Abstract: An improved CNM community detection algorithm is proposed to detect communities in a complex network with certain nodes already partitioned. These nodes are defined as the source nodes in this algorithm which uses the modularity increment maximization as the objective function to group the unpartitioned nodes. Communities are identified by the proposed algorithm when the highest total modularity is reached. An example based on the mobile phone data collected in Shenzhen, China is applied to the algorithm. The base stations in the planed urban centers are treated as the source nodes in the algorithm to detect the communities in Shenzhen. The results indicate that the proposed algorithm is effective in detecting the boundaries of major urban centers as well as other urban communities.

Key words: complex network, community detection, modularity

摘要： 提出了一种改进的带源节点的CNM快速社区发现算法，用于含有部分已知社区属性节点的复杂网络的社区结构划分。算法中将这部分节点作为源节点，采用模块度增量最大化为目标函数对待划分节点进行聚合，并在取得最大全局模块度值时得到社区划分结果。以深圳市手机基站用户流量网络为例，将位于各规划城市中心的基站点作为源节点引入，对城市进行区域划分。结果表明，该算法不但能够发现基于各城市中心的服务边界，也能发现一些隐含的城市区域。

关键词: 复杂网络, 社区发现, 模块度