A typical learning task of a data mining system is to define an expert's knowledge in terms of properties available to the learner. This means that the classification an expert would give to an object should be available to the learner in concepts he will understand, and thus be able make the same classification for such objects himself. This is done by having rules which on the basis of condition attribute values give a classification, i.e. a value for the decision attribute.
The following form is used for the rules (decision rules):
where is condition attribute number i, is a value within the domain of , D is the decision attribute and is a classification (value in the domain of D). means a conjunction and means implication. This is a formula of propositional logic, and we will describe its meaning.
If such a rule is meant to cover all cases, it is called a definite
rule, but default rules could also be represented in this way. The standard
definition of implication is used in a decision rule, and the meaning is the
following:
Whenever is true for an object, then . This means that the
expert would say that for an object satisfying the preconditions of the rule,
he would classify that object as belonging to the class .