Improving Knowledge-Based System Performance by Reordering Rule Sequences

May 2001

Proceedings of the Fourteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2001)

Volume

Proceedings of the Fourteenth International Florida Artificial Intelligence Research Society Conference (FLAIRS 2001)

All Papers

In this paper, we argue that KBS validation should not be limited to testing functional properties of the system, such as its input - output behavior, but must also address its dynamic properties, such as its run-time performance. We describe an automated procedure, which under certain limitations can recognize relations between rules, typically expressed as "meta-rules" or "control heuristics" and hard-wired in the KBS’s control strategy. The presented procedure takes as an input the operational version of the knowledge base theory generated by a CTMS-based verification tool, and returns a reordered set of rules which when applied to a specific problem generates possible solutions of that problem in order according to a specified criterion, for example the length of the path leading to the solution. We show an example to illustrate the proposed procedure.