A User Influence Mining Algorithm In Online Social Network

Social network analysis has received enormous attention in recent years, owing to the success of online social networking sites. These sites have gained huge number of data about friendship relationships and contents generated by users(such as user opinions, comments, photos and so on). This trend leads to the generation of a wealth of social network data. Researchers begin to focus on search engine, data relationship and behaviors of users. As the amount of data grows, the regular interactions among social network users grow to be a huge dataset with millions of actions. The usual processing methods seem to be useless. There are differences between the techniques developed by the research community and their deployment to mass data in real-world. Therefore, the potential research impact of these techniques is still largely unexplored. In academia and industry, social network analysis and mining have become hot research areas. User behavior analysis has become an important aspect of social network analysis and also received a great deal of attention. However, most of the existing methods are oriented to the global networks, the algorithm for given node's ego-centric network is very lack; existing methods mainly consider the network link structure, but ignore the closeness, influence, user interactions and other social features.In this article, we address the problem of behavior analysis of huge amounts of data produced in social network. Such a problem arises naturally in data analysis industry where one aims to understand users' taste with multiple traces from his history of surfing the net as correctly as possible. The general version of this problem can be regarded as a Page Rank-Based algorithm which could be efficiently utilized by big data collected. In each step, a brief overview of the problem is presented, and classic approaches are described. Then the model is transformed to deal with massive data examples, then map each of the topics to a behavior analysis framework. The Page Rank-Based algorithm is used to search for the important node in the network. The effectiveness of Page Rank-Based algorithm is proved via using the node importance degree coefficient. Furthermore, two probability analysis methods are compared to handle the situations that what are really the users' interest and to what extent that users' privacy via online social network could be disclosed. We then investigate into applications of our algorithm to community user tastes analysis. D-S evidence theory has been used to cope with the decision process. The proposed Page Rank-Based algorithm is also utilized to detect community structure. A closeness function is applied to get rid of the node that may cause overlapping community. In addition, experimental results on challenging real-world datasets show that our proposed algorithm is effective.