Vous êtes sur la page 1sur 42

Lecture 08

AnaIytic Hierarchy Process


(ModuIe 1)
ndustrial Systems Engineering Dept.- U
Office: Room 504
Learning Objecties
Students will be able to:
1. Use the multifactor evaluation process in making
decisions that involve a number of factors, where
importance weights can be assigned.
2. Understand the use of the analytic hierarchy process
in decision making.
3. Contrast multifactor evaluation with the analytic
hierarchy process.
4/ule Outline
1.1 Introduction
1.2 MuItifactor EvaIuation Process
1.3 AnaIytic Hierarchy Process
Intr4/ucti4n
Multifactor decision making involves individuals
subiectively and intuitively considering various
Iactors prior to making a decision.
Multifactor evaluation process (MFEP) is a
quantitative approach that gives weights to each
factor and scores to each alternative.
nalytic hierarchy process (P) is an approach
designed to quantiIy the preIerences Ior various
factors and alternatives.
ultiact4r Laluati4n Pr4cess
actor mportance
(weight)
AA
Co.
EDS,
LTD.
PW,
nc.
Salary 0.3 0.7 0.8 0.9
Career
Advancement
0.6 0.9 0.7 0.6
Location 0.1 0.6 0.8 0.9
Example: $teve. M.: considering employment with 3 companies.
determined 3 factors important to him, assigned each Iactor a weight.
Weights should sum to 1
Steve evaIuated the various factors on a 0 to 1 scaIe for each
of these jobs.
$core Table
Laluati4n 4 AA (4
actor actor actor Weighted
Name Weight EvaIuation EvaIuation
SaIary 0.3 0.7 0.21
Career 0.6 0.9 0.54
Location 0.1 0.6 0.06
TotaI 0.81
Factor Factor Weighted
Weight Evaluation Evaluation
X

