Information Dissemination Of Group Behavior Based On The Hotspot

With the large-scale popularization and application of mobile communication, Internet, intelligent terminal, social network and other technologies, the information and communication engineering has been breakthrough growing. And social network has become an important platform for users to get, share and spread information. Therefore, analyzing user behavior with social data provides the opportunity to study the topic evolution, the information dissemination and the group relationship.This thesis discusses the information diffusion and users group behavior from two levels of topic evolution and user relationship by model, algorithm and experiment. Particularly, the main work of this thesis are as follows:1. In the aspect of topic evolution, using user behavior data from hotspots in social network, author studies how different user groups play different roles for a hotspot topic. Firstly, by analyzing users' behavior records, author mines group situation that promotes the hotspot. Secondly, several major attributions in a hotspot outbreak, such as individual, peer and group triggers, are defined formally according to the viewpoint of social identity, social interaction, retweet depth and opinion leader. Thirdly, for the problem of the uneven and sparse data in each stage of hotspot topic's life cycle, author proposes a dynamic influence model based on grey system to formalize the effect of different groups. Lastly, the process of hotspot evolution driven by distinct crowd is showed dynamically. The experimental result confirms that the model is able not only to qualify users' influence on a hotspot topic but also to predict effectively an upcoming change in a hotspot topic.2. In the user relationship level, author researches the implication link on account of interests and conformity between discrete users with user data from hotspots in social network. First of all, author proposes three behavior factors about behaviors of user from social data. The factors include the common behavior, followed behavior, and following behavior. Secondly, according to the topic relation data between users and their friends, it constructs an users relationship network to mining the common friends. Thirdly, the implication attributions are formed by combining user's behavior attributes with relationship structure attributes. Then author proposes an implication link model based on multivariate analysis theory. Finally, the adaptive boosting classifier is obtained from implication link model and algorithm. It can find out the implication link with the discrete users group and help this group recommend friends. The experiment shows that the model not only can classify users' direct relation, but also can effectively find out the implication link from hotspot topic.