| Nodes with similar attributes in the network can form a community.And community structures are ubiquitous in complex networks.Facing today's ultra-large-scale networks,quickly detecting the community structure in the network helps to discover the inherent properties and laws of complex networks.The Label Propagation Community Detection Algorithm(LPA)has the advantages of simple thinking,near-linear time complexity,etc.But it also has the problems of strong randomness and poor stability of results.In this paper,aiming at the problems of LPA,the influence of node influence on label propagation is studied,and the stability of LPA detection results is improved.The main works of this paper include:(1)A label propagation community detection algorithm based on node influence is proposed.When the label is initialized,only some nodes with high influence are selected to allocate labels,which reduces the number of labels.If there are multiple labels to be selected when the node updates the label,the selection is based on the influence of the corresponding node of the candidate labels.So the randomness of the classical LPA algorithm is avoided.The use of real network datasets proves that the proposed algorithm improves the stability of the algorithm without increasing the complexity,improves the quality of community detection,and reduces the number of algorithm iterations.(2)The application of node influence in the community detection label propagation algorithm is proposed.At the initial stage of propagation,the node influence maximization algorithm is used to select the nodes with strong influence in the network as the propagation source,which improves the efficiency of label propagation.Secondly,based on the comprehensive consideration of the node influence index,a new measurable node influence index is proposed and then the label update order is guided.Finally,the real network data set is used to verify the reduced randomness and improved stability of the improved algorithm. |
PDF Full Text Request