Composite Network Weibo User Behavior Based On Feature

Micro-blogging users network is a complex network which is involved in many factors such as ordinary users, celebrities, attention, tweet, retweet, comment and other behavior relationships. Ignoring the existence of correlations between behaviors of users, previous research on micro-blogging user network usually only pay attention to the discussion of one relationship. However, users behaviors are often associated with each other. Therefore, the analysis and discovery of relations among multiple behaviors of micro-blogging users are of great practical significance for revealing the law of the micro-blogging network user behaviors and understanding the complex mechanisms of micro-blogging network user behaviors.Based on the above, this paper uses task1of the KDD CUP2012of tencent micro-blogging data to extract the attention of the user and relationships of the same keyword in tweet, retweet and comment. Based on the complex network model, the paper builds the users’ attention subnet and the users’ keyword subnet. Based on subnet load operation of this model, the paper constructs the complex network of users’ attention subnet and users’ keyword subnet. Through the analysis of several topological properties of the complex network and the subnet, this paper finds some interesting phenomena that are contributed to understanding complex behaviors of micro-blogging network users.The main works of this paper are as follows:(1)This paper builds the users’ attention subnet and the users’ keyword subnet and analyzes their topological properties. This paper sees the users as nodes and sees the attention relationship between the users as edges abstractly to construct the attention subnet; The degree distribution of attention subnet is approximated power-law distribution. Celebrities own high degree, in other words, Celebrities are more paid attention by others. This paper sees users as nodes and sees users who have≥k(k∈N+) keywords relationship as edges to build many keyword subnets. The degree distribution of keyword subnets are also approximated power-law distribution, but ordinary users’ degrees are much higher than celebrities’ degrees. It shows that there are much more topics which ordinary users participate in and much less topics which celebrities take part in which may be associated with celebrity professional field.(2)This paper builds complex network of the user attention subnet and the user keyword subnet and analyzes the nature of complex network boundary nodes. It finds that with the increase of keyword subnet degrees of boundary nodes, the degrees of attention subnet become a little higher. It shows that the keyword similarity subnet has small positive correlation with attention subnet. It turns out that that the user who owns much more topics, the more likely he or she should be concerned about by others.(3)For discovery user type is whether or not affecting the nature of the boundary nodes, this paper extracts celebrities and their attention relationships and keyword relationships data to construct complex network and analyzes the nature of boundary nodes. This paper found that as celebrity keyword similarity subnet degrees become higher gradually, degree of celebrity attention subnet has a small decreasing trend. In other words, the relationship between the degree of celebrity keyword similarity subnet and celebrity attention subnet is a negative one, which is just the opposite law compared to ordinary users. And it means that celebrities are more likely to pay attention to celebrities who are more specialized and authoritative in some fields.