(425aris4n 4 Results
actor AA Co. EDS,LTD. PW,nc.
Salary 0.21 0.24 0.27
Career 0.54 0.42 0.36
Location 0.06 0.08 0.09
Weighted
Evaluation
0.81 0.74 0.72
ecision is Co: ighest weighted evaluation
8
1he Analvtic lierarchv Pr4cess AlP,
ounded by Saaty in 1980.
t is a popular and widely used method
for multi-criteria decision making.
Allows the use of qualitative, as well as
quantitative criteria in evaluation.
Wide range of applications exists:
Selecting a car for purchasing
Deciding upon a place to visit for vacation
Deciding upon an A program after graduation.
.
r. Thomas L. Saaty
istinguished Prof. at U. of Pittsburgh
9
AlP-General I/ea
Develop an hierarchy of decision criteria and define the
alternative courses of actions.
AHP algorithm is basically composed of two steps:
1. Determine the relative eights of the decision criteria
2. Determine the relative rankings (priorities) of
alternatives
oth qualitative and quantitative information can be
compared by using informed judgments to derive
weights and priorities.
Ste5s
Step 0: Construction of Hierarchy Structure
(including: Goal, actors, Criteria, and Alternatives)
Step 1: Calculation of actor Weight
Step 1-1: Pairwise Comparison atrix
Step 1-2: Eigenvalue and Eigenvector (Priority vector)
Step 1-3:Consistency Test
Consistency ndex
Consistency Ratio
Step 2:Calculation of Level Weight
Step 3: Calculation of Overall Ranking
C
1
C
2
C
3
C
11
C
12
C
13
Mare !eetiit
k|teraatives
Mare 0eaera|
0aa|
C
21
C
22
C
31
C
32
C
33
!e|-triteria at the
|awest |eve|
ierarchy Tree
LeveI 0
LeveI 1 (factors)
LeveI 2 (criteria)
LeveI ..
Tom Saaty suggests that hierarchies be limited to six levels and nine items per Tom Saaty suggests that hierarchies be limited to six levels and nine items per
level. level.
This is based on the psychological result that people can consider 7 +/ This is based on the psychological result that people can consider 7 +/- - 2 2
items simultaneously (iller, 1956). items simultaneously (iller, 1956).
Pairise Comparisons
Size
Apple A Apple Apple C
Size
Comparison
AppIe A AppIe B AppIe C
AppIe A 1 2 6 6/10 A
AppIe B 1/2 1 3 3/10 B
AppIe C 1/6 1/3 1 1/10 C
ResuIting
Priority
Eigenvector
ReIative Size
of AppIe
The image cannot be displayed. Your computer may not have
enough memory to open the image, or the image may have
been corrupted. Restart your computer, and then open the file
again. !f the red x still appears, you may have to delete the
image and then insert it again.
The image cannot be displayed. Your
computer may not have enough memory to
open the image, or the image may have
been corrupted. Restart your computer,
and then open the file again. !f the red x
still appears, you may have to delete the
image and then insert it again.
The image cannot be
displayed. Your
computer may not
have enough memory
to open the image, or
the image may have
been corrupted.
Restart your
computer, and then
open the file again. !f
the red x still
appears, you may h.
The image cannot be displayed. Your computer may not have
enough memory to open the image, or the image may have
been corrupted. Restart your computer, and then open the file
again. !f the red x still appears, you may have to delete the
image and then insert it again.
The image cannot be displayed. Your
computer may not have enough memory to
open the image, or the image may have
been corrupted. Restart your computer,
and then open the file again. !f the red x
still appears, you may have to delete the
image and then insert it again.
The image cannot be
displayed. Your
computer may not
have enough memory
to open the image, or
the image may have
been corrupted.
Restart your
computer, and then
open the file again. !f
the red x still
appears, you may h.
The image cannot be displayed. Your computer may not have
enough memory to open the image, or the image may have
been corrupted. Restart your computer, and then open the file
again. !f the red x still appears, you may have to delete the
image and then insert it again.
The image cannot be displayed. Your
computer may not have enough memory to
open the image, or the image may have
been corrupted. Restart your computer,
and then open the file again. !f the red x
still appears, you may have to delete the
image and then insert it again.
The image cannot be
displayed. Your
computer may not
have enough memory
to open the image, or
the image may have
been corrupted.
Restart your
computer, and then
open the file again. !f
the red x still
appears, you may h.
The image cannot be displayed. Your computer may not have
enough memory to open the image, or the image may have
been corrupted. Restart your computer, and then open the file
