Sie sind auf Seite 1von 9

3.

MULTI-CRITERIA DECISION ANALYSIS

3.1 Definition :

Multipe criteria decision making (MCDM) or multiple criteria decision analysis (MCDA) is a sub-
discipline of operations research that explicitly evaluates different conflicting criteria when it comes to
decision making either in daily life or settings like business or medicine. Conflicting criteria are typical
in evaluating options : cost or price is usually one of the main criteria and another criterion that is related
to quality.
Therefore, the decision-maker is faced with several possible scenarios, and he defines a set of criteria
for judging those scenarios.
It is important to determine the structure of the problem and explicitly evaluate all the various criteria.
For example, in building a nuclear power plant, certain decisions have been taken based on different
criteria. There are not only very complex issues involving multi criteria, some criteria may have effect
toward some problem, but over all to have an optimum solution, all the alternatives must have common
criteria which clearly lead to more informed and better decisions [3].

Multi criteria deision making steps :


At first, the potential alternatives are identified. In a second step, the preferences of the decision maker
are modeled, which will differentiate potential actions. Finally, a process of mathematical calculations
is launched. In other words, an aggregation procedure, is defined or chosen in order to provide elements
of the answer to the decision problem [24].

- Identification of potential alternatives:


The identification of potential alternatives consists in defining all the alternatives that need to be
examined. This set of alternatives is not always definitively delimited. It may be subject to revisions.
Alernatives may appear or disappear.

- Modeling the decision maker preferences:


Comparing alternatives from their consequences is often difficult because of inaccuracy, uncertainty
and bad determination of these consequences. This is why the notion of criterion is introduced. The
criteria make it possible to compare the alternatives with each other. A criterion is a function C with real
values defined on all the potential alternatives. This function is defined so that two alternatives a1 and
a2 can be compared based on the numbers C (a1) and C (a2). Decision maker preferences can be
represented by a set of criteria. A weight can also be associated with each criterion, depending on the
importance of the criterion in relation to others. To ensure a good representation of these preferences,
the set of criteria must meet the requirements of comprehensiveness, cohesion and non-redundancy.
When these requirements are verified, the family of criteria is said to be consistent. The evaluation of
alternatives is done on the family of criteria. A table which presents the results of this evaluation on
each criterion gives a view of all the alternatives. The criteria are on columns and alternatives are on the
lines. The box Cj (ai) gives the performance of the alternative ai according to criterion Cj.
Fig 9 : Evaluation matrix alternatives/Criteria

- The aggregation procedure:


A multi-criteria aggregation procedure must be defined in order to answer the decision-making problem
posed by relying on the performance chart which characterizes the potential alternatives to be evaluated.

3.2. Methods :

A very large part of the decision problems is characterized by diverse points of view that are often
contradictory and that measure things of different nature. The resolution of this type of problem must
take into account simultaneously from all points of view deemed relevant by the decision maker. The
Research into multicriteria decision support aims to develop more or less formalized models with a
view to improving, facilitating and to assist the manager in the decision-making process.
There are many MCDA methods, and the most commonly used ones are : AHP, TOPSIS,
ELECTRE, SkyRule and below is the explanation of each one.
AHP is an analytical hierarchy method. This method is both powerful and flexible way of
implementing and allocating resources for product portfolio management. It was developed by
Thomas Saaty in 1970 . It allows to decompose a complex problem into a hierarchical system, in
which binary combinations are established at each level hierarchy. This method is very useful for
determining relative weights to criteria. By classifying hierarchically the situations that the company
encounters, the decision-maker can deduce relative priorities, synthesize them more easily and use
them to efficiently allocate resources and / or define the priority objectives in a better coherence .
Classification is done in three levels, namely, the environment, the objectives of the the company and
the courses of activities with which different criteria are associated. Thus, it is possible to determine
the most relevant alternative, depending on the priority given to each of the criteria taken into
consideration [20].
TOPSIS (technique for Order of Preferene by Similarity to Ideal Solution) is a method whose
purpose is to be able to classify in order of choice a number of alternatives based on a set of favorable
or unfavorable criteria. This method is a part of the techniques used in the area of multicriteria
decision support MCDM. It has been developed by Hwang and Yoon in 1981 [27] . Its principle is to
determine for each alternative a coefficient between 0 and 1 on the basis of distances (Euclidean)
between each alternative on the one hand and the ideal solutions. This method will be explained in
depth on this chapter to follow step by step. An alternative is called 'ideal favorable' if it is the farthest
from the worst alternative and the closest to the best alternative. An alternative is said 'Unfavorable
ideal' if it is closest to the worst alternative and furthest from the best alternative [27].
Electre (Elimination et choix Traduisant la Réalité) is a family of multicriteria analysis methods
developed in Europe, ELECTRE I was developed by Bernard Roy in 1968 , with the help of P.
BERTIER [23].

