基于影响簇选择模型和MCMC采样的社交圈子识别算法

提出一种新的紧密度公式和一种影响簇发现模型,并在此基础上设计基于局部社团探测的采样算法MCMCS_LCD,以及基于MCMCS_LCD的社交圈子自动识别算法SCD_MCMCS_LCD,算法综合考虑局部模块度和节点间紧密度.在真实数据集上的实验表明,SCD_MCMCS_LCD算法在具有较快收敛速度的同时还具有较好的社交圈子识别效果.

An automatic detection algorithm for social circles based on influential cluster selection model and MCMC sampler

Discovering social circles in ego networks is a new research direction in social network data mining. Currently, the newer social circles automatic recognition algorithms have high time complexity. This paper proposes a new expression of the close affinities and the influential cluster selection model, and on this basis to design an efficient sampler algorithm, called MCMCS_LCD, and further design an automatic detection method called SCD_MCMCS_LCD. The algorithm takes account into both local modularity M and close affinities. Our experiments demonstrate that, SCD_MCMCS_LCD has a faster convergence speed while still maintains a good social circle recognition effect.