Sie sind auf Seite 1von 74

p 


   

@   



 
  
Multi Criteria Decision making
O =    
 =  == is a systematic
procedure for transforming complex decision problems by
a sequence of transparent steps assists the decision
maker in arriving at a rational decision.


 

‡ MCDM enables multiple stakeholder preferences to be
modeled
‡ MCDM offers improved coordination and collaboration
‡ MCDM can be implemented to integrate spatial
information
`ow does MCDM works

O 2 Multi Criteria
Decision Making
method for 2 powerful and
complicated and understandable
unstructured methodology that allows
problems. groups or individuals to
O 2n approach that uses combine qualitative and
a hierarchical model quantitative factors in
having levels of goal, decision making process
criteria, possible sub-
criteria, and
alternatives.

Analytic hierarchy process (AHP) was proposed by Saaty (1977,


1980) to model subjective decision-making processes in a
hierarchical system.
he heoretical Foundation of the
2`
ased on pairwise comparison

2nalyzes the criteria + alternatives in a structured manner and ranks


alternatives
2ssumes decision maker can provide paired comparisons ± based on
knowledge, intuition and decision maker never judges one alternative
infinitely over other.
2ll decision problems are considered as a hierarchical structure in the
2`
he applications of 2` can refer to corporate planning, portfolio
selection, and benefit/cost analysis by government agencies for
resource allocation purposes.
he general form of the 2`
he main four steps of the 2`

!  Set up the hierarchical system by decomposing the


problem into a hierarchy of interrelated elements.

!  Compare the comparative weights between the


attributes of the decision elements to form the reciprocal matrix.

!  Synthesize the individual subjective judgments and


estimate the relative weights.

!  Aggregate the relative weights of the decision elements


to determine the best alternatives/strategies.
rimary Scales of Measurement
Scale
Nominal Numbers Finish
2ssigned
M 
to Runners

Ordinal Rank Order Finish


of Winners
hird Second First
place place place

Interval erformance
Rating on a .2 9.1 9.6
0 to 10 Scale

Ratio ime to 15.2 14.1 1 .4


Finish, in
Seconds
2 Classification of Scaling echniques

Scaling echniques

Comparative Noncomparative
Scales Scales

aired Rank Constant Q-Sort and Continuous Itemized


Comparison Order Sum Other Rating Scales Rating Scales
rocedures

Semantic Stapel
Likert
Differential
aired Comparison
o make tradeoffs among the many objectives and
many criteria, the judgments that are usually made in
qualitative terms are expressed numerically. o do this,
rather than simply assigning a score out of a person¶s
memory that appears reasonable, one must make
reciprocal pair wise comparisons in a carefully designed
scientific way.
Õ  ! 
1 Equal Importance
3 Moderate importance of one over
another
5 Strong or essential importance
7 Very strong or demonstrated
importance
9 Extreme importance
2,4,6,8 Intermediate values

Îse reciprocals for inverse comparisons


Comparison Matrix
ü  

     
    
  
     
  
        
 
    
 
     

               

    


  
 
      

        
 
    
 

    
 

        


       
     

  
 
   
    
 
    
     
 

    
 
  

    
  

     

2 2 2 2 

Π      !"   


!
 
  

 S1/S1 S1/S2 S1/S

 S2 / S1 S2 / S2 S2 / S

 S / S1 S / S2 S /S
airwise Comparisons
!
  
! 
    
  
    
   
        
 
    
 
  

#
 
   

        

#&!
       

 
    
  

$ 
 
      

        
 
    
 

    
 

        


        '
    
  
    
   
        
 
    
 
  
   

        


     

  
 
   
    
 
    
     
% &  
         

    
 
  

    
  

     

     ! 
    
  
 
      

        
 
    
 

    
 

        


       

   " " !


     

  
 
   
    
 
    
     
 

    

  "   !


 
  

    
  

     

When the judgments are consistent, as they are here, any


normalized column gives the priorities.
Consistency
  


! #    

$Œ
         
   


% 

 !&$ '()*#! %% 


 +%     ,  
%%      

- %   ,        . 


