Coping with Uncertainty in Expert Systems: A Comparative Study

This paper suggests a way to represent the uncertainty of evidence in expert systems. It is proposed that the degree of certainty will be represented by a fuzzy interval rather than a scaler. The degree of necessity represents the lower bound of this fuzzy interval, while the degree of possibility is its upper bound. Belief functions and plausibility measures, are subsets of the set of fuzzy measures. If certain constraints are imposed upon the universe of discourde, e.g. for finite domains, then necessity and possibility are respectively consonant belief functions and possibility measures, thus subsets of the belief functions and plausibility measures. In this study the authors propose to evaluate functions, based on triangular norms and conorms, and to perform all the aggregations required in the matching of assertions and premise and in the weighting and aggregation of conclusions.