Currently,the number of Intemet users,along with complex networks including online social networks and electronic commerce networks,is growing explosively.To effectively and efficiently detecting overlapping community structure from complex network,big data plays an essential role in point of interest recommendation and hotspot propagation.In this study,a new algorithm over complex networks is proposed to detecting overlapping communities with a time complexity of O(nlog2(n)).The algorithm applies a new method for updating node and edge modularity based on the techniques of modularity clustering and graph computing.Balanced binary tree is used to index the modularity increment,and an overlapping community detection approach is provided based on the idea of modularity optimization to reduce the frequency of node analysis compared to traditional approaches.Experiments are conducted on real complex network big data,and the results show that the DOC algorithm can effectively detect overlapping communities with high accuracy,the normalized mutual information (NMI) can reach to 0.97 in large-scale LFR benchmark datasets,and the overlapping community detecting standard F-score value is averagely higher than 0.91.In addition,the runtime efficiency beats traditional approaches in complex network big data.%提出一种新的面向复杂网络大数据的重叠社区检测算法DOC(detecting overlapping communities over complex network big data),时间复杂度为O(nlog2(n)),算法基于模块度聚类和图计算思想,应用新的节点和边的更新方法,利用平衡二叉树对模块度增量建立索引,基于模块度最优的思想设计一种新的重叠社区检测算法.相对于传统的重叠节点检测算法,对每个节点分析的频率大为降低,可以在较低的算法运行时间下获得较高的识别准确率.复杂网络大数据集上的算法测试结果表明:DOC算法能够有效地检测出网络重叠社区,社区识别准确率较高,在大规模LFR基准数据集上其重叠社区检测标准化互信息指标NMI最高能达到0.97,重叠节点检测指标F-score的平均值在0.91以上,且复杂网络大数据下的运行时间明显优于传统算法.
展开▼
