User talk:HelpUsStopSpam

Cluster Analysis page edited by SPA

The Cluster Analysis page is being overedited/overhauled by a special purpose account Glokc (talk · contribs · deleted contribs · logs · filter log · block user · block log). It deserves a look and some scrutiny. He keeps reverting me and ignoring my signals. Limit-theorem (talk) 18:04, 2 February 2019 (UTC)

The Cluster Analysis page has been extended by me and all the made statements have references to the state-of-the-art scientific papers published in the peer reviewed journals and conferences. I have not deleted any of the original material, just restructured it a bit, significantly extended and modified some statements (citing the trusted scientific papers). I believe that the article has been improved with the made changes and would be glad to receive some feedback and refinements for the made edits but the account Limit-theorem (talk · contribs · deleted contribs · logs · filter log · block user · block log) just "undo"-ed my extensions multiple times instead of refining them further. --Glokc (talk) 18:15, 2 February 2019 (UTC)

@Limit-theorem: while the edits by Glokc are a bit too excessive for my liking (and therefore should probably have been discussed on the talk page first, at least after the first revert), and a bit suspicious because the user has not edited anything before, I am not convinced that they are spam. I don't like them because much of the addition is a rather useless list of abbreviated algorithms. And some of the "survey" articles are pretty bad these days, too... Some of the rewriting contradicts my shallow understanding of the matter (e.g., "Nowadays ... clustering is used for various applied tasks including ... construction of the numerical taxonomy, botryology", when numerical taxonomy is where clustering was actually originally used first? Like in the 60s, not so much "nowadays".) There is one reference in there that I consider suspicious (Lutov 2018, no citations), but apart from that it does not look like the usual spam to me, so I leave this to the domain experts to handle. @Glokc: when you get reverted once, using the Talk page is usually a good idea. Try to find consensus, not try to "win" by repeated reverting. HelpUsStopSpam (talk) 22:21, 2 February 2019 (UTC)

Thanks. We will watch. Limit-theorem (talk) 00:08, 3 February 2019 (UTC)

@Limit-theorem: With the latest additions, it seems that indeed that Glokc may have a WP:COI, and tries to promote Clubmark, which has 0 citations and now plays a prominent role here. In fact, it very much looks like a WP:COPYVIO with at least page 2 of: https://exascale.info/assets/pdf/icdm18_clubmark.pdf IMHO it is also mostly misplaced, because there is a separate article on community detection... what do you think? HelpUsStopSpam (talk) 09:58, 4 February 2019 (UTC)

@Limit-theorem:, @HelpUsStopSpam: I extended the Cluster Analysis page with all published benchmarking frameworks for the clustering algorithms evaluation I'm aware about. Please, refine the article restructuring or modifying it to make it better. Yes, Clubmark is the newest benchmark, it was presented in ICDM this November and either does not have any citations or such publications have not been indexed yet but this benchmark does have a published paper and provides both the executables and sources as most of other mentioned benchmarking frameworks.

Community structure detection is a subset of the clustering task from the perspective of the graphs clustering specified by the pairwise relations and the superset of the clustering task from the perspective of the extensions the clustering with additional heuristics to model and recover the actual social communities instead of (based on) the found clusters by some (intrinsic) statistical properties. All the mentioned benchmarking frameworks and toolkits perform evaluation of the pure clustering algorithms and/or clusters formed by these algorithms. --Glokc (talk) 10:28, 4 February 2019 (UTC)