NAME: SOHOMJIT GANGULY

ROLL NO: 36

SUBJECT: DATA ANALYTICS (C20)

PROJECT TOPIC: ANALYTICAL HIERARCHY PROCESS

 INTRODUCTION

Analytic hierarchy process

The multi-criteria programming made through the use of the analytic hierarchy
process is a technique for decision making in complex environments in which
many variables or criteria are considered in the prioritization and selection of
alternatives or projects.
AHP was developed in the 1970s by Thomas L. Saaty and has since been
extensively studied, and is currently used in decision making for complex
scenarios, where people work together to make decisions when human perceptions,
judgments, and consequences have long-term repercussions (Bhushan & Rai,
2004).
The application of AHP begins with a problem being decomposed into a hierarchy
of criteria so as to be more easily analyzed and compared in an independent
manner (Exhibit 2). After this logical hierarchy is constructed, the decision makers
can systematically assess the alternatives by making pair-wise comparisons for
each of the chosen criteria. This comparison may use concrete data from the
alternatives or human judgments as a way to input subjacent information (Saaty,
2008).
AHP transforms the comparisons, which are most often empirical, into numerical
values that are further processed and compared. The weight of each factor allows
the assessment of each one of the elements inside the defined hierarchy. This
capability of converting empirical data into mathematical models is the main
distinctive contribution of the AHP technique when contrasted with other
comparing techniques.
After all the comparisons have been made, and the relative weights between each
of the criteria to be evaluated have been established, the numerical probability of
each alternative is calculated. This probability determines the likelihood that the
alternative has to fulfill the expected goal. The higher the probability, the better the
chances the alternative has to satisfy the final goal of the portfolio.
The mathematical calculation involved in the AHP process may at first seem
simple, but when dealing with more complex cases, the analyses and calculations
become deeper and more exhaustive.

The Comparison Scale (SAATY scale)

The comparison between two elements using AHP can be done in different ways
(Triantaphyllou & Mann, 1995). However, the relative importance scale between
two alternatives as suggested by Saaty (SAATY, 2005) is the most widely used.
Attributing values that vary from 1 to 9, the scale determines the relative
importance of an alternative when compared with another alternative, as we can
see in Exhibit 3.
Exhibit 3 – Saaty's Scale of Relative Importance (Saaty, 2005)

It is common to always use odd numbers from the table above to make sure there is
a reasonable distinction among the measurement points. The use of even numbers
should only be adopted if there is a need for negotiation between the evaluators.
When a natural consensus cannot be reached, it raises the need to determine a
middle point as the negotiated solution (compromise) (Saaty, 1980).

The comparison matrix is constructed from the Saaty scale (Exhibit 4).

Exhibit 4- Comparison Matrix (presuming that Criterion 1 dominates over

Criterion 2)
An example of the application of ahp in a portfolio
In order to serve as an example of the AHP calculations for a prioritization of
projects, the development of a fictitious decision model for the ACME
Organization has been chosen. As the example is further developed, the concepts,
terms, and approaches to AHP will be discussed and analyzed.
The first step in building the AHP model lies in the determination of the criteria
that will be used. As already mentioned, each organization develops and structures
its own set of criteria, which in turn must be aligned to the strategic objectives of
the organization.
For our fictitious ACME organization, we will assume that a study has been made
together with the finance, strategy planning, and project management areas of the
criteria to be used. The following set of 12 (twelve) criteria has been accepted and
grouped into 4 (four) categories, as shown on the hierarchy depicted in Exhibit 5.

Exhibit 5 – Hierarchy of Criteria for the Fictitious ACME Organization

Determining the Comparison Matrix, the Priority Vector

(eigenvector), and the Inconsistency
After the hierarchy has been established, the criteria must be evaluated in pairs so
as to determine the relative importance between them and their relative weight to
the global goal.
The evaluation begins by determining the relative weight of the initial criteria
groups (Exhibit 6). Exhibit 7 shows the relative weight data between the criteria
that have been determined by ACME's decision makers.
Exhibit 6 – ACME's Initial Group of Criteria/Objectives

Exhibit 7 – Comparison Matrix for ACME's Group of Criteria

In order to interpret and give relative weights to each criterion, it is necessary to
normalize the previous comparison matrix. The normalization is made by dividing
each value by the total column value (Exhibits 8 and 9).

