Talk:Predictive Model Markup Language

A suggestion about the controversial "PMML Tools" section

The "PMML Tools" section has been inserted and deleted several times in recent days. Here is one suggestion: Name the section something else, maybe "Industry Support for PMML", and in the section describe the support that several software packages provide for PMML. I think it is important to material in this section about KNIME, STATISTICA, SAS-Enterprise Miner, and other packages that provide PMML support. One possible citation would be http://journal.r-project.org/2009-1/RJournal_2009-1_Guazzelli+et+al.pdf. Other citations would be needed as well. I think in this context, a description of PMML Converter and Transformations Generator would be appropriate. I think a good citation would be http://kdd11pmml.files.wordpress.com/2011/09/p5_kdd2011_pmmlworkshop_guazzelli_et_al1.pdf. For SAS-EM's PMML support, one possible reference would be Bill Frank's http://www.amazon.com/Taming-Data-Tidal-Wave-Opportunities/dp/1118208781/ref=sr_1_3?s=books&ie=UTF8&qid=1354001370&sr=1-3&keywords=pmml. I think that Knime can convert R models into PMML (again I don't know a citation). Augustus also reads and writes PMML (http://rgrossman.com/2011/12/30/pmml-version-4-1-and-augustus-version-0-5-released/), but a better citation is needed. The 2nd edition of the book "PMML in Action" might be an appropriate source for some of the points made in this section: http://www.amazon.com/PMML-Action-2nd-Edition-Unleashing/dp/1470003244. Karl (talk) 07:35, 27 November 2012 (UTC)[reply]

No. A workshop presentation by Zementis employees about a tool published by Zementis is not a good source to establish that it is due weight to mention that tool in this article. - MrOllie (talk) 15:47, 27 November 2012 (UTC)[reply]

OK. I certainly see your point that a conference presentation is not as weighty a publication as a paper in Nature or Science. But I think two of the other sources I listed would be a good step up from a conference presentation: 1) "The R Journal" (ISSN 2073-4859) is a peer-reviewed journal, and 2) The "PMML in Action" book is notable enough that a 2nd edition version has been released.

Please note, that I am not saying that I think the "PMML Tools" section should be reinstated as it was. I am trying to suggest that it was a good start, but that a full section should include more citations and material about other tool vendor's support for the PMML standard. I am hopeful that a discuss might help, and that maybe User:Sunsetsky would not just try re-posting the same material again, but instead post some enhanced material with more citations, and a framework of other material too, perhaps with a few citation needed tags where he/she doesn't have proper citations, and then other wikipedians could assist in adding more material and citations. Karl (talk) 23:20, 27 November 2012 (UTC)[reply]

Thanks Krexer for your input. Although the validity of a workshop may be questioned (although it was organized by the DMG, the makers of PMML, and part of KDD), I would emphasize that the tools in questions are sanctioned by the DMG. That is the reason both tools are available through the DMG website (note that no other tool is). I believe such sanctioning gives the tools in questions the due weight to be mentioned in the article. Would really appreciate to find out why that is not perceived to be the case. Thanks. Sunsetsky (talk) 23:48, 27 November 2012 (UTC)[reply]