An efficient algorithm for link prediction in temporal uncertain social networks

Due to the inaccuracy, incompleteness and noise in data from real applications, uncertainty is a natural feature of real-world networks. In such networks, each edge is associated with a probability value indicating its existence in the network. Predicting links in uncertain networks is computationally more challenging and differs semantically from predicting connections in deterministic networks. This paper presents a method for link prediction in temporal uncertain networks. In our method, the predicting problem is formalized by designing a random walk in temporal uncertain networks. The algorithm first transforms the link prediction problem in uncertain networks to a random walk in a deterministic network. Then, the similarity scores between a node and its neighbors are computed within a sub-graph around this node to reduce the computational time. The proposed method integrates temporal and global topological information in temporal uncertain networks and can obtain more accurate results. Experimental results on real social networks show that our method can predict future links efficiently in temporal uncertain social networks and achieves higher quality results than other similar methods.