Exhibit 8 – Calculating the Total of Each Column

Exhibit 9 – Comparison Matrix for ACME's Group of Criteria after Normalization
The contribution of each criterion to the organizational goal is determined by
calculations made using the priority vector (or Eigenvector). The Eigenvector
shows the relative weights between each criterion; it is obtained in an approximate
manner by calculating the mathematical average of all criteria, as depicted in
Exhibit 10. We can observe that the sum of all values from the vector is always
equal to one (1). The exact calculation of the Eigenvector is determined only in
specific cases. This approximation is applied most of the times in order to simplify
the calculation process, because the difference between the exact value and the
approximate value is less than 10% (Kostlan, 1991).

Exhibit 10 – Eigenvector Calculation (ACME)

For comparison purposes, a mathematical software application has been used to
calculate the exact value for the Eigenvector through the use of potential matrices.
The results are shown in Exhibit 11.

Exhibit 11 – Comparative Results for the Calculation of the Eigenvector –

Approximate and Exact
It can be observed that the approximate and exact values are very close to each
other, so the calculation of the exact vector requires a mathematical effort that can
be exempted (Kostlan, 1991)
The values found in the Eigenvector have a direct physical meaning in AHP; they
determine the participation or weight of that criterion relative to the total result of
the goal. For example, in our ACME organization, the strategic criteria have a
weight of 46.04% (exact calculation of the Eigenvector) relative to the total goal. A
positive evaluation on this factor contributes approximately 7 (seven) times more
than a positive evaluation on the stakeholder commitment criterion (weight
6.84%).

The next step is to look for any data inconsistencies. The objective is to capture
enough information to determine whether the decision makers have been consistent
in their choices (Teknomo, 2006). For example, if the decision makers affirm that
the strategic criteria are more important than the financial criteria and that the
financial criteria are more important than the stakeholder commitment criteria, it
would be inconsistent to affirm that the stakeholder commitment criteria are more
important than the strategic criteria (if A>B and B>C it would be inconsistent that
A<C).

The inconsistency index is based on maximum Eigenvalue, which is calculated by

summing the product of each element in the Eigenvector (see Exhibits 10 and 11)
by the respective column total of the original comparison matrix (see Exhibit 8).
Exhibit 12 demonstrates the calculation of maximum Eigenvalue ( ).

Exhibit 12 – Calculation of Maximum Eigenvalue (using the exact Eigenvector

values)
The calculation of the consistency index (Saaty, 2005) is given by the following
formula

Where CI is the consistency index and n is the number of evaluated criteria.

For our ACME organization, the consistency index (CI) is
In order to verify whether the consistency index (CI) is adequate, Saaty (Saaty,
2005) suggests what has been called the consistency rate (CR), which is
determined by the ratio between the consistency index and the random consistency
index (RI). The matrix will be considered consistent if the resulting ratio is less
than 10%.

The RI value is fixed and is based on the number of evaluated criteria, as shown in
Exhibit 13.

Exhibit 13 – Exhibit of Random Consistency Indices (RI) (SAATY, 2005)

For our ACME organization, the consistency rate for the initial criteria group is

Since its value is less than 10%, the matrix can be considered to be consistent. The
priority criteria results for the first level can be seen in Exhibit 14.

Exhibit 14 – Results of the Comparison Matrix for ACME's Criteria Group,

Demonstrating the Contribution of Each Criterion to the Goal Defined for ACME
By looking at Exhibit 14 and the Eigenvector values, it is evident that the strategic
criteria have a contribution of 46.04% to the goal, whereas the stakeholder
commitment criterion contributes with 6.84% to the goal.
So armed with this knowledge we shall solve the given problem.
Our goal is to purchase a new car. Our purchase is based on different criteria such as cost,
comfort, and safety. We could evaluate several alternatives but let us assume that we have
only two: Car 1 and Car 2.