  %    
  % /0
      
1         
%            
 
If we wish to compare a set of à attributes pairwise according to their relative
importance weights, where the attributes are denoted by  , ,..., à and the
weights are denoted by  ,  ,..., à , then the pairwise comparisons can be

represented as:
>  [  [ à
 
 
2 [ [ 
  , È  .
   à
 , where  È   ,      
 
 
 à [ à [ àà 

y multiplying 2 by yield
>   
 [ [ 
    à 
  >  > 
    
        
, or ( 2 â à )   w.
2   [ [     à      à

 
 à      
    
   à   
 à
  à à 
 [ [
 
 à 
Since solving the above equation is the eigenvalue problem, we can
derive the comparative weights by find the eigenvector with respec
to @a  Which satisfies 2  @a   , where @a  is the largest
eigenvalue of the matrix 2.

Furthermore, in order to ensure the consistency of the subjective


perception and the accuracy of the comparative weights, two indices,
including the consistency indexes (¢ .) and the consistency ratio
(¢..), are suggested. he equation of the ¢.. can be expressed as:

¢  † (@a ± à)  (à-),

where @a  is the largest eigenvalue, and à denotes the numbers


of the attributes.
Saaty (190) suggested that the value of the ¢  should not exceed 0.1
for a confident result. On the other hand, the ¢.. can be calculated as:

¢ . .
¢ . . È ,
 . .

where .. refer to an random consistency index which is derived from


a large sample of randomly generated reciprocal matrices using the
scale
 , ,K ,,K ,,.
The .. with respect to different size a tri is shown s

R aer of
eleaents 3 4 5 6 7      3

.. .5 . . . 5 .35 .4 .45 .4 .5 .54 .56

 ¢   


  w   
  w 
  
ypical 2` roblem

[  

 

  


[
   
 

2 
[ 
 

!

 
üoal

Price Mileage Prestige Comfort


¢  

  
Car 1 Car 2 Car 3

 

Comparing the Cluster node with respect to üoal node

@   @   



@  1 3 4 3

 1/3 1 2 1/3

@   1/4 1/2 1 1/4



1/3 3 4 1


    
Normalizing the matrix
@   @   


@  w 522 w 4ww w 364 w 655
 w 174 w 133 w 182 w w73
  !"#$"#"#$
@   w 13w w w67 w w91 w w55
"#"#$%!


w 174 w 4ww w 364 w 218
R   


@   † w 522+w 4+w 364+w 655)/4

  † w 174+w 133+w 182+w w73)/4

@    † w 13+w w67+w w91+w w55)/4



 † w 174+w 4+w 364+w 218)/4

i1
—max = 1.92 (0.48525) + (7.50 )(0.1405) + (11)(0.08575) + 0.289 *
4.58 = 4.249

¢  
         

     

¢ ¢   

  
Comparing wrt Price node in ͚Alternative͛ cluster

     &
R 
  1 ww 7 ww 4 ww   w 67w
  1/7 1 ww 1/5   w w73
 & 1/4 5 1 ww  & w 2w3

Rormalizing and Equating the above table we get A1


Comparing wrt Mileage node in ͚Alternative͛ cluster

     &
R 
  1 ww 1 ww 1 ww
  w 328
  1 ww 1 ww 2 ww   w 411
 & 1 ww 1/2 1 ww  & w 261

Rormalizing and Equating the above table we get


A2
      &

  w 33 w4 w 25 f = 3.0556 , Consistency Index = 0.02778 .


Relative index for R= 3 matrix is 0.52 . So
  w 33 w4 w5 consistency ratio is 0. 0534
 & w 33 w2 w 25
Comparing wrt Prestige node in ͚Alternative͛ cluster

     &
R 
  1 ww 3 ww 7 ww
  w 643
  1/3 1 ww 5 ww   w 283
 & 1/7 1/5 1 ww  & w w74
Rormalizing and Equating the above table we get
A3
      &

  w 675676 w 714286 w 538462 f = 3.093 , Consistency Index = 0.0469 .


Relative index for R= 3 matrix is 0.52 . So
  w 225225 w 238w95 w 384615 consistency ratio is 0. 09
 & w w96525 w w47619 w w76923
Comparing wrt Comfort node in ͚Alternative͛ cluster

     &
R 
  1 ww 3 ww 1/5
  w 193
  1/3 1 ww 1/7   w w83
 & 5 ww 7 ww 1 ww  & w 724

Rormalizing and Equating the above table we get


A4
      &

  w 157978 w 272727 w 149254 f = 3.114 , Consistency Index = 0.057 .


