The discovery of communities in dynamic networks is a critical

Table 4.
    The parameters used for WP-1066 generation of synthetic networks.nkmaxkmincmaxcOnOnOmOmμ500020502050000.1–0.8Full-size tableTable optionsView in workspaceDownload as CSV
Table 5.
    Real networks used for the evaluation of DLPA+.DatasetReferenceVerticesEdgesprotein–protein (PP)[23]26146379word_association (WA)[23]720731,784PGP[24]10,68024,316Amazon (AM)[25]334,863925,872Full-size tableTable optionsView in workspaceDownload as CSV
Table 6.
    Core parameters used for the generation of the 4 dynamic networks.snkmaxkmincmaxcOnOnOmOmμ10100010502050000.3Full-size tableTable optionsView in workspaceDownload as CSV
Table 7.
    Custom parameters used for the generation of the 4 dynamic networks.Birth/deathExpand/contractMerge/splitSwitchbirthdeathexpandcontractmergesplitp1010101010100.3Full-size tableTable optionsView in workspaceDownload as CSV
Then, we evaluate the performance of DLPAE on two real datasets Enron and DBLP, both contain overlapping communities. As for Enron, we use the same Enron data described by Tang et al. [17]. The data provides email information collected from 150 senior executives in the Enron company. We only consider the time period from January, 2000 to October, 2000 and thus 10 months data. As for DBLP, from the original DBLP data we extract papers published between 2000 and 2009 in the field of database, data mining and artificial intelligence. Scholars contributing to one work are connected, and users without edges to others are removed. Information about the two datasets are shown in Table 8.