Anda di halaman 1dari 5

IPASJ International Journal of Computer Science(IIJCS)

Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm


A Publisher for Research Motivation ........ Email: editoriijcs@ipasj.org
Volume 1, Issue 1, June 2013 ISSN 2321-5992


Volume 1, Issue 1, June 2013 Page 1


ABSTRACT
Linear weighted method converts the multi-objective programming problems to single objective programming problems, and
we can get Pareto optimal solutions. But there isnt any comparison between these Pareto optimal solutions, so its difficulty
for decision makers to select a satisfied solution. Based on this idea, the concept of trade off rate between different objective
functions is proposed. It can show that one objective function decreases one unit, and how many units of another objective
function increases. Two examples are given to illustrate the significance of trade off rate for decision makers selecting satisfied
solution.
Keywords: weights, multi-objective, Pareto optimal solution
1. INTRODUCTION
In multi-objective programming problems, the decision maker searches a compromising solution, instead of finding an
optimal solution of every objective function. Based on this idea, the concepts of Pareto optimal solution is introduced
into multi-objective programming problems [1]. There are various approaches to solve multi-objective programming
problems. These methods generally convert the multi-objective programming problems to single objective
programming problems, and some Pareto optimal solutions can be got by solving corresponding single objective
programming problems. In order to provide solutions satisfied the decision makers demand, many methods involve
interactive process with decision makers.
B.Y.H. Wong, M. Luque and J. Yang establish an equivalence model between data envelopment analysis (DEA) and
multiple objective linear programming (MOLP) and show how a DEA problem can be solved interactively without any
prior judgements by transforming it into a MOLP formulation. The decision makers (DM) can then search along the
efficient frontier to locate the most preferred solution where resource allocation and target levels based on the DM
value judgements can be set [2]. N. Xiao and D.A. Bennett create a unified conceptual framework that can be used to
address a broad range of multiobjective spatial decision problems. In this framework, evolutionary algorithms are
employed to generate optimal, or near optimal, solutions to a problem being addressed. Alternatives created are then
displayed in an interactive visual support system that can be used by decision makers to discover the competing nature
of multiple objectives and to gain knowledge about the tradeoffs among alternatives [3]. M. Luque, F. Ruiz, and R.E.
Steuer describe an interactive procedural algorithm for convex multiobjective programming. At each iteration, DM
have the option of expressing his objective-function aspirations in the form of a reference criterion vector. Also, the
DM has the option of expressing minimally acceptable values for each of the objectives in the form of a reservation
vector [4].
L. Qin, Y. Liu and Z. Lan put up an interactive decision-making support model to improve the efficiency of analyzing
and evaluation the Pareto optimal solutions. This model is composed of 3 pivotal parts: geographic brush mechanism,
similarity querying operators as well as interactive searching method. And they design a prototype system on the base
of the model [5]. F. Lu, D. Yan and M. Wang propose a new interactive decision-making method for nonlinear multi-
objective programming. In the method, the distance between objective function and ideal solution is denoted by
Tchhebycheff distance, and the mini-max decision rule and multi-agent genetic algorithm are applied to find the set of
preferred solutions. The preference of decision makers such as expected objective function and weight can be gradually
included to narrow down the final solution [6]. G. Dong and L. Shu present a new improved method of interactive
multi-objective decision-making based on extreme value. The function of the single-objective extreme value and
interactive decision-making model are introduced. It can reflect the wish of the decision makers, realize the interaction
between the decision-maker and decision-making model, and avoid the difficulty determining the single-object or
Weights Analysis of Multi-objective
Programming Problem

Hua Zuo

and Guoli Zhang



Information Processing and Control Institute, North China Electric Power University, Baoding 071003, China

IPASJ International Journal of Computer Science(IIJCS)
Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm
A Publisher for Research Motivation ........ Email: editoriijcs@ipasj.org
Volume 1, Issue 1, June 2013 ISSN 2321-5992


Volume 1, Issue 1, June 2013 Page 2

single-objective satisfaction degrees weight coefficients during the conversion from multi-object to single-object [7]. X.
Zhang present an improved interactive multi-objective decision-making method, which can reflect the wish of the
policy-maker, carry out the information switching with the subsystems and avoid disadvantage from being luck of
experience of some weighted coefficients [8]. W. Wang and X. Wang implement the interior point algorithm into the
interactive method to solve MOLP problems. The utility function is taken as weighted sum of the objective functions,
whose coefficients depend on the DMs reference and is updated successively. The DM will be asked to provide the
locally relevant preference information when the iteration is still on the process rather than the optimal solution has
been calculated with the current coefficient. Thus the weighting vector can be updated and the interior path can be
adjusted in time [9].
While these methods include interactive process with decision makers, any criteria hasnt been proposed to distinguish
Pareto optimal solutions generated.
In linear weighted method, multi-objective programming problems are converted to single objective programming
problems, and Pareto optimal solutions can be got. When changing the weights, different Pareto optimal solutions can
be got. As we known, the improvement of one objective function needs the sacrifice of another objective function, but
what extent the sacrifice is not clear. So we introduce trade off rate and give a graph to explain it. In practical
application, the graph is provided to the decision makers to measure whether it is worthy to improve one objective
function at the cost of the sacrifice of another objective function.
2. PRELIMINARY KNOWLEDGE
The general multi-objective programming problem can be formulated as the form below

