358 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 33, NO. 3, MAY 2003
Fuzzy Compromise Programming for
Group Decision Making
Predrag Prodanovic and Slobodan P. Simonovic
Abstract—A multicriteria technique named fuzzy compromise
programming is combined with a methodology known as group
decision making under fuzziness to come up with a new technique
that supports decision making with multiple criteria and multiple
participants (or experts). All criteria (qualitative and quantitative)
are modeled by way of fuzzy sets, utilizing the fact that criteria
values in most water resources problems are vague, imprecise
and/or ill defined. The involvement of multiple experts in the
decision process is achieved by incorporating each participant’s
perception of criteria weights, best and worst criteria values,
relative degrees of risk acceptance, as well as other parameters
into the problem. The proposed methodology is illustrated with
a case study taken from the literature, combined with the input
of four expert individuals with diverse backgrounds. After pro-
cessing the input from the experts, a group compromise decision
is formulated.
Index Terms—Compromise decision, fuzzy set ranking methods,
multiple criteria, multiple decision makers, risk preferences, water
resources systems.
I. INTRODUCTION
MULTIPLE criteria multiple expert decision making prob-lems arise quite often in the field of water resources man-
agement. Examples of this type include work of Borsuk et al.
[1], where a decision support system is proposed to reduce ni-
trogen content in a river in North Carolina, with presence of
multiple stakeholders with conflicting interests. Research by
Hämäläinen et al. [2] works toward development of a water-
level management policy for a lake-river system in Finland, con-
sidering stakeholders input from the initial stage of problem
structuring, through the stage of group consensus, to final stage
of seeking public support. Raj and Kumar [3], [4] propose and
apply a methodology for the purpose of finding the most suit-
able planning of reservoirs aimed for the development of Kr-
ishna River Basin in India. Additional literature on multiple cri-
teria methodologies applied to environmental issues include the
work of Agrell et al. [5], Fukuyama et al. [6], Haimes et al. [7],
Rajabi et al. [8], Ridgley et al. [9], Tecle et al. [10], Hipel et al.
[11], and others.
Manuscript received December 9, 2002. This work was supported in part by
grants from the Natural Sciences and Engineering Research Council, the Social
Sciences and Humanities Research Council and the Community University Re-
search Association. This paper was recommended by Associate Editor T. Sud-
kamp.
P. Prodanovic is with the Department of Civil and Environmental En-
gineering, University of Western Ontario, London, ON, N6A 5B9 Canada
(e-mail: pprodano@uwo.ca).
S. P. Simonovic is with the Department of Civil and Environmental
Engineering and Institute for Catastrophic Loss Reduction, The University of
Western Ontario, London, ON, N6A 5B9 Canada (e-mail: simonovic@uwo.ca).
Digital Object Identifier 10.1109/TSMCA.2003.817050
Stakeholders participation is a key issue in planning and man-
agement of complex systems, such as those encountered in the
field of water resources. An operational framework that involves
managing evolving relations between players (regulators, de-
cision makers, stakeholders, general public) and their values
is still missing in water resources management. Therefore, a
methodology that includes active participation of stakeholders,
in multiple criteria decision-making setting, is one that is sought
after in this paper. Furthermore, a particular emphasis shall be
placed on modeling uncertainties, as well as preferences of de-
cision makers, with the aid of the theory of fuzzy logic. Water
resources decision-making demands interaction between stake-
holders with conflicting interests and/or stakeholders and the
environment. Using the tools of fuzzy logic these types of inter-
actions can be understood and modeled rather accurately. In this
paper, an original technique for water resources group decision
making (with multiple objectives) is proposed. The technique
integrates a methodology named group decision making under
fuzziness (Kacprzyk and Nurmi [12]), and a multicriteria tech-
nique called fuzzy compromise programming, FCP for short
(Bender and Simonovic [13], [14]).
Classical compromise programming (or compromise pro-
gramming) has extensively been documented and studied in the
literature. Goicoechea et al. [15] use compromise programming
to evaluate a set of water quality management alternatives
subject to multiple (conflicting) criteria. Simonovic [16] uses
compromise programming to evaluate alternative options in
the context of long-term water resources planning. Tkach and
Simonovic [17] extended compromise programming, together
with geographical information systems, to come up with spatial
compromise programming—a methodology able to model
spatial variability of criteria values. Tecle et al. [10] use com-
promise programming to formulate a decision support system
for analysis of multiresource forest management problems.
