完全子图的邻域重叠社团结构探测

摘要

For the purpose of detecting the overlapping community more precisely, the algorithm of community detection based on complete subgraph is improved. While merging the complete subgraphs, the amount of overlapping nodes between the pair of subgraphs is calculated and the threshold value is set. Then the complete subgraphs are merged if the amount of overlapping nodes is larger than the threshold value. The rule of scale factor is used to divide the nodes which are not in any cluster. The algorithm is used in Zachary’s karate club network and scientific collaboration network, which can detect the overlapping community structure more accurately.%　　为了能够更准确地对邻域重叠网络进行社团结构探测，对基于完全子图的社团探测算法进行了改进。在合并完全子图团簇时，计算每一对完全子图的重叠节点个数，设置合并完全子图的阈值，如果大于阈值，则合并。当处理不在团簇内的其他节点时，按照比例系数大小为划分规则进行划分。该算法可以应用于空手道俱乐部和科学家合作网当中，其验证算法可以更准确地探测邻域重叠社团结构。