again. !f the red x still appears, you may have to delete the
image and then insert it again.
The image cannot be displayed. Your
computer may not have enough memory to
open the image, or the image may have
been corrupted. Restart your computer,
and then open the file again. !f the red x
still appears, you may have to delete the
image and then insert it again.
riteria #1 riteria #2
1
ntensity of
mportance
2 3 4 5 6 8 7 9 9 8 7 6 5 3 4 2
Pairwise Comparison Matrix
Pairwise Comparison atrix A = ( a
ij
)
a
33
a
32
a
31
A3
a
32
a
22
a
21
A2
a
13
a
12
a
11
A1
A3 A2 A1 to
Values for a
ij
:
NumericaI vaIues VerbaI judgment of preferences
1 equaIIy important
3 eakIy more important
5 strongIy more important
7 very strongIy more important
9 absoIuteIy more important
2,4,6,8 =>
reciprocals =>
intermediate
values
reverse
comparisons
Ranking 4 (riteria an/ Alternaties
#anking Scale for Criteria and Alternatives
(a) a
ii
= 1 A comparison of criterion i ith itseIf: equally important
(b) a
ij
= 1/ a
ji
a
ji
are reverse comparisons and must be the reciprocaIs of a
ij
14
Lxa25le 1: (ar Selecti4n 115,
Objective
Selecting a car
Criteria
Style, Reliability, uel-economy Cost?
Alternatives
Civic Coupe, Saturn Coupe, ord Escort, azda iata
15
ierarchy tree
Style Reliability uel Economy
Selecting
a New Car
ivic $aturn Escort Miata
Lxa25le 1: (ar Selecti4n 215,
16
#anking of Criteria
$tyle Reliability Fuel Economy
$tyle
Reliability
Fuel Economy
1/1 1/2 3/1
2/1 1/1 4/1
1/3 1/4 1/1
Lxa25le 1: (ar Selecti4n 315,
1
#anking of Priorities
Consider [Ax = 2x] where
A is the comparison matrix of size nn, for n criteria, also called the priority
matrix.
x is the Eigenvector of size n1, also called the priority vector.
2 is the Eigenvalue, 2 ZT > n.
To find the ranking of priorities, namely the Eigen Vector X:
1) Normalize the column entries by dividing each entry by the sum of the column.
2) Take the overall row averages.
.3 .29 .38
.6 .57 .5
.1 .14 .13
CoIumn sums 3.33 1.75 8.00 1.00 1.00 1.00
A=
1 .5 3
2 1 4
.33 .25 1.
NormaIized
CoIumn Sums
Ro
Averages
0.3196
0.5584
0.1220
Priority vector
X=
Lxa25le 1: (ar Selecti4n 415,
#anking of Priorities
Lxa25le 1: (ar Selecti4n 415,
Pairise Comp. Matrix Norm. Pairise Comp. Matrix
18
Criteria weights
Style .3196 = .3
Reliability .5584 = .6
uel Economy .1220 = .1
Style
.3196
Reliability
.5584
uel Economy
.1220
Selecting
a New Car
1.0
irst important criterion
Second most important criterion
Here is the tree of criteria ith the criteria eights
The Ieast important criterion
#anking of Priorities (cont.)
Lxa25le 1: (ar Selecti4n 515,
19
Checking for Consistency
Consistency Ratio (CR): measure how consistent the judgments have been
relative to large samples of purely random judgments.
AHP evaluations are based on the asumption that the decision maker is
rational, i.e., if A is preferred to and is preferred to C, then A is preferred
to C.
Suppose we judge apple A to be twice as large as apple and apple Suppose we judge apple A to be twice as large as apple and apple
to be three times as large as apple C. to be three times as large as apple C.
To be perfectly consistent, apple A must be six times as large as apple To be perfectly consistent, apple A must be six times as large as apple
C. C.
f the CR is greater than 0.1 the judgments are untrustworthy because they
are too close for comfort to randomness and the exercise is valueless or
must be repeated.
Lxa25le 1: (ar Selecti4n 615,
The image cannot be
displayed. Your
computer may not
have enough memory
to open the image, or
the image may have
been corrupted.
Restart your
computer, and then
open the file again. !f
the red x still
appears, you may h.
The image cannot be displayed. Your computer may not have
enough memory to open the image, or the image may have
been corrupted. Restart your computer, and then open the file
again. !f the red x still appears, you may have to delete the
image and then insert it again.
The image cannot be displayed. Your
computer may not have enough memory to
open the image, or the image may have
been corrupted. Restart your computer,
and then open the file again. !f the red x
still appears, you may have to delete the
image and then insert it again.
20
Calculation of Consistency #atio
.3
.6
.1
1 .5 3
2 1 4
.333 .25 1.
.9
1.6
.35
=
A x Ax x
=
A x Ax x
Consistency index (C) is found by
The next stage is to calculate 2, Consistency ndex (C) and the
Consistency Ratio (CR).
Consider [Ax = 2x] where x is the Eigenvector.
=
= 2
0.30
0.60
0.10
A x Ax x
Lxa25le 1: (ar Selecti4n 15,
Consistency Vector =
0.90/0.30
1.60/0.60
0.35/0.10
6 . 3
3
5 . 3 67 . 2 . 3


