Research Of Social Network Data Mining Algorithm Based On Graph Clustering

In recent years, with the wide application of biochemistry, social relation network, text retrieval and web analysis, some more complex structures cannot be displayed by sequence and itemset any more. Graph, however, as a kind of universal model, can be used to show all kinds of complex systems in the real world. The complex problems in reality in form of figure can be better analyzed and understood. In real life, a variety of social relations among people constitute a social network which describes the social members and their mutual relations in society. From the perspective of data analyzing, this network can be demonstrated by figure. The nodes in the figure refer to the individuals studied while the lines to the mutual-relationships between individuals.Combining with structure similarity proposed from SCAN(structural clustering algorithm for networks, SCAN) algorithm, GN(Girvan Newman algorithm, GN) algorithm, as the classic community mining approach, is studied in this paper, according to which, GNSCAN(Girvan-Newman algorithm based on SCAN algorithm, GNSCAN), a community mining clustering algorithm based on structure similarity, and its improved algorithm IGNSCAN(the improved Girvan-Newman algorithm based on SCAN algorithm, IGNSCAN), are elaborated. At the same time, combining with GN algorithm and RA, an indicator of resource allocation in the link, GNRA(Girvan-Newman algorithm based on Resource Allocation, GNRA), a community mining algorithm based on link prediction, and its improved algorithm IGNRA( the Improved Girvan-Newman algorithm based on Resource Allocation, IGNRA), are given as well. The split hierarchical clustering creation are adopted in the above two. The similarity measurement with a lower time complexity, rather than edge betweenness calculation, reduces time complexity of GN algorithm. The implementation of these two validates the proposed algorithm with a lower time complexity.Graph clustering is applied in community detection in this paper. The social networks, viewed as a figure, can be divided by figure clustering. The research in this paper would bring certain significance to theoretical expansion and practical application of statistics mining.