中文摘要：

随着社会媒体的发展，在线社会网络成为重要的信息传播平台．信息传播实则反映用户影响力的传播，传播的过程促使网络拓扑结构的动态演变，也逐渐形成社团．同时，社团的形成又反过来影响信息传播．节点影响力描述了节点影响其他节点的能力，而节点影响力受所处的位置和活动表现所影响．本文定义了结构中心度和活动权威性，并计算出节点之间的相互影响概率．再综合考虑影响概率、潜在重叠节点及标签选择策略，提出了基于标签传播的改进算法．在同构和并构网络上进行实验，实验结果证明，该算法能够有效地发现异构社会网络q-重叠社团．得到的社团不仅社团内紧致度高，社团间分散度高，同时社团内节点之间社交行为相似性和互动性也高．

英文摘要：

As the fast development of social media, online social network has became an important platform for information propagation. Information propagation reflects the propagation of user influence in real life, which leads to the dynamic evolution of network topology and formation of community structure. Also, community structure reflects on propagation. We propose an overlapping community detection algorithm based on node influence propagation. Node influence describes ability of node to influence other nodes, which includes two factors-the location and activity behavior of node. We introduce a novel definition about structure centrality and activity authority, and compute a new influence probability considering the structure and activity influence. Also, we propose and optimized label propagation algorithm over heterogeneous network considering the influence probability potential overlapping vertices and corresponding label selection heuristic. The experiments on the homogeneous and heterogeneous social network verify the effecttiveness of our algorithm to detect the overlapping communities. The extracted communities not only have high closeness intra-community, and high dispersion inter-community,but also have high social behavior similarity and interactivity intra-community.