2
3.00
2.67
3.50
=
3 .
1 3
3 6 . 3
1

3
3

2
Note. This is iust a3 approximate method to determi3e value of
Consistency Index
reflects the consistency of
one's judgement
#andom Index (#I)
the CI of a randomly-generated
pairwise comparison matrix
Tabulated by size of matrix (n):
(given by author)
n #I
2 0.0
3 0.58
4 0.90
5 1.12
6 1.24
7 1.32
8 1.41
9 1.45
10 1.51
Lxa25le 1: (ar Selecti4n 815,
1

3
3

2
Consistency #atio
n practice, a CR of 0.1 or below is considered acceptable.
Any higher value at any level indicate that the judgements
warrant re-examination.
#

#
Lxa25le 1: (ar Selecti4n 915,
n the above example:
so, the evaluations are consistent
1 . 52 .
58 .
3 .

#

#
23
#anking Alternatives
$tyle
Civic
$aturn
Escort
1 1/4 4 1/6
4 1 4 1/4
1/4 1/4 1 1/5
Miata 6 4 5 1
Civic $aturn Escort Miata
Miata
Reliability
Civic
$aturn
Escort
1 2 5 1
1/2 1 3 2
1/5 1/3 1 1/4
Miata 1 1/2 4 1
Civic $aturn Escort Miata
.13
.24
.7
.56
Priority vector
.38
.29
.7
.26
Lxa25le 1: (ar Selecti4n 1015,
24
Fuel Economy
Civic
$aturn
Escort
Miata Miata
34
27
24
28
113
Miles/gallon
Normalized
.30
.24
.21
.25
1.0
#anking Alternatives (cont.)
Since fuel economy is a quantitative measure, fuel consumption ratios
can be used to determine the relative ranking of alternatives; however
this is not obligatory. Pairwise comparisons may still be used in some
cases.
Lxa25le 1: (ar Selecti4n 1115,
#anking Alternatives (cont.)
Lxa25le 1: (ar Selecti4n 1115,
25
Civic 0.13
Saturn 0.24
Escort 0.07
iata 0.56
Civic 0.38
Saturn 0.29
Escort 0.07
iata 0.26
Civic 0.30
Saturn 0.24
Escort 0.21
iata 0.25
Style
0.30
Reliability
0.60
uel Economy
0.10
Selecting a New Car
1.00
#anking Alternatives (cont.)
Lxa25le 1: (ar Selecti4n 1215,
Car StyIe(0.3) ReIiabiIity(0.6) ueI Economy(0.1) TotaI
Civic 0.13 0.38 0.30 0.30
Saturn 0.24 0.29 0.24 0.27
Escort 0.07 0.07 0.21 0.08
Miata 0.56 0.26 0.25 0.35
Iargest
26
#anking of Alternatives (cont.)
S
t
y
l
e


R
e
l
i
a
b
i
l
i
t
y



u
e
l
E
c
o
n
o
m
y
Civic
Escort
Miata Miata
$aturn
.13 .38 .30
.24 .29 .24
.07 .07 .21
.56 .26 .25
x
.30
.60
.10

.30
.27
.08
.35
actor Weights Priority matrix
Lxa25le 1: (ar Selecti4n 1315,
2
Including Cost as a Decision Criteria
CVC $12K .22 .30 0.73
SATURN $15K .28 .27 1.04
ESCORT $ 9K .17 .08 2.13
ATA $18K .33 .35 0.94
Cost
Normalized
Cost
Cost/Benefits
Ratio
Adding "cost as a a new criterion is very difficult in AHP. A new column
and a new row will be added in the evaluation matrix. However, whole
evaluation should be repeated since addition of a new criterion might
affect the relative importance of other criteria as well
nstead one may think of normalizing the costs directly and calculate the
cost/benefit ratio for comparing alternatives
Benefits
Lxa25le 1: (ar Selecti4n 1415,
Methods for including Cost Criterion
Use graphical representations to make trade-offs.
Calculate cost/benefit ratios
Use linear programming
Use seperate benefit and cost trees and then combine the results
28
Civic
Escort
Saturn
iata
Lxa25le 1: (ar Selecti4n 1515,
0
5
10
15
20
25
30
35
40
0 5 10 15 20 25 30 35

e
n
e
f
i
t
Cost
Civic
Escort
Saturn
iata
29
(425lex Decisi4ns
WMany levels oI criteria and sub-criteria exists Ior
complex problems.
Goal: uying the best car
There are three criteria:
-Cost
-Quality
Safety
Comfortability
-aintenance
nsurance
Services
Three alternatives: Honda, ercedes, Hyundai
Lxa25le 2: Buving the best car
LeveI 0
LeveI 1
Criteria
LeveI 2
Sub-criteria
AIternatives
%he Hierarchy Ior prooblem uvi3 the best car
Lxa25le 2: Buving the best car
Step 1: Criterion comparison
Criterion comparison
Normalize values:
ind Column vector
The process is repeated for the sub-criteria until the evaluation for all other
alternatives. This example will be supported by Expert Choice software
Price antenance Quality
Price 1 3 5
aintenance 1/3 1 2
Quality 1/5 1/2 1

