Practical Applications of Data Mining

36 Chapter 2 assoCiation rules

Furthermore, the generated rules can be used for intelligent query process-

ing, cooperative query answering, and semantic query optimization.

2.3.1 Concept Hierarchies

Since a database may consist of multiple relations, each with a varying num-

ber of attributes, containing a huge amount of data, not all the data in the

database may be relevant to a specific learning task. Rule induction in data-

bases relies mainly on generalization. To derive a characteristic rule by gen-

eralization, the task-relevant data are collected into a class called the target

class, whereas to derive a discrimination rule, the data are collected into two

classes, the target class and the contrasting class.

Concept hierarchies are