An Attribution Reduction Method for Weighted Approximation Representation Space

This paper first introduces the definition and decision condition for weighted consistent approximation representation spaces and its data characteristics. It then presents a method for data fusion and attribute reduction as well the confidence calculation. The proposed method compares the difference between un-weighted and weighted methods and shows that the proposed attribution reduction and confidence calculation are simple and promising with a clear physical meaning. The method can better represent the impacts of attribute weights on the confidence of rules. A new representation method is proposed in order to describe the un-weighted approximation representation space, which extends the application of rough set application.