Link prediction algorithm in signed networks based on clustering coefficient and sign influence

Abstract:

In order to achieve the dual goals of link prediction and sign prediction in signed social networks quickly and accurately, a link prediction algorithm is proposed based on the clustering coefficient of common neighbor nodes and the influence of the sign of edges. With the structural balance theory, the similarity of the two nodes based on their first-order common neighbors and the second-order common neighbors is defined respectively by using the degree, clustering coefficient, intermediate transitive nodes, and the influence of the sign of the edge, the total similarity score of the two nodes is finally obtained and its absolute value is used to measure the possibility to establish a link of the two nodes, then its sign is the sign prediction result of the link. Accordingly, the link prediction and sign prediction are realized in signed networks. Experiments have been carried out on six representative signed network datasets, with evaluation indicators such as AUC, adjusted precision' and accuracy. The experiment results are compared with several link prediction algorithms in signed networks the sensitivity of adjustable step size parameters is also analyzed. Experimental results show that the proposed algorithm can achieve good performance in both link prediction and sign prediction, and its accuracy is higher than other algorithms for both sparse networks and the prediction of negative links.