A computational paradigm that integrates rule-based and model-based reasoning in expert systems

Abstract

This article presents a new computational paradigm that integrates rule-based and model-based reasoning in expert systems. Our experience in expert systems research and development indicates that the rule-based technique is simple, elegant, and efficient; whereas the model-based approach is complex but powerful, CPU-consuming but robust. Combining both the rule-based and the model-based methods into one paradigm means having the best of both worlds. to achieve this goal, we have extended the Prolog unification algorithm to accommodate semantic unification. the resulting computational procedure is named R.M. This new inference procedure uses rule-based reasoning by default, and it automatically invokes model-based reasoning when all the rules become inapplicable, but it returns to rule-based reasoning whenever the rules become usable again. the idea behind this problem-solving strategy is to achieve maximum efficiency as well as robustness in expert systems. Examples are used throughout the article to illustrate our notions. the article also sketches an application in the domain of telecommunication networks maintenance and describes our experimental results.