Inter Domain Community Mining Algorithm Research Based On Spectral Clustering

Internet high-end technology innovation and development has brought great changes to human society. Human society has entered an era of information technology. In many fields, the way people get information is changing with each passing day, the number of information is increasing rapidly and the information of the same thing is described. Kinds of social network carrying people in life and production in the formation of the complex relationship. How to find the potential value of a relationship from the jumble of social networks? It attracts more and more people attention. With the development of Web 2.0 technology, the internet is moving in the direction of community development, users want to participate, interact to get more interest information. Some potential structural relations exist between the social network users, general structure diagram said graph nodes represent individuals in the network, and edges represent certain relationship between individuals (such as the participation of the same activity relationship). In general links within the same community are dense, and the links between different communities are sparse.Traditional community found mainly in the face of a single field, due to a single factor, not through the effective information of other areas to improve the accuracy of clustering, this paper presents a based on spectral clustering iterative community discovery algorithm, using k-means algorithm as the foundation, on the spectral clustering algorithm for the solution, in the field of similarity matrix is updated continually, in order to achieve between the field and promote each other. According to the analysis of the results of the algorithm by modify the similarity and iteration termination condition is used to optimize the algorithm, and on the optimization results with the previous results were compared, can intuitively obvious optimization effect.Under the premise of iterative domain of mining community and results in different fields of fusion of clustering, clustering fusion scheme combined with label merging and voting method, effective in a number of areas clustering results are fused and the algorithm can adapt to various fields entities are not all the same and experiments verify the effectiveness of clustering results.