Link prediction in complex networks: a clustering perspective

Link prediction is an open problem in the complex network, which attracts much research interest currently. However, little attention has been paid to the relation between network structure and the performance of prediction methods. In order to fill this vital gap, we try to understand how the network structure affects the performance of link prediction methods in the view of clustering. Our experiments on both synthetic and real-world networks show that as the clustering grows, the accuracy of these methods could be improved remarkably, while for the sparse and weakly clustered network, they perform poorly. We explain this through the distinguishment caused by increased clustering between the score distribution of positive and negative instances. Our finding also sheds light on the problem of how to select appropriate approaches for different networks with various densities and clusterings.