An Approach to Evaluation on the Accuracy of the Knowledge Bases Applied to Diagnosis Systems

The main contributions of this paper consist in stating a proposal for the evaluation and the improvement of the knowledge bases in order to assure their quality.

This paper addresses the issue of analysing selected qualitative properties of the diagnosis systems in a systematic way and designing the set of rules in an appropriate manner. We investigate a common model for the representation of a base knowledge called Object Attribute Table.

The evaluation of a given knowledge base, is measured using the following steps : first we obtain its complete extensional representation, next we deduce the characteristics that an optimal rule base should exhibit. We consider criteria that measure the accuracy of different knowledge bases describing the concepts.