The Marketing Book 5th Edition

218 The Marketing Book

classification ambiguity with each subset of
partition:

G(PF)=

k

i=1

W(EiF)G(EiF),

where G(EiF) is the classification ambiguity
with fuzzy evidence Ei  F, w(Ei|F) is the
weight which represents the relative size of
subsetEiFinF:

W(EiF)=

uU

min (Ei(u),F(u))

k

j=1

[^

uU

min (Ej(u),F(u))]

The fuzzy decision tree method considered
here utilizes these functions. In summary, attri-
butes are assigned to nodes based on the lowest
level of ambiguity. A node becomes a leaf node
if the level of subsethood (based on the con-
junction (intersection) of the branches from the
root) is higher than some truth value assigned
to the whole of the decision tree. The classifica-
tion from the leaf node is to the decision class
with the largest subsethood value.
The results of the decision tree are classifi-
cation rules, each with an associated degree of
truth in their classification. These rules are
relatively simple to read and apply.

Artificial intelligence

Artificial intelligence (AI) models have
emerged in the last few years as a follow-up to
simulation, attempting to portray, comprehend
and analyse the reasoning in a range of situa-
tions. Although the two methods of artificial
intelligence (expert systems and neural net-
works) are, in a certain sense, ‘simulations’,
because of the importance and the potential of
these methods, we have introduced them under
a separate stand-alone heading.

Expert systems

Simply defined, an expert system is a computer
program which contains human knowledge or
expertise which it can use to generate reasoned

advice or instructions. The knowledge base is

usually represented internally in the machine

as a set of IF... THEN rules and the ‘inference

engine’ of the expert system matches together

appropriate combinations of rules in order to

generate conclusions.

In determining whether a particular mar-

keting domain is suited for this methodology

the following checklist is useful:

 Are the key relationships in the domain logical

rather than computational? In practical terms,

the answer requires an assessment of whether

the decision area is knowledge-intensive (e.g.

generating new product areas) or

data-intensive (e.g. allocating an advertising

budget across media).

 Is the problem domain semi-structured rather

than structured or unstructured? If the

problem is well structured, a traditional

approach using sequential procedures will be

more efficient than an expert system approach.

This would be true, for example, when the

entire problem-solving sequence can be

enumerated in advance.

 Is knowledge in the domain incomplete? If the

problem is well structured, a traditional

approach using sequential procedures will be

more efficient than an expert system approach.

This would be true, for example, when the

entire problem-solving sequence can be

enumerated in advance. Moreover, for highly

unstructured domains, expert system

performance may be disappointing because the

available problem-solving strategies may be

inadequate.

 Is knowledge in the domain incomplete? In

other words, is it difficult to identify all the

important variables or to specify fully their

interrelationships? Expert systems are

particularly applicable in domains with

incomplete knowledge.

 Will problem solving in the domain require a

direct interface between the manager and the

computer system? A direct interface may be

necessary in situations calling for on-line

decision support. Such situations are generally