Link Prediction On Social Network

With the development of Internet technology, social network has become more and more popular, and it helps people make friends easier. Meanwhile, with the rapid development of social network analysis, social network has been widely used in sociology, anthropology, economics, biology and so on. Everyone has the motion to expand their relationship, but has no enough time to find the potential friends, so as the subtask of social network analysis, link prediction has become more and more important.In this paper, we review the basic concepts of social network and the basic methods of social network analysis, and propose the result as follows.We find some limitation of traditional link prediction models which are limited to single-type link prediction, while there are multiple relation types between people. So we try to consider the impact of different relationships between people.We use link pattern to predict the link in the social network, and also user Tensor to figure out the multiple relationship. The experiments on the Enron email dataset and the YouTube dataset shows that this method outperforms traditional link prediction models.As a part of social network analysis, link prediction can help other application of social network analysis to improve their accuracy. We apply the link prediction method into the social mining, and the experiment of the Enron email dataset shows that the link prediction method can improve the accuracy of association mining, so link prediction plays an important role of social mining and social network analysis.