Link Prediction Algorithm Based on Node Structure Similarity Measured by Relative Entropy

Abstract To solve the problem that link prediction method based on local information ignores influence of neighbor structure similarity measurement nodes, a relative entropy and nodes is proposed. Firstly, second-order network introduced to describe nodes; then, structural between described by redefining entropy; finally, measured entropy, index node proposed considering neighbor. Experimental results 7 real data sets show can achieve better be applied networks with small average aggregation coefficient compared other indexes global information, also have performance large-scale networks.