The Improvement And Application Of Dynamic Network Node Influence Index Based On PageRank Algorithm

In the real world,people are connected in an interactive way in society(the theory of six-degree segmentation),the person with the highest contact density(nodes,centers)is a key participant responsible for the largest propagation process.It can be seen that among the many factors of rapid and widespread dissemination,influential spreaders play an important role in determining the influential "spreaders"(key nodes)in the network,which is necessary to understand the process of network information dissemination.Compared with the simple and direct computational central algorithm,PageRank algorithm expands the concept of network center and node influence computing application range,but it is still a static algorithm for the non-right network,more and more not applicable to such nodes,edges and edge contact strength are variable dynamic network.In practice,the influence size of dynamic network nodes is closely related to such factors as particle size,time span and contact strength between nodes.Based on this,the dynamic network node influence algorithm based on PageRank algorithm,EWTPR algorithm,is proposed.First of all,this paper puts forward a dynamic network division method that does not require any prior information of the real network structure and does not contain the parameters that need to be set,divides a complete dynamic network into independent static network snapshots according to the time node,and then overlays the network in turn;Extending the non-static network PageRank algorithm to calculate node influence at each moment in the weighted overlay network,then introducing a time factor(parameter ?)to adjust the degree of influence of node influence on the node at the current moment,taking into account the time smoothness of dynamic network node influence,and finally,a step-by-step experiment in simulated network and largescale real network,which showed that compared to PageRank and the central algorithm(Degree centrality,Betweenness centrality,Closeness centrality),If EWTPR algorithm do not consider the introduction of time factors,only consider dynamic slicing and edge weight,although further improve the accuracy of dynamic network node influence ordering,but cut off the time relationship of node impact,there are certain limitations,and the introduction of time factor EWTPR algorithm to the different cutting network node influence ordering and real sorting results of Kendall coefficient and Spearman coefficient are much higher than other algorithms.And its mean square error and standard error are far lower than other algorithms,which fully demonstrates the accuracy and validity of the EWTPR algorithm in the ordering of dynamic network node influence.