A Survey of Dynamic Network Link Prediction

Link prediction has been one of the hot topics in the field of dynamic network research in recent years. Its research focuses on predicting the link state at future times based on historical network link structures and node states, among other information. This approach can help deepen our understanding of the internal operation mechanisms of complex systems and provide support for upper-layer protocols such as routing. Building upon previous research, this paper systematically categorizes recent link prediction methods on dynamic networks. We first introduce the relevant concepts of dynamic networks and link prediction, then classify the link prediction methods on dynamic networks in detail. We summarize different types of link prediction methods and introduce typical representative methods of each type in detail,, including link prediction methods based on graph neural networks that have received extensive attention recently. Finally, we summarize the problems that need to be addressed in further research in this field and discuss possible future directions.