Evaluation of approximate reasoning methodologies

A methodology and architecture have been developed to compare the effectiveness and efficiency of evidential reasoning techniques based on the precision and completeness of domain information and observations. A generic classification tool (GCT) is developed to analyze the effectiveness of various evidential reasoning techniques. To compare the reasoning paradigms, the representation of the domain and evidential knowledge must be robust as possible without favoring any of the methods. Due to the ability to transform the representation into those required by each method, a variant of the fuzzy relational database model is chosen to represent domain knowledge. This generic evidential reasoning system is then used to compare the generation of support by the Dempster-Shafer theory and a fuzzy reasoning technique on two domains using both precise and imprecise information.<
>