Abstract: Attribute reduction is one of important problems in rough set theory, and attribute reductions in ordered information systems are also concerned in recent years. We first proposed a new information granulation based on dominance classes, which can be used to measure uncertainty of an attribute set in the context of ordered information systerns. Using this information granulation, a definition of attribute significance was then given in ordered information systams. At last, a heuristic reduction algorithm was put forward to ordered information systems, and the validity of this algorithm was exemplified by an illustrative example.

Key words: Ordered information systems, Dominance classes, Heuristic reduction algorithm