You are on page 1of 36

Business Decision ProcessCourse 2

Decision Making using


Analytic Hierarchy Processes

Decision making
in the economic framework of utility theory

a brief history of the development of the utility theory


a reconceptualization of the basic sources of utility.
intangible factors are brought together into
consideration in the formulation of the utility functions,
in the mainstream of economic theory.

Decision problem

Single-attribute
Decision problem

Each decision alternative


can be assigned
one number

Multiattribute Decision
problem

Multiple attributes which


are
typically uncomensurable

Multiattribute utility
theory

Axiom of additive
independence

Analytic Hierarchic
Processes

Ordinal Utility
consumers
had preferences
for one combination of goods
over another
Cardinal Utility
original concept of utility
moral philosopher Jeremy Bentham
formally integrated
Jevons, Menger and Walras-in the 1870s
cardinally measurable
psychological
flow of satisfactions
attached to goods and services

A properly conceptualized utility


theory implies measurements
that cannot be inferred from
behavior
but must be obtained directly from
consumers
and this conceptualization can help
to predict behavior that is not
otherwise understandable.

In AHP
order of preferences
preferences

intensity of

Given two alternatives : A and B


Which of A or B do you prefer ?
On a scale from 1 to 9
(1-equally preferred, 9-extremely
preferred)
by how much you
prefer
the alternative you chose
over the other one?

A hierarchy
Decision Problem

Criterion 1 (C1)

Criterion 2 (C2)

Criterion 3 (C3)

(Criterions)

Alternative 1 (A1)
(Alternatives)

Alternative 2 (A2)

Decision matrix for a


problem with 2, 3,..n
criterions

decision
decision

2 decision criterions C1, C2

What criterion is more important among C 1


and C2 ?
Answer :C1
On a scale from 1 to 9 (1-equally important,
extremely important) by how much you
think is more important the criterion you
chose to the other one ?
Answer : 3
Decisional Matrix
C1
C2
C1
1
3
C
1/3
1

3 decision criterions C1, C2 ,C3


What criterion you think is more important
(Preference)
C1 or C2 ?

Answer :C2

On a scale from 1 to 9 (1-equally important, 9-extremly important)


(Intensity)
by how much you consider is more important the criterion you chosen to the
other one ?
Answer : 5
What criterion you think is more important,
C2 or C3 ?

Answer :C2

On a scale from 1 to 9 (1-equally important, 9-extremly important)


by how much you consider is more important the criterion you chosen to the
other one ?
Answer :3
What criterion you think is more important,
C1 or C3 ?

Answer :C3

On a scale from 1 to 9 (1-equally important, 9-extremly important)


by how much you consider is more important the criterion you chosen to the
other one ?
Answer : 3

Decision matrix
C1
C 2 C3
C1

1/5

1/3

C2

C3

1/3

In general
Decision matrix

Reciprocity

aij=1/aji
C1

C2

C1

1
= aik

a12

C2

a21

C3

a31

a32

C3
a13
a23
1

Consistency

aij * ajk

Decision vector-associated with a


reciprocal, consistent decision matrix is the
normalized
form
of
the
eigenvector
associated with the highest eigenvalue
Matricea decizionala (reciproca, consistenta)
decizie
C1
C 2 C3
C1
1
a12 a13
C2 a21
1
a23
C3
a31 a32
1

cu a12 =w1/ w2 , a13 =w1/ w3 , a23 =w2/ w3


(in general, aij =wi/ wj )

Vector de

w1
w2
w3

If the decision matrix is not consistent, the associated


decision vector is the eigenvector associated with the
highest eigenvalue of the normalized matrix

Decision matrix

C1
C2
C3

C1
C 2 C3
1
1/5
1/3
5
1
3
3
1/3 1

Decision vector

w1=0.105
w 2=0.637
w3 =0.258

In general, a positive,reciprocal, square matrix- C- is


consistent if and only if max=dim(C)

The computation of a priority vector corresponding


to the final alternatives in an hierarchy

C1

C2

C3

Cn

w1

w2

w3

wn

A1 v11

v12 v13

v1n

w1v11+w2v12+..wnv1n

A2 v21

v22 v23

v2n

w1v21+w2v22+..wnv2n

Am vm1

vm2 vm3

vmn

Priority vector for the


alternatives
A1,..Am

w1vm1+w2vm2+..wnvmn

Conclusion ?

How do we aggregate the results ?


How many experts are needed to
validate the results ?
In the AHP original framework
experts meet together, discuss,
compromise and agree to a common
point of view
is this all the time feasible ?
Is a groups point of view
superior to an individuals one ?

Disputable points in the AHP theory


numerical scale (1-9, or 1-5 or others)
individual scale:
an individual who prefers C1 to C2 with
intensity 3
may not be comparable with another
individual delivering the same answer.
Since different individuals may have
different standards coming from different
experiences

IF:
It would be a way to decide whether
- the Xs response- declared 3actually means 1.24
-the Ys response-declared 3actually means 3.71
so that the decision matrices corresponding to the two
respondents are more consistent than the initial ones
and this could be somehow extended to the both
hierarchies associated with the two respondents
then the arithmetic mean of the two correspondent
components in the decision vector would be a
representative approximation for the true value, thus
making the statistical representatives useless

Saatys most famous experiments

This section reproduces four examples


from Saaty (2010) in which the numerical
scale using integers from 1 to 9 is applied
and the true vector of priorities is known.

