You are on page 1of 34

IME634: Management

Decision Analysis
Raghu Nandan Sengupta
Industrial & Management Department
Indian Institute of Technology Kanpur

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 1


Analytic Hierarchy Process (AHP)
 The Analytic Hierarchy Process
(AHP) is a structured technique for
organizing and analyzing complex
decisions
 It was developed by Thomas L.
Saaty in the 1970s
 Application in group decision
making.
IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 2
Analytic Hierarchy Process (AHP) (contd…)

 Decision analysis problems involving finite number


of alternatives arise frequently in practical situations
 One must remember that the type of data available
for analysis, based on which one has to draw some
conclusions can be deterministic, probabilistic or
uncertain
 When the data is uncertain, then one of the many
tools used for analysis is Analytical Hierarchy
Process (AHP)

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 3


Analytic Hierarchy Process (AHP) (contd…)

 In AHP, subjective judgement is quantified


in logical manner and then utilized to reach
some meaningful conclusions
 One must remember that the decision
makers assessment towards risk and his/her
attitude towards return or average benefit
reflects the decision makers overall outlook
about any decision process

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 4


Example: AHP
 Consider Ram has received the final calls from IIMA,
IIMB and IIMC. His main criterion based on which he
will take the decision is
1. Academic reputation
2. Placement potential
 For his academic reputation is two (2) more important
than placement potential. Thus placement potential is
1/3, while academic reputation is 2/3

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 5


Example: AHP (contd…)
 Thus Ram ranks this as
  Percent weight estimates
Criterion

IIMA IIMB IIMC

Academic reputation 0.40 0.25 0.35

Placement potential 0.30 0.40 0.30

 IIMA: (0.30*1/3+0.40*2/3)
 IIMB: (0.40*1/3+0.25*2/3)
 IIMC: (0.30*1/3+0.35*2/3)

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 6


Example: AHP (contd…)
 Now consider Rams brother Shyam, also has got calls from the same three
institutes and both want to be in the same place, so that their parents can
reduce their overall cost of expenditure
Decision: Select IIM
Hierarchy # 1: Placement potential Academic Reputation
¼ ¾
Placement potential
Alternatives: IIMA IIMB IIMC
0.25 0.25 0.50
Academic Reputation
Alternatives: IIMA IIMB IIMC
0.35 0.35 0.30

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 7


Example: AHP (contd…)
 IIMA: (0.25*1/4+0.35*3/4)
 IIMB: (0.25*1/4+0.35*3/4)
 IIMC: (0.50*1/4+0.30*3/4)

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 8


Example: AHP (contd…)
So Rams and Shyams collective hierarchy is as
given
Decision Select IIM
Hierarchy # 1 Ram Shyam
0.5 (p) 0.5 (q)
Hierarchy # 2: PP AR PP AR
1/3 2/3 ¼ ¾
(p1) (p2) (q1) (q2)

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 9


Example: AHP (contd…)
Alternatives: IIMA IIMB IIMC 0.30 0.40 0.30
(p11) (p12) (p13)
Alternatives: IIMA IIMB IIMC 0.30 0.40 0.30
(p21) (p22) (p23)
Alternatives: IIMA IIMB IIMC 0.25 0.25 0.50
(q11) (q12) (q13)
Alternatives: IIMA IIMB IIMC 0.35 0.35 0.30
(q21) (q22) (q23)

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 10


Example: AHP (contd…)
So
IIMA: p*p1*p11 + p*p2*p21+q*q1*q11+q*q2*q21
IIMB: p*p1*p12 + p*p2*p22+q*q1*q12+q*q2*q22
IIMC: p*p1*p13 + p*p2*p23+q*q1*q13+q*q2*q23

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 11


Example: AHP (contd…)
Wide range of applications exists:
Selecting a car for purchasing
Deciding upon a place to visit for
vacation
Deciding upon an MBA program
after graduation

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 12


Example: AHP (contd…)
AHP algorithm is basically composed of two
steps:
Determine the relative weights of the decision
criteria
Determine the relative rankings (priorities) of
alternatives
Both qualitative and quantitative information
can be compared by using informed judgments to
derive weights and priorities.

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 13


Example: AHP (contd…)
 Objective: Selecting a car
 Criteria: Style, Cost, Fuel-
economy
 Alternatives: Civic , i20 ,
Escort, Alto

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 14


Example: AHP (contd…)
S e lec ting
a N e w C ar

S tyle C ost F u e l E c on o m y

Civic i20 Escort Alto


Alternative courses of action

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 15


Example: AHP (contd…)
Ranking Scale for Criteria & Alternatives

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 16


Example: AHP (contd…)
Ranking Scale for Criteria & Alternatives
Style Cost Fuel Economy

Style 1 1/2 3

Cost 2 1 4

Fuel Economy 1/3 1/4 1

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 17


Example: AHP (contd…)
Ranking Scale for Criteria & Alternatives

Row
Normalized 0.30 0.28 0.37
1 0.5 3 Column Sums
averages 0.32
A= 2 1 4 0.60 0.57 0.51 X= 0.56
0.33 0.25 1.0 0.10 0.15 0.12 0.12
Priority vector
Column sums 3.33 1.75 8.00 1.00 1.00 1.00

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 18


Example: AHP (contd…)
Criteria Weights
Criteria weights
 Style 0.32
 Cost 0.56
 Fuel Economy 0.12
Selecting a New Car
1.00

Style Cost Fuel Economy


0.32 0.56 0.12

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 19


Example: AHP (contd..)
Checking for consistency
 The next stage is to calculate a Consistency Ratio
(CR) to measure how consistent the judgments have
been relative to large samples of purely random
judgments.
 AHP evaluations are based on the assumption that
the decision maker is rational, i.e., if A is preferred
to B and B is preferred to C, then A is preferred to C.
 If 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.

