基于高阶信息局部策略的符号网络社区检测

Community Detection in Signed Network Based on High-Order Information Local Search Strategy

现有符号网络社区检测方法中,局部搜索策略作为符号网络社区检测算法的重要组成部分,可加速算法收敛速度,但符号网络局部搜索策略大多仅利用连边及节点等低阶结构信息,忽略了可挖掘符号网络更深层、更丰富信息的高阶结构.为提升现有符号网络社区检测的局部搜索策略性能,提出了一种基于符号模体的局部搜索策略,设计了一种基于符号模体进行社区迁移的新方法,将传统社区编号在二元组之间的迁移扩展到了三元组,综合利用符号网络低阶和高阶拓扑结构信息来优化节点的结构平衡性,提升算法收敛速度和检测性能.在模型网络和实证网络上的实验表明,设计的局部搜索策略相对于现有算法表现出更高的精度和质量.

In existing signed network community detection methods,the local search strategy serves as an important component of the signed network community detection algorithm and can accelerate the convergence rate of the algorithm.However,most of the local search strategies in signed networks only utilize low-order structural information such as edges and nodes,thereby,neglecting the deeper and richer high-order structure that can be mined from signed networks.To enhance the performance of the existing signed network community detection local search strategy,this article proposes a local search strategy based on signed motifs.A new method based on signed motif for community migration was designed,which extends the traditional community numbering migration between binary tuples to ternary tuples.By comprehensively utilizing the low-order and high-order topological structural information of the signed network,the structural balance of nodes is further optimized,thereby,improving the convergence rate and detection performance of the algorithm.Experiments on synthetic and real-world networks demonstrated that the designed local search strategy exhibited good accuracy and quality.