1 2
min{ ( ), ( ), , ( )}
k
x S
f x f x f x
e
(1)
where x is an n dimensional vector of decision variables,
1 2
( ), ( ), , ( )
k
f x f x f x are k distinct objective functions and
S is the constrained set defined by { ( ) 0, 1,2, , }
n
j
S x R g x j m = e s = .
Definition 1.
*
x S e is said to be a Pareto optimal solution of problem (1), if and only if there does not exist another
x S e such that
*
( ) ( ), 1,2, ,
i i
f x f x i k s = , with strict inequality holding for at least one i .

3. INTERACTIVE LINEAR WEIGHTED METHOD
Linear weighted method is one of the most widespread methods to solve multi-objective programming problems, and it
can convert the original problem (1) to the following form

1 1 2 2
min{ ( ) ( ) ( )}
k k
x S
w f x w f x w f x
e
+ + + (2)
where
i
w represents the importance of corresponding objective function for decision makers, and
1
1, 0
k
i i
i
w w
=
= >

.
According to the literature [1], Pareto optimal solutions of original multi-objective programming problem can be got by
solving corresponding single objective programming problem (2). When weights change, different Pareto optimal
solutions can be got. So a series of Pareto optimal solutions may be provided to decision makers. However, it is difficult
to make an ideal option, because there is not any criterion to judge which Pareto optimal solution is more superior.
We just consider the special condition that there are only two objective functions, thus problem (2) is simplified to the
form below

1 1 2 2
min ( ) ( )
x S
w f x w f x
e
+ (3)
where
1 2
1 w w + = .
We know that if the weight
1
w increases, the value of
1
( ) f x decreases, and the value of
2
( ) f x increases.
1
( ) f x
decreases one unit, and how many units of
2
( ) f x will increase. Discussion will be unfolded in the following. First, the
concept of trade off rate is given.
IPASJ International Journal of Computer Science(IIJCS)
Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm
A Publisher for Research Motivation ........ Email: editoriijcs@ipasj.org
Volume 1, Issue 1, June 2013 ISSN 2321-5992


Volume 1, Issue 1, June 2013 Page 3

Definition 2. Suppose
1
w obtains the value
*
w , and
*
x is the optimal solution of corresponding problem (3). Then
1
w
increases to
** * **
( ) w w w < , and
**
x is the optimal solution of corresponding problem (3).
*
x and
**
x are all Pareto
optimal solutions.
** *
1 1 1
** *
2 2 2
( ) ( )
( ) ( )
f f x f x
f f x f x
A
=
A
is called trade off rate from
*
w to
**
w .
With the change of weights, a series of
1
2
f
f
A
A
can be got, which help us draw a graph. Because one objective function
increases and another objective function decreases,
1
2
f
f
A
A
is constant negative value. From the graph, decision makers
can judge whether it is worthy to improve one objective function at the cost of sacrificing another objective function and
what extent of sacrifice is reasonable.

4. ILLUSTRATE EXAMPLES
Consider the multi-objective programming problem

2 2 2 2
1 2 1 2 1 2
1 2
min{ 4 4 , }
. . 2 4
x x x x x x
s t x x
+ +
+ s
(4)
For convenience, denote that
2 2 2 2
1 1 2 1 2 2 1 2
4 4 , f x x x x f x x = + + = .
By linear weighted method, problem (4) is converted into single objective programming problem (5).

2 2 2 2
1 1 2 1 2 2 1 2
1 2
min ( 4 4 ) ( )
. . 2 4
w x x x x w x x
s t x x
+ + +
+ s
(5)
Changing weight
1
w , different Pareto optimal solutions can be got. The weight
1
w increases from 0.05 to 0.95 with
step 0.05, and record corresponding to the value of trade off rate
1
2
f
f
A
A
. For the convenience of making and viewing the
graph, use
1
2
f
f
A
A
to replace
1
2
f
f
A
A
. Based on these data, we draw the following the changing graph of trade off rate with
weight.


Figure 1 The changing graph of trade off rate with weight

The abscissa represents weight
1
w , and the ordinate represents
1
2
f
f
A
A
.
IPASJ International Journal of Computer Science(IIJCS)
Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm
A Publisher for Research Motivation ........ Email: editoriijcs@ipasj.org
Volume 1, Issue 1, June 2013 ISSN 2321-5992


Volume 1, Issue 1, June 2013 Page 4

From the graph, we can see that at the beginning,
1
f decreases one unit, and
2
f increases less than one-tweifth unit.
If
1
w obtains around 0.55,
1
f decreases one unit, and
2
f increases the same unit. If
1
w obtains over 0.55,
1
f decreases
one unit, and
2
f increases more than one unit. And the bigger weight
1
w increases, the more
2
f increases.
In practical application, the graph above should be provided to decision makers to help them judge whether it is
worthy to improve one objective function at the cost of sacrificing another objective function and what extent of
sacrifice is reasonable.
Now, see another example.

