The Study Of Attribute Reduction Method Based On User-Oriented

Attribute reduction is one of the core research issue in rough set theory, Which deletes the redundancy(or dispensable) attribute and promise that the ability of classification un-changed. Many learning algorithms about attribute reduction have such a bias:all attributes are equally important. However, this assumption is unreasonable, and is not completely practical.Reduct is a kind of special significance attribute reduction. Many scholars obtain Reduct through the heuristic algorithm. If the attributes are considered as having different importance according to actual condition. The model will be more realistic.This paper selects the method of attribute reduction based on user-Oriented as topics. Firstly attributes are roughly sorted according to the user’s different preference in practice, then we constitute the attribute order to represent the different importance of attributes. After that the paper selects information system;decision tables and incomplete information system as the research objects,and applies the set covering method to dealing with the prob-lem of solving Reduct and in the establishment of reasonable model based on the attribute order.The main contents of this paper are as follows:(1) A reduction method based on attribute order in the information system:First of all, the information system will be divided into equivalence classes through equivalence relation,and equivalence classes discernibility matrix will be established according to the difference between two different equivalence classes.So the paper will build the comnection between the attribute reduction of information system and the set covering. The attribute order reflects the user’s preferences of attributes,and I solve the problem of attribute reduction by using the set covering reduction method.And I will design the algorithm with the help of attribute order to reflect user’s preference.So the results may be more realistic,practical and effective. Finally I will use an example to verify the feasibility and correctness of the algorithm (2)A reduction algorithm based on the attribute order in decision table. The literature [19] proved that is some connection the connection between the attribute reduction and set covering reduction problem in information systems. This paper proves that the connection between the decision table and covering;and establishes the model through the attribute order which reflects user’s preference.Then the algorithm solves the Reduct problem in the decision table through the set covering reduction method. Finally the algorithm is analyzed in detail and verified by the use of examples.(3) A reduction method based on attribute order in incomplete information system. The literature [37] proved the connection between the attribute reduction of incomplete in-formation system and the set covering reduction problem.This paper uses tolerance relation to classify incomplete information system, and forms a covering in it.This paper also struc-tures the discernibility matrix of tolerant kind according to two different tolerant classes and builds the connection between attribute reduction and the set covering through the discernibility matrix. We get the Reduct of the incomplete information system by a set covering reduction method. And with practical examples we prove that this method can get only Reduct. Compared with the literature [37] in this paper, the method is more practical.