在大规模复杂网络社区划分中,标签传播算法已经被证实为一种速度极快的算法,被广泛应用。但是标签传播算法还存在一些缺陷,比较突出的是社团划分结果的不稳定,鲁棒性较差。通过某些指标来计算节点在网络中的影响力,在节点第一次更新时,有效地将影响力较大的核心节点标签值传播出去,准确形成各个社区的基本框架,大幅改善了传统标签传播算法的鲁棒性,同时取得了更好的社区划分效果。
Label propagation proves itself an extremely fast algorithm for community detection of large-scale complex network, and thus is widely applied. However, some flaws still exist in this algorithm. With some parameters to calculate the influence factor of all nodes in the network and effectively propagate the label of core nodes with high influence in the first iteration, the basic frame of each community is thus exactly formed. Experimental results indicate that all this could significantly improve the robustness of traditional label propagation algorithm while raising the performance of community detection.