Bardossy et al. [18] modified compromise programming to
form composite programming—a methodology that deals with
problems of hierarchical nature (i.e., when certain criteria con-
tain a number of subcriteria). In addition, composite program-
ming was further modified into fuzzy composite programming,
which instead of crisp input variables considers fuzzy variables.
However, fuzzy composite programming (as of now) models
only criteria values as fuzzy sets, while keeping other param-
eters of the equation (such as weights, and deviation measure-
ment) crisp. Applications of fuzzy composite programming in-
clude that of Lee et al. [19], and [20], and Hagemeister et al.
[21].
No one in the literature (to the best knowledge of the authors)
attempted use of compromise programming for group decision-
1083-4427/03$17.00 © 2003 IEEE

PRODANOVIC AND SIMONOVIC: FUZZY COMPROMISE PROGRAMMING 359
making. The objective of this paper is to include group support
within the framework of compromise programming. The pro-
posed approach includes a combination of fuzzy compromise
programming (Bender and Simonovic [13], [14]) and group de-
cision making under fuzziness (Kacprzyk and Nurmi [12]). In
other words, two techniques are merged together and a new mul-
tiple criteria multiple expert decision support methodology is
formulated.
The rest of the paper is organized as follows: Section II will
present the mathematics needed to evaluate discrete alterna-
tives using fuzzy compromise programming (FCP), a technique
which uses a single decision maker in a multicriteria setting. Al-
gorithms of Kacprzyk and Nurmi [12] shall then be put forward,
to illustrate ways of aggregating opinions and preferences of dif-
ferent experts. A procedure for combining the above-mentioned
methodologies will be discussed next. Section III will show the
utility of the proposed methodology through a case study taken
from the literature, in combination with the input from individ-
uals with various backgrounds (these individuals can be thought
of as different stakeholders or decision makers involved in the
case study). Lastly, Section IV will present concluding remarks,
and recommendations for further research will be given in Sec-
tion V.
II. MERGING OF FUZZY COMPROMISE PROGRAMMING WITH
GROUP DECISION MAKING UNDER FUZZINESS ALGORITHMS
A. Fuzzy Compromise Programming
Classical compromise programming is a multicriteria deci-
sion analysis technique used to identify the best compromise so-
lution from a set of solutions by some measure of distance. The
measure of distance, referred to as a distance metric, determines
the closeness of a particular solution to a generally infeasible
(ideal) solution. Therefore, obtaining a compromise solution is
analogous to obtaining a solution that is as close as practically
possible to the ideal solution. To see the meaning of the concept
of compromise programming, consider the following example.
Suppose two objectives are to be met for a maximization
problem. Assume that the objectives are “protection of the en-
vironment” and “development possibility,” for some set of al-
ternatives. Consider four available alternatives, from which one
is to be chosen for implementation. Now, the ideal point (or the
ideal alternative) would be one where both objectives are maxi-
mized. This point is in most practical cases infeasible, and so a
compromise must be sought (i.e., if we are to have well a pro-
tected environment, the chance is that development will not be
able to proceed, and vice versa.)
Graphically, compromise programming is illustrated in
Fig. 1, where it is clear that is an alternative closest to the
ideal solution, assuming equal weighting of both objectives (its
distance metric is the smallest). Mathematically, compromise
programming distance metric in its discrete form can be
presented as
(1)
Fig. 1. Graphical illustration of compromise programming.
where
and represents criteria or objectives;
and represents alternatives;
distance metric of alternative ;
corresponds to a weight of a particular criteria or
objective;
parameter ;
and best and the worst value for criteria , respec-
tively (also referred to as positive and negative
ideals);
actual value of criterion .
