Research On Attribute Reduction Of Discernibility Information Tree And Multi-source And Multi-granulation Rough Set Method

In recent years,with the progress of information technology and the development of statistical technology,the amount of data and information faced by human beings has increased significantly.It is particularly important to deal with the massive and complex data from biomedical,economic and financial,artificial intelligence and other fields reasonably and to mine valuable knowledge from these data.The attribute reduction technology in rough set theory can delete the valueless(redundant)knowledge from the original knowledge base while keeping the classification ability of the original knowledge base unchanged.Discernibility matrix is an important medium for attribute reduction.However,there are many duplicate information in the discernibility matrix,which leads to the increase of storage cost.The discernibility information tree(DIT)can solve the problem of repeated information in discernibility matrix.It is worth noting that the researches on DIT are all based on the equivalence relation without considering the preference order relation between data,and only one reduction of information system can be obtained through DIT.In this paper,the interval-valued ordered information system(Iv OIS)is taken as the research background.Firstly,the construction of DIT under the dominance relation is studied,and then the corresponding attribute reduction method is proposed.Then the core attribute and attribute importance are introduced to improve the discernibility information tree under the dominant relation,and then the attribute reduction method based on the improved discernibility information tree(IDIT)is proposed.Finally,in order to solve the problem of knowledge discovery in multi-source fuzzy information system(Ms FIS),the multi-granulation rough set model of Ms FIS and its uncertainty by combining the idea of multi-granulation are studied in this paper.The main innovations of this article are divided into the following three points.1.Under the background of IvOIS,combined with the relevant theoretical knowledge of discernibility matrix and DIT,this paper puts forward the construction method of DIT based on discernibility matrix,thus realizing the construction of DIT under dominance relation.Furthermore,the related theorems of DIT under dominance relation are studied.Finally,a new complete attribute reduction method is proposed,which enriches the research results of attribute reduction.2.Considering that core attribute and attribute importance play an important role in the process of attribute reduction,this paper studies the construction method of IDIT based on the core attribute and attribute importance in Iv OIS,and further proposes the knowledge reduction method based on IDIT.At the same time,it is found that all the reduction of Iv OIS can be found by constructing IDIT under different attribute order.3.In the context of MsFIS,combined with the relevant knowledge of granular computing,each source is regarded as a granularity.Firstly,the support characteristic function of Ms FIS is defined.Then three kinds of multi-granulation rough set models for Ms FIS are proposed by the defined support characteristic function,and the related properties of the models are studied.Secondly,in order to study the uncertainty of Ms FIS,this paper presents five kinds of uncertainty measures and designs a corresponding algorithm.Finally,four datasets are used to verify the algorithm.The experimental results show that the generalized multi-granulation rough set model of Ms FIS is more applicable in practical application.