Research On Social Influence Prediction Based On Preference Propagation

In recent years,with the rapid development of Internet technology such as big data and cloud computing,various social networks begin to fill people's daily life.With the explosive growth of user scale,the research on the influence of social network users began.User social impact refers to a person's behavior or consciousness changes under the influence of other people.It can provide personalized recommendations for users and effective marketing strategies for enterprises by deeply mining the social impact among users.At present,how to predict the social impact of users more accurately is a very valuable and practical research.Traditional social impact prediction methods rely on manual features set by domain experts,which are directly related to the effectiveness of the model,so it is difficult to transplant them on different types of social networks.Recently,thanks to the development of network representation learning technology,users in the network can be mapped to a low-dimensional vector,so as to learn the potential characteristics of users to predict social impact.But this method only considers the structural factors in the network,and whether the actual users are affected is closely related to their own preferences.At present,some researches use the content information of social network to express users' preferences,but this method needs to crawl the social content on the network,and different methods are needed to express the content information of different types of network,which has poor generalization.In view of the above problems,this paper proposes a social impact prediction algorithm RippleInf-GAT based on user propagation,which does not rely on manual features,and does not need to crawl social network content information.The work of this paper is as follows:First of all,this paper uses advanced network embedding technology node2 vec to represent network nodes,and gives the concept of using the state of neighbor nodes to simulate the initial user preference,defines the process of preference propagation on social networks,and calculates the preference probability of central nodes to neighborhood nodes according to the propagation.The preference probability is incorporated into the representation of network nodes to get the representation of nodes with preference information.Then,the nodes with preference information are represented by graph convolution network and graph attention network learning,and a social network impact prediction method based on preference propagation is proposed.This method does not rely on manual features and content information to predict the social behavior of users.Finally,the model RippleInf-GAT proposed in this paper is validated on the real social network data set,and compared with the existing algorithm by using three indicators of Precision,Recall and F1-score.The results show that the method proposed in this paper is better than the comparison algorithm to some extent.