A formal equivalence between propositional expert systems and decision tables is proved, and a practicable procedure given to perform the transformation between propositional expert systems and decision tables. The method gave an order of magnitude speed increase for a well-known expert system in routine use. The method is very general: adaptations are shown for forward and backward chaining inferencing engines, inexact reasoning, and systems where some facts have a high cost and must be determined only if necessary. A particular application for the decision table representation is in real-time expert systems, since a simple hardware implementation is available which gives further orders of magnitude increase in performance. Finally, the decision table representation greatly simplifies the problem of completeness and consistency checking.
Robert M. Colomb, Charles Y. C. Chung