Research And Application Of Three-way Clustering Based On Set Pair Information Granule

Aiming at the problem of uncertain data clustering,the idea of set pair information granule is introduced into the traditional clustering algorithm.At the same time,in order to better express the relationship between samples and clusters,a set pair three-way clustering algorithm is constructed,and apply it to network community mining.Firstly,analyze the characteristics of incomplete data in depth,consider the three dimensions of positive degree,difference degree,and negative degree to construct a set pair distance measurement method,which can better represent uncertain information.Furthermore,this method is applied to the k-means algorithm,and a set pair three-way kmeans clustering algorithm is constructed.The parameters and feasibility of the algorithm are discussed through experimental design.Secondly,the problem of multi-level and multi-granularity of clustering is studied,and a set pair three-way hierarchical clustering algorithm is constructed.Based on the CURE algorithm,the distance formula is improved based on the set pair information granule theory,so that it can effectively deal with missing values.In addition,considering the uncertain relationship between the samples and clusters,the definition of the intra-cluster average distance of each cluster is given,and then the set pair clustering results composed of positive region,boundary region and negative region is formed.Finally,the set pair three-way clustering algorithm is applied to community detection.Considering the topological information between nodes,weight and degree of nodes,a method of set pair similarity measurement between nodes is given.Furthermore,the improved set pair three-way clustering algorithms are used to assign nodes and generate three-way community structures.And an overlapping node set is generated according to the calculation results of community node belongingness.Through experimental design on artificial network and real network,the results show that the proposed algorithm has good performance on multiple indicators.Figure 14;Table 24;Reference 63...