Research And Application Of Recommended Algorithm For Social Network Friends

Development of information technology has brought more and more information, and also changed people’s social way. The emergence of social networks not only broaden the social circle, but also bring commercial profits. Therefore, it has become a hot research topic on how to promote personalized recommendation to social network users.This paper has conducted in-depth study based on the recommended technology network architecture. As the random walk is expensive, off-target and it does not consider the importance of vertex and other issues, the paper has established a restart model based on local random tour connectivity (Connectivity Local Random walk With restart recommendation Algorithm, called C-LRWR). Experiments show that the improved model on a given set of data has ensured the accuracy of certain recommendations, effectively reduced the computational overhead migration process. The random walk model ignoring the starting vertex farther distance by limiting the search range, is a "local" scope of migration. And it also introduced the importance of network of vertices, the vertex differences reflected different network structures.The paper has proposed an improved label-based recommendation algorithm can’t tap the potential users on the traditional recommendation algorithm. The experimental results show that the improved algorithm can not only recommend similar user and the target user, but also tap the potential users. Firstly, the algorithm find a similar pre-M users through the user’s tag type, and according to the trust phenomenon existed in social network, to calculate the similarity among user’s friends with the target user, and finally recommend the first N users calculated from trust descending.In this paper, the experimental data come from the International Knowledge Discovery and Data Mining Contest released in Tencent micro blog in 2012. The data set contains the user attributes, user behavior, user social connections and other data. This paper has applied the proposed algorithm model on the data set, and conducted the experiment on C-LRWR algorithms and algorithm based label experiments. Finally, the systematic experiment has verified the correctness and validity of the proposed algorithm.