A New Rough Set Model for Knowledge Acquisition in Incomplete Information System

Rough set models based on the tolerance and similarity relations, are constructed to deal with incomplete information systems. Unfortunately, tolerance and similarity relations have their own limitations because the former is too loose while the latter is too strict in classification analysis. To make a reasonable and flexible classification in incomplete information system, a new binary relation is proposed in this paper. This new binary relation is only reflective and it is a generalization of tolerance and similarity relations. Furthermore, three different rough set models based on the above three different binary relations are compared and then some important properties are obtained. Finally, the direct approach to certain and possible rules induction in incomplete information system is investigated, an illustrative example is analyzed to substantiate the conceptual arguments.