基于博弈的有向时序网络两阶段社区发现算法

文中提出一种新的用于时序有向网络的两阶段社区发现算法。第1个阶段通过节点距离、源节点影响力、目标节点影响力和节点分解度4个矩阵，确定节点在网络中的重要程度；第2阶段根据节点重要程度确定社区划分的核心节点，接着分别令核心节点作为源头进行级联传播，其他节点通过博弈确定与各个领导节点的跟随度，最后选择跟随度最高的节点所领导的社区。将算法在国际贸易网络中进行实证研究表明，该算法能够适用于现实世界时序有向加权网络中。

Two-stage Community Discovery Algorithm for Directed Dynamic Network Based on Game Theory

A new two-stage community detection algorithm for time series directed networks is presented. In the first stage, the importance of nodes in the network is determined by four matrices: node distance, source node influence, target node influence and node decomposition degree. The second stage determines the core node of the community division according to the importance of the nodes, then makes the core node as the source for cascade dissemination, other nodes determine the degree of follow with each leading node through the game, and finally the community led by the node with the highest degree of follow is selected. An empirical study on the algorithm is carried out in the International trade network show that the algorithm can be applied to the real world time series weighted network.