【摘要】 社区发现旨在挖掘复杂网络的社区结构,现有的社区发现方法普遍存在着划分速度和精度不均衡的问题.商空间理论是一种粒度计算理论,通过粒度变换来降低问题求解复杂度,同时保持问题求解精度.提出一种基于商空间的多层粒化社区发现方法(multilayer granulation community detection method based on quotient space,MGQS).该方法首先通过快速粒化操作对复杂网络进行多层次粒化,形成逐层粒化、逐层抽象的多粒度商空间,再依据所求问题选择最佳粒层作为最终划分结果.在公用数据集上的系列实验结果表明,相比于其他算法,该方法既能快速划分不同类型和规模的网络,也能获取多粒度的社区结构并根据所求问题选择最佳粒层,取得较高的模块度值和NMI值.更多还原

【Abstract】 Community detection aims at mining the community structures of complex networks.The existing community detection methods don’t have a tradeoff between speed and accuracy.In order to reduce the complexity of problem and hold the accuracy of results,the quotient space theory is introduced in this paper.Quotient space theory is one of the granular computing theories in which different granular spaces can be transformed for problem solving.A method based on quotient space,named MGQS(multilayer granulation community detection method based on quotient space),is proposed for multilayer granulation community detection.Firstly,fast granulation operation for network is given.Fast granulation operation includes discourse domain(refers to node in the network)granulation and structure(refers to edge in the network)granulation.The complex network is granulated into different granular net-works from thin to coarse,and a multi-granularity quotient space with layer by layer granulation and layer by layer abstraction is formed.Then,according to the object of problem,the optimal granular layer is selected as the final results.Compared to other algorithms,the results of a series of experiments on the public data sets show that,the proposed method MGQS not only can quickly partition the network of different types and scale,but also can obtain the multi-granularity community structure.The optimal granular layer,higher modularity and NMI values can be obtained according to the object of problem.更多还原