alternative method is proposed to solve it. In the
proposed method, after determining the -cuts of objective function and constraints,
the fuzzy triangular numbers are converted to crisp intervals.
In most of the existing methods for possibilistic linear programming, where the
-cut is used, the solution is obtained by comparing the intervals in left and right
hand side of the constraints (see Bezdek (1987), Buckley (1988, 1989), Delgado,
Verdegay and Vila (1990), Fuller (1986), Lai and Hwang (1992, 1993), Negi (1989),
Ramik and Rimanek (1985), Rommelfanger et al (1989), Sengupta (1992), Shaocheng
(1994), Tanaka, Ichihashi and Asai (1984)). Different methodologies have been
suggested for the comparison of the intervals. In some of these methods simply
the end points of the interval are considered for justification that makes the model
very simple and hence a lot of information might have been lost. In the others the
complexity of the algorithm may cause computational inefficiency (see Modarres and
Sadi-Nezhad (2001)).
A new idea is to find a point in each interval, satisfying the set of constraints and at
the same time maximizing (minimizing) the objective function. This is defined as a
variable in the suggested method.
DEA assigns an efficiency score less than one to inefficient DMUs and equal to one to
efficient DMUs. So, for inefficient DMUs a ranking is given but for efficient ones no
ranking can be given. Some methods for ranking efficient DMUs with crisp data are
proposed (see Andersen and Petersen (1993), Hougaard (1999), Mehrabian, Alirezaee
and Jahanshahloo (1999)). In this paper, by considering fuzzy DMUs, an alternative
ranking method based on comparing the best part of DMUs by the inner part of
efficiency frontier is proposed.
The paper is organized as follows: Section 2 is concerned with a fuzzy version of
CCR model with triangular fuzzy numbers. The suggested method to solve fuzzy CCR
model is presented in section 3. Section 4 deals with a ranking model for fuzzy DMUs.
To demonstrate the concept, numerical examples are given in section 5. Section 6 closes
with conclusion.
EFFICIENCY ANALYSIS AND RANKING OF DMUS 257
Assume that there are n decision-making units (DMUs) to be evaluated. Each DMU
consumes varying amounts of m different inputs to produce s different outputs.
Specifically, DMUj consumes amounts xij (i ¼ 1, . . . , m) of inputs and produces
amounts yrj (r ¼ 1, . . . , s) of outputs. In the model formulation, xip (i ¼ 1, . . . , m) and
yrp (r ¼ 1, . . . , s) denote, respectively, the nonnegative crisp vectors of input and output
values for DMUp.
The primal and dual linear programming statements for the (input oriented) CCR
model are:
X
n X
m
s:t : hxip kj xij 0 8i; s:t : vi xip ¼ 1 ð1Þ
j¼1 i¼1
X
n X
s X
m
kj yrj yrp 8r; ur yrj vi xij 0 8j;
j¼1 r¼1 i¼1
kj 0 8j: ur ; vi 0 8i; r:
X
s
max Wp ¼ ur y~rp
r¼1
X
m
s:t : vi x~ip ¼ 1~ 8i; ð2Þ
i¼1
X
s X
m
ur y~rj vi x~ij 0 8j;
r¼1 i¼1
ur ; v i 0 8i; r:
Figure 1 shows the efficiency frontier of fuzzy CCR model in the simplest case of one
input and one output. In Figure 1.a, one DMU uses fuzzy input x̃ ¼ (xm, xl, xu) to
produce yo. Production frontier when the input and output of DMU are xo and ỹ ¼ ( ym,
yl, yu), respectively is shown in Figure 1.b.
Among the various types of fuzzy numbers, triangular fuzzy numbers are of more
importance. In the sequel, we consider the inputs and outputs of DMUs as triangular fuzzy
numbers.
