Opportunity network link prediction based on graph-BERT

There are many methods of opportunistic network link prediction, and scholars at home and abroad have also fully studied them, and many analysis methods can be well applied today. In order to improve the message delivery efficiency of opportunistic networks, it is necessary to determine the appropriate relay node for message forwarding. Link prediction can discover implicit links between nodes and help in the selection of relay nodes. In order to improve the link prediction efficiency of opportunistic network nodes, this paper proposes a link prediction method based on graph-bert. In this paper, the comparative learning objective of link prediction is designed as a fine-tuning task, in which the positive sample of each training is the neighbor node and the negative sample is the non-neighbor node, which can efficiently discover the topology of the opportunistic network. In the experimental part, the proposed method is compared with LSTM, RNN and classical link prediction models. The results show that this method has good accuracy for link prediction of opportunistic networks.