You are on page 1of 4

Fig. 1. A simple AHP model.

The analytic hierarchy process

We have a tourism database which the user can access by selecting his preferable hotels to say
in.
That choice isn’t so easy if we have to consider diverse criteria and dealing with its weight in
case if we have more preference for any of criteria
This technique is employed for ranking a set of alternatives or for the selection of the best in a
set of alternatives. The Ranking/selection is done with respect to an overall goal
(OBJECTIVE), which is broken down into a set of criteria ((LOCATION, QUALITY and
PRICE), and forwarding to the possible alternatives (HOTEL IPANEMA, HOTEL
MADEIRA, HOTEL PESTANA).

Application of AHP to a decision problem involves four steps:

Step 1: Structuring of the decision problem into a hierarchical model.


It includes decomposition of the decision problem into elements according to their common
characteristics and the formation of a hierarchical model having different levels. A simple AHP
model (Fig. 1) has three levels (Objective, criteria and alternatives). Though the simple model
with three levels shown in Fig. 1 we try to describe what our model should approximate.

The criteria can be divided further into sub-criteria, such as LOCATION has sub criteria (SEA,
MOUNTAIN, LAKE, etc) and each of them has sub-sub-criteria (HOTEL IPANEMA, HOTEL
MADEIR, etc). The same happens with the other criteria (PRICE, QUALITY) which can be
divided in many sub categories for the user to choose for a better goal achievement.

Step 2: Making pair-wise comparisons and obtaining the judgment matrix.


In this step, the elements of a particular level are compared with respect to a specific element in
the immediate upper level. The resulting weights of the elements may be called the local
weights (to be contrasted with final weights, discussed in Step 4). The opinion of a decision-
maker (DM) is elicited for comparing the elements. Elements are compared pair-wise and
judgments on comparative attractiveness of elements are captured using a rating scale (1–9)
scale in traditional AHP). Usually, an element receiving higher rating is viewed as superior (or
more attractive) compared to another one that receives a lower rating. The comparisons are
used to form a matrix of pair-wise comparisons called the judgment matrix A. Each entry aij of
the judgment matrix are governed by the three rules: aij >0; aij =1/aji ; and aii =1 for all i.
If the transitivity property holds, i.e., aij =aik ∗akj , for all the entries of the matrix, then the
matrix is said to be consistent. If the property does not hold for all the entries, the level of
inconsistency can be captured by a measure called consistency ratio (see next step).

For our model, shown in Fig. 1, four judgment matrices need to be elicited from DM—one for
estimating the local weights of criteria with respect to the goal (see Table 2A for an illustrative
judgment matrix), and one each for computing the local weights of alternatives with respect to
each of the three criteria.
For our case study we must consider three criteria (LOCATION, QUALITY and PRICE), they
accomplish what we consider the most important criteria for choosing the most approximate
goal.

All the entries in Table 2A, except the local weights, are random values.

Table 2
Illustration of AHP calculations for the model shown in Fig. 1

A: Comparison of criteria with respect to goal

LOCALIZATION PRICE QUALITY

C1 C2 C3 Local weights

C1 1 5 4 0.400
C2 3 1 5 0.394
C3 1/4 1/5 3 0.128

Consistency ratio = 0.088

B: Comparison of alternatives with respect to C1 (LOCATION)


A1 A2 A3 Local weights
A1 1 1/3 5 0.279
A2 3 1 7 0.649
A3 1/5 1/7 1 0.072

Consistency ratio = 0.055

C: Comparison of alternatives with respect to C2 (PRICE)


A1 A2 A3 Local weights
A1 1 1/9 1/5 0.060
A2 9 1 4 0.709
A3 5 ¼ 1 0.231

Consistency ratio = 0.061

D: Comparison of alternatives with respect to C3 (QUALITY)


A1 A2 A3 Local weights
A1 1 2 5 0.582
A2 ½ 1 3 0.309
A3 1/5 1/3 1 0.109

Consistency ratio = 0.003

F: Final weights of alternatives A1 0.261


A2 0.590
A3 0.148
Step 3: Local weights and consistency of comparisons.

In this step, local weights of the elements are calculated from the judgment matrices using the
eigenvector method (EVM). The normalized eigenvector corresponding to the principal
eigenvalue of the judgment matrix provides the weights of the corresponding elements.
When EVM is used, consistency ratio (CR) can be computed. For a consistent matrix CR=0.
A value of CR less than 0.1 is considered acceptable because human judgments need not be
always consistent, and there may be inconsistencies introduced because of the nature of scale
used. If CR for a matrix is more than 0.1, judgments should be elicited once again from the DM
till he gives more consistent judgments. Local weights computed for the illustrative judgment
matrices using EVM and the corresponding CR values are also shown in Table 2A–E.

Step 4: Aggregation of weights across various levels to obtain the final weights of
alternatives.

Once the local weights of elements of different levels are obtained as outlined in Step 3, they
are aggregated to obtain final weights of the decision alternatives (elements at the lowest level).
For example, the final weight of alternative A1 is computed using the following hierarchical
(arithmetic) aggregation rule in traditional AHP:

By definition, the weights of alternatives and importance of criteria are normalized so that they
sum to unity.

The final weights computed using the Final weight formula above for the illustration is shown
in Table 2F. Thus, given the
Hypothetical judgments of Table 2A–E, alternative A2 is the most preferred alternative,
followed by alternatives A1 and then A3.

Local weights for consistent judgment matrices


When applied to a consistent matrix, for which weights are known, we can correctly estimate
the true weights. Specifically, suppose that a predetermined set of n weights, say w1,
w2. . . Wn, is used to create a consistent judgment matrix of size n given by
Traditional AHP View

The Elements N are referred to the alternatives that are possible to choose for (Hotel
IPANEMA, Hotel MADEIRA and Hotel PESTANA).

This process is done for every pair .This matrix has positive entries everywhere and satisfies
the reciprocal property .

It is called a reciprocal matrix. We note that if we multiply this matrix by the column vector (
) we obtain the vector NW. That is:

You might also like