Abstract: Two methods of knowledge discovery based on rough set (FCKD and SCKD), unlike those based on reduction of attributes, are presented in this paper. Their function is to find new determinative attributes which the original rules do not possess by reducing vagueness of the rough set when information system has varied dynamically. The structure of attributes is accordingly re constructed and new rules formed. Two operators for finding new attributes are defined, two theorems and an inference formulated. Examples for illustrating the new methods are also given.