Let x̃ij ¼ (xijm, xijl , xiju ) and ỹrj ¼ ( yrjm, ylrj, yrju ). Therefore, (2) can be written as follows:
X
s
max Wp ¼ ur ðym l u
rp ; yrp ; yrp Þ
r¼1
X
m
s:t : vi ðxm l u l u
ip ; xip ; xip Þ ¼ ð1; 1 ; 1 Þ 8i; ð3Þ
i¼1
X
s X
m
ur ðym l u
rj ; yrj ; yrj Þ vi ðxm l u
ij ; xij ; xij Þ 0 8j;
r¼1 i¼1
ur ; v i 0 8i; r:
3. Proposed Method
To solve (3), we apply the concept of -cut. By introducing -cuts of objective function
and constraints, the following model is obtained:
X
s
max Wp ¼ ur ½ym l m u
rp þ ð1 Þyrp ; yrp þ ð1 Þyrp
r¼1
Xm
s:t : vi ½xm l m u l u
ip þ ð1 Þxip ; xip þ ð1 Þxip
¼ ½ þ ð1 Þ1 ; þ ð1 Þ1
8i;
i¼1
X
s
ur ½ym l m u
rj þ ð1 Þyrj ; yrj þ ð1 Þyrj
r¼1
Xm
vi ½xm l m u
ij þ ð1 Þxij ; xij þ ð1 Þxij
0 8j;
i¼1
ur ; vi 0 8i; r:
ð4Þ
Note that all of the coefficients are stated as intervals. This implies that (4) involves
treating an interval problem and, hence, cannot be treated by standard methods without
further transformations. There are many methods for solving (4). In most of these methods,
the main idea is based on comparison of intervals. In this section, instead of comparing the
intervals, a new approach is suggested in which, a variable is defined in the intervals not
only satisfying the set of constraints, but also maximizes the objective function.
s:t : ½x2 ; y2
½x3 ; y3
Step 1. Interval alterations, that substitute the resulted intervals by the following
variables:
^xij 2 ½xm l m u
ij þ ð1 Þxij ; xij þ ð1 Þxij
^yrj 2 ½ym l m u
rj þ ð1 Þyrj ; yrj þ ð1 Þyrj
L 2 ½ þ ð1 Þ1l ; þ ð1 Þ1u
X
s
max Wp ¼ ur ^yrp
r¼1
Xm
s:t : vi ^xip ¼ L
i¼1
X
s X
m
ur ^yrj vi ^xij 0 8 j; ð5Þ
r¼1 i¼1
xm l
xij xm
ij þ ð1 Þxij ^
u
ij þ ð1 Þxij 8i; j;
ym l
yrj ym
rj þ ð1 Þyrj ^
u
rj þ ð1 Þyrj 8r; j;
þ ð1 Þ1l L þ ð1 Þ1u
ur ; v i 0 8i; r:
X
m
s:t : xip ¼ L
i¼1
X
s X
m
yrj xij 0 8j; ð6Þ
r¼1 i¼1
vi ðxm l m u
ij þ ð1 Þxij Þ xij vi ðxij þ ð1 Þxij Þ 8i; j;
ur ðym l m u
rj þ ð1 Þyrj Þ yrj ur ðyrj þ ð1 Þyrj Þ 8r; j;
þ ð1 Þ1l L þ ð1 Þ1u
ur ; v i 0 8i; r:
EFFICIENCY ANALYSIS AND RANKING OF DMUS 261
The objective function of this model like the objective function of CCRID model reflects
the value of efficiency of the DMU under consideration. In CCRID model the efficiency is
evaluated by the summation of all outputs multiplied by respective weights. Same concept
has been extended in model (6). Here the ȳrp, (r ¼ 1, . . . , m) are variables in the intervals
m l m u
ur[yrp þ (1 )yrp , yrp þ (1 )yrp ], (r ¼ 1, . . . , m). Hence the efficiency value is
represented in term of interval. The optimal value of these variables ȳrp, (r ¼ 1, . . . , m) is
the optimizing point where satisfies the set of constraints and at the same time the
summation of these variables evaluates the efficiency of DMUp.
