基于圈结构的LPANNI优化算法

针对重叠社区发现准确率提升问题，提出了一种基于圈结构的LPANNI优化算法CLPANNI（cycle label propagation algorithm with neighbor node influence）。该算法通过挖掘节点的最小圈信息，依据圈比指标衡量节点的重要性并按升序进行标签更新，增加了标签传播过程的稳定性，按照邻居节点影响力大小加权接收邻居节点传递的标签。与四种基准算法在NMI_LFK、NMI_MGH、MOV指标下进行测试比较，CLPANNI算法在社区发现准确率方面表现较好。实验结果表明，该算法能够有效探测网络重叠社团结构，发现网络的紧密子团，识别的社团分布与真实网络结构更为接近。

LPANNI optimization algorithm based on cycle structure

In order to improve the accuracy of overlapping community detection, this paper proposed an LPANNI optimization algorithm CLPANNI based on cycle structure by mining the minimum circle information of nodes, measuring the importance of nodes according to the circle ratio index and updating labels in ascending order. This algorithm increased the stability of label propagation process, and received the labels transmitted by neighbor nodes according to the influence of neighbor nodes. Compared with four benchmark algorithms in NMI_ LFK, NMI_ MGH and Mov indicators, CLPANNI algorithm performed well in the accuracy of community discovery. Experimental results show that the algorithm can effectively detect the overlapping community structure of the network, find the close sub clusters of the network, and the identified community distribution is closer to the real network structure.