Research On Model Extension And Algorithm Based On Neighborhood Rough Set

Rough set theory is an important mathematical tool proposed by the Poland scholar Pawlak in 1982,which can effectively carry out uncertain information processing.At present,it has been widely used in data mining,artificial intelligence,pattern recognition and other research fields.With the development of science and the improvement of automation,there are large scale and complex information systems in many application fields,including all kinds of data,such as symbolic and numerical data.Classic rough sets can only handle symbolic data and can't handle other types of data.In this case,experts and scholars extend the classic rough set,and the neighborhood rough set is one of the most important extensions.The neighborhood rough set is based on the neighborhood relation,and the similar objects are divided into one class by granulation,and the numerical data can be handled effectively.With the continuous progress of science and technology,many new problems have emerged in the research of data intelligent processing.There are many new challenges in the further research and application of neighborhood rough sets.For example,how to solve the problem of inconsistent and incomplete neighborhood decision system with neighborhood rough set,how to define an uncertainty measure function that is more suitable for dealing with numerical data,and how to design a more efficient attribute reduction algorithm and so on.In order to solve these problems,this paper takes the neighborhood decision system as the research object,and uses the model extension of the neighborhood rough set and the attribute reduction based on this as the research purpose.The specific contents are summarized as follows:(1)In this paper,first analyze the shortcomings of neighborhood rough set to deal with inconsistent neighborhood decision system.In order to deal with inconsistent data better,an inconsistent neighborhood rough set model is established.Then we consider that the classical uncertainty measurement method is difficult to be applied to the inconsistent neighborhood rough set model.Combined with the characteristics of the model,we give a heuristic function of neighborhood conditional entropy,and prove that it accords with the basic condition of uncertainty measurement.At the same time,the degree of association between conditional attributes will also affect the reduction result.In this case,the knowledge of rank correlation coefficient is introduced in this paper,and an attribute reduction algorithm based on the correlation coefficient is proposed.Finally,the experimental results are compared with other algorithms,and the results show the validity of the proposed model and algorithm.(2)The neighborhood of incomplete decision system may lead to the emergence of new information coordination,in recent years,experts and scholars have proposed many methods to deal with incomplete data,which many of the distance measure without considering the potential of existing data in information,has certain limitations.In order to overcome these limitations,this paper constructed a new distance formula of numerical and categorical data,and in the process to fully consider the characteristics of data distribution on the influence of the distance is established based on incomplete neighborhood rough set model,and carried out scientific analysis examples prove the proposed model.Then analyzes the algebra and information entropy based uncertainty measure method based on their respective advantages and shortcomings,combining the advantages of both proposed heuristic function neighborhood hybrid entropy for uncertainty measure;then put forward the attribute neighborhood entropy reduction algorithm based on Hybrid.Finally,the superiority of the model in attribute reduction and classification performance is explained by the experiment.The innovation of this article mainly includes the following points:(1)Constructed inconsistent neighborhood rough set model can effectively deal with inconsistent data;measure function neighborhood condition entropy is proposed on this basis;put forward the attribute reduction algorithm based on correlation coefficient,correlation between the concept of rank correlation coefficient in the algorithm attribute,which can further eliminate the redundant attribute.(2)Considering the incomplete distribution of existing data neighborhood decision system,set up different distance formula for numeric and symbolic attributes,established the incomplete neighborhood rough set model;combined with the advantages of heuristic function algebra and information entropy respectively based on the establishment of a new uncertainty measure:mixed neighborhood function entropy;put forward the attribute reduction algorithm based on neighborhood entropy of mixing,can effectively eliminate the redundant attributes and get better classification accuracy.