X
s X
m
yrp xip 0 ð7Þ
r¼1 i¼1
If 1u > 1 then, some of the DMUs may obtain efficiency greater than 1. So, it must be
equal to 1. Then the last constraint of (6) can be written as:
þ ð1 Þ1l L 1 ð8Þ
With reference to the objective function, the first constraint of (6) and (7), (8) is
redundent and L ¼ 1. So, the model (6) can be equivalently written as:
X
s
max W ¼ yrp
r¼1
X
m
s:t : xip ¼ 1
i¼1
X
s X
m
yrj xij 0 8 j; ð9Þ
r¼1 i¼1
vi ðxm l m u
ij þ ð1 Þxij Þ xij vi ðxij þ ð1 Þxij Þ 8i; j;
ur ðym l m u
rj þ ð1 Þyrj Þ yrj ur ðyrj þ ð1 Þyrj Þ 8r; j;
u r ; vi 0 8i; r:
The standard DEA models assign an efficiency score less than one to inefficient DMUs,
from which a ranking can be derived. However, efficient DMUs all have an efficiency of 1,
so that for these units no ranking can be given. Two models for ranking efficient DMUs in
DEA were proposed in Andersen and Petersen (1993) and Mehrabian, Alirezaee and
Jahanshahloo (1999). These models, like other standard DEA models, do not consider the
inexact and fuzzy data in evaluations.
Fuzzy DEA also may show some DMUs efficient. So, there is a need to rank these
efficient DMUs. A method for ranking the efficient DMUs is to develop a mathematical
model that will be elaborated below.
The fuzzy CCRIP model is as follows:
min Z ¼ h
X
n
s:t : hx~ip kj x~ij 8i;
j¼1
X
n
y~rp kj y~rj 8r; ð10Þ
j¼1
kj 0 8j:
The notation ‘f’ indicates that the data are fuzzy numbers.
Let x̃ij ¼ (xijm, xijl, xiju ) and ỹrj ¼ ( yrjm, yrjl , yrju ). Therefore, (10) can be written as follows:
min Z ¼ h
!
X
n X
n X
n
s:t : ðhxm l u
ip ; hxip ; hxip Þ kj x m
ij ; kj xlij ; kj xuij 8i;
j¼1 j¼1 j¼1
!
X
n X
n X
n
ð11Þ
ðym l u
rp ; yrp ; yrp Þ kj y m
rj ; kj ylrj ; kj yurj 8r;
j¼1 j¼1 j¼1
kj 0 8 j:
To solve this problem, we apply the concept of -cut. By introducing -cuts of constraints
and summation of triangular fuzzy numbers, we will have the following problem:
min Z ¼ h
s:t : ½hðxm l m u
ip þ ð1 Þxip Þ; hðxip þ ð1 Þxip Þ
" #
X
n X
n
kj ðxm
ij þ ð1 Þxlij Þ; kj ðxm
ij þ ð1 Þxuij Þ 8i;
j¼1 j¼1
EFFICIENCY ANALYSIS AND RANKING OF DMUS 263
½ym l m u
rp þ ð1 Þyrp ; yrp þ ð1 Þyrp
" #
X
n X
n
kj ðym
rj þ ð1 Þylrj Þ; kj ðym
rj þ ð1 Þyurj Þ 8r; ð12Þ
j¼1 j¼1
kj 0 8j:
To rank DMUs, for each DMU the lower level of inputs and upper level of outputs (that
is the best part of DMU) are compared by the inner part of efficiency frontier. If in this
case, the best part of DMU goes out of this part of frontier, then an efficiency score more
than one will assign to it. So, by this idea, the efficient DMUs will be ranked.
In (12), the best part of DMUp is (X mp þ (1 )Xpl, Ypm þ (1 )Ypu ) and the inner
part of the frontier is ðnj¼1 kj ðXjm þ ð1 ÞXju Þ; nj¼1 kj ðYjm þ ð1 ÞYjl ÞÞ. By this
explanation, the suggested model for ranking DMUs can be written as follows:
min Z ¼ h
X
n
s:t : hðxm l
ip þ ð1 Þxip Þ kj ðxm u
ij þ ð1 Þxij Þ 8i;
j¼1
X
n ð13Þ
ym u
rp þ ð1 Þyrp kj ðym l
rj þ ð1 Þyrj Þ 8r;
j¼1
kj 0 8 j:
5. A Numerical Example
As an example, first consider the presented example in Guo and Tanaka (2001). The data
are listed in Table 1 and each DMU consumes two symmetric triangular fuzzy inputs to
produce two symmetric triangular fuzzy outputs.
Fuzzy efficiencies of DMUs with the proposed method in Guo and Tanaka (2001) with
different values are listed in Table 2.
Using (6), the efficiencies are summarized in the first part of Table 3.
