Management of Uncertainty in an Expert System

The objective of this paper is the application of two numeric approaches used in the treatment of uncertainty and its respective validation. This research accomplished a study on the probabilistic reasoning in an Expert System: Bayesian Networks and Certainty Factors, presenting as application domain the Clinical Diagnosis of the Epileptic Seizures. The system was developed with KADS methodology and is based on the classification of the type of seizures of the International League Against Epilepsy /ILAE81 [1]. The concurrent validation showed that the Sensibility and Negative Predicted Value was more precise through the Bayesian Networks while Specificity and Affirmative Predicted Value showed similarity in both approaches.