| The rapid development of network science brings human lives great convenience,but it also brings about a lot of questions to be explored. In recent ten years, the study of the link prediction has been greatly promoted by theoretical achievements on complex network, at the same time, which is closely related to network evolution mechanism.This paper presents two kinds of link prediction algorithm based on different network properties and builds a mathematical model to describe evolution mechanism of some real-world networks. The major work of this paper includes:(1) Base on local similarity index, we bring in node degree and build a new linkage probability metric index, which we based on to implement a kind of link prediction algorithm.(2) We come up with a timevarying random geometric graph model, which is used to imitate evolution mechanism of some real-world networks. By theoretical calculation and sample analysis, we prove its rationality.(3) Based on spring algorithm, we implements a mapping method from graph to geometric graph. On the basis of connectionism of the time-varying random geometric graph model, we build a linkage probability metric index to implement a kind of link prediction algorithm based on geometric metrics.It turned out that the algorithm based on node degree improves the prediction accuracy than that based on Commom Neighbors and keeps low computational complexity. It also turned out that the algorithm based on geometric metrics is feasible for the prediction of the networks which has high clustering coefficient or clear community structure. It also suggests that the time-varying random geometric graph model is an effective representation of the evolution mechanism on some real-world networks. |