With reference to Tables 2 and 3, it can be noted that the efficiency value of the proposed
model is greater than Guo and Tanaka’s model. It is because the obtained optimizing point
provides in the best situation for each DMU in contrast to Guo and Tanaka’s model, which
264 SAATI, MEMARIANI AND JAHANSHAHLOO
DMUs D1 D2 D3 D4 D5
I1 (4.0, 3.5, 4.5) (2.9, 2.9, 2.9) (4.9, 4.4, 5.4) (4.1, 3.4, 4.8) (6.5, 5.9, 7.1)
I2 (2.1, 1.9, 2.3) (1.5, 1.4, 1.6) (2.6, 2.2, 3.0) (2.3, 2.2, 2.4) (4.1, 3.6, 4.6)
O1 (2.6, 2.4, 2.8) (2.2, 2.2, 2.2) (3.2, 2.7, 3.7) (2.9, 2.5, 2.3) (5.1, 4.4, 5.8)
O2 (4.1, 3.8, 4.4) (3.5, 3.3, 3.7) (5.1, 4.3, 5.9) (5.7, 5.5, 5.9) (7.4, 6.5, 8.3)
D1 D2 D3 D4 D5
0 (0.81, 0.66, 0.99) (0.98, 0.88, 1.09) (0.82, 0.60, 1.12) (0.93, 0.71, 1.25) (0.79, 0.61, 1.02)
0.5 (0.83, 0.75, 0.92) (0.97, 0.94, 1.00) (0.83, 0.71, 0.97) (0.97, 0.85, 1.12) (0.82, 0.72, 0.93)
0.75 (0.84, 0.80, 0.88) (0.99, 0.96, 1.02) (0.83, 0.77, 0.90) (0.98, 0.92, 1.05) (0.83, 0.78, 0.89)
1 0.85 1.00 0.86 1.00 1.00
Efficiency Rank
D1 D2 D3 D4 D5 D1 D2 D3 D4 D5
0 1.00 1.00 1.00 1.00 1.00 1.11 1.24 1.28 1.52 1.30
0.5 0.96 1.00 1.00 1.00 1.00 0.96 1.11 1.03 1.26 1.16
0.75 0.90 1.00 0.93 1.00 1.00 0.90 1.05 0.93 1.13 1.10
1 0.85 1.00 0.86 1.00 1.00 0.85 1.00 0.86 1.00 1.00
is based on the comparison of the intervals. It must be noted that in the proposed model only
one linear programming problem is sufficient for efficiency evaluation whereas the Guo
and Tanaka’s model needs two linear programming problems for the same evaluation. In the
first part of Table 3, as seen, the efficiencies are decreased by increasing , but D2, D4 and
D5 are efficient for all 2 (0, 1]. The last row of this table shows efficiencies by ¼ 1. In
this case, (9) is equivalent to the standard CCRID model (1). By ¼ 1, the efficiencies are
calculated by mean values of data, without considering the left or right spreads.
Consider ¼ 0.5. By the proposed model, all DMUs except D1 are efficient, for which a
ranking cannot be obtained. Now, for ranking these DMUs, the model (13) is being used.
The ranking is presented in the second part of Table 3. The ranking order of inefficient
DMUs does not change by proposed model and the efficient DMUs are ranked as D4, D5,
D2 and D3.
The above example is for symmetric fuzzy numbers. The next example is provided for
asymmetric fuzzy numbers. This is the ability of proposed method vs. Guo and Tanaka
(2001) in tackling asymmetric fuzzy numbers.
Consider 10 DMUs as Table 4, which each DMU consumes two asymmetric fuzzy
inputs to, produces two asymmetric fuzzy outputs.
