基于联合建模下一类新的网络社区检测方法  

A New Class of Web Community Detection Methods Based on Joint Modeling Framework

作　　者：丁烨青 付英姿[1,2] 唐玉萍 DING Ye-qing;FU Ying-zi;TANG Yu-ping(Faculty of Science,Kunming University of Science and Technology,Kunming Yunnan 650500,China;Center for Applied Statistics,Kunming University of Science and Technology,Kunming 650500,China)

机构地区：[1]昆明理工大学理学院,云南昆明650500 [2]昆明理工大学应用统计学研究中心,云南昆明650500

出　　处：《计算机仿真》2025年第1期377-381,共5页Computer Simulation

基　　金：国家自然科学基金项目(11201200,11561035)。

摘　　要：社区检测是理解社交网络结构的重要工具,然而对于大型复杂社交网络而言,单一的建模策略难以精准刻画其全局稀疏和局部稠密的特征,于是提出了联合建模框架下一类新的网络社区检测方法,有针对性地对社区间和社区内的网络结构分别进行拟合。在以上基础上,建立起模型参数估计的后验贝叶斯算法,并以一组印度村庄的公开数据为例进行模型构建以及方法验证。实验结果表明,上述法能够准确得到参与者的社区划分,且拟合精度高,计算复杂度适中。在实际应用中,可以灵活嵌套不同的建模手段以合理描述大型社交网络的异质性和局部特征,具有广泛的应用推广前景。Community detection is an important tool for understanding the structure of social networks,however,for large complex social networks,it is difficult for a single modeling strategy to accurately portray their globally sparse and locally dense characteristics,so we propose a new class of network community detection methods under the joint modeling framework to fit the inter-community and intra-community network structures respectively.On this basis,we establish the posterior for model parameter estimation.The experimental results show that this method can accurately obtain the community division of participants with high ftting accuracy and moderate computational complexity.In practical applications,different modeling tools can be flexibly nested to reasonably describe the heterogeneity and local characteristics of large social networks,which have wide application prospects.

关 键 词：社区检测 联合建模 潜在空间模型 随机块模型 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]