PMML support for universal deployment needed...

[Atabarezz]

Idea:

Let's add the universal PMML predictive markup language to rapidly deploy our predictive models. 

Rationale:

Most of the predictive models created are essentially built to be deployed into well known enterprise level decisioning tools.

These decisioning tools are working at banks, non bank financial institutions, credit card issuers, insurers on a daily basis.

If Alteryx want's to replace SAS, SPSS and similar competitor this PMML addition seems to be a must.

They all have a common markup language and even with the newer version we can deploy almost all data pre-processing steps too...

Current status in Alteryx as I'm aware of:

Only case where Alteryx and PMML comes close is the following link; http://www.alteryx.com/topic/pmml promising including PMML in the upcoming releases in 2012. And then no word of it. The writer is retired already and no word since 2012 (3 yrs is a long time span). 

What to do:

On easy way is already done for R, this is a Java library for converting R models to PMML https://github.com/jpmml/jpmml-converter

Clients:

FICO is one of the well known credit bureaus worldwide and also the vendor of decisioning software. They also have an Analytics Modeler Suite. All FICO systems are PMML compliant. Chek it out: http://www.decisionmanagementsolutions.com/2014/10/17/first-look-fico-analytic-modeler-suite-jtonedm...

Also Experian is one of the other big software vendors in the Analytics space with decision management tools and they are actually a partner of Alteryx too... Their well known widely implemented Powercurve tool has PMML language import too... Check it out: http://jtonedm.com/2014/05/13/first-look-experian-powercurve-update/

IBM ODM is a nother vendor for decisioning tools in the enterprise level. They also have PMML integration, look; http://www.prolifics.com/blog/odm-integration-spss-predictive-analysis-suite-part-1-pmml-import

Competitor status:

SAS supports PMML export from SAS Enterprise Miner

IBM SPSS products (Modeler and Statistics) offer a great deal of support for PMML

KNIME is one of the leaders in processing PMML documents 

RapidMiner has a PMML extension

MicroStrategy strongly supports the standard

[chris_love]

Is there a standard on PMML yet? I know in the past each vendor had their own implementation making it more or less useless.

[Atabarezz]

Hi Chris,

Actually PMML markup is maintained by data mining group as you may be aware and with every release there are huge developments going on. So it may not be stable but it is definately essential. Most up to date version is 4.2.1.

I believe vendor based implementation like somebody integrating PMML version 3 in their enterprise wide decisioning tool

while a predictive modeling advanced analytics app like KNIME updating to version 4.0.1 creates obvious mismatches.

Here are some developments since PMML 4.0 

• Improved Pre-Processing Capabilities
• Boolean operations and
• If-Then-Else function
• Time Series Models
• Model Explanation: Saving of evaluation and model performance measures to the PMML file itself.
• Multiple Models: Capabilities for  ensembles
• Extensions of multi-class classification for Support Vector Machines,
• improved representation for Association Rules
• Addition of Cox Regression Models
• New model elements for representing Scorecards,
• k-Nearest Neighbors (KNN) and Baseline Models.
• A new attribute that identifies for each model element if the model is ready or not for production deployment. 
• Enhanced post-processing capabilities (via the Output element) 
• Transformations: New elements for implementing text mining 
• New built-in functions for implementing regular expressions: matches, concat, and replace

Here is a nice recent article: http://www.datasciencecentral.com/profiles/blogs/pmml-a-good-citizen-of-the-analytics-ecosystem

[Atabarezz]

DrDan posted recently a nice news here, saying;

"Alteryx will be launching a new macro in the Public Gallery called the Output Model tool that will enable a predictive model created in Alteryx to be written to disk in PMML format... Models created with the Linear, Logistic, Count, and Gamma Regression, Decision Trees, Random Forest, Naive Bayes, Neural Network and Support Vector Machine will be supported. An announcement will be made on the EngineWorks blog when it is released."

FYI

[JohnJPS]

At Inspire, someone asked about this, and it was mentioned they are working more towards having broad support for PFA, and not so much PMML.  The vision is that PFA will unseat PMML anyway.  Not sure when it will all happen, but that's the word on the street.  http://dmg.org/pfa/docs/motivation/

For the time being, you might be able to use this R package to import some PMML.

[Atabarezz]

Looking forward for the Outlook Model Tool... Client's ask for a method to export the workflow if possible...

[mceleavey]

Hi All,

Does anyone know if there's been any movement on this?

I have a client request to output PMML from SAS for consumption in Alteryx, and vice versa. From what I can gather, PMML presents as XML, does this mean I can simply read in the file as native XML?

I've looked at Dr. Dan's Output Model, and this seems to package the predictive model as XML for output.

I may be completely off on this, but if anyone can help it would be greatly appreciated!

[Atabarezz]

"Model Output: A tool that allows a predictive model created in Alteryx to be written to a file in either R's native binary file format or PMML, allowing the model to be used in an external scoring application."

Check out the gallery; https://gallery.alteryx.com/#!app/Output-Model/56bbd338aa690a0e9ca53c65

[Atabarezz]

It's been a while since the PMML discussion.

You may check PMML version 4.3 from here by the way... http://dmg.org/pmml/v4-3/GeneralStructure.html

Changes since version 4.2.1 are here http://dmg.org/pmml/v4-3/Changes.html

[Community_Admin]

The status of this idea has been changed to 'Inactive'. This status indicates that:

1. The idea has not had activity in the form of likes or comments in over a year.

2. The idea has not reached ten likes.

3. The idea is still in the 'New Idea' status. 

However, this doesn't mean your idea won't be implemented! The Community can still like and comment on this idea. With enough renewed interest, this idea can be brought back into the 'New Idea' status. 

Thank you for contributing to the Alteryx Community and the Alteryx Product Idea Boards!