Research on the Link Prediction Model of Dynamic Multiplex Social Network Based on Improved Graph Representation Learning

In the natural and social systems of real world, various network can be seen everywhere. The world where people live as a combination with different dimensions. Link prediction formalizes interaction behavior between people. Traditional link methods mainly study user static network. This article studied dynamic graph representation learning so to put forward an improved model in Besides, interactions multiple, links at moments may have meanings. proposed firstly solved problem on multiple kinds edges. whole embedding each node is separated into two parts, basic edge embedding. Then selected time slices for get embeddings snapshots. What's more, t+1 step vector was used validate t effect performed better than traditional methods.