Price aintenance Quality
Price 0.652 0.667 0.625
aintenance 0.217 0.222 0.25
Quality 0.131 0.111 0.125
Lxa25le 2: Buving the best car
Price
Price 0.648
Mainternance 0.23
QuaIity 0.122
Step 2: Determining the Consistency Ratio - CR
etermining the Consistency vector
We begin by determining the weighted sum vector. This is done by
multiplying the column vector times the pairwise comparison matrix.
Column vector: Pairwise comparison matrix:
Price 0.648
ainternance = 0.230
Quality 0.122
1 3 5
1/3 1 2
1/5 1/2 1
Lxa25le 2: Buving the best car

1.948
0.690
0.366
Weighted sum vector
Consistency vector =
Weighted sum vector/ CoIumn vector
Consistency vector
etermining 2 and the Consistency Index-CI
2 (3.006+3.0+3.0) / 3 = 3.002
The C is:
C = (3.002 - 3) / (3 - 1) = 0.001
etermining the Consistency #atio-C#
with n = 3, we get R = 0.58
CR = 0.001 / 0.58 = 0.0017
Since 0< CR < 0.1, we accept this result and move to the lower
level. The procedure is repeated till the lowest level.
Lxa25le 2: Buving the best car
Continue for other levels:
or subcriteria nsurance Service:
nsurance Service
nsurance 1 3
Service 1/3 1

HONDA 25000
ER. 60000
HUYNDA 15000

Honda er Huyndai
Honda 1 1/3 1/4
er 3 1 2
Huyndai 4 1/2 1

W or ost
W or Insurance:
Honda er Huyndai
Honda 1 3 4
er 1/3 1 2
Huyndai 1/4 1/2 1

W or $ervice
Honda er Huyndai
Honda 1 1/4 1/5
er 4 1 1/2
Huyndai 5 2 1

W or Quality
O And muke vour IInuI evuIuuLIon (sLudenLs seII deveIop LIIs evuIuuLIon)
1) Weights are defined for
each hierarchical level...
2) ...and multiplied down to get
the final lower level weights.
0.6 0.4
0.7 0.3 0.2 0.6 0.2
0.6 0.4
0.7 0.3 0.2 0.6 0.2
MuItipIy
0.42 0.18 0.08 0.24 0.08
otes:
n general, the evaluation scores are collected from
many experts and the average scores is used in the
pairwise comparison matrix.
The AHP solving is computer-aided by Expert Choice
(EC) software.
- uilding structure of problem
- Enter judgments (Pairwise Comparisons)
- Analysis the weights
- Sensitivity Analysis
- Advantages and disadvantages
- iscellaneous
4re ab4ut AlP: Pr4s an/ (4ns
41
There are hidden assumptions like consistency.
Repeating evaluations is cumbersome.
Difficult to use when the number of criteria or
alternatives is high, i.e., more than 7.
Difficult to add a new criterion or alternative
Difficult to take out an existing criterion or
alternative, since the best alternative might
differ if the worst one is excluded.
Users should be trained to use
AHP methodology.
Use cost/benefit ratio if
applicable
P
r
o
s
C
o
n
s
t allows multi criteria decision making.
t is applicable when it is difficult to formulate
criteria evaluations, i.e., it allows qualitative
evaluation as well as quantitative evaluation.
t is applicable for group decision making
environments
l42ew4rk 08
Due: next class,
1.4, 1.10, 1.11

Vous aimerez peut-être aussi