Bernard Roy is widely recognized as the father of the ELECTRE method, which was one of the
earliest approaches in what is sometimes known as the French School of decision making. It is usually
classified as an “outranking method” of decision making.
There are two main parts to an ELECTRE application: first, the construction of one or several
outranking relations, which aims at comparing in a comprehensive way each pair of actions; second,
an exploitation procedure that elaborates on the recommendations obtained in the first phase. The
nature of the recommendation depends on the problem being addressed: choosing, ranking or sorting.
Usually the Electre Methods are used to discard some alternatives to the problem, which are
unacceptable [23] .

Skyrule is a method whose purpose is to extract un-dominated rules from a relational table. This
method is part of the techniques used in the field of multicriteria decision support. It has been
developed by Slim Bouker in 2015. A rule r is said to be dominated by another rule r ', if for all
measurements, r is less relevant than r ' [22].
The figure 12 shows the percentage of the widely applied MCDM techniques, and TOPSIS is the
most used one.
Fig 12 : Multicriteria decision support methods and its contributions [22]

Comparative analysis :

Most multicriteria decision support methods belong to one of the following two operational
approaches (Roy, 1985) [23] :
In the first approach: American inspired, local preferences (level of each attribute) are aggregated into
a function (value, utility) unique then it is to optimize. Work on multicriteria methods belonging to this
approach study the conditions of aggregation, the particular forms of aggregate function and
construction methods of these functions (both local and global). The main methods belonging to this
approach are: MAUT, SMART, UTA, TOPSIS, AHP and G. P.
The second approach: inspired French, it aims at first to build binary relationships, called outranking
relationships, to represent the preferences of decision makers, given the information available.
In some of the multicriteria methods that follow this path, before build these outranking relationships,
we introduce discrimination thresholds (indifference, preference) and even veto, at the level of each of
the criteria, to locally model the decision maker's preferences. These relationships do not are, in general,
neither transitive nor complete. In a second step, these relations are used to help formulate a
recommendation that can be to answer the decision problem. The wording is formulated in taking into
account the chosen decision-making problem. Indeed, help to decide not necessarily or only solve the
problem of choosing the best solution. Decision support can relate to other issues than that of choice
(Bana e Costa, 1996) . This approach contains methods who may not have a very good axiomatic basis
but who are imprints of a good pragmatic realist given the decision-making contexts frequently
encountered. The main methods or families of methods belonging to this approach are: ELECTRE,
PROMETHÉE, ORESTE, QUALIFLEX, some of these methods are purely ordinal [20].
Fig 13 : The most commonly used multicriteria decision support methods [29]

The figure 13 shows the percentage of each MCDA method in terms of the use.