Relative index for R= 3 matrix is 0.52 . So
  w w52659 w w9w9w9 w 1w661 consistency ratio is 0. 109
 & w 789889 w 636364 w 746269
Õ   '
()@ 
 

@   @  



  * 

  * & *& &

  &  & &

 & & *  

  
 +!%*$%&$*%*&$%&!
  
 +!%&$%$*%&$%&!
 & 
 +!%&$%*$*%$%!&

,
,    --- 
 +"@R
 .-

/ / -
//+' . 
    .- #
! 
   
  
he 2nalytic Network rocess (2N )
involves Dependence and Feedback
‡ Real life problems involve dependence and feedback.
Such phenomena can not be dealt with in the framework
of a hierarchy but we can by using a network with
priorities.
‡ With feedback the alternatives can depend on the criteria
as in a hierarchy but may also depend on each other.
‡ he criteria themselves can depend on the alternatives
and on each other as well.

‡ Feedback improves the priorities derived from judgments


and makes prediction more accurateu
2nalytical `ierarchy rocess

2 methodology that allows groups or individuals to deal with the


interconnections(dependence and feedback) between factors
of complex structure in decision making process

2 Multi Criteria Decision Making method for complicated and


unstructured problems

2n approach that uses a network model having clusters of


elements(criteria and alternatives)
NEWORKS, DE ENDENCE 2ND
FEED2CK
O 2 network has clusters of elements, with the elements in one
cluster being connected to elements in another cluster (outer
dependence) or the same cluster (inner dependence).

O 2 network or feedback has sources and sinks. 2


  node is
an origin of paths of influence and never a destination of such
paths. 2
 node is a destination of paths of influence and
never an origin of such paths. 2 full network can include source
nodes; intermediate nodes that fall on paths from source nodes,
lie on cycles, or fall on paths to sink nodes; and finally sink
nodes. Some networks can contain only source and sink nodes.
Still others can include only source and cycle nodes or cycle
and sink nodes or only cycle nodes. 2 decision problem
involving feedback arises often in practice.
Feedback Network with components having
Inner and Outer Dependence among heir Elements
) 2rc from component C4 to C2 indicates
the outer dependence of the elements in
C2 on the elements in C4 with respect
to a common property.



*+ 
" No arrow enters are
  components such as C1
No arrow leaves are known as
 components
arrows both enter and exit leave are known as 
 
components such asC2, C and C4
C2and C forms a  
[       
    ' 
 
  
     (
Steps Of 2nalytical Network
rocess (2N )
‡! -Deconstructing a problem into a complete
set of hierarchical or network models

‡! ,Generating pairwise comparisons to


estimate the relative importance of various
elements at each level and Derive the local weights
using the 2` .
Steps Of 2nalytical Network
rocess (2N ) (continued)
!"- Formulate the supermatrix according to the
results of the local weights and the network
structure.

!)- Raise the supermatrix to limiting powers for


obtaining the final results

.
Super Matrix
O he influence of elements in the network on other elements
in that network can be represented in „   

  [ a
 [ à   [  à [ a [ aàa
 Where ¢ denotes the th cluster,
 à denotes the àth element in the
> Œ Œ [ Œa th cluster, and  is the local
  
à   priority matrix of the influence of the
  elements compared in the th cluster
  
   to the th cluster. In addition, if the
 ΠΠ[ Π th cluster has no influence to the th
Π 
a

à   cluster, then 0.
 
 
 
a  
a  
 Œa Œa [ Œaa 
a
aàa
Super Matrix : Example
R  
   ,  
    
Step 1: Define the decision problem
O u   
   
  
 
   
   

   
       

 u
O 
      
    

    
Step 2: Determine Control and ubCriteria
O petermine the control criteria and subcriteria in the
four control hierarchies one each for the benefits,
opportunities, costs and risks of that decision
O obtain their priorities from paired comparisons matrices.
O If a control criterion or subcriterion has a global priority
of 3% or less, you may consider carefully eliminating it
from further consideration.
Step . Determine the most general network
of clusters

