基于结构稳定性校准的在线式社区识别

本文探讨在线社会网络的社区识别问题, 重点研究网络演变特性对社区结构产生的影响. 首先基于节点的邻域倾向性提出社区稳定性的概念并给出稳定社区的快速识别算法, 然后设计了一种基于事件的社区稳定性校准算法以此识别新网络的社区结构. 由于算法的局部搜索策略, 该方法无需在新时间片段重复执行, 并且可以在无参数条件下识别加权网络中具有任意形状的社区结构. 在人工合成网络和真实网络上的实验结果验证了算法的可行性和有效性.

Identifying Online Communities by Calibrating Structure Stability

This paper investigates into the problem of identifying communities in online social networks. It focuses on the effect of evolution behavior of network on the community structure. In the work, the notion of community stability and a high-speed algorithm for mining stable community are first proposed according to the tendency between nodes and its neighborhood. Then a community stability calibrating method is designed, which takes advantage of the evolution event of network to recognize the new snapshot communities. Thanks to the local search strategy of the algorithm, the new method does not require to execute repeatedly in new snapshot and is able to recognize community structures with arbitrary shapes under the condition of no input parameters. Finally, competitive experiments on both synthesized and real-world social networks demonstrate the effectiveness and efficiency of the proposed algorithm.