3.3. Description of the selected method :

TOPSIS (Technique for Order Preference by Similarity to Ideal Solution) is a multi-attribute decision
making method to identify solutions from a finite set of alternatives. TOPSIS method chooses an
alternative, which respectively has the shortest and highest distance from the positive and negative
ideal solutions. The positive and negative ideal solutions are respectively composed of the best and
worst performance values. Steps of the TOPSIS are as follow: Formation of the decision matrix,
Normalization of the decision matrix, Calculation of the weighted normalized decision matrix,
Determination of the positive and negative ideal solutions, Calculation of the separation measures of
each alternative from the ideal solutions using the m-dimensional Euclidean distance, Calculation of
the relative closeness of each alternative to ideal solutions and rank of alternatives [25].
This method considers three types of attributes or criteria [28]:
-Qualitative benefit attributes/criteria
-Quantitative benefit attributes
-Cost attributes or criteria

In this method two artificial alternatives are hypothesized [28] :


- Ideal alternative : The one which has the best level for all attributes considered .
- Negative ideal alternative : The one which has the worst attribute values.
TOPSIS selects the alternative that is the closest to the ideal solution and farthest from negative ideal
alternative, each criterion can be taken into consideration in making a final ranking.
The concept of TOPSIS is rational and the computation involved is easy, it allows objective weights to
be incorporated into the comparison process.
TOPSIS method while determining « ideal » and « anti-ideal » solutions computes the weighted
distances to measure the relative distances away from the ideal and anti-ideal solutions for each
alternative (i.e., decision rules).
Not only the best alternative should be as close as possible to the ideal solution but also it should be as
far as possible away from the anti-ideal solution.

Fig 14 : Idea of TOPSIS method [28]

Components of TOPSIS :

TOPSIS is composed of two components [28] :


- Weights
- Distances
The most commonly used weights
- Mean weight
- Standard deviation weight
Among the most commonly used distances is Euclidean distance.

Positive and negative ideal solution :

•Positive ideal solution [25]:


–Maximizes the benefit criteria and minimizes the cost criteria
–Best values attainable from the criteria
•Negative ideal solution [25]:
–Maximizes the cost criteria and minimizes the benefit criteria
–Worst values attainable from the criteria

TOPSIS assumes that we have m alternatives (options) and n attributes/criteria and we have the score
of each option with respect to each criterion.
Let Xij score of option/alternative i with respect to criterion j. We have a matrix X = (xij) mxn matrix.
Let J be the set of benefit attributes or criteria (more is better)
Let J’ be the set of negative attributes or criteria (less is better) [26]

Fig 15 : Decision matrix for ranking [26]


STEPS OF TOPSIS [38] :

1- Construct normalized decision matrix :


this step transforms various attrivure dimensions into non-dimensional attributes, which allows
comparisons across criteria.
Then normalize scores or data as follows :
rij = xij /( ∑ x²ij) for i=1,…,m; j=1,…,n (1)

Fig 16 : Normalized decision matrix

2- Construct the weighted normalized decision matrix :


Assume we have a set of weights for each criteria wj for j = 1,…n and then multiply each column of
the normalized decision matrix by its associated weight.
An element of the new matrix is: Vij = Wj rij (2)
Fig 17 : Weighted normalized decision matrix

3- Determine the ideal and negative ideal solutions :

•Ideal solution :
A* = { v1*, …, vn*} (3)
Where :
vj*={ max (vij) if j ∈ J ; min (vij) if j ∈ J’ } (4)

•Negative ideal solution :


A' = { v1', …,vn' } (5)
Where :
v' = {min (vij) if j ∈ J ; max (vij) if j∈ J’} (6)

4- Calculate the separation measure for each alternative :


•The separation from the ideal alternative is:
Si *= [ ∑ (vj*–vij)2 ] ½ i= 1, …, m (7)

•Similarly, the separation from the negative ideal alternative is:


S'i= [ ∑ (vj'–vij)2 ] ½ i= 1, …, m (8)

5- Calculate the relative closeness to the ideal solution Ci*


Ci*= S'i/ (Si*+S'i) , 0 < Ci* <1 (9)
Select the option with Ci* closest to 1 (Cj*index value lies between 0 and 1)
The larger the index value means the better the performance of the alternatives

Das könnte Ihnen auch gefallen