Sunteți pe pagina 1din 35

DECISION MAKING WITH

MULTIPLE OBJECTIVES:
ELECTRE AND TOPSIS
METHODS

STQM6034 Decision and Game Analysis


A Taxanomy of MCDM Methods
Type of Information Major Classes of
from the Decision Salient Feature of Methods
Maker Information

Dominance
No Information Maximin
Maximax

Multi-Attribute
Decision Conjunctive Method
Making (Satisfying Method)
Standard Level
Disjunctive Method
Information On
The Attributes
Elimination by Aspect
Ordinal Lexicographic Semi Order
Lexicographic Method

Chen and Hwang, “Fuzzy Multiple Weighted Sum Method


Attribute Decision Making: Methods Cardinal Weighted Product Method
and Applications”, Lecture Notes in Analytic Hierarchy Process
Economics and Mathematical ELECTRE
Systems, No. 375, Springer-Verlag, TOPSIS
Berlin, Germany
ELIMINATION AND CHOICE
TRANSLATING REALITY
(ELECTRE)

STQM6034 Decision and Game Analysis


Basic Concept
 Elimination et choix traduisant la realite
(Elimination and choice expressing reality)

 To deal with “outranking relations” by using


pairwise comparisons among alternatives under
each one of the criteria separately.
Basic Concept
 The outranking relationship of two decision
alternatives Ai and Aj, denoted by AiAj,
describes that even when the ith decision
alternative does not dominate the jth decision
alternative quantitatively, then the decision
maker may still take the risk of regarding Ai as
almost surely better than Aj.

 Decision alternatives are said to be


DOMINATED, if there is another decision
alternative which excels them in one or more
criteria and equals in the remaining criteria.
MODEL FOR ELECTRE

STQM6034 Decision and Game Analysis


Model
 Given
◦ A set of m decision alternatives A1, A2, …, Am
◦ A set of n decision criteria C1, C2, …, Cn
◦ A set of n criteria weights w1, w2, …, wn
◦ An m × n matrix A with aij performance values
(i = 1,2,…,m and j = 1,2,…,n)

Criteria
Dec C1 C2 C3 … Cn
Alts (w1 w2 w3 … wn)
A1 a11 a12 a13 … a1n
A= A2 a21 a22 a23 … a2n
: : : : : :
Am am1 am2 am3 … amn
Method
 Begins with pairwise comparisons of decision
alternatives under each criterion.

 Use physical or monetary values to denote


gi(Aj) and gj(Ak) of the decision alternatives Aj
and Ak respectively.

 Take gi(Aj) and gj(Ak) and by decision


alternatives, a CONCORDANCE INDEX and a
DISCORDANCE INDEX are obtained.
The Indices
 A CONCORDANCE INDEX is defined as the
amount of evidence to support the conclusion
that alternative Aj outranks, or dominates,
alternative Ak.

 A DISCORDANCE INDEX is defined as the


amount of evidence to support the conclusion
that alternative Ak outranks, or dominates,
alternative Aj.

Discordance Index is the counterpart of Concordance Index


Method continued
 Finally, the method yields a system of binary
outranking relations between the decision
alternatives.
SOME NOTES

STQM6034 Decision and Game Analysis


Advantages
1. The method has a clearer view of decision
alternatives by eliminating less favorable
ones.

2. The method is especially convenient when


there are decision problems that involve a few
criteria with a large number of decision
alternatives.
Disadvantages
1. The system is not necessarily complete,
therefore ELECTRE is sometimes unable to
identify the most preferred decision
alternative.

2. Only a core of leading decision alternatives is


produced.
Variants of ELECTRE
 ELECTRE I
 ELECTRE II
 ELECTRE III
 ELECTRE IV
 ELECTRE IS

 Self-study:
 Find how these variants are different from
each other.
 Check for conditions where these variants
are suitable to use in.
STEPS

STQM6034 Decision and Game Analysis


Step 1: Normalizing the
Decision Matrix
 Transform the entries of the decision matrix A
into dimensionless comparable entries by using
the following equation:

aij
xij =
√ k=1 akj
m 2
Step 1: Normalizing the
Decision Matrix
 Therefore the normalized matrix X is defined
as follows:

x11 x12 x13 … x1n


x21 x22 x23 … x2n
X= : : : : :
xm1 xm2 xm3 … xmn

where m is the number of decision alternatives,


n is the number of criteria, and xij is the
normalized preference measure of the ith
decision alternative in terms of the jth criterion.
Step 2: Weighting the
Normalized Decision Matrix

 Each one of the columns of the previous X


matrix is multiplied by the associated weight of
importance of the corresponding decision
criterion.

