Resolution of incomplete conditions in rule-based expert systems

Abstract

The authors address the complicated problem of incorporated conditions in rule-based expert systems. Incomplete conditions are defined as having insufficient conditions in the inference chain. The authors propose a precise representation and interpretation of the rule-based system with clear guidelines given for the meaning of the uncertain conditions and the resulting conclusions. To facilitate the accurate evaluation of the incomplete conditions, weighting factors are used and assigned to each antecedent relative to other antecedents in the same rule. These weighting factors are determined by the domain experts and knowledge engineers during knowledge acquisition periods based on three proposed elements, the significance of which is discussed.<>