EFFICIENCY ANALYSIS AND RANKING OF DMUS 265
I1 I2 O1 O2
D1 (7.0, 6.0, 8.0) (30.0, 29.0, 32.0) (38.0, 35.5, 41.0) (411.0, 409.0, 416.0)
D2 (6.0, 5.5, 6.5) (35.0, 33.0, 36.5) (40.0, 39.0, 43.0) (480.0, 478.0, 484.0)
D3 (9.0, 7.5, 10.5) (45.0, 43.0, 48.0) (35.0, 32.0, 38.0) (299.0, 297.0, 301.0)
D4 (8.0, 7.0, 10.0) (39.0, 37.5, 42.0) (31.0, 28.0, 31.0) (352.0, 347.0, 360.0)
D5 (11.0, 9.0, 12.0) (44.0, 43.0, 45.0) (35.0, 33.0, 38.0) (411.0, 406.0, 415.0)
D6 (10.0, 10.0, 10.0) (55.0, 53.0, 57.5) (38.0, 36.0, 40.0) (286.0, 282.0, 289.0)
D7 (12.0, 10.0, 14.0) (110.0, 107.0, 113.0) (36.0, 34.5, 38.0) (400.0, 396.0, 405.0)
D8 (13.0, 9.0, 16.0) (100.0, 95.0, 101.0) (41.0, 37.0, 46.0) (393.0, 387.0, 402.0)
D9 (14.0, 12.0, 15.0) (125.0, 120.0, 131.0) (27.0, 24.0, 28.0) (404.0, 400.0, 406.0)
D10 (8.0, 5.0, 10.0) (38.0, 35.0, 39.0) (50.0, 48.0, 51.0) (470.0, 470.0, 470.0)
D1 D2 D3 D4 D5 D6 D7 D8 D9 D10
0 1.25 1.30 0.84 0.76 0.78 0.69 0.63 0.85 0.46 1.70
0.2 1.19 1.24 0.79 0.74 0.75 0.67 0.59 0.75 0.44 1.48
0.4 1.12 1.17 0.75 0.72 0.73 0.65 0.55 0.66 0.42 1.30
0.6 1.08 1.11 0.71 0.70 0.71 0.63 0.51 0.59 0.40 1.20
0.8 1.04 1.05 0.66 0.68 0.69 0.60 0.48 0.53 0.38 1.10
1 1.00 1.00 0.61 0.66 0.68 0.58 0.45 0.47 0.36 1.00
To evaluate the efficiency of illustrated DMUs, the model (9) is used. The results are
summarized in Table 5.
In Table 5, as seen, the efficiencies are decreased by increasing , but D1, D2 and D10 are
efficient for all 2 (0, 1]. The last column of this table shows efficiencies by ¼ 1. By
¼ 1, the efficiencies are calculated by the mean values of fuzzy data (the value for which
the membership is 1), without considering the left or right spreads and (9) is same as
standard CCRID model (1).
266 SAATI, MEMARIANI AND JAHANSHAHLOO
By this model, 3 DMUs are efficient for all 2 [0, 1], for which a ranking cannot be
obtained. The efficient DMUs are ranked by suggested model (13) as Table 6.
The efficient DMUs are ranked as D10, D2 and D1, respectively.
6. Conclusion
In this paper, CCR model is suggested for fuzzy data. Using fuzzy data, standard CCR
model is converted to a possibilistic-programming problem. We propose a method for
converting this problem into a crisp linear programming model based on -cut.
In most -cut based methods, the resulting model is solved by means of comparing two
intervals, i.e. interval of LHS and interval of RHS of each equality/inequality constraint.
This comparison either oversimplifies the model or makes it difficult to solve.
The proposed approach assumes that the solution lies in the interval and defines suitable
variables for this solution. The substitutions of these variables make the model non-linear.
By further suitable substitutions the model is linearized. Hence, by solving a linear
programming problem for a given -cut, it is possible to generate a reliable and robust
solution for possibilistic mathematical programming problems.
In contrast to the Guo and Tanaka’s models, which solve two linear programming
problems to obtain the efficiency of a given DMU, our method by solving a single linear
programming problem for each DMU, determines its efficiency. On the other hand, the
Guo and Tanaka’s models are designed for symmetrical triangular fuzzy numbers
whereas the proposed method in the present paper is able to tackle asymmetrical
triangular fuzzy numbers as well.
The suggested ranking model in this paper is an application of fuzzy set theory in
DEA. This model not only dissolves the difficulties encountered in AP and MAJ
models, but also is capable to rank the DMUs that consume fuzzy inputs to produce
fuzzy outputs.
Acknowledgments
The authors would like to thank Professor Nuttle whose comments improved the quality of
the paper.