O petermine the most general network of clusters or components)


and their elements that applies to all the control criteria.
O Rumber and arrange the clusters and their elements in a
convenient way (perhaps in a column).
O Îse the identical label to represent the same cluster and the same
elements for all the control criteria.
O petermine the most general network of clusters
Step 4. Determine Clusters and Elements

O Õor each control criterion or subcriterion, determine the


clusters of the general feedback system with their
elements
O Connect them according to their outer and inner
dependence influences.
O An arrow is drawn from a cluster to any cluster whose
elements influence it.
O pescribe the decision problem in detail including its
objectives, criteria and subcriteria, actors and their
objectives and the possible outcomes of that decision.
Step 5. Determine the approach
O petermine the approach you want to follow in the
analysis of each cluster or element, influencing (the
preferred approach) other clusters and elements with
respect to a criterion, or being influenced by other
clusters and elements.
O he sense (being influenced or influencing) must apply
to all the criteria for the four control hierarchies for the
entire decision.
Step 6. Supermatrix Construction
O Õor each control criterion, construct the supermatrix by
laying out the clusters in the order they are numbered
and all the elements in each cluster both vertically on the
left and horizontally at the top.
O Enter in the appropriate position the priorities derived
from the paired comparisons as subcolumns of the
corresponding column of the supermatrix.
Step M. erform aired Comparisons
O Perform paired comparisons on the elements within the
clusters themselves according to their influence on each
element in another cluster they are connected to (outer
dependence) or on elements in their own cluster (inner
dependence).
O Comparisons of elements according to which element
influences a given element more and how strongly more
than another element it is compared with are made with
a control criterion or subcriterion of the control
hierarchy in mind.
Step . aired Comparisons on the Clusters
O Perform paired comparisons on the clusters as they
influence each cluster to which they are connected with
respect to the given control criterion.
O he derived weights are used to weight the elements of
the corresponding column blocks of the supermatrix.
Assign a zero when there is no influence. hus obtain
the weighted column stochastic supermatrix.
Step 9. Compute Limit riorities of the
Stochastic Supermatrix

Compute the limit priorities of the stochastic supermatrix according


to whether it is
O irreducible (primitive or imprimitive [cyclic]) or
O reducible with one being a simple or a multiple root and whether
the system is cyclic or not.

wo kinds of outcomes are possible.


O In the first all the columns of the matrix are identical and each
gives the relative priorities of the elements from which the
priorities of the elements in each cluster are normalized to one.
O In the second the limit cycles in blocks and the different limits are
summed and averaged and again normalized to one for each
cluster.
Step 10. Synthesize the Limiting riorities
O 10. Synthesize the limiting priorities by weighting each idealized
limit vector by the weight of its control criterion and adding the
resulting vectors for each of the four merits: Benefits (B),
Opportunities (O), Costs (C) and Risks (R).
O here are now four vectors, one for each of the four merits. An
answer involving marginal values of the merits is obtained by
forming the ratio BO/CR for each alternative from the four
vectors. he alternative with the largest ratio is chosen for some
decisions.
O Companies and individuals with limited resources often prefer this
type of synthesis.
Step 11. Determine the strategic criteria and
their priorities
O petermine strategic criteria and their priorities to rate
the four merits one at a time. Rormalize the four ratings
thus obtained.
O Õor each alternative, subtract the costs and risks from the
sum of the benefits and opportunities.
O At other times one may add the weighted reciprocals of
the costs and risks.
O Still at other times one may subtract the costs from one
and risks from one and then weight and add them to the
weighted benefits and opportunities.
O In all, we have four different formulas for synthesis.
Step 12. Sensitivity 2nalysis

O Perform sensitivity analysis on the final outcome and


interpret the results of sensitivity observing how large or
small these ratios are.
O Can another outcome that is close also serve as a best
outcome? Why?
O By noting how stable this outcome is. Compare it with
the other outcomes by taking ratios. Can another
outcome that is close also serve as a best outcome? Why?
[ 
 

 -
/R
.  - 
/
'+ 
1(+ R .
2@
3- 2

/  

-  
4(   4
5   4
üoal

Price Mileage Prestige Comfort


¢  

  0/+ 

  Car 1 Car 2 Car 3


 
  - &. 
  -. 

