Multi-Criteria Decision Marking

1
Analytic Hierarchy Process (AHP)

It was developed by Thomas L. Saaty in the 1970s and has been

extensively studied and refined since then.

It has application in group decision making and is used around the

world in a wide variety of decision situations, in fields such as
government, business, industry, healthcare, and education.

Method for ranking decision alternatives and

selecting the best one given multiple criteria.
2
 Some examples
Goal:
To select a house

Decision alternatives:
Location A, B, C & D

Criteria:
Cost
Proximity to school
Proximity to Metro
Closer to shopping places
Closer to hospitals

Goal:
To select a best car

Decision alternatives: http://en.wikipedia.org/wiki/Analytic_hierarchy_process

i10, Swift, Figo

Criteria:
Price
Mileage
Appearance
3
Reliability
 Analytical Hierarchy Process
• Designed to solve multicriteria decision problems

• Model under conditions of certainty where we

know all of the info, but we need to trade-off
many factors.

• We are comparing several alternatives based on

the same set of attributes.

4
 AHP - Steps
1. Develop model hierarchy
2. Make pairwise comparisons
3. Synthesize judgments
4. Check for consistency
Synthesize Judgement – the process of calculating the
priority of each criterion in terms of its contribution to the
overall goal of achieving your goal

5
 1. Develop model Hierarchy

• Develop a visual representation of the problem in terms of

the overall goal, the criteria to be used and the decision
alternatives

6
 Example: Car Selection Problem

• Hierarchy for the Car Selection Problem

Overall Goal To buy a Car that fits my preferences

Criteria Cost Reliability Fuel Economy

Decision i10 Swift Figo

Alternatives

7
 2. Make Pairwise Comparisons &
3. Synthesize judgment
Preference Scale for the Pairwise Comparisons

Extremely
Equal Moderate Strong Very Strong Preferred

1 3 5 7 9 8
• Forming the Pairwise Comparison Matrix for Cost

Swift is moderately preferred to i10, i10 in the Swift row is 3 and

Swift entry in the i10 row is 1/3.

Swift is very strongly preferred to Figo, Figo's entry in the Swift row
is 7 and Swift's entry in the Figo row is 1/7.

i10 is strongly to very strongly preferred to Figo, Figo's entry in the

i10 row is 6 and i10's entry in the Figo row is 1/6.

• Pairwise Comparison Matrix for Cost

i10 Swift Figo

i10 1 1/3 6
Swift 3 1 7
Figo 1/6 1/7 1

9
• Normalized Matrix for Cost
Divide each entry in the pairwise comparison matrix by its
corresponding column sum. For example, for i10, the column
sum = 1 + 3 + 1/6 = 25/6. This gives:

i10 Swift Figo

i10 6/25 7/31 6/14

Swift 18/25 21/31 7/14
Figo 1/25 3/31 1/14

10
• Priority Vector For Cost
The priority vector is determined by averaging
the row entries in the normalized matrix. Converting
to decimals we get:

i10: ( 6/25 + 7/31 + 6/14)/3 = .298

Swift: (18/25 + 21/31 + 7/14)/3 = .632
Figo: ( 1/25 + 3/31 + 1/14)/3 = .069

11
• Forming the Pairwise Comparison Matrix for Reliability

It is determined that for reliability of i10 is very

strongly preferable to Swift and equally to moderately
preferable to Figo. Also, Swift is strongly preferable to Figo.

• Pairwise Comparison Matrix for Reliability

i10 Swift Figo

i10 1 7 2
Swift 1/7 1 5
Figo 1/2 1/5 1

12
• Normalized Matrix for Reliability
Divide each entry in the pairwise comparison
matrix by its corresponding column sum. For example,
for i10, the column sum = 1 + 1/7 + 1/2 = 23/14. This
gives:
i10 Swift Figo

i10 14/23 35/41 2/8

Swift 2/23 5/41 5/8
Figo 7/23 1/41 1/8

13
• Priority Vector For Reliability
The priority vector is determined by averaging the
row entries in the normalized matrix. Converting to
decimals we get:

i10: (14/23 + 35/41 + 2/8)/3 = .571

Swift: ( 2/23 + 5/41 + 5/8)/3 = .278
Figo: ( 7/23 + 1/41 + 1/8)/3 = .151

14
• Forming the Pairwise Comparison Matrix for Fuel Economy

It is determined that for Fuel Economy, i10 is equally

preferable to Figo. Both i10 and Figo are very strongly
to extremely preferable to Swift.

