Subscribe to Dr. Granville's Weekly Digest

Information

PMML

PMML (Predictive Model Markup Language) provides a standard way to represent data mining models so that these can be shared between different statistical applications.

Website: http://www.zementis.com
Members: 90
Latest Activity: Mar 21

The purpose of this group is to promote the understanding and the use of PMML as a standard to represent data-mining models. Questions and answers as well as your point of view about PMML are welcome.

If you want to learn more about PMML or see a list of PMML resources, click HERE.

ATTENTION: Please, only post discussion topics that are directly related to PMML.

Comment Wall

Comment

You need to be a member of PMML to add comments!

Comment by Alex Guazzelli on June 11, 2009 at 12:04pm
If you are attending KDD, please note that Zementis has organized a panel around Open Standards and Cloud Computing. Among the main topics to be discussed is PMML. Also, representatives from the DMG and vendors supporting the standard will be part of the panel. Make sure you attend if you make it to Paris in two weeks. More info here: http://www.sigkdd.org/kdd2009/panels.html
Comment by Alex Guazzelli on March 12, 2009 at 2:24pm
Just a quick comment. We finally moved our R to PMML movie to YouTube. If you want to find out how easy it is to export models from R, check it out:

http://www.youtube.com/watch?v=3t8LiXlBL40
Comment by Alex Guazzelli on February 17, 2009 at 4:18pm
Hi Ross! That's something Zementis has in mind for a while now given that such a tool does not exist. We have actually started to collect requirements for it. If you would like to contribute your list of requirements, please contact me directly (Alex.Guazzelli@zementis.com). As for debugging a PMML producer/consumer, you can find that functionality with the Zementis tools available for free on the web. We offer now the PMML Converter tool which besides converting older version of PMML to version 3.2 also checks for schema violations. If any are detected, the converter pin-points the exact location together with a meaningful message. On the other hand, the ADAPA scoring engine also checks for errors other than schema violations. The paid version of ADAPA allows you to debug any score mismatches that are found during the score matching test which in this case is used to make sure the model behaves as expected. For more, please take a look here.
Comment by Ross Gayler on February 17, 2009 at 1:05pm
I think the last part of Ramesh's question could be construed as asking whether there is a PMML-aware editor that could be used to create/modify PMML files. This could be useful for: Creating PMML data transforms; Learning more about PMML; Debugging a PMML producer/consumer; Modifying PMML transforms/models to cope with conditions that are anticipated in deployment but were not present in the training data.
Comment by Alex Guazzelli on February 17, 2009 at 10:27am
Hi Ramesh! You raised a very good point. We are working with the open source community so that they provide PMML export (and import) functionality. The latest release of KNIME is able to import and export many PMML models. Check the KNIME website at www.knime.org. Also, weka is importing certain PMML models. For a complete list of PMML consumers and producers, check here or go directly to the DMG website (www.dmg.org).
Comment by Ramesh Hariharan on February 12, 2009 at 11:36pm
I'm an enthusiastic advocate of PMML and actively promote it to my clients. However, most tools that generate PMML are expensive (except R). For those clients that do not have (or unwilling to invest in) SAS Enterprise Miner or other high-end tools, are there any alternative ways to generate the PMML files? Is there a visual way of developing a PMMLs outside of these tools?
Comment by Jim Baker on November 12, 2008 at 1:00pm
Has anyone worked out a way to encrypt PMML files to protect IP (e.g., model input variables, weights, factors etc.) when installing and deploying models at client sites?
Comment by idris on October 17, 2008 at 6:55am
Hello. I work for Fair Isaac India.

We do use PMML for Segmentation ART (Decision Trees) and Decision Tables

I have developed a tool, used to Parse PMML and to migrate Decision Trees from one Product to Other. We do no Strictly follow the schema for PMML suggested in DMG group.
Comment by Alex Guazzelli on October 10, 2008 at 3:29pm
Welcome to all new members. Please feel free to post your comments on PMML and let's start the discussion.
 

Members (89)

 
 
 

© 2014   AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC

Badges  |  Report an Issue  |  Terms of Service