The Research Of Community Search Algorithm Based On Graph Structure Clustering

In recent years,with the rapid development of information technology,various industries have formed their own huge amounts of data,mass communication data formed in the telecommunications industry,molecular relationship data in biochemistry,architecture information in transportation networks,and social networks data information.How to fully mining the potential value of these data is the hot issue discussed by academia and industry at present.For the huge complex network system formed by these data,we can clearly show the information by using graph,and clustering the data in the graph is a basic tool for digging the data of the graph.However,the data in the real world are constantly updated dynamically.The traditional structural-based clustering algorithm SCAN(A Structural Clustering Algorithm for Networks)does not effectively process and maintain the real-time updated data.Therefore,this paper proposes an Incremental Structural Clustering for Dynamic Networks(ISCAN)algorithm based on BFS-tree(breadth first search-tree).When the data is updated,we do not need to recalculate the whole graph with this algorithm.We only need to update a small number of edges and maintain the existing clustering results through the breaking and merging of the breadth-first tree.In addition,in order to reduce the time consumed in calculating the similarity between the vertices in the graph,this paper proposes how to use the multi-core to calculate the structural similarity and proposes two effective load balancing strategies.The traditional structure clustering algorithm is very sensitive to given threshold parameters.Subtle changes will have a great impact on the clustering results.In this paper,we propose a triangle-based structural clustering model.Compared with the traditional structural poly The class algorithm can not only reduce the sensitivity of clustering results to parameters,but also get closer to the community.Finally,we make a large number of comparative experiments based on the massive graph data formed in the real world.The experimental results demonstrate the validity and correctness of the incremental graph structure clustering algorithm and parallel computing structure similarity proposed by us.At the same time,through experiments,we found that the triangle-based structure clustering proposed by us not only can obtain a more compact community,but also when the threshold changes,the clustering structure is more stable.