Link Prediction Based on Path Force for Weighted Network

In view of the fact that most of the existing link prediction algorithms only consider the local or global characteristics of the network, it is difficult to balance the prediction accuracy and computational complexity, and the link prediction research on the path information of the weighted network does not fully mine the information on the path. To this end, for undirected weighted networks, the event that there is an edge between any two nodes of the network is defined firstly, and the calculation method of the event force and the path force is given in this paper. Secondly, the path similarity contribution and the path average cluster coefficient are defined. Finally, the event force, path force, path similarity contribution and path average cluster coefficient are fused, and a weighted network link prediction algorithm, called WPF, based on path force is proposed in this paper. The experimental results show that the proposed WPF algorithm has higher prediction performance by comparing with the relevant algorithms on twelve real weighted network datasets, with AUC as the main evaluation index and Precision as the auxiliary evaluation index.