These examples refer to single matrix


validation example.

You are strongly encouraged to


reproduce these experiments and present
their findings.

Optics Example

In this experiment participated Saatys kids of 5 and


7 years old, to provide one decision matrix and
Saatys wife to provide the second decision matrix.

Four identical chairs were placed on a line from a


light source at the distances of 9,15,21 and 28 yards.
The decision maker is placed by the light and she/he
has to look at the chair and compare their relative
brightness in pairs, fill the decision matrix and
obtain a relationship between the chairs and their
distance from the light source.

This experiment was repeated twice and the


correspondent decision matrices are given below.

DM1
C1
C2
C3
C4

C1
1

C2
5
1

C3
6
4
1

C4
7
6
4
1

DM2

C1

C2
C3
C4

C1

C2

C3

C4

3
1

4
2
1

The corresponding prioritys


vectors (PV)for the relative
brightness are shown in the next
table:

Relative
brightness PV1

Relative
brightness PV2

0.61
0.24
0.1
0.05

0.62
0.22
0.2
0.06

These results should be compared


with the so-called inverse square
law in optics according to which
the brightness is inverse
proportional with the square of
the distance to the source of light
(see Table 2).

Dista
nce

Normali
zed
distanc
e

Square of
normalize
d
distance

Recipr
ocal of
previo
us
column

Normali
zed
recipro
cal

Rounding
of

0.12
3
0.20
5
0.28
8
0.38
4

0.015
129
0.042
025
0.082
944
0.147
456

66.0
98
23.7
9
12.0
5
6.78

0.60
79
0.21
88
0.11
08
0.06
23

0.61

15
21
28

0.22
0.11
0.06

Relative consumption of drinks


In this experiment 30 people, using consensus arrived at the
same judgment regarding the dominance of the consumption
of drinks in the United States (which drink is consumed more
in the US and how much more than another drink?).

The priority vector (PV ) corresponding to a single decision


matrix is compared with the actual consumption, from
statistical sources .

Maybe, after contemplating the amazing match of the results,


one would want to think at the diferences among consensus
and individual opinion (expressed through a survey,
eventually) and redo the experiment.

The decision matrix corresponding to the joint evaluation of 30


individuals for the drink consumption in the US (DMDrink
Consumption) is presented below:

DMDrink Consumption

Cofe
Wine
Tea
Beer
Sodas
Milk
Water

Cofe

Wine

Tea

Beer

Sodas

Milk

9
1
9
9
9
9
9

1
1
2

Water

1
3
4
3
9

1
2
1
3

1
2

1
2
1
3


The comparison among the
priority vector (PV) corresponding
to the previous decision matrix
and the actual consumption (from
statistical sources) is shown in the
next table:

PV

Cofee
Wine
Tea
Beer
Sodas
Milk
water

0.177
0.019
0.042
0.116
0.190
0.129
0.327

Actual
consumpti
on
0.180
0.010
0.040
0.120
0.180
0.140
0.330

Pairwise comparisons of the weights of five objects

This experiment gives the


estimated pairwise comparisons of
the weights of the five objects
lifted by hand, made by a friend of
Saaty.

The decision matrix (DMWeight)is


presented below:

DMWeight

Radio
Typewriter
Large Attach Case
Projector
Small Attach Case

Radio

1
5
3
4

Typewriter

Large
Attach
Case

2
1
2

Projector

Small Attach
Case

4
8
4

The comparison among the


priority vector (PV) and the actual
relative weights is shown in the
next table:

PV

Radio
Typewrit
er
Large
Attach
Case
Projector
Small
Attach
Case

0.09
0.4

Actual
Relative
Weights
0.10
0.39

0.18

0.20

0.29
0.04

0.27
0.04

Relative Wealth of Seven Nations

This exercise was done by Saaty


and one of his collaborators while
traveling by plane.
Using common knowledge about
the relative power and standing of
seven countries, in the next is
shown the decision matrix
(DMRelative Wealth) deduced by
pairwise comparisons of the
relative wealth of seven nations.

DMRelative Wealth

U.S.

Russia

China

France

UK

Japan

U.S.
Russia
China
France
UK
Japan
Germany

9
1
9
9
9
9
9

1
1
2

Germany

1
3
4
3
9

1
2
1
3

1
2

1
2
1
3

The comparison among the


priority vector (PV) and the
normalized GNP values is shown
in the next table:

PV

Actual
GNP
(1972)

U.S.
Russia
China
France
UK
Japan
Germa
ny

0.427
0.23
0.021
0.052
0.052
0.123
0.094

1167
635
120
196
154
294
257

Norma
lized
GNP
Values
0.413
0.225
0.043
0.069
0.055
0.104
0.091

You might also like