2
1 2 1 2
2
1 2
1 2
2
min{ 2 ,2 }
0
. . 20
0
x x x x
x x
s t x x
x
+
>

+ s

>


(6)
For convenience, denote that
2
1 1 2 2 1 2
2 , 2 f x x f x x = + = .
By linear weighted method, the problem (4) is converted into single objective programming problem (7)

2
1 1 2 2 1 2
2
1 2
1 2
2
min ( 2 ) (2 )
0
. . 20
0
w x x w x x
x x
s t x x
x
+ +
>

+ s

>


(7)
The weight
1
w increases from 0.05 to 0.95 with step 0.05, and record corresponding to the value of trade off rate
1
2
f
f
A
A
, and the following graph is drawn based on these data.

Figure 2 The changing graph of trade off rate with weight

The abscissa represents weight
1
w , and the ordinate represents
1
2
f
f
A
A
.
From the graph, we can see that at the beginning,
1
f decreases one unit, and
2
f increases less than one-sixteenth
unit. If
1
w obtains around 0.5,
1
f decreases one unit, and
2
f increases the same unit. If
1
w obtains over 0.5,
1
f
decreases one unit, and
2
f increases more than one unit. And the bigger weight
1
w increases, the more
2
f increases.
But we should notice that when
1
w is more than 0.75, the solution of problem (7) doesnt change any more, which
means no new Pareto optimal solutions can be got.
Then, the graph is offered to decision makers to judge how many units of
2
f is sacrificed in exchange of one unit
improvement of
1
f . And then choose the reasonable weights to get Pareto optimal solution what they want.

IPASJ International Journal of Computer Science(IIJCS)
Web Site: http://www.ipasj.org/IIJCS/IIJCS.htm
A Publisher for Research Motivation ........ Email: editoriijcs@ipasj.org
Volume 1, Issue 1, June 2013 ISSN 2321-5992


Volume 1, Issue 1, June 2013 Page 5

5. CONCLUSION
In linear weighted method, when weights change, new Pareto optimal solution can be got, and accordingly, one
objective function is improved at the cost of a degradation in another objective function. The concept of trade off rate is
proposed. The changing graph of trade off rate with weight shows that one objective function decreases one unit, and
how many units of another objective function increases. Here, we only consider the case of two objective functions
involving. If there are more than two objective functions, how to define trade off rate will be discussed in another paper.

Acknowledgement
This work is supported by the Natural Science Foundation of Hebei Province under Grand A2012502061.
REFERENCES
[1] Y. Hu., Effectiveness Theory of Multi-objective Programming, Shanghai Science and Technology Press, Shanghai,
1994.
[2] B.Y.H. Wong, M. Luque and J. Yang, Using Interactive Multiobjective methods to solve DEA problems with
value judgements, Computers & Operations Research, Vol. 36, pp. 623-636, 2009
[3] N. Xiao, D.A. Bennett, and M.P. Armstrong, Interactive Evolutionary approaches to multiobjective spatial
decision making: A synthetic review, Computer, Environment and Urban Systems, Vol. 31, pp. 232-252, 2007
[4] M. Luque, F. Ruiz, and R.E. Steuer, Modified Interactive Chebyshev Algorithm for Convex multiobjective
programming, European Journal of Operational Research, Vol. 204, pp. 557-564, 2010
[5] L. Qin, Y. Liu and Z. Lan, Interactive Decision-making support model in MOSD, Science of Surveying and
Mapping, Vol. 36, No. 2, pp. 152-155, Mar. 2011
[6] F. Lu, D. Yan and M. Wang, Application of Interactive Multo-objective Decision-making Method to Hydraulic
Project Planning, Water Resource and Power, Vol. 28, No. 8, pp. 121-122 & 145, Aug. 2010
[7] G. Dong, L. Shu, N. Chen and J. Tian, Method of interactive multi-objective decision-making based on extreme
value, Journal of Liaoning Technical University (Natural Science), Vol. 27, No. 1, pp. 137-140, Feb. 2008
[8] X. Zhang, An improved Method of Interactive Multi-objective Decision-making Based on Objective Satisfaction
Degree, Systems Engineering, Vol. 22, No. 9, Sep. 2004
[9] W. Wang and X. Wang, An Interactive Interior Algorithm for multiobjective linear programming based on
weighting utility function, Journal of Guizhou University (Natural Science), Vol. 2, No. 1, pp. 30-35, Feb. 2004

AUTHOR
Hua Zuo received the B.S. and M.S. degrees in HeiBei Normal University and North China Electric Power
University in 2010 and 2013, respectively. Her research interests include Pareto optimal solution of multi-
objective programming and fuzzy multi-objective programming.

Guoli Zhang received the B.S. and M.S. degrees in Mathematics Science, from Jilin University and Hebei
University in 1982 and 1988, respectively. He received Ph.D. degree in Electrical engineering, from North
China Electric Power University in 2006. Since 1999 he has working as a professor in School of Mathematics
and Physics at North China Electric Power University. His research interests include intelligent computation
theory and application in electric power system.

Anda mungkin juga menyukai