The parameter is used to represent the importance of the max-
imal deviation from the ideal point. If , all deviations are
weighted equally; if , the deviations are weighted in pro-
portion to their magnitude. Typically, as increases, so does the
weighting of the deviations. As Tecle et al. [10] put it, “varying
the parameter from 1 to infinity, allows one to move from mini-
mizing the sum of individual regrets (i.e., having a perfect com-
pensation among the objectives) to minimizing the maximum
regret (i.e., having no compensation among the objectives) in
the decision making process. The choice of a particular value of
this compensation parameter depends on the type of problem
and desired solution. In general, the greater the conflict between
players is, the smaller the possible compensation becomes.”
The weight parameter characterizes decision makers’
preference concerning the relative importance of criteria.
Simply stated, the parameter places emphasis on the criteria
the decision maker deems important. The parameter is needed
because different participants in the decision-making process
can/will have different viewpoints concerning the important
criteria.
Many criteria in water resources planning and management
problems are subjective in nature, so using the theory of fuzzy
logic seems appropriate. This is because the theory of fuzzy
logic is able to address subjective uncertainties rather well.
Thus, fuzzy compromise programming instead of using crisp
input parameters (i.e., ) uses fuzzy sets (i.e.,
), where the italicized word describes
subjective uncertainty. It is important to note that Fuzzy Com-
promise Programming considers all input parameters as fuzzy

360 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 33, NO. 3, MAY 2003
sets, not just criteria values (as fuzzy composite programming
does).
Advantages of adopting the fuzzy compromise programming
approach are plentiful, particularly when dealing with criteria
weights, , deviation parameter, , and positive and neg-
ative ideals. These parameters are usually provided by ex-
perts or decision makers, and are thus, inherently subjective. Use
of fuzzy sets in representation of these parameters insures that
as much as possible of relevant information is used. The more
certain the expert is (about a particular parameter value), the less
fuzziness is assigned to the fuzzy number.
However, the fuzzy approach is not without its downside. In
compromise programming (where crisp inputs are used), best
compromise solution is simply one with the smallest distance
metric value, . In fuzzy compromise programming on the
other hand, obtaining the smallest distance metric values is not
easy, because the distance metrics are also fuzzy. (This means
that distance metrics values in Fig. 1 may not have fixed length;
there may, in fact, exist a range of lengths that are somewhat
valid). To pick out a smallest fuzzy distance metric, from a
group of distance metrics, fuzzy set ranking methods have to
be used. A study by Prodanovic and Simonovic [22] compared
fuzzy set ranking methods for use in fuzzy compromise
programming, and recommended using the method of Chang
and Lee [23]. This recommendation was founded on the fact
that Chang and Lee’s [23] method gave most control in the
ranking process—with degree of membership weighting and
the weighting of the subjective type. The overall existence
ranking index (OERI) has the following mathematical form:
(2)
where the subscript stands for alternative , while represents
the degree of membership. and are the subjective type
weighting indicating neutral, optimistic or pessimistic prefer-
ences of the decision maker, with the restriction that
. Parameter is used to specify weights which are to be
given to certain degrees of membership (if any). For example,
if it is wished to have certain membership values be counted
for more than others, an equation for could be formulated
to reflect that. For this study, all degrees of membership were
weighed equally, namely . Lastly, represents
an inverse of the left part, and the inverse of the right
part of the membership function.
For values greater than 0.5, the left side of the member-
ship function is weighted more than the right side, which in
turn makes the decision maker more optimistic. Of course, if
the right side is weighted more, the decision maker is more of a
pessimist (this is because he/she prefers larger distance metric
values, which means the farther solution from the ideal solu-
tion). In summary, the risk preferences are: if , the
user is a pessimist (risk averse); if , the user is neu-
tral; and if , the user is an optimist (risk taker). Simply
stated, Chang and Lee’s [23] overall existence ranking index is
a sum of the weighted areas between the membership axis and
the left and right inverses of a fuzzy number.
B. Group Decision Making Under Fuzziness Algorithm
Kacprzyk and Nurmi [12] present a methodology which takes
in opinions of individuals concerning crisp alternatives,
and then outputs an alternative (or a set of alternatives) that are
preferred by most individuals. Each individual is required to
make a pairwise comparison between the alternatives; then a
fuzzy preference relation matrix is constructed for each expert,
results aggregated, and a group decision made.