• Pairwise Comparison Matrix for Fuel Economy

i10 Swift Figo

i10 1 8 1
Swift 1/8 1 1/8
Figo 1 8 1

15
• Normalized Matrix for Fuel Economy
Divide each entry in the pairwise comparison
matrix by its corresponding column sum.

i10 Swift Figo

i10 8/17 8/17 8/17

Swift 1/17 1/17 1/17
Figo 8/17 8/17 8/17

16
• Priority Vector For Fuel Economy
The priority vector is determined by averaging the
row entries in the normalized matrix. Converting to
decimals we get:

i10: (8/17 + 8/17 + 8/17)/3 = .471

Swift: (1/17 + 1/17 + 1/17)/3 = .059
Figo: (8/17 + 8/17 + 8/17)/3 = .471

17
• Forming the Pairwise Comparison Matrix for Criteria

It is determined that in terms of criteria, cost is

extremely preferable to Fuel Economy and very
strongly preferable to reliability, and that reliability is
very strongly preferable to Fuel Economy.

• Pairwise Comparison Matrix for Criteria

Cost Reliability Fuel Eco.

Cost 1 7 9
Reliability 1/7 1 7
Fuel Eco. 1/9 1/7 1

18
• Normalized Matrix for Criteria
Divide each entry in the pairwise comparison
matrix by its corresponding column sum.

Cost Reliability Fuel Eco.

Cost 63/79 49/57 9/17

Reliability 9/79 7/57 7/17
Fuel Eco. 7/79 1/57 1/17

19
• Priority Vector For Criteria
The priority vector is determined by averaging the
row entries in the normalized matrix. Converting to
decimals we get:

Cost: (63/79 + 49/57 + 9/17)/3 = .729

Reliability: ( 9/79 + 7/57 + 7/17)/3 = .216
Fuel Eco.: ( 7/79 + 1/57 + 1/17)/3 = .055

20
• Overall Priority Vector
The overall priorities are determined by multiplying the
priority vector of the criteria by the priorities for each decision
alternative for each objective.

Priority Vector
for Criteria [ .729 .216 .055 ]

Cost Reliability Fuel Eco.

i10 .298 .571 .471
Swift .632 .278 .059
Figo .069 .151 .471

21
• Overall Priority Vector (continued)
Thus, the overall priority vector is:

i10: (.729)(.298) + (.216)(.571) + (.055)(.471) = .366

Swift: (.729)(.632) + (.216)(.278) + (.055)(.059) = .524
Figo: (.729)(.069) + (.216)(.151) + (.055)(.471) = .109

Swift appears to be the overall recommendation.

22
 4. Check for Consistency
• A key step in the making of several pairwise comparisons is
considering the consistency of the pairwise judgements.
– Example: If A compared to B = 3 and B compared to C = 2 then
A compared to C should = 3x2 = 6. If it wasn’t, some
inconsistency would occur.

• This type of inconsistency increases as the number of pairwise

comparison increases.

• Certain level of inconsistencies are expected; however, if the

amount inconsistencies is too high, then the results of AHP is
questionable.

• With AHP, we can measure the degree of consistency; and if

unacceptable, we can revise pairwise comparisons
23
Checking Consistency…
• Inconsistencies are measured by Consistency Index (CI) and
Consistency Ratio (CR)

• CR = CI / RI

(RI is Random Index - this is the consistency index of a randomly

generated pairwise comparison matrix.

If CR > 0.1, then there may be serious inconsistencies in the

pairwise comparison.)

24
Checking Consistency for Cost
•Multiply pair-wise comparison matrix by its
priority:

1 1/3 6 .298 .923

3 1 7 .632 = 2.009
1/6 1/7 1 .069 .209

•Divide these number by their priorities to get:

.923/.298 = 3.097
2.009/.632 = 3.179
.209/.069 = 3.029

• Average the above results to get max.

max = (3.097 + 3.179 + 3.029)/3 = 3.102 25
Checking Consistency

• Compute the consistence index, CI, for two terms.

CI = (max - n)/(n - 1) = (3.102 - 3)/2 = .051
n = number of items being compared

• Compute the consistency ratio, CR, by CI/RI, where RI = .58

for 3 factors:
CR = CI/RI = .051/.58 = .088
Since the consistency ratio, CR, is less than 0.10, this is well within
the acceptable range for consistency.

Random Index Values for the Comparison of n items

Super Decisions Software for Decision-Making

The Super Decisions software

implements the Analytic
Network Process developed
by Dr. Thomas Saaty. The
program was written by the
ANP Team, working for
the Creative Decisions
Foundation.

Creative Decisions Foundation 4922 Ellsworth

Avenue Pittsburgh, PA 15213 USA e-mail:
support@superdecisions.com

http://www.superdecisions.com/
27
AHP with Spreadsheets

28
Back-up

29