Link Prediction in Weighted Networks: The Role of Structure Correlation

Abstract

Recently, the problem of missing link prediction in complex networks has attracted much attention. Among these studies, the weight of links and the structure weight between pairs of nodes are rarely taken into account. In this paper, we propose a method to analyze the structure correlation between pairs of nodes and apply four different arithmetic-operators to measure the accuracy of the link prediction in two real networks. The results show that the structure correlation from the global network plays an important role in the link prediction. Further experimental study shows the optimal distance of the structure correlation between pairs of nodes is the key factor, otherwise it would have the opposite effect.