Number of alternatives are denoted by subscripts
and number of individuals by subscript
. In order to construct a fuzzy preference
relation matrix for each individual, we must ask that person
to compare every two alternatives in the system. For example,
if there are three alternatives in the system ( and ),
the individual must compare to to , and to
, and tell us, for each comparison, what alternative he/she
prefers and to what degree. The options given to the individual
are (from Kacprzyk and Nurmi [12]):
(3)
With the restrictions above, each individual is to construct a
fuzzy preference relation matrix. For our three alternative ex-
ample, a sample matrix for individual 1 may be
(4)
Note, our individual 1 said that he/she preferred to both
and , and to , only slightly. Clearly, our individual
thinks that is the best option.
Once we obtain the fuzzy preference relation matrix from
each individual, the aggregation of the results is performed in
the following way. First, is calculated to see weather Ai de-
feats (in pairwise comparison) Aj or not .
(5)
Then, we calculate
(6)
which is the extent, from 0 to 1, to which individual is not
against alternative Aj, where 0 stands for definitely not against
and 1 stands for definitely against, through all intermediate
values.
Next, we calculate
(7)

PRODANOVIC AND SIMONOVIC: FUZZY COMPROMISE PROGRAMMING 361
which expresses to what extent, from 0 to 1, all individuals are
not against alternative Aj. Then, we compute
(8)
which represents to what extent, from 0 to 1 as before, Q (most)
individuals are not against alternative Aj. Q is a fuzzy linguistic
quantifier, (in our case meaning “most”) which is defined, after
Zadeh [24]
(9)
Lastly, the final result (fuzzy Q-core) is expressed as
(10)
and is interpreted as a fuzzy set of alternatives that are not de-
feated by Q (most) individuals.
Similarly, fuzzy /Q-core and fuzzy s/Q-core can be deter-
mined. The former is obtained by changing (5) into
r
k
ij (11)
and then performing all above steps as before. represents
a degree of defeat to which Ai defeats Aj; as such it is taken
between [0, 0.5]. The final result in this case is interpreted as
a fuzzy set of alternatives that are not sufficiently [at least to a
degree ] defeated by Q (most) individuals. The parameter
was arbitrarily chosen at 0.3. Fuzzy s/Q-core is determined by
changing (5) to
(12)
and, again, performing all above steps as before. With (12)
above, strength is introduced into the defeat (parameter
stands for strength), and the final result interpreted as a fuzzy
set of alternatives that are not strongly defeated by Q (most)
individuals.
It should be noted that (8) represents Zadeh’s [24] way to
evaluate a fuzzy linguistic quantified statements. An alternate
way to perform the same thing would be via Ordered Weighted
Averaging operators of Yager [25]. In this paper, only the
former method is implemented—noting that similar results
would be obtained via the latter methodolody as well, according
to Kacprzyk and Nurmi [12]. Also, there exists a modification
to the above algorithm which can assign a different experts
(and/or alternatives) different levels of importance. For the pur-
poses of this paper, all experts (and all alternatives) contained
an equal level of importance.
C. Combining Fuzzy Compromise Programming With Group
Decision Making Under Fuzziness
The following is a proposed algorithm for including mul-
tiple experts in a multicriteria decision-making process that uses
fuzzy compromise programming.
1) Each decision maker is to specify his/her fuzzy weights,
, deviation parameter, , as well as positive
and negative ideals concerning the criteria of the
problem. Also, experts overall degree of risk is to be
specified here as well (parameter ). It should be noted
that these parameters are entirely subjective and are
based on the preferences of the expert.
2) Then, for each expert, a set of fuzzy alternatives is gener-
ated via fuzzy compromise programming equation. This
means that the fuzzy compromise programming equation
takes in (fuzzy) criteria (for each alternative, for each
expert), and produces one (fuzzy) distance metric—one
distance metric for every alternative of the problem, for
each expert. (It should be mentioned that alternatives are
the same for each expert.)
3) After this, for each individual, a fuzzy preference relation
matrix is generated.
4) Finally, after everyone’s fuzzy preference relation matrix
is obtained, Q-core, /Q-core and s/Q-core algorithms are
performed, and a group decision is made.
