Using paths to detect redundancy in rule bases

One of the major causes of inefficiency in rule-based systems is rule base redundancy; hence, its detection is an actively pursued research topic. Most of the existing methods consider rule base redundancy at two levels: at the rule level and at the level of rule sequences or rule chains. The computational complexity and the extent of redundancy that can be detected are two measures of performance that can be used to compare different methods of redundancy detection. We propose the use of "paths" for redundancy detection. Starting from a specification mechanism to abstract the problem solving knowledge of the domain, called goal specification, we have developed software tools for efficient extraction of paths. Based on such paths, we identify rule situations called "rule aberrations" which form the core of the algorithm proposed for redundancy detection. For illustration, the algorithm is applied to an example rule base, and redundant rules and atoms are detected. Our method of redundancy detection is also compared with the existing approaches.<<ETX>>