聚类分析是web社区发现的主要方法之一,近年来基于谱图理论的谱聚类由于其诸多优点而备受关注,在介绍谱聚类基本理论的基础上,着重分析了包括SM、NJW、NystormCut与KASP在内的4种典型谱聚类方法的基本流程与优缺点,在聚类有效性与时间效率上对4种算法进行实验比较,实验结论为web社区发现工程实践提供了借鉴.
Clustering is an important approach to discovering web community, spectral clustering based on spectral graph theory has been increasingly a major concern. On the base of introduction of principle theory in spectral clustering, procedures, advantages and disadvantages of four typical spectral clustering algorithms, i.e. SM, NJW, NystormCut and KASP, are analyzed, conclusion from experiments on synthesized data and real data can be reference for engineering practice in web community discovery.