中文摘要：

传统的社区发现方法多是基于同构网络和拓扑结构,为此,提出基于异构信息网络信息维统计量的社区发现算法,该算法通过对异构信息网络进行信息维上卷后构建概率模型,采用引入模块度最大化曲线的层次化狄利克雷过程自适应地确定社区数目,然后通过狄利克雷分布来表征某个社区,再通过计算最大后验概率来进行社区发现。实验表明,所提出算法相比基于拓扑结构的算法具有更好的社区发现效果和稳定性。

英文摘要：

As traditional community discovery methods are based on the homogeneous network and topology structure, presents a community discov-ery algorithm based on information dimension statistics of heterogeneous information network. This algorithm rolls up information dimen-sion to build probabilistic model, uses hierarchical Dirichlet process with modularity maximization curve to adaptively determine the num-ber of communities, characterizes communities with Dirichlet distribution and discovers communities by calculating the maximum posterior probability. Experiments show that compared with the algorithm based on topological structure, the proposed algorithm has better commu-nity discovery effect and stability.