Community detection method is significant to character statistics of complex network.Community detection in inhomogeneous structured network is an attractive research problem while most previous approaches attempted to divide networks into communities which are based on node or edge measurement.The label propagation algorithm (LPA)adopts semi-supervised machine learning and implemented community detection in an intelligent way with automatic convergent process of network node label iteration but it often results in low efficiency in the final convergent process.In this article,ai-ming to promote low efficiency and stagnant converging rate of LPA in network with overlapping communities,we propose a new method (ESLPA)for community detection using epidemic spreading model combined with network connection ma-trix’s largest Eigenvalue as label propagation threshold.Extensive experiments in synthetic signed network and real-life large networks derived from online social media are conducted to explore optimal mechanism of the most suitable community de-tecting virus infection threshold.Experiments result prove that our method is more accurate and faster than several traditional modularity based community detection methods such as COPRA,GN,FastGN and CPM.%社团结构划分对于分析复杂网络的统计特性非常重要。在非均匀社交网络的信息传播中，社团结构划分更是一个广泛关注的研究热点，相关研究往往侧重于研究紧密连接的社团结构对于信息传播所产生的关键影响。传统社团划分方法大多基于点和边的相关特性进行构建，如标签传播算法LPA（Label Propagation Algorithm）通过半监督机器学习方法，基于网络节点标签的智能交换和社团融合过程进行社团划分，但运行效率较低。为提高LPA类算法的运行速度，使其快速收敛，并提高社团划分精度，特别是重叠社团划分精度，针对LPA算法划分中的低运行效率和低融合收敛速度，本文从标签传播的网络连接矩阵本质出发，将该矩阵的最大非零特征值与网络标签信息传播的阀值相结合，提出了新的基于传染病传播模型的社团划分方法（简称ESLPA算法，Epidemic Spreading LPA）。通过经典LFR Benchmark模拟测试网络、随机网络以及真实社交网络数据上的算法验证，结果表明该算法时间复杂度大幅优于经典LPA算法，在重叠社团划分上精确度优于基于LPA模型的经典COPRA算法，特别是在重叠社团较明显时，划分精度接近精度较高GA、N-cut和A-cut算法，明显优于GN、FastGN和CPM等经典算法。
展开▼
