An Improved Model of Rough Sets on Incomplete Information Systems

Abstract

According to rough set theory, there are two approaches used to handle incomplete information systems. One is to transfer the incomplete information systems into complete information systems that can be tackled by the classical rough set theory. The other is to deal directly with the incomplete information systems through improving the classical rough set theory. But basically the first approach will change what the original incomplete information systems implicate. Therefore, the better approach is to utilize the second one. In this paper, the second approach is discussed. Through an analysis to the limited tolerance relation on incomplete information systems, the concept of "similarity degree" is introduced for improving the limited tolerance relation into a new model named after the limited tolerance relation based on similarity degree. The new model is more flexible on processing incomplete information systems than the limited tolerance relation is.