Link Prediction In Complex Network

A complex network is a network with small worlds and scale-free features.The explosive growth of the Internet allows all areas of life to be abstracted into complex networks,such as social networking networks,ecological networks,collaborative paper networks,and criminal networks.Link prediction helps to understand and understand the network structure.Therefore,the research of link prediction in complex networks is extremely important.Aiming at the problem that link prediction only relies on the local information of nodes,the prediction accuracy is low.This paper proposes a link prediction algorithm based on node degree for secondary neighbor nodes and a link prediction algorithm based on weighted average clustering coefficients.Firstly,in the network without weight,a link prediction algorithm based on node degree for secondary neighbor nodes is proposed for the problem that the accuracy of the existing link prediction algorithm is low.By adding the calculation of the secondary neighbor node,the importance of the secondary neighbor node in the network is solved.At the same time,when calculating the connection probability of two nodes that are not connected,considering the contribution degree of different common neighbor nodes,the node degree is introduced as a penalty factor.Secondly,in the network without weight,in order to distinguish the interaction between nodes and the different contributions of multi-level nodes,a link prediction algorithm based on weighted average clustering coefficients is proposed.By calculating the average clustering coefficient based on the same degree as the edge weight,the problem of the interaction strength between two nodes is solved.At the same time,considering the propagation between nodes,two nodes that are not connected can propagate messages through neighboring nodes.Since the longer the path,the weaker the propagation strength,only the propagation of the second-order path and the third-order path are considered.Finally,the node-based second-level neighbor node algorithm and the weighted average clustering coefficient algorithm are compared with other classical algorithms onthe real dataset,and the correctness and effectiveness of the proposed algorithm are verified.