欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
在真实世界网络中，一个人可以属于多个兴趣小组，一个蛋白质可以属于多个蛋白复合体，因此发现网络中的重叠社团结构更能准确地反映网络中真实信息。与现有一些发现重叠社区结构的算法相比，多标签传播算法具有时间复杂度低的优点，但当节点含有多个邻居社区且属于这些邻居社区的隶属度相等且都要小于设定的阈值参数时，多标签传播算法随机地从邻居节点标签中选择社区，这严重影响了算法稳定性。为此，本文提出了一种基于贡献度改进的多标签传播算法。在真实基准网和计算机生成网的测试结果表明该标签传播算法具有较好的社区发现性能，我们将该算法应用在科学网博客中“图书馆、情报与文献学”领域用户的好友关系网上，能有效地发现该领域中存在的重叠社区结构。
英文摘要：
In real-world networks, a man can belong to many different interest groups, a protein can belong to different complexes, so detecting overlapping communities is very important to achieve accurate structures about real-world networks. Compared with existing overlapping community detection algorithms, multi-label propagation algorithm has low algorithm complexity. However, when a node has muhiple belonging neighbor communities and its belonging coeffcients are equal and less than the threshold parameter, that muhi-label propagation algorithm randomly chooses one from these neighbor communities seriously affects the stablity of the multi-label propagation algorithm. Hence, we put forward an optimization of multi-label propagation algorithm based on contribution degree in this paper. The tests on real-world networks and synthetic networks show that our algorithm has stronger validity and stability, then our algorithm is applied at the social network of science bloggers in ＂library,＂ information and bibliography＂ field , and it detects the overlapping communities in these social network effectively.
同期刊论文项目
同项目期刊论文
期刊信息
位置：