References
Andersen, P. and N. C. Petersen. (1993). ‘‘A Procedure for Ranking Efficient Units in Data Envelopment
Analysis,’’ Management Science 39, 1261 – 1264.
Bezdek, J. C. (1987). ‘‘Analysis of Fuzzy Information.’’ (Vol. III): Application in Engineering and Sciences. Boca
Raton: CRC Press, 241 – 263.
Buckley, J. J. (1988). ‘‘Possibilistic Linear Programming with Triangular Fuzzy Numbers,’’ Fuzzy Sets and
Systems 26, 135 – 138.
EFFICIENCY ANALYSIS AND RANKING OF DMUS 267
Buckley, J. J. (1989). ‘‘Solving Possibilistic Linear Programming Problems,’’ Fuzzy Sets and Systems 31,
329 – 341.
Charnes, W., W. Cooper, and E. Rhodes. (1978). ‘‘Measuring the Efficiency of Decision Making Units,’’ EJOR 2,
429 – 444.
Cooper, W. W., K. S. Park, and G. Yu. (1999). ‘‘Models for Dealing with Imprecise Data in DEA,’’ Management
Science 45, 597 – 607.
Delgado, M., J. L. Verdegay, and M. A. Vila. (1990). ‘‘Relating Different Approaches to Solve Linear Program-
ming Problems With Imprecise Costs,’’ Fuzzy Sets and Systems 37, 33 – 42.
Fuller, R. (1986). ‘‘On a Special Type of Fuzzy Linear Programming,’’ Colloquia Mathematica Societatis Janos
Bolyai 50, 511 – 519.
Guo, P. and H. Tanaka. (2001). ‘‘Fuzzy DEA: A Perceptual Evaluation Method,’’ Fuzzy Sets and Systems 119,
149 – 160.
Hougaard, J. L. (1999). ‘‘Fuzzy Scores of Technical Efficiency,’’ European Journal of Operational Research 115,
529 – 541.
Kao, C. and Shiang-Tai Liu. (2000). ‘‘Fuzzy Efficiency Measures in Data Envelopment Analysis,’’ Fuzzy Sets
and Systems 113, 427 – 437.
Lai, Y. J. and C. L. Hwang. (1992). Fuzzy Mathematical Programming, Berlin: Springer.
Lai, Y. J. and C. L. Hwang. (1993). ‘‘A New Approach to Some Possibilistic Linear Programming Problem,’’
Fuzzy Sets and Systems 49.
Mehrabian, S., M. R. Alirezaee, and G. R. Jahanshahloo. (1999). ‘‘A Complete Efficiency Ranking of Decision
Making Units in Data Envelopment Analysis,’’ Computational Optimization and Applications 14, 261 – 266.
Modarres, M. and S. Sadi-Nezhad. (2001). ‘‘Ranking Fuzzy Numbers by Preference Ratio,’’ Fuzzy Sets and
Systems 118, 429 – 436.
Negi, D. S. (1989). Fuzzy Analysis and Optimization. Ph.D. Dissertation, Dept. of IE., Kansas State University.
Ramik, J. and J. Rimanek. (1985). ‘‘Inequality Relation Between Fuzzy Numbers and Its Use in Fuzzy Opti-
mization,’’ Fuzzy Sets and Systems 16, 123 – 138.
Rommelfanger, H. R., R. Hanuscheck, and J. Wolf. (1989). ‘‘Linear Programming With Fuzzy Objectives,’’ Fuzzy
Sets and Systems 29, 31 – 48.
Seaver, B. and K. Triantis. (1992). ‘‘A Fuzzy Clustering Approach Used in Evaluating Technical Efficiency
Measures in Manufacturing,’’ Journal of Productivity Analysis 3, 337 – 363.
Sengupta, J. K. (1992). ‘‘A Fuzzy System Approach in Data Envelopment Analysis,’’ Computers Math. Applic.
24, 259 – 266.
Shaocheng, T. (1994). ‘‘Interval Number and Fuzzy Number Linear Programming,’’ Fuzzy Sets and Systems 66,
301 – 306.
Tanaka, H., H. Ichihashi, and K. Asai. (1984). ‘‘A Formulation of Fuzzy Linear Programming Problems Based on
Comparison of Fuzzy Numbers,’’ Control and Cybernetics 13, 186 – 194.