Agent-Based Group Behavior Induction In Social Networks

The essence of social networks is that it can quickly form public opinions,and affects people's thinking and behavior.However,negative group behavior affects the stability of social.Popular behavior induction approaches adopted in social networks include political restriction,using mainstream media to induct and employing people to publish positive information.However,most of these approaches are too simple and inefficient,easily leading to reverse recognization,feeling offensive and political negligence and more.Therefore,the study of group behavior induction in social networks has received more and more attention.Based on this,the main work of this dissertation is as follows:(1)Summarized the background and research significance of group behavior induction,analyzed the research status of group behavior in social networks,and expatiated the concepts and approaches of related field in detail.(2)According to the research of group behavior in social networks,an approache of group semantic behavior extraction is proposed.Firstly,we use the topic model to construct the semantic social network by using the Pearson correlation coefficient to calculate the similarity between the users,and then use the community discovery algorithm to divide the semantic social network into community.Finally,the semantic behavior of the community is extracted.The experimental results show that the proposed approache can correctly extract the group semantic behavior.(3)According to the result of the above semantic behavior extraction,the semantic behavior center is used as the agent,and an approache of agent-based behavior induction is proposed.Specifically,the behavior of users is initialized based on the group semantic behavior extraction result.Then we use the agent to induct the behavior of other users in the social networks according to a dynamic control mechanism,so that users in the social network as little as possible to participate in negative behavior.The simulation results show that the proposed approache can effectively induct the behavior of the social network.