Research Of Attribute Reduction Algorithm Based On Neighborhood Rough Set

Rough set theory is a mathematical theory for data analysis and it can deal with imprecise,inconsistent and uncertain knowledge effectively.Pawlak rough set model is the most classical model in rough set theory,which granulates the universe through equivalence relation and is well suitable for dealing with discrete data.However,it can't handle common numeric data directly.The neighborhood rough set model generalizes the equivalence relation of Pawlak rough set model into neighborhood relation,solving the problem that Pawlak rough set model can't directly process the numeric data and widening the application scope of rough set theory greatly.Attribute reduction plays a key role in the application of neighborhood rough set model,so it is very important to research and design attribute reduction algorithm based on neighborhood rough set.The focus of this paper research content is on the basis of the existing attribute reduction algorithms based on neighborhood rough set to improve against some of these deficiencies and validates the feasibility.In addition,this paper also builds a new variable precision rough set model and studies it deeply.The main work of this paper is as follows:Firstly,to solve the problems that the attribute reduction algorithm based on dependency degree model in neighborhood rough set exists,this paper proposes an attribute reduction algorithm based on improved attribute significance in neighborhood rough set.The algorithm based on improved attribute significance comprehensively considers the change of dependency degree and neighborhood knowledge granularity after adding the attribute,which can measure the importance of condition attributes fully and improve the performance of classification.Secondly,aiming at the problem that variable precision rough set model cannot deal with infinite set,on the basis of variable precision rough set model,this paper proposes a variable precision rough set model based on Lebesgue measure combining with measure theory and defines the upper and the lower approximation sets of this model.Moreover,this paper fully studies the properties for approximation sets and analyzes the validity of the model in theory.Finally,based on the Lebesgue measure,an neighborhood approximate conditional entropy model based on infinite set is presented in this paper.Meanwhile,a heuristic attribute reduction algorithm based on neighborhood approximate conditional entropy is also designed in this paper,which solves the problem that the attribute reduction algorithm based on approximate conditional entropy in neighborhood rough set can't deal with neighborhood decision systems without core.This paper analyzes the algorithm and the two typical attribute reduction algorithms based on neighborhood rough set under the view of algebra and information theory through five data sets of UCI database.The experimental results show that the algorithm is effective and can obtain better attribute reduction results.