Automated Discovery of Hierarchical Ripple-Down Rules (HRDRs)

F.M. Ba-Alwi and K.K. Bharadwaj (India)


Data Mining, specificity, generality, RDR, HRDR, hierarchy.


Ripple-Down Rule (RDR) is a machine learning technique that represents the exception rules in a concise manner in order to keep all general rules consistent. Logically, RDRs have exactly the same representational and deductive capabilities as standard production rules. Where they differ is in the strong context provided by the `if-true' ('Except') and `if-false' ('Else' part) links. A RDR is however, unable to capture the taxonomical structure inherent in the knowledge about the real world, and hence is not able to impart control over the specificity part of a precision in decision making. In order to incorporate hierarchical structure into RDR, we have proposed Hierarchical Ripple-Down Rule (HRDR) as an extension of RDR. An HRDR is a RDR augmented with Generality and Specificity operators. In this paper a novel algorithm is proposed that integrates discovery of RDRs and the process of hierarchy generation for automated discovery of HRDR. Examples are given to demonstrate the performance of the proposed algorithm.

Important Links:

Go Back