Follower Recommendation Based On Time-aware Hybrid Graph Embedding

In recent years,social networks have developed tremendously and have become an essential part of people's real life.Graph structure information mining has become a hot research topic.In order to make social network information more widely and more frequently,we need to increase the number of links in the network.A very important issue is to recommend to users who they may be interested in.There are a lot of research in the low-dimensional representation of static graph nodes,which have achieved good results in link prediction.However,most of the graphs in real life are dynamic,with timestamp information.This paper first introduces the various existing methods in social network link prediction.We propose a model called CIGAT(Contents and Interests Graph Attention model).Then we can get low-dimension representations of nodes in different aspects to better recommend users.Then on this basis,we propose a model called CIGTAT(Contents and Interests Graph Time-aware Attention model),to learn the influence of contents and interests between users better,to better balance the user's long-term interests and short-term interests and to achieve better link prediction,so as to better recommend users to users.