Thanks for the background. Agreed with your approach, however I think that being able to
go from PMML->{Guvnor,spreadsheet} should also be a target feature. Although direct
PMML->DRL is useful for deployment, it's rare that customers will deploy a model
without at least having a human take a look. As well they can run tests within Guvnor or a
spreadsheet to see if their assumptions about the model are accurate by testing with
historical data (coming more from the operational side of things).
In the present situation, spreadsheets are already in use, so we can gently nudge them
toward a direct deployment of PMML by using spreadsheets to raise the comfort level. So,
while taking the long way 'round, what we need to do here is something like SAS Model
-> Spreadsheet -> PMML -> DRL. I need to specify a spreadsheet format that we can
then get SAS to output. Once everyone realises and is comfortable with this, we can point
out that SAS can deliver the model in PMML for a direct import.
I briefly saw a message from Michael pointing out the location of a spreadsheet that
I'll investigate this morning. What would be very useful at this stage is a document
describing how the XLS -> PMML spreadsheet is specified, i.e. what specific, repeatable
format can I produce to start the processing pipeline to deploy scorecards in the engine.
Regards,
- Steve
-----Original Message-----
From: rules-users-bounces(a)lists.jboss.org [mailto:rules-users-bounces@lists.jboss.org] On
Behalf Of Davide Sottara
Sent: Friday, 11 January 2013 4:52
To: rules-users(a)lists.jboss.org
Subject: Re: [rules-users] Location of ScoreCard Spreadsheet File?
Let me tell you the story of scorecards, to help you understand what is going on.
Scorecards are one of the possible predictive models covered by the PMML standard - which
also supports other models such as neural networks and decision trees. However, PMML is an
XML interchange format rather than a user-friendly authoring format.
So, we decided to adopt a number of stages for representation and
evaluation:
- XLS (or a guvnor-based UI) for authoring
- PMML for internal representation (or to import scorecards generated with other tools)
- DRL for execution
So, XLS is converted in PMML; the UI, instead, generates PMML directly. PMML is always
converted in DRL.
This, at least, in the ideal world :)
Scorecards were ready for release way before PMML will be - being much broader, PMML
requires much more effort. However, being strategical, scorecards could not wait to be
released together with the rest of PMML. So, scorecards were "split" and
released with the main
product: drools-scorecards
is part of drools. drools-pmml, instead, is still being incubated in drools-chance, which
is NOT part of the official release. What you have found in drools-chance is an example,
once originated in an XLS but further modified manually for testing purpose, so there is
no "source"
XLS anymore.
Again, notice that the pmml support in drools-chance is more powerful (and lets you import
PMML directly among other things), but is not currently integrated with the BRMS and the
spreasheet sources. The two will be merged together again in 6.0, when pmml will become
part of the official release too.
Sorry for the confusion generated by this situation, but feel free to ask for more
clarifications or help
Best regards
Davide
--
View this message in context:
http://drools.46999.n3.nabble.com/rules-users-Location-of-ScoreCard-Sprea...
Sent from the Drools: User forum mailing list archive at
Nabble.com.
_______________________________________________
rules-users mailing list
rules-users(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/rules-users