A prediction rule is a rule where certain properties of the data in the body of the rule are related to a value for a designated target attribute. The semantics of such a rule is that all data examples that satisfy the conditions in the rule body are predicted to have the target value specified in the rule head.
Depending on the type of predicted value, we can discriminate between classification rules where the value of a discrete attribute is predicted, and regression rules where the value of a numeric attribute is predicted.
We can also identify descriptive rules where the focus of the rule is to provide a good characterization of the target value, and discriminative ruleswhere the focus is to be able to discriminate the predictive value from other values. For example, a discriminative rule that identifies an elephant among all animals would only need to specify that it has a tusks and a trunk, whereas a descriptive rule would also specify that...