Block Discernibility Matrix Based On Decision Classification And Its Attribute Reduction Algorithm

Attribute reduction is a basic approach of rough set theory to make data mining.The attribute reduction algorithm based on discernibility matrix is one of the important attribute reduction algorithms,and it is intuitive and easy to understand.The heuristic reduction algorithm based on attribute core gains more attention.The discernibility matrix representation and correlation calculation are significant,but the existing discernibility matrix and its algorithm have limitations of time and space.According to the sparse and large scale of the discernibility matrix,this article constructs a block discernibility matrix based on decision classification and discusses the core algorithm and attribute reduction algorithm.The specific contents of this article are as follows.First,the existing discernibility matrix usually has a lot of null values(that is,the element of the discernibility matrix is ?),and for this sparse and large-scale nature,the information of the core attribute is concentrated in the non-empty value part.In this paper,matrix segmentation strategy is used to extract key information and reduce dimension effectively,and block discernibility matrix algorithm based on decision classification is established.Secondly,according to the block discernibility matrix based on decision classification,the connotation of the core is determined,and the core algorithm is given.The correctness and high efficiency of the algorithm are verified by comparing the experiments with the documents in 5 types of UCI data sets.Thirdly,the importance of the traditional attribute only considers the direct influence of the single conditional attribute on the decision attribute.In this paper,the importance of the improved attributes is given,and this importance takes into account the direct and indirect influence of attributes.Finally,a heuristic attribute reduction algorithm based on block discernibility matrix is designed to select the improved attribute importance,and the attribute reduction is obtained.This paper applies the algorithm to practical problems based on the example of meteorological condition.Finally,the validity and efficiency of attribute reduction algorithm are verified by using 5 types of UCI data sets to compare the experiments.In summary,the block discernibility matrix and its attribute reduction algorithm based on decision classification reduce the time and space complexity of the attribute reduction algorithm based on discernibility matrix greatly,and the basically needed information is integrated into the formal structure and problem solving more directly.The relevant research is innovative and effective.