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 20


Example: AHP (contd..)
Calculation of consistency ratio
 The next stage is to calculate max so as to lead to the Consistency
Index (CI) and the Consistency Ratio.
 Consider [Ax = max x] where x is the Eigenvector.
A x Ax x
1 0.5 3 0.32 0.98 0.32
2 1 4 0.56 = 1.68 = max 0.56
0.33 0.25 1.0 0.12 0.36 0.12

 λmax= average{0.98/0.32, 1.68/0.56, 0.36/0.12}=3.04


 CI = (λmax-n)/(n-1)=(3.04-3)/(3-1)= 0.02

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 21


Example: AHP (contd..)
 CR = CI/RI where RI is the random index

n 1 2 3 4 5 6 7
R.I. 0 0 0 0.52 0.88 1.11 1.25
1.35

 C.I. = 0.02
n=3
RI = 0.50 (from table)
 Hence: CR = (CI/RI) = 0.02/0.52 = 0.04
 CR ≤ 0.1 indicates sufficient consistency for decision.

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 22


Example: AHP (contd..): Ranking Alternatives
Priority vector
Style Civic i20 Escort Alto
Civic 1 1/4 4 1/6 0.13
i20 4 1 4 1/4 0.24
Escort 1/4 1/4 1 1/5 0.07
Alto 6 4 5 1 0.56

Cost Civic i20 Escort Alto


Civic 1 2 5 1 0.38
i20 1/2 1 3 2 0.29
Escort 1/5 1/3 1 1/4 0.07
0.26
Alto 1 1/2 4 1
IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 23
AHP (contd..) Ranking alternatives
Kilometer/litre Priority Vector

Fuel Economy Civic 34 0.30


i20 27 0.24
Escort 24 0.21
Alto 28 0.25
1.0
113
Since fuel economy is a quantitative measure, fuel
consumption ratios can be used to determine the
relative ranking of alternatives.

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 24


Example: AHP (contd..): Ranking Alternatives
Selecting a New Car
1.00

Style Cost Fuel Economy


0.32 0.56 0.12

Civic 0.13 Civic 0.38 Civic 0.30


i20 0.24 i20 0.29 i20 0.24
Escort 0.07 Escort 0.07 Escort 0.21
Alto 0.56 Alto 0.26 Alto 0.25

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 25


Example: AHP (contd..): Ranking Alternatives

Civic 0.13 0.38 0.30 0.28


0.32
i20 0.24 0.29 0.24 0.25
Escort 0.07 0.07 0.21 x 0.56 = 0.07
Alto 0.56 0.26 0.25 0.12 0.34

Priority matrix Criteria Weights

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 26


AHP (contd..): Including Cost as a Decision Criteria

 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!
 Instead one may think of normalizing the costs directly and calculate
the cost/benefit ratio for comparing alternatives!
Normalized Cost/Benefits
Cost Benefits
Cost Ratio
Civic 620000 0.22 0.28 0.78
i20 900000 0.28 0.25 1.12
Escort 540000 0.17 0.07 2.42
Alto 1080000 0.33 0.34 0.97

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 27


Example: AHP (contd..)
 Escort is the winner with the
highest benefit to Cost Ratio,
hence it is 1 st

 2nd position is that of i20


 At 3rd is Alto
 While 4th position goes to Civic

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 28


AHP (contd..) (Pros)
 It allows multi criteria decision making
 It is applicable when it is difficult to
formulate criteria evaluations, i.e., it
allows qualitative evaluation as well as
quantitative evaluation
 It is applicable for group decision
making environments

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 29


AHP (contd..) (Cons)
 There are hidden assumptions like consistency
 Repeating evaluations is cumbersome
 It is difficult to use when the number of criteria
or alternatives is high, i.e., more than 7
 It is difficult to add a new criterion or
alternative
 It is difficult to take out an existing criterion or
alternative,
alternative since the best alternative might
differ if the worst one is excluded

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 30


AHP (contd..)

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 31


AHP (contd..)
Now if the matrix is consistent,
consistent then its form
will be
 w1 w1  w1 
 
 w2 w2  w2 
N 
   
 
w
 n wn  wn   1 a12  a1n 
 
 a21 1  a2 n 
Such that we have: A
   
 
a an 2  1 
 n1

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 32


AHP (contd..)
and:

1 a12 a1n
w1    
1  a21    an1  a12  1    an 2  a1n  a2 n    1

1 a12 a1n
   

1 
1 1 


 a12  1   
1 

a1n  a2 n    1
 
 a12 a1n   a2 n 

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 33


AHP (contd..)
Thus we have: 
 1
w1

w1 

 w2 wn 
 w2 w2 
w 1 
wn 
A 1 
    
 
w wn 
 n  1 
 w1 w2 
Hence: A(nXn)w(nX1) = nw(nX1) iff A is consistent and in case
of inconsistency we try to find w  1 w
n

i 
n j 1
ij  i  1,2, , n

IME634:MDA RNSengupta,IME Dept.,IIT Kanpur,INDIA 34

You might also like