Graph modularity maximization as an effective method for co-clustering text data
作者：
Highlights：
• We describe and prove the convergence of the CoClus co-clustering algorithm, which is based on the direct maximization of graph modularity.
• In rigorous comparative experiments, we demonstrate the effectiveness of Coclus to co-cluster document × term matrices.
• We show that the modularity measure can be used to identify good candidate numbers of co-clusters.
• We provide a systematic and comprehensive benchmark comparing most known block diagonal (as well as several well-known non-diagonal) coclustering algorithms.
摘要
•We describe and prove the convergence of the CoClus co-clustering algorithm, which is based on the direct maximization of graph modularity.•In rigorous comparative experiments, we demonstrate the effectiveness of Coclus to co-cluster document × term matrices.•We show that the modularity measure can be used to identify good candidate numbers of co-clusters.•We provide a systematic and comprehensive benchmark comparing most known block diagonal (as well as several well-known non-diagonal) coclustering algorithms.
论文关键词：Co-clustering,Modularity
论文评审过程：Received 10 January 2016, Revised 30 June 2016, Accepted 1 July 2016, Available online 4 July 2016, Version of Record 3 September 2016.
论文官网地址：https://doi.org/10.1016/j.knosys.2016.07.002
