Research And Application Based On Improved Binary Discernibility Matrix Attribute Reduction Algorithm

Rough set theory is a kind of research method of uncertainty,incomplete knowledge,which is widely used in data mining,artificial intelligence and pattern recognition and many other fields.Attribute reduction means that to express the decision table information by minimum attribute number without affecting the quality of original classification of the decision table.Attribute reduction is one of the core of rough set theory.At present,In the algorithm for attribute reduction of rough set,the attribute reduction algorithm which based on binary discernibility matrix attribute reduction method get the attention of many scholars because of low storage,simple and easy understanding.At present,In the algorithm for attribute reduction based on binary discernibility matrix generally use the following ideas: firstly,to set up binary discernibility matrix according to the known decision table,and taking the complete attribute as the initial set,secondly,choose a non-nuclear attribute to delete one by one according to a certain measure until the last of the collection is a reduct,which is called elimination.however,the method choose a redundant attribute to delete each time,which lead to attribute reduction has lower efficiency.Therefore,The focus of this article research content is to improve the efficiency of the elimination of attribute reduction algorithm which based on binary discernibility matrix attribute reduction method and get the minimum reduction,in this paper,the main work which have been done of this thesis is as follows:1.Aiming at the shortcomings of the elimination based on binary discernibility matrix method,this paper presents an improved algorithm of attribute reduction based on simplified binary discernibility matrix,which consider the two dimensions of binary discernibility matrix such as row and column,to delete more than one redundant attributes by a new measurement of the attribute reduction and simplify binary discernibility matrix after the delete operation until the last is a reduct.This paper prove the feasibility of the measure.Moreover,the experiment and test results prove the correctness of the method.2.In this paper,improved algorithm of attribute reduction based on simplified binary discernibility matrix is applied to the sales statistics analysis function module of order collaboration system,to reduce the attribute which related to the first quarter of 2013 bluetooth speaker sales data and obtain the corresponding decision rules.Thedecision rules coincide better with the actual sales of the second quarter of 2013 with bluetooth speakers,further illustrate the effectiveness of the algorithm.