A novel approach to computing rule confidence in rough sets theory

Rough sets theory deals with representation, learning and generalization of uncertain knowledge. It can abstract underlying rules from data. Confidence is the criterion to scaling the reliability of rules. This paper synthesize and improve the prevenient researches on rule confidence in rough sets theory, and then give a novel and relatively complete approach to computing confidence, which considers effects on confidence caused by both imperfect and incompatible information. A medical example prove the validity of this approach.