An individual fuzzy preference relation matrix is obtained via
available ranking methods. Each individual’s set of alternatives
is ranked with a selected ranking method, and from the ranking
values, the fuzzy preference relation matrix is obtained. These
matrices were obtained in the following way:
First, a ranking method is called to rank the alternatives for
each expert. Then, from all the ranking values for that expert, a
difference is found for every two alternatives compared. To see
what this means, consider the following. Suppose that a ranking
method produces a vector of ranking values for each particular
alternative, that is . Then,
a difference is found for every pair of and . From
these differences in the ranking values, a fuzzy preference re-
lation matrix is constructed. Then, if is large and
negative, that means that is much more preferred than .
Therefore, a fuzzy preference relation for this pair is given a
value close to (or just less than) 1.0. Similarly, if the difference
is large and positive, meaning that is much more preferred
to , a value close to zero is assigned for that particular pair.
The IF statements implemented in the code cover all interme-
diate ranges, and thus assign values between [0, 1] within the
fuzzy preference relation matrix. However, one must be cau-
tioned when defining the meaning of small and large differences
in the ranking values, as these can have a profound effect on the
results produced by the methodology. The authors suggest that
their precise meaning [small and large differences] be defined
each time the methodology is implemented.
It is worthwhile noting that even thought group decision
making under fuzziness requires pairwise comparisons be-
tween the alternatives, the experts themselves do not have
to perform this comparison [it is embedded into the code
of the proposed methodology]. This is extremely important,
since it has been documented in the literature (Iz [26], and
Davey and Olsen [27]) that pairwise comparisons between
the alternatives are difficult and as well as burdensome for
the experts. Even though pairwise comparisons are an integral
part of the proposed methodology, they are hidden from the
decision makers—this is, in the opinion of the authors, a strong
argument in favor of the proposed methodology.

362 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 33, NO. 3, MAY 2003
TABLE I
TISZA RIVER BASIN INPUT DATA AFTER DAVID AND DUCKSTEIN [30]
TABLE II
EXPERT WEIGHTS
III. CASE STUDY
To demonstrate the usefulness of the proposed multiple
criteria multiple decision makers methodology, the Tisza
River Basin (David and Duckstein [28]; Keeney and Wood
[29]; and Goicoechea et al. [15]) case study was used.
The case study consists of five distinct alternative water
resources systems designed for meeting long-range goals.
The specifications of the problem included twelve criteria,
eight being subjective. The subjective criteria were as-
signed a value from one of the following linguistic scales:
excellent very good good fair bad very easy easy
fairly dicult dicult and very sensitive sensitive
. Table I shows the alterna-
tives, together with their corresponding criteria values. Data
from this table was fuzzified into triangular fuzzy numbers by
Bender and Simonovic [13], which was used as base data for
the case study.
For the purposes of this paper, four individuals were asked
to give their input concerning the criteria of the problem. Using
a scale from one to five, each expert was asked how important
is each criteria to him or her, with one being least important to
five being most important. These weights are shown in Table II.
Fuzzy weights were constructed from their responses, giving
everyone the same level of fuzziness. To keep things straight-
forward, the following additional simplifications were imple-
mented.
Fig. 2. Definition of a triangular fuzzy number.
1) Only triangular fuzzy numbers were used (see Fig. 2), as
these seemed to capture the essence of the data provided
by the experts.
2) Experts were asked to give only their criteria weights for
the problem, while keeping fuzzy , as well as fuzzy posi-
tive and negative ideals, constant (for all experts). In addi-
tion, each expert had an equal level of importance, along
side with a neutral risk preference—this means that each

PRODANOVIC AND SIMONOVIC: FUZZY COMPROMISE PROGRAMMING 363
Fig. 3. Expert 1 fuzzy distance metric.
Fig. 4. Expert 2 fuzzy distance metric.
expert was neither a risk taker, nor someone wishing to
avert risk.
3) Only method of Chang and Lee [23] was used to rank
fuzzy distance metrics.
