Professional Documents
Culture Documents
The Analytic Hierarchy Process: Value of Attribute 2
The Analytic Hierarchy Process: Value of Attribute 2
Value of
x1 x'1 attribute 1
Now choose the job offer with the highest overall score. Note that the overall score gives
more weight to a job offer’s score on the more important objectives. Computing each job’s
overall score, we obtain
Job 1 overall score .5115(.571)
.0986(.159)
.2433(.088)
.1466(.069) .339
Job 2 overall score .5115(.286)
.0986(.252)
.2433(.669)
.1466(.426) .396
Job 3 overall score .5115(.143)
.0986(.589)
.2433(.243)
.1466(.506) .265
Thus, the AHP would indicate that Jane should accept job 2.
Suppose there are n objectives. We begin by writing down an n n matrix (known as the
pairwise comparison matrix) A. The entry in row i and column j of A (call it aij) indi-
cates how much more important objective i is than objective j. “Importance” is to be mea-
sured on an integer-valued 1–9 scale, with each number having the interpretation shown
in Table 14. For all i, it is necessary that aii 1. If, for example, a13 3, objective 1 is
weakly more important than objective 3. If aij k, then for consistency, it is necessary
that aji 1k. Thus, if a13 3, then a31 13 must hold.
Suppose that Jane has identified the following pairwise comparison matrix for her four
objectives (SAL high salary; QL high quality of life; IW interest in work; NF
nearness to family):
SAL QL IW NF
SAL 1 5 2 4
1
QL 1 1 1
5 2 2
1
IW 2 1 2
2
1
NF 2 1 1
4 2
Suppose there are n objectives. Let wi the weight given to objective i. To describe
how the AHP determines the wi’s, let’s suppose the decision maker is perfectly consistent.
Then her pairwise comparison matrix should be of the following form:
w w w
A 1 1 1 (13)
w1 w2 wn
w2 w2 w
2
A w1 w2 wn (13)
wn wn wn
w1 w2 wn
For example, suppose that w1 12 and w2 16. Then objective 1 is three times as impor-
tant as objective 2, so
w
a12 1 3
w2
Now suppose that a consistent decision maker has a pairwise comparison matrix A of the
form (13). How can we recover the vector w [w1 w2 wn] from A? Consider the
system of n equations
AwT wT (14)
.5128 .5000 .5000 .5333
.1026 .1000 .1250 .0667
Anorm
.2564 .2000 .2500 .2667
.1282 .2000 .1250 .1333
Step 2 To find an approximation to wmax (to be used as our estimate of w), proceed as
follows. Estimate wi as the average of the entries in row i of Anorm. This yields (as previ-
ously stated)
.5128
.5000
.5000
.5333
w1 .5115
4
.1026
.1000
.1250
.0667
w2 .0986
4
.2564
.2000
.2500
.2667
w3 .2433
4
.1282
.2000
.1250
.1333
w4 .1466
4
Intuitively, why does w1 approximate the weight that objective 1 (salary) should be given?
The percentage of the weight that SAL is given in pairwise comparisons of each objec-
tive to SAL is .5128. Similarly, .50 represents the percentage of total weight that SAL is
given in pairwise comparisons of each objective to QL. Thus, we see that the four num-
bers averaged to obtain w1 each represents in some way a measure of the total weight at-
tached to SAL. Thus, averaging these numbers should give a good estimate of the per-
centage of the total weight that should be given to SAL.
1 5 2 4 .5115 2.0775
1 1 1 1 .0986 0.3959
AwT 5
1
2 2
2 1 2 .2433 0.9894
2
1
2 1 1 .1466 0.5933
4 2
Step 2 Compute
in
ith entr y in AwT
1
n
i1 ith en try in w
T
2 0
3 .58
4 .90
5 1.12
6 1.24
7 1.32
8 1.41
9 1.45
10 1.51
Job 1 1 2 4
Job 2 1 1 2
2
1 1
Job 3 1
4 2
.571 .571 .571
Anorm .286 .286 .286
.143 .143 .143
This yields w [.571 .286 .143]. These weights indicate how well each job “scores” with
respect to the SAL objective. As previously stated in Table 13, we obtain
Job 1 salary score .571
Job 2 salary score .286
Job 3 salary score .143
Since all three columns of the pairwise comparison matrix for salary are identical, Jane’s
pairwise comparisons for salary exhibit perfect consistency.
Suppose Jane’s pairwise comparison matrix for quality of life (QL) is as follows:
Job 3 Job 1 Job 2 Job 3
Job 1 1 1 1
2 3
Job 2 2 1 1
3
Job 3 3 3 1
Then
1 1
1
6 9 5
Anorm 1
3
2
9
1
5
1 6 3
2 9 5
and we obtain
1
1
1
6 9 5
Job 1 quality of life score .159
3
1
2
1
3 9 5
Job 2 quality of life score .252
3
1
6
3
2 9 5
Job 3 quality of life score .589
3
For interest in work, suppose the pairwise comparison matrix is as follows:
Job 3 Job 1 Job 2 Job 3
Job 1 1 1 1
7 3
Job 2 7 1 3
Job 3 3 1 1
3
Job 1 1 1 1
4 7
Job 2 4 1 2
Job 3 7 2 1
Routine calculations yield
Job 1 score for nearness to family .069
Job 2 score for nearness to family .426
Job 3 score for nearness to family .506
As described earlier, we can now “synthesize” the objective weights with the scores of
each job on each objective to obtain an overall score for each alternative (in this case,
each job offer). As before, we find that job offer 2 is most preferred, followed by job of-
fer 1, with job offer 3 the least preferred.
We close by noting that AHP has been applied by decision makers in countless areas,
including accounting, finance, marketing, energy resource planning, microcomputer se-
lection, sociology, architecture, and political science. See Zahedi (1986) and Saaty (1988)
for a discussion of applications of AHP.