Link Prediction and Unlink Prediction on Dynamic Networks

Link prediction on dynamic networks has been extensively studied and widely applied in various applications. However, existing methods only consider either the network structure or temporal information, ignoring potentialities of using both types information together to comprehend complex behaviors networks. Moreover, unlink prediction, which also plays an important role evolution social networks, not paid much attention. Accurately predicting links unlinks future greatly contributes analysis that uncovers more latent relations between nodes. In this work, we assume there are two kinds nodes, namely, long-term short-term relations, propose effective algorithm called LULS for link based such relations. Specifically, each snapshot a network, first collects higher order structures as topological matrices by applying short random walks. Then, initializes optimizes global matrix sequence temporary all snapshots nonnegative factorization (NMF) matrices, where denotes represent snapshots. Finally, calculates similarity predicts network. addition, further improve results graph regularization constraints enhance matrix, resulting contains wealth information. The conducted experiments real-world illustrate outperforms other baselines tasks.