Link Prediction Algorithms Based On Local Topology And Path Influence

In recent years,the rapid development of computer science has brought great convenience to human life.At the same time,the complex relationships between people,people and things,and things have formed various complex networks.The rapid development of network science has provided powerful help for understanding and exploring these networks.As one of the popular research contents of network science,link prediction is mainly used to explore unknown or missing edges in the network.The research of link prediction can not only help to accelerate the development and evolution of network science,but also has a wide range of application value in biological networks,social networks and other fields.Classical link prediction algorithms are mostly based on the nature of the network topology.However,most of these algorithms ignore the importance of nodes and links themselves.Therefore,this article proposes a new chain from three perspectives.Road prediction algorithm.Firstly,since the common neighbor index does not take into account the importance of the degree of the common neighbor node,this paper starts from the node degree centrality(DC),betweenness centrality(BC)and proximity centrality(CC),combined with the common neighbor index,and proposes Introduced new link prediction indicators CDC,CBC and CCC.Experimental analysis on multiple real networks shows that the CCC and CDC indicators have a good predictive effect,which verifies the effectiveness of the algorithms.Secondly,in order to improve the prediction accuracy,based on the above three indicators,combined with the compactness of the edges between the predicted node pairs,this paper proposes new indicators ECDC,ECBC and ECCC,and experiments on multiple real networks,The results show that these indicators have improved compared to the original indicators.Finally,due to the limited amount of information,the index proposed based on the common neighbor information may cause the constructed network structure to be incomplete,thereby affecting the prediction accuracy.For this reason,this article combines the path between nodes and proposes a path-based influence(PBI)link prediction algorithm,which weakens the inefficient influence of long paths on link prediction.Experiments on multiple real networks show that the PBI indicator has a good predictive effect and certain applicability.