Social Network Links Prediction Based On User Content Information Transfer

Social network is a social structure composed of social actors,ie individuals or organizations,social relations and other social interactions between actors.It is used in the social sciences to study the relationships between individuals,groups,organizations and even a society.With the rapid development of internet,the cost of information generation and dissemination is greatly reduced and the number of information increases in geometric multiple.Since the huge amounts of data on social networks has some obvious characteristics such as high quality,big data,semi-structure and direct reflection of real human society,many researchers from different areas or disciplines pay more and more attention to social networks.Social network analysis is a branch of sociological research,developed in the fields of mathematics,social science,anthropology,biology,communication science and so on.They are highly dynamic,that might lead the nodes and edges to appear or disappear in the future.Therefore,predicting the missing or unobserved links in current social networks and newly added or deleted links in future social networks is very important.Traditional prediction methods based on node similarity of network topology do not take the characteristics of user-generated content into account.In this paper,user-generated content is introduced based on the traditional social network link prediction method,and the transfer entropy is used to quantify the transfer of information among users as a feature of similarity among users.Then,three kinds of link prediction methods are defined by linear combination of information transfer features and topological features.In this paper,we conduct experiments on the above three kinds of link prediction methods respectively based on LDA model topic vector representation content and distributed word vector representation content,and compare the link prediction based information transfer with several classic traditional link predictions to verify the impact of information transfer on link prediction.In the experiment,we find that the network with information transfer has more realistic social network and better link prediction performance in social network link prediction,which has certain advantages in social network analysis.