| Knowledge reduction and knowledge discovery in the information systems are important topics of rough set theory.The indistinguishable relation is the basis of rough set theory,which is used to describe the similarity of objects in information systems.In decision system,with respect to global attribute reduction,reduction based the specific decision class can reveal more compact decision rules.As a mathematical tool of uncertaintyinformation processing,attribute reduction has been paid more and more attention in academia.The main research work of this paper is shown as follows:1.This paper designed a heuristic reduction algorithm based on the attribute frequency in the distinguishing matrix for λ-reduction,which can reduce the complexity of reduction calculations.Furthermore,the feasibility and effectiveness of the proposed algorithm is verified by examples.In the next,we explored the relationships between λ-reduction,maximal distribution reduction and distribution reduction in decision table.2.Based on upper approximation,the notion of upper approximation attribute reduction of specific decision class was proposed.The reduction approach by using discernibility functions is presented.On the basis of kernel,boundary and usage attributes,some relationships between upper approximation reduction and assignment reduction based on the specific decision classes were discussed.The relationships among assignment reduction,upper approximation attribute reduction of specific decision class and upper approximation attribute reduction of specific object are surveyed. |
