Vous êtes sur la page 1sur 5

Pair-wise Comparison Method:

Compare two attributes at a time and give a preference ratio

between them (ratio wi/wj indicates how much attribute i is

preferred to relative to attribute j);

Pair-wise preference ratios are presented in the cells of a

table;

Consistency checks between weightings are possible and

necessary;

Weights can also be normalized (put on a 0 to 1 scale).


Example: Vehicle selection

j = 1 2 3

i Criteria Ride quality Purchase Maintenance

price Expense

1 Ride quality w1/w1 = 1 w1/w2 = 1.5 w1/w3= 2.0

2 Purchase price w2/w2 = 1 w2/w3 = ?

3 Maintenance w3/w3 = 1

expense

Values for the cells shown as empty are inverses from other

cells

Assess w2/w3, but recognize that for consistency it should be:

(w1/w3) / ( w1/w2) = (w2/w3) = (2.0) / (1.5) = 1.33

If the assessed value of w2/w3 doesn't equal 1.33, then

the analyst should iterate on the assessment of all wi/wj ;

Thus, consistency checks are possible.


Weightings of individual attributes can be determined by

assigning w1 = 1 and then directly determining other relative

weightings (e.g., w2 = 0.66); Further, these weightings can

be normalized (i.e., put on a 0 to 1 scale) by prorating over

the sum of all weightings (e.g., w1= 1.0/2.16 = 0.463);

Analytic Hierarchy Process

Builds on the previously discussed Pair-Wise Comparison

Model for determining attribute weights and includes a

similar pair-wise comparison of alternatives relative to and for

every unique attribute and sub-attribute.

New key step: Assess the relative superiority of one

alternative compared to another for a given specific

attribute (or sub-attribute)

Example: For Attribute = Vehicle ride quality

j = 1 2 3
i Alternative Lexus Mercedes BMW

Benz

1 Lexus s1/s1 = 1 s1/s2 = 1.2 s1/s3= 1.4

2 Mercedes Benz s2/s2 = 1 s2/s3 = ?

3 BMW s3/s3 = 1

As with pair-wise comparison of attribute weights,

consistency checks are again possible

Assess s2/s3, but recognize that for consistency it should be:

(s1/s3) / ( s1/s2) = (s2/s3) = (1.4) / (1.2) = 1.17

Score values (for each alternative given an attribute) can be

determined by starting with s1 = 1 and then determining

other s values; These values should then be normalized or

put on a 0 to 10 scale (or percentage).

Overall Preference Index values for each alternative can

then be determined (as before) by multiplying attributes

weights by attribute scores and summing these up.


AHP software is now available that makes use of matrix

algebra, Eigen values, and consistency checks in reaching

conclusions.

AHP is growing in popularity and acceptance.

Another source on AHP: ASTM E 1765-98: Standard

Practice for Applying Analytical Hierarchy Process (AHP) to

Multiattribute Decision Analysis of Investments Related to

Buildings and Building Systems (originally published in

1995)

Group Assignment:
Set up the exercise by identifying four alternative choices for analysis and
six selection criteria for analysis. Develop a pair-wise weighting matrix
for the six different selection criteria. Then, for one of the selection
criteria, develop a pair-wise score matrix that characterizes how the four
alternative choices compare on the basis of the selected criterion. Finally,
give the priority matrix with the overall priority vector (weights) of each
alternative choice. Put all your work on the PowerPoint and calculation
on Excel and be prepared to present your effort in class!

Vous aimerez peut-être aussi