Based on the given information, analyze the decision of purchasing a car using the analytic
hierarchy.

(1) Develop a model for the decision: Break down the decision into a hierarchy of goals,
criteria, and alternatives.

(2) Derive priorities (weights) for the criteria.

(3) Derive local priorities (preferences) for the alternatives.

(4) Derive Overall Priorities (Model Synthesis).

(5) Making a Final Decision

Given:

a. Pairwise comparison matrix for criteria

Buying a car Cost Comfort Safety

Cost 1 7 3

Comfort 1/7 1 1/3

Safety 1/3 3 1

b. Comparison with respect to cost

Cost Car1 Car 2

Car 1 1 7

Car 2 1/7 1

c. Comparison with respect to comfort

Comfort Car1 Car 2

Car 1 1 1/5

Car 2 5 1
 d. Comparison with respect to safety
Safety Car1 Car 2

Car 1 1 1/9

Car 2 9 1

Now,

Let us create the Prioritization matrix first:

Cost Comfort Safety

Cost 1 7 3
Comfort 1/7 1 1/3
Safety 1/3 3 1
Sum 1.476 11 4.33

So we add the respective columns to a column total (Sum).

Now we find the normalized matrix. That is divide each element of the column with the Sum
of each column.

Cost Comfort Safety

Cost 0.677 0.636 0.692
Comfort 0.096 0.090 0.076
Safety 0.225 0.272 0.23

Now to get the Priority vector we need to add the row values and average them accordingly:

Priority Vector
Cost (0.677+0.636+0.692)/3 0.668
Comfort (0.096+0.09+0.076)/3 0.087
Safety (0.225+0.272+0.23)/3 0.242

Therefore, arranging the ranks:

Criteria Priorities/Rank
Safety 1 (weight 0.242)
Cost 2 (weight 0.668)
Comfort 3 (weight 0.087)
Now comparing with respect to cost:

Cost Car 1 Car 2

Car 1 1 7
Car 2 1/7 1
Sum 1.142 8

So, now like the previous working we find the priority vector:
Car 1 Car 2 Priority Vector
Car 1 0.875 0.875 0.875
Car 2 0.125 0.125 0.125

Now ranking them accordingly:

Cost Rank/Weight
Car 1 1 (0.875)
Car 2 2 (0.125)

Comparing with respect to comfort:

Comfort Car 1 Car 2

Car 1 1 1/5
Car 2 5 1
Sum 6 1.2

So, now like the previous working we find the priority vector:
Car 1 Car 2 Priority Vector
Car 1 0.166 0.166 0.166
Car 2 0.833 0.833 0.833

Now ranking them accordingly:

Cost Rank/Weight
Car 1 2 (0.166)
Car 2 1 (0.833)

Comparing with respect to Safety:

Safety Car 1 Car 2

Car 1 1 1/9
Car 2 9 1
Sum 10 1.111

So, now like the previous working we find the priority vector:
 Car 1 Car 2 Priority Vector
Car 1 0.10 0.10 0.10
Car 2 0.90 0.90 0.90

Now ranking them accordingly:

Cost Rank/Weight
Car 1 2 (0.10)
Car 2 1 (0.90)

Comparing of overall priorities:

Criteria Weights - 0.668 0.087 0.242

Car 1 0.875 0.166 0.10
Car 2 0.125 0.833 0.90

Comparing the weights of Car 1,

 0.875 x 0.668 + 0.166 x 0.087 + 0.10 x 0.242 = 0.623

Comparing the weights of Car 2,

 0.125 x 0.668 + 0.833 x 0.087 + 0.90 x 0.242 = 0.373

So from here we can conclude that:

Our Goal was to purchase a car. The best option among these two cars is Car 1 which has all
the priorities we need. To arrive at this decision, we compare the weights of both cars and see
that Car 1 is our final choice aided by AHP for the decision making.