A Novel Approach to Fuzzy Rough Set-Based Analysis of Information Systems

This paper presents an approach to analysis of crisp and fuzzy information systems. It is based on comparison of elements of the universe to prototypes of condition and decision classes instead of using binary crisp indiscernibility or fuzzy similarity relations. We introduce several notions, such as dominating linguistic values, linguistic labels, characteristic elements, which lead to a new definition of fuzzy rough approximations. The presented method gives the same results as the original rough set theory of Pawlak, in the special case of crisp information systems. Furthermore, fuzzy information systems can be analyzed more efficiently than in the standard fuzzy rough set approach. Moreover, interpretation of results is quite natural and intuitive. Analysis of information systems will be illustrated with an extended example.