基于谱聚类的二分网络社团检测算法

Community Detection Algorithm for Bipartite Networks Based on Spectral Clustering

针对二分网络社团检测算法存在精度不高和丢失原始网络信息等问题,设计了一种新的融合奇异值分解的谱聚类(SVD-MS)算法.该方法是将Barber的二分网络模块度最大化问题映射到奇异值向量分解上,并结合启发式算法快速求解向量划分问题.在3个真实世界的网络中对比SVD-MS算法与7种算法的模块度,结果表明,在保留原始网络信息的情况下,SVD-MS算法能更有效地划分二分网络的社团结构.

In order to solve the problem of low precision and original network information lose in community detection of bipartite network,a new spectral clustering algorithm named SVD-MS is proposed.This method maps Barber's problem of maximizing the module size of bipartite networks to the problem of singular value vector decomposition,and combines heuristic algorithms to quickly solve vector partitioning problems.Experimental results show that,the SVD-MS algorithm can effectively partition the community structure of bipartite networks and preserve the original network information.