Implementing the case study with above limitations proved ad-
equate for the present purpose—to show the usefulness and the
reliability of the proposed approach. It is quite clear that for
real decision-making problems, each expert should be providing
his/her deviation parameter, positive and negative ideals, as well
as a degree of relative risk. The degree of fuzziness would also
be requested from each expert, for every piece of input data pro-
vided. Of course, other shapes of fuzzy numbers (other than tri-
angular) could also be used as input data, if that is deemed ap-
propriate.
First two experts considered in the case study truly wished
that a best compromise solution be found, while the other two
were not as considerate. Third expert had a mindset of someone
who places emphasis on the protection of the environment, with
Fig. 5. Expert 3 fuzzy distance metric.
Fig. 6. Expert 4 fuzzy distance metrics.
little consideration on such issues as development possibility.
The fourth expert was exactly opposite of expert 3, and pos-
sessed strong opinions in favor of the development, and very
little concern on the protection of the environment. Such diverse
experts were chosen to simulate a conflict amongst the decision
makers, as this is usually the case in real situations.
A. Results of the Case Study
Figs. 3 to 6 show fuzzy distance metrics for each expert,
which resulted from applying the fuzzy compromise program-
ming equation. The differences in the distance metrics are due
to differences in weights provided by the experts. By applying
Chang and Lee’s [23] fuzzy set ranking method (with neutral
viewpoints) to distance metrics for each expert (see Table III), a
set of individual fuzzy preference relation matrices is generated
(see Table IV). These fuzzy preference relation matrices were
put through an algorithm of Section II-B, which produced the
following results:

364 IEEE TRANSACTIONS ON SYSTEMS, MAN, AND CYBERNETICS—PART A: SYSTEMS AND HUMANS, VOL. 33, NO. 3, MAY 2003
TABLE III
EXPERT RANKING VALUES, CHANG AND LEE [23] WITH  = 0:5
TABLE IV
INDIVIDUAL PREFERENCE RELATION MATRICES
Q-core: ; which represents the degrees
that alternatives and were not at all defeated (in
pairwise comparison).
/Q-core: ; which gives the de-
grees that alternatives and were not sufficiently de-
feated (to a degree of 0.7).
s/Q-core: ; which expresses the
degrees that alternatives and were not strongly
defeated.
Results obtained by this methodology concern only the best
compromise alternatives, or ones that were not defeated in pair-
wise comparison. As such, no information is given about the
three worst alternatives.
IV. CONCLUSION
A new multiple criteria multiple expert decision making
technique is proposed. The methodology is formulated by
merging Kacprzyk and Nurmi’s [12] group decision making
under fuzziness method with Bender and Simonovic’s [13],
[14] fuzzy compromise programming. Since fuzzy compromise
programming is a more general case of compromise program-
ming, and compromise programming is a technique that has
been proven useful in the water resources literature, it makes
sense to try to broaden it even further to allow it to include
multiple decision makers.
The methodology presented in this paper could prove useful
in situations where conflicting criteria, as well as decision
makers with conflicting interests are present. Many water
resources planning, management and operations applications
quite often involve various types of criteria (objective and
subjective), and a diverse group of stakeholders (economists,
politicians, developers, engineers, etc.). The methodology
presented in this paper is able to formulate a solution that
satisfies all parties involved.
V. RECOMMENDATIONS FOR FUTURE WORK
It is important to note that this research investigated only one
methodology that gives support to multiple decision makers in a
fuzzy environment. Others are also found in the literature, such
as Cheng [30] as well as Ghyym [31], which should be inves-
tigated as well. Techniques which employ aggregation opera-
tors (which combine preferences of many into a single group
preference) could also be effectively used and combined with
fuzzy compromise programming. It is suggested that these (and
other) methodologies be investigated so that multiple decision
makers may be included into an already powerful multicriteria
technique (fuzzy compromise programming).
ACKNOWLEDGMENT
The authors would like to thank Dr. Nirupama for her assis-
tance throughout this research, and Dr. S. Ahmad for his nu-
merous suggestions and comments. Lastly, the comments made
by the anonymous reviewers were greatly appreciated.
REFERENCES
[1] M. Borshuk, R. Clemen, L. MacGuire, and K. Reckhow, “Stakeholder
values and scientific modeling in the Neuse River watershed,” Group
Decis. Negotiation, vol. 10, pp. 355–373, 2001.
