A NEW RIPPLE DOWN RULES INDUCTION ALGORITHM AND ITS LATTICE EXTENSION

摘要

Ripple down rules (RDR), rules with hierarchically structured exceptions, are effective knowledge representation schemes used in knowledge acquisition (KA). In this paper, a new ripple down rules induction algorithm LRI is proposed. It is originally developed to cope with continuous-valued attributes. After analyzing the central process in LRI in detail, we extend LRI to deal with attributes having lattice structure.Then, we slightly weaken the lattice definition and widely enlarge LRI's application range. After that, we demonstrate the application of LRI to induce RDR from standard datasets with various kinds of attributes. The experimental results show that the proposed approach is promising. Finally, we conclude our paper with a summary and outlook on future work.