Abstract: Attribute weighting is an important approach in management decision and evaluation. The common weighting methods are always relied on the expert's transcendental knowledge. In rough set weighting method,no transcendental knowledge about the disposing data sets is needed,which reflects the objectivity of the data. So this method is more and more applied in management decision,with the shortage that redundant attribute weight could not be ascertained in current studies. Based on the characteristic that the information expression of rough set is more comprehensive than the algebraic expression,attribute importance represented by rough set information entropy is studied deeply in this paper,and a new method of ascertaining attribute weight is put forward based on rough set conditional information entropy,and it's rationality is proved. Finally,a case is used to prove that new weighting method could solve the existing problem and is more applicable and explicable.

Key words: weight, rough set, attribute significance degree, information entropy