[2] R. P. Hämäläinen, E. Kettunen, and H. Ehtamo, “Evaluating a framework
for multi-stakeholder decision support in water resources management,”
Group Decis. Negotiation, vol. 10, pp. 331–353, 2001.
[3] P. A. Raj and D. N. Kumar, “Ranking multi-criterion river basin planning
alternatives using fuzzy numbers,” Fuzzy Sets Syst., vol. 100, pp. 89–99,
1998.
[4] , “Ranking alternatives with fuzzy weights using maximizing set
and minimizing set,” Fuzzy Sets Syst., vol. 105, pp. 365–375, 1999.
[5] P. J. Agrell, B. J. Lence, and A. Stam, “An interactive multicriteria de-
cision model for multipurpose reservoir management: The shellmouth
reservoir,” J. Multicriteria Decis. Anal., vol. 7, pp. 61–86, 1998.
[6] K. Fukuyama, D. M. Kilgour, and K. W. Hipel, “Systematic policy de-
velopment to ensure compliance to environmental regulations,” IEEE
Trans. Syst., Man, Cybern., vol. 24, pp. 1289–1305, Sept. 1994.
[7] Y. Y. Haimes, W. A. Hall, and H. T. Freedman, Multi-Objective Opti-
mization in Water Resources Systems. New York: Elsevier, 1975.
[8] S. Rajabi, K. W. Hipel, and D. M. Kilgour, “Multiple criteria water
supply planning,” in Proc. IEEE Int. Conf. Systems, Man, Cybernetics,
vol. 4, Orlando, FL, Oct. 1997, pp. 3484–3489.
[9] M. A. Ridgley, D. C. Penn, and L. Tran, “Multicriterion decision sup-
port for a conflict over stream diversion and land-water reallocation in
Hawaii,” Appl. Math. Comput., vol. 83, pp. 153–172, 1997.
[10] A. Tecle, B. P. Shrestha, and L. Duckstein, “A multiobjective decision
support system for multiresource forest management,” Group Decision
Negotiation, vol. 7, pp. 23–40, 1998.
[11] K. W. Hipel, D. M. Kilgour, L. Fang, and X. Peng, “The decision support
system GMCR in environmental conflict management,” Appl. Math.
Comput., vol. 83, pp. 117–152, 1997.
[12] J. Kacprzyk and S. H. Nurmi, “Group decision making under fuzziness,”
in Fuzzy Sets in Decision Analysis, Operations Research and Statistics,
R. Slowinski, Ed. Norwell, MA: Kluwer, 1998, pp. 103–136.
[13] M. Bender and S. P. Simonovic, “Fuzzy Compromise Programming,”
Univ. Manitoba, Winnipeg, MB, Canada, Water Resources Research Re-
port no. 034, 1996.
[14] , “A fuzzy compromise approach to water resource systems plan-
ning under uncertainty,” Fuzzy Sets Syst., vol. 115, pp. 35–44, 2000.
[15] A. Goicoechea, D. R. Hansen, and L. Duckstein, Multiobjective Deci-
sion Analysis with Engineering and Business Applications. New York:
Wiley, 1982.

PRODANOVIC AND SIMONOVIC: FUZZY COMPROMISE PROGRAMMING 365
[16] S. P. Simonovic, “Application of water resources systems concepts to
the formulation of a water master plan,” Water Int., vol. 15, pp. 37–51,
1988.
[17] R. J. Tkach and S. P. Simonovic, “A new approach to multi-criteria deci-
sion making in water resources,” J. Geograph. Inform. Dec. Anal., vol.
1, pp. 25–43, 1997.
[18] A. Bardossy, I. Bogardi, and L. Duckstein, “Composite programming
as an extension of compromise programming,” in Mathematics of Mul-
tiple Objective Optimization, P. Serfini, Ed. Vienna, Austria: Springer-
Verlag, 1985, pp. 375–408.
[19] Y. W. Lee, I. Bogardy, and J. Stansbury, “Fuzzy decision making in
dredged-material management,” J. Environ. Eng., vol. 117, pp. 614–630,
1991.