 The weights, (w1, w2, …, wn) are determined


by the decision maker.
Step 2: Weighting the
Normalized Decision Matrix
 Therefore the weighted matrix, Y is
Y=XW

y11 y12 y13 … y1n


y21 y22 y23 … y2n
Y= : : : : :
ym1 ym2 ym3 … ymn
Step 2: Weighting the
Normalized Decision Matrix

w1x11 w2x12 w3x13 … wnx1n


w1x21 w2x22 w3x23 … wnx2n
Y= : : : : :
w1xm1 w2xm2 w3xm3 … wnxmn

where w1 0 0 … 0 , and
0 w2 0 … 0 n
W= : : : : : wi = 1
0 0 0 … wn i=1
Step 3a: Determining the
Concordance Set

 The concordance set Ckl of two decision


alternatives Ak and Al, where m ≥ k, l ≥ 1, is
defined as the set of all the criteria for which
Ak is preferred to Al.

i.e:
Ckl = {j, ykj ≥ yij}, for j = 1, 2, 3, …, n
Step 3b: Determining the
Discordance Set

 The discordance set Dkl of two decision


alternatives Ak and Al, where m ≥ k, l ≥ 1, is
defined as the set of all the criteria for which
Ak is NOT preferred to Al.

i.e:
Dkl = {j, ykj < yij}, for j = 1, 2, 3, …, n
Step 4a: Construct the
Concordance Matrix
 The relative value of the elements in the
concordance matrix C is calculated by means
of the concordance index.

 The concordance index ckl is the sum of the


weights associated with the criteria contained
in the concordance set.

i.e:
ckl = wj= 1 , for j = 1, 2, 3, …, n
j Ckl
Step 4a: Construct the
Concordance Matrix
 The concordance index ckl indicates the relative
importance of decision alternative Ak with
respect to decision alternative Al.

 The concordance matrix C is defined as


follows:
- c12 c13 … c1n
c21 - c23 … c2n
C= : : : : :
cm1 cm2 cm3 … -
Step 4b: Construct the
Discordance Matrix
 The discordance matrix D expresses the degree
that a certain decision alternative Ak is worse
than a competing decision alternative Al.

 The elements dkl of the discordance matrix is


defined as
max | ykj – ylj |
j Dkl
dkl = , for j = 1, 2, 3, …, n
max | ykj – ylj |
j
Step 4b: Construct the
Discordance Matrix

 The discordance matrix D is defined as follows:

- d12 d13 … d1n


d21 - d23 … d2n
D= : : : : :
dm1 dm2 dm3 … -
Step 5a: Determine the
Concordance Dominance Matrix
 The concordance dominance matrix is
constructed by means of a threshold value
for the concordance index.

 A decision alternative Ak will only have a


chance to dominate decision alternative Al,
if its corresponding concordance index ckl
exceeds at least a certain threshold value c.
i.e:
ckl ≥ c
Step 5a: Determine the
Concordance Dominance Matrix
 The threshold value c can be determined as
the average concordance index.
i.e:

1 m m

c=
m(m – 1)
 c
k=1 l=1
kl

and and
kl lk
Step 5a: Determine the
Concordance Dominance Matrix
 Based on the threshold value c , the
elements of the concordance dominance
matrix F are determined as follows:

fkl = 1, if ckl ≥ c
fkl = 0, if ckl < c
Step 5b: Determine the
Discordance Dominance Matrix
 The discordance dominance matrix is
constructed by means of a threshold value
d for the discordance index.

 The threshold value d can be determined as


the average discordance index.
i.e:
1 m m

d=
m(m – 1)
 d
k=1 l=1
kl

and and
kl lk
Step 5b: Determine the
Discordance Dominance Matrix
 Based on the threshold value d , the
elements of the discordance dominance
matrix G are determined as follows:

gkl = 1, if dkl ≥ d
gkl = 0, if dkl < d
Step 6: Determine the
Aggregate Dominance Matrix

 The elements of the aggregate dominance


matrix E are defined as:

ekl = fkl × gkl


Step 7: Eliminate the Less
Favorable Decision Alternatives

 From the aggregate dominance matrix E, a


partial preference ordering of the decision
alternatives can be derived.

 If ekl = 1, this means decision alternative Ak


is preferred to decision alternative Al, by
using BOTH concordance and discordance
criteria.
Step 7: Eliminate the Less
Favorable Decision Alternatives
 If any column of the aggregate dominance
matrix E has at least one element equal to 1,
then this column is “ELECTREally” dominated by
the corresponding row.

 Therefore, a decision maker can simply


eliminate any column(s) which have an element
equal to 1.

 Hence, the best decision alternative is the one


which dominates all other decision alternatives
in this manner.
Exercise: Assignment 4
 Find an article from any journal that uses ELECTRE
in a real problem.

 Do a critical review on the article. For example:


1. Report on the problem statement and objective.

2. Suitability of the method to be used with the


particular problem.
3. If data is given, use Excel or any other suitable
program to check the results.
4. Report calculation errors (if any).

 Submit latest by 1st April, 2010 to my room


BR2126.

S-ar putea să vă placă și