 

  

@   @   



@     
 1/5 1 ww 1/2 1/5
R 
@   1/7 2 ww 1 ww w 14
@  w 495


2 5 ww 7 ww 1 ww
 w w69
Rormalizing and Equating the above table we get
@   w w85

@   @   





w 352
@  w 543478 w 384615 w 451613 w 5988w2
B1
 w 1w8696 w w76923 w w32258 w w5988

@   w w7764 w 153846 w w64516 w w42772 Consistency ratio is less than 10 %




w 271739 w 384615 w 451613 w 2994w1
  - &. 
  -. 

 

  

@   @   



@    & 6


 1/2 1 ww 1 ww 1/3
R 
@   1/3 1 ww 1 ww 1/3
@  w 288


2 ww 3 ww 3 ww 1 ww
 w 138
Rormalizing and Equating the above table we get
@   w 127

@   @   





w 447
@  w 261w97 w 285714 w 375 w 23w415
B2
 w 13w548 w 142857 w 125 w 15361
@   w w87w32 w 142857 w 125 w 15361 Consistency ratio is less than 10 %


w 522193 w 428571 w 375 w 46w829
  - &. 
  -. 

 

  "

@   @   



@   6* 6 6


 6 ww 1 ww 3 5
R 
@   5 ww 1 ww 1 2
@  w w68


2 ww 3 ww w 2w 1
 w 561
Rormalizing and Equating the above table we get
@   w 25w

@   @   





w 121
@  w w71429 w w98w39 w w42553 w w58824
B3
 w 428571 w 588235 w 638298 w 588235
@   w 357143 w 196w78 w 212766 w 235294 Consistency ratio is less than 10 %


w 142857 w 117647 w 1w6383 w 117647
R & '  ! /
  
& 


     & @   @   



  w w w

  w w w ( ( (& (

 & w w w

@  w w w w

 w w w w
7 7 7&
@   w w w w



w w w w
     & @   @   

 
w w w w 67w w 328 w 643 w 193
 
w w w w w73 w 411 w 283 w w83
 & w w w w 2w3 w 261 w w74 w 724
@  w 495 w 288 w w68 w w w w
 w w69 w 138 w 561 w w w w
@   w w85 w 127 w 25w w w w w


w 352 w 447 w 121 w w w w
'R    ' [ &  '! 
=/(

0

'   +  /  1 .

  " $  = $


  ' 
 0.22192 0.22192 0.22192 0.22192 0.22192 0.22192 0.22192
 0.096M9 0.096M9 0.096M9 0.096M9 0.096M9 0.096M9 0.096M9
" 0.1129 0.1129 0.1129 0.1129 0.1129 0.1129 0.1129
$  0.150M2 0.150M2 0.150M2 0.150M2 0.150M2 0.150M2 0.150M2
= 0.1 059 0.1 059 0.1 059 0.1 059 0.1 059 0.1 059 0.1 059
$
 0.0M465 0.0M465 0.0M465 0.0M465 0.0M465 0.0M465 0.0M465
 '  0.14404 0.14404 0.14404 0.14404 0.14404 0.14404 0.14404
R ' '  R$ 
 '- &
.  
R 
 &


  Calculations : Car 1 †
w 44384
w 22192) /
  w 19358 w 22192+w w9679+w 1
8129)
 & w 36258

ANP ith Outer


AHP dependies)

Car 1 w 482 w 44384


Car 2 w 141 w 19358
Car 3 w 351 w 36258

 
 '
 R$ (Outer dependance) : Model to be selected still
remains the same but the priorities have changed he weight age for Car 1 has
decreased by  % and Car 2 weightage has increased drastically
 -
/   - 
/(R@". - 
///
  //#8

Inner Dependency and Outer üoal


dependency exists in this
case

Price Mileage Prestige Comfort


¢  

Car 1 Car 2 Car 3


  
!  
+  

'    

!2 
R $ 
 '  +$
3 ' 3$


 '  +$ 3 ' 

1. Comparison of restige and comfort with respect to rice Rormalized matrix

$
  '  $
  ' 

$
 1 $
 0.M5 0.M5
 '  1/ 1  '  0.25 0.25


 „

  

  „ 
 „ 
R
9.