[20] Y. W. Lee, M. F. Dahab, and I. Bogardy, “Nitrate risk management under
uncertainty,” J. Environ. Eng., vol. 118, pp. 151–165, 1992.
[21] M. E. Hagemeister, D. D. Jones, and W. E. Woldt, “Hazard ranking of
landfills using fuzzy composite programming,” J. Environ. Eng., vol.
122, pp. 248–258, 1996.
[22] P. Prodanovic and S. P. Simonovic, “Comparison of fuzzy set ranking
methods for implementation in water resources decision-making,” Can.
J. Civil Eng., vol. 29, pp. 692–701, 2002.
[23] P. T. Chang and E. S. Lee, “Ranking of fuzzy sets based on the concept
of existence,” Comput. Math. Applicat., vol. 27, pp. 1–21, 1994.
[24] L. A. Zadeh, “A computational approach to fuzzy quantifiers in natural
languages,” Comput. Math. Applicat., vol. 9, pp. 149–184, 1983.
[25] R. R. Yager, “On ordered weighted averaging aggregation operators in
multicriteria decision-making,” IEEE Trans. Syst., Man, Cybern., vol.
18, pp. 183–190, Jan./Feb. 1988.
[26] P. H. Iz, “Two multiple criteria group decision support systems based on
mathematical programming and ranking methods,” Int. J. Manage., vol.
18, pp. 595–604, 1992.
[27] A. Davey and D. Olson, “Multiple criteria decision making models in
group decision support,” Group Decision Negotiation, vol. 7, pp. 55–75,
1998.
[28] L. David and L. Duckstein, “Multi-criterion ranking of alternative
long-range water resources systems,” Water Resources Bull., vol. 12,
pp. 731–754, 1976.
[29] R. L. Keeney and E. F. Wood, “An illustrative example of the use of
multiattribute utility theory theory for water resource planning,” Water
Resources Research, vol. 13, pp. 705–712, 1977.
[30] C. H. Cheng, “A simple fuzzy group decision making method,” in Proc.
IEEE 8th Int. Fuzzy Systems Conf., vol. 2, 1999, pp. 910–915.
[31] S. H. Ghyym, “A semi-linguistic fuzzy approach to multi-actor deci-
sion-making: Application to aggregation of experts’ judgments,” Ann.
Nuclear Energy, vol. 26, pp. 1097–1112, 1999.
Predrag Prodanovic was born in Novi Sad,
Yugoslavia, in 1978. He received the B.E.Sc. degree
in civil and environmental engineering from the Uni-
versity of Western Ontario, London, ON, Canada, in
2002 and is currently pursuing the M.E.Sc. degree at
the same university.
His research interests are in modeling and analysis
of complex systems through optimization and simu-
lation.
Mr. Prodanovic held Natural Sciences and Engi-
neering Research Council’s (NSERC) undergraduate
student research awards, in 2000 and 2001. He has received a number of other
awards dealing with research involving both, experimental, and numerical
topics, most notably a post-graduate scholarship from NSERC and an award
from the Gordon M. MacNabb Foundation. He is a student member of
Canadian Society for Civil Engineering.
Slobodan P. Simonovic was born in Belgrade, Yu-
goslavia, in 1949. He recieved the B.Sc. degree in
civil engineering and water resources engineering, in
1974, the M.Sc. degree in interdisciplinary studies
and water resources systems, in 1976, from the Uni-
versity of Belgrade, and the Ph.D. degree in engi-
neering and water resources systems from the Uni-
versity of California, Davis, in 1981.
He has over 25 years of research, teaching,
and consulting experience in water resources
engineering. He is a Professor with the Department
of Civil and Environmental Engineering, and the Engineering Research Chair
with the Institute for Catastrophic Loss Reduction, University of Western
Ontario, London, on Canada. He teaches courses in civil engineering and water
resources systems. His primary research interest focuses on the application
of systems approach, and development of the decision support tools for
management of complex water and environmental systems. Most of his work
is related to the integration of risk, reliability, uncertainty, simulation and
optimization in hydrology and water resources management. He has over 250
professional publications.
Dr. Simonovic is a member of ten national and international professional or-
ganizations. He has received a number of awards for excellence in teaching,
research, and outreach.