Research On Visualization Of Social Network Based On Graph Mining

With the rise of microblogging and social network rapidly, the research of social network has become a hot topic, how to make full use of modern network tools to find the information in the network quickly is an important direction. Based on the graph mining,the clustering of node in network has been one of the important research topic in large-scale network, researchers in the community discovery has made a lot of achievements, but there are still many problems to be solved: the algorithm's efficiency is not so high, very little attention to overlapping community structure, cannot eliminate noise interference. The development of information visualization technology provides a very effective means to let people understand the structure of the network and mining effectively, but for the traditional large-scale social network visualization layout is not clear.Aiming at these problems, this thesis takes Spark as the platform, using the graph mining, launched a study of community discovery and visualization of social networks, the specific contents are as follows:(1) the design of network community discovery algorithm based on edge graph and its parallelization. The GN algorithm can discover overlapping communities and its complexity is high, the network diagram to edges of the graph, the edges in the graph nodes similarity instead of edge betweenness in GN algorithm, improved GN algorithm, and using MapReduce model, the algorithm for parallel processing, improve the efficiency(2) proposed community discovery algorithm based on user influence and its parallelization scheme. According to the local modularity of community discovery algorithm based on [26] stability problem, using PageRank algorithm, the research of users in a social network influence, the influence of user nodes as the initial community, making the community classification results are stable and more accurate, and studies the parallelization of the community discovery algorithm.(3) the design of force directed layout algorithm based on MapReduce. Analysis of force directed placement algorithm of serial memory computing framework, combined with Spark, realized the parallelization of the algorithm, to speed up the network layout.(4) proposed a social network visualization method based on community structure, improved force directed layout algorithm, a separate distribution in every community, finally get the layout of the entire network, which can be used in large scale social network visualization and layout.The innovative in this thesis:proposed community discovery algorithm based on user influence, to avoid the instability of the local community, to improve the accuracy of the classification of community.