Approximate reasoning as a basis for rule-based expert systems

The theory of approximate reasoning developed by L.A. Zadeh provides a natural format for representing the knowledge and performing the inferences in the rule-based expert systems by providing a new structure for including the rules that only require the satisfaction to some subset of the requirements in its antecedent.

Abstract

The concept of a rule-based expert system that includes data and production rules is discussed. It is shown how the theory of approximate reasoning developed by L.A. Zadeh (New York, Wiley, 1979) provides a natural format for representing the knowledge and performing the inferences in the rule-based expert systems. The representation ability of the systems is extended by providing a new structure for including the rules that only require the satisfaction to some subset of the requirements in its antecedent. This is accomplished by use of fuzzy quantifiers. A methodology is also provided for the inclusion of a form of uncertainty in the expert system associated with the belief attributed to the data and production rules.