Link Prediction Based on 3D Convolutional Neural Network
Jian Shu, Jiahao Li, XuePei Zhang · 2022 IEEE/CIC International Conference on Communications in China (ICCC) · 2022
The dynamic evolution of networks opportunities is very complex, which brings challenges to link prediction. By analyzing the mobile characteristics of nodes, a new link prediction algorithm of opportunistic networks based on a 3D convolutional neural network (3D-CNN) was proposed. It uses the evolution information of the whole network topology to predict the possible changes of local links. In this paper, network attributes under different dimensions, such as time, space, correlation between nodes, are combined to represent opportunistic networks. The pattern classification method is adopted to predict future link combinations between multi-nodes. Using a 3D convolutional neural network to extract spatiotemporal features from multi-dimensional attributes can affect local link status. According to the extracted features, the evolution trend of future links can be inferred, and the multi-node link prediction of opportunistic networks can be realized. We have designed experiments under different conditions to find the best parameters. The experimental results demonstrate that our model achieves impressive results compared to the state-of-the-art competitors.