Knowledge discovery in databases: A rule-based attribute-oriented approach

An attribute-oriented induction has been developed in the previous study of knowledge discovery in databases. A concept tree ascension technique is applied in concept generalization. In this paper, we extend the background knowledge representation from an unconditional non-rule-based concept hierarchy to a rule-based concept hierarchy, which enhances greatly its representation power. An efficient rule-based attribute-oriented induction algorithm is developed to facilitate learning with a rule-based concept graph. An information loss problem which is special to rule-based induction is described together with a solution suggested.