An improved attribute significance measure based on rough set

摘要

Attribute significance is an important concept of rough set theory. It plays an important role in attribute reduction and decision making. However, many attributes have the same significance value based on the traditional definition of attribute significance. This causes an important problem that one cannot determine which attribute is more important when the aforementioned phenomenon occurs. To overcome this shortcoming, an improved definition of attribute significance, which considers the change of the number of equivalence classes, is proposed in this paper. Moreover, an algorithm which can compute the new attribute significance by using the representation method of binary granule of the finite set is presented, and a computation program of attribute significance by MATLAB is given. Finally, based on the improved definition of attribute significance, the attribute reduction algorithm AR_IAS is proposed.