Research Of Knowledge Reduction Based On Binary Discernibility Matrix

The rough set theory is an effectively data analysis tool to handle uncertain information which is imprecise,inconsistent,incomplete and so on.Knowledge reduction is the kernel of rough set theory,the critical technique of the application of rough set theory,and also an important research direction of knowledge discovery.Through the reduction of information systems to remove redundant information,and obtain the reduction of knowledge.This paper surrounds two important problems of knowledge reduction:attribute reduction and value reduction,new algorithms are presented in consistent and inconsistent decision tables,experimental results show the validity of algorithms.The main research and innovation are as follows:(1)The paper mainly in the study of reduction algorithm based on binary discernibility matrix of rough set theory,propose a new construction method of binary discernibility matrix,reduce the storage space effectively,and speed up the scanning speed.It summarized and generalized some relational nature and theorem in the binary discernibility matrix,and give it corresponding prove.Algorithms only to take a simple arithmetic on the binary discernibility matrix can be completed on the reduction of decision table operation;(2)The heuristic attribute reduction algorithm based on the binary discernibility matrix is presented,take attribute significance as inspiration information.Given a fast and simple method of calculating the negative domain in inconsistent decision table for the attribute reduction;(3)Analysis of the process and the nature of value reduction,propose the concept of the binary discernibility matrix based on object,and on this matrix to take value reduction.New inspiration information is defined in value reduction,take inductive of attributes as reduction rules of the metrics;It gives simple calculation methods in the accuracy of rules and a new method to judge and handle the repeat or redundant rules is given.According the accuracy of rules and inductive of attributes to extract the concise rules in the decision table,the speed reduction have been effective improved.