Efficient Link Prediction in Continuous-Time Dynamic Networks Using Optimal Transmission and Metropolis Hastings Sampling

—Effective link prediction in continuous time dynamic networks is a challenging issue that has received much research attention in recent years. A widely used method for dynamic network link prediction is to extract the local structure of the target link through random walks on the network, and then employ encoder models to learn node features. However, this method often incorporates prior information, assuming that candidate neighbors adhere to predefined criteria for spatially adjacent, without considering the temporally proximate principles and its consistence with spatial similarity and high-order correlation. To address this limitation, we propose a framework in Continuous-time dynamic networks based on Optimal Transmission and Metropolis Hastings sampling (COM). Specifically, we utilize optimal transmission theory to calculate the distribution similarity between the current node and the time-valid candidate neighbors, aiming to minimize the loss of temporal information during node information propagation. Then we couple spatial structural information by applying the Metropolis-Hastings algorithm, which captures the local structure of target links and global spatial correlations within the network. We realize the fusion of spatiotemporal information using an encoder model. The extensive experiments performed on 14 datasets from different fields demonstrate the superiority of COM compared with representative and state-of-the-art competitors.