Research On User Behavior Of Social Media

Social media refers to a network platform for content production and exchange based on user relationships.User behavior on social media is the sum of the user 's willingness to use social network services based on a comprehensive assessment of their needs,social impact,and social network technology,as well as the resulting use activities.Studying user behavior on social media can provide theoretical guidance for user-defined services on social platforms from the user's perspective to achieve personalized recommendations for users.It is also useful for revealing the evolution of online social network structure,the law of information dissemination,and effective Monitoring the sudden public opinion on the network has an important role,so it has extremely high commercial value and scientific research value.The research area of social media user behavior analysis covers a wide range.This article mainly researches the influence of users on social media and the organization of user identification.The traditional influence analysis is based on the manual and manual establishment of features,and relies on the accuracy of feature engineering,and usually researches based on a single kind of network,without integrating more information about the relationship network.An organizational account is the official account of an organizational entity in a social network on a social network.Generally,there are specialized personnel of the organization to run the organizational account,and use this account to speak for the organization.The current research on organizational accounts is still blank.In view of the above problems,based on the existing research,this paper proposes a user influence analysis framework based on heterogeneous networks and an organization account identification method based on user anomaly consistency and influence.The main work is summarized as follows:(1)Propose a user influence analysis framework based on heterogeneous network.First,the framework builds heterogeneous networks using a variety of social relationships that are reflected by social behavior.Second,it introduces graph embedding technology to represent user social network relationships in the form of embedded vectors.Finally,based on user feature vectors,the social behavior of social media users in the future.Make predictions to assess the impact of this user.Compared with traditional methods,the method proposed in this paper evaluates the impact to the user level accurately,while using social relationships to enrich the information dimension and improve the accuracy of the algorithm.In addition,the semi-supervised learning method used by the framework can achieve better performance results with less training data,reducing the time-consuming and labor-intensive problem that traditional methods need to manually mark a large amount of data;(2)An organization account classification method based on user anomaly consistency and influence is proposed.This method uses the social relationship of user accounts on social media and published text content to extract the abnormal consistency scores and influence scores of accounts as features,which can distinguish organizational accounts from non-organization accounts,and fill the current gap in the research field.