Exploiting neighbors' latent correlation for link prediction in complex network

Link prediction, which seeks to explore missing links between nodes, is an important task in complex network analysis. Although this problem has attracted much attention recently, there are still several challenges that have not been addressed so far, even for the most popular one: similarity link prediction based on common neighbors. Most existing algorithms focus on how to enhance neighbors' role to the candidate pair, and takes the neighbors' role as the sole contribution. For this reason, these algorithms seldom pay attention to how neighbors may influence to others since neighbors may link together in real network. To address this issue, in this paper, we investigate the problem of defining the latent correlation between common neighbors and improve several similarity-based methods via two modified naive Bayesian models. The experimental results on several real-world networks demonstrate the effectiveness of our models.