Computation of Maximal Characteristic Neighborhoods for Incomplete Information Systems

In rough set approach, incomplete information systems have been used to represent data tables with missing values. Similarity relation is one of the most popular ways to represent approximation space of incomplete information systems. Maximal consistent blocks have been used to improve approximation accuracy. In this paper, we introduce an algorithm for computing maximal characteristic sets represented by binary neighborhood systems. Characteristic relation is a generalization of similarity relation, and it's been shown that the adopted approach can further improve approximation accuracy. The time complexity of our algorithm is in 0(N · M2) where M is the average size of characteristic sets and N is the number of objects in a data table.