@ 

@ 

@    @  w 833 w 833



6 

w 1667 w 1667
-
 -,   

     & @  @   




  w w w
  w w w A1 A3 A4
 & w w w
@  w w 8333
 w w w
1 2 3
@   w 75 w w


w 25 w 1677 w

Values have come up in this


cluster due to inner dependency
between criterias
'
 & 
4 ! /  

  " $  = $


 ' 
 0 0 0 0.6M0 0. 2 0.64 0.19
 0 0 0 0.0M 0.411 0.2 0.0
" 0 0 0 0.20 0.261 0.0M4 0.M24
$  0.495 0.2 0.06 0 0 0.

= 0.069 0.1  0.561 0 0 0 0

$
 0.05 0.12M 0.250 0.M5 0 0 0

 '  0. 52 0.44M 0.121 0.25 0 0.16MM 0


R
  !   

     &
@   @   

  w w w w 34354 w 32748 w 32456 w 18839

  w w w w w3459 w 4126 w 13948 w w8w96

 & w w w w 12187 w 25992 w w3596 w 73w64

@  w 49942 w 28795 w w6632 w w w 41667 w

 w w6591 w 13766 w 56614 w w w w

@   w w7945 w 1258 w 24737 w 375 w w w



w 35522 w 44859 w 12w18 w 125 w w w8333 w
Õ  
 
          
 „„

 „    
 „
  " $  = $
 ' 
 0.1M02 0.1M02 0.1M02 0.1M02 0.1M02 0.1M02 0.1M02
 0.05 5 0.05 5 0.05 5 0.05 5 0.05 5 0.05 5 0.05 5
" 0.16M6 0.16M6 0.16M6 0.16M6 0.16M6 0.16M6 0.16M6
$  0.1M551 0.1M551 0.1M551 0.1M551 0.1M551 0.1M551 0.1M551
= 0.11M9 0.11M9 0.11M9 0.11M9 0.11M9 0.11M9 0.11M9
$
 0.1 156 0.1 156 0.1 156 0.1 156 0.1 156 0.1 156 0.1 156
 '  0.151 0.151 0.151 0.151 0.151 0.151 0.151
R

 R  
 „
    „ 
R
  
„  „  

 
  
 
& & &
 &
** ** **
R
  
    R  
 „„

  w 392145
  w 211663 ,
,   &
 & w 396191
 


ANP ith Outer ANP ith Inner


AHP dependies) decencies
  w 482 w 44384 w 392145
  w 141 w 19358 w 2w1663
 & w 351 w 36258 w 396191

"
 &   
(((! 
 
"
2dvantages of the 2nalytic `ierarchy rocess and the
2nalytic Network rocess over other Multi Criteria
Decision Making methods
‡ 2s compared to other MCDM approaches, 2` /2N is not
proportionately complicated, and this helps improve
management understanding and transparency of the modeling
technique(

‡ hey have the supplemental power of being able to mix


quantitative and qualitative factors into a decision.

‡ his approach can be fit together with other solution approach


such as optimization, and goal programming.

‡ 2` /2N may use a hierarchical structuring of the factors


involved. he hierarchical structuring is universal to the
composition of virtually all complex systems, and is a natural
problem-solving paradigm in the face of complexity.
‡ In 2` /2N , judgment elicitations are completed using
a decompositional approach, which has been shown in
experimental studies to reduce decision-making errors.

‡ 2` /2N is a technique that can prove valuable in


helping multiple parties (stakeholders) arrive at an
agreeable solution due to its structure, and if
implemented appropriately can be used as a
consensus-building tool

‡ 2` has also been validated from the decision makers


perspective as well in recent empirical studies
Limitations of 2N :

O ARP is more complex than the AHP and it increases the


effort his can be overcome by using software ² Super
pecisions, Expert Choice

O 2. In case, if there are several alternatives in the decision


making a number of pair wise comparisons would be
quite demanding
ÿ  :

Saaty, T  2ww5) Theory and Applications of the Analytic


Network Process Pittsburgh, PA: RS Publications, 4922 Ellsworth
Avenue, Pittsburgh, PA 15213

Decision making with dependence and feedback The Analytic


Network Process Thomas  Satty

The Analytic Hierachy process ʹ An Exposition by Ernest h Forman


, !chool of B siness nd P lic M n eaent , Geore W shinton
University , W shinton D C 5
 

Das könnte Ihnen auch gefallen