R help - Jun 2009 - Question on implementing RandomForest

Hi - I am new to RandomForest and want to be able to predict for a set of values
outside of R.  So for example, if I create the model in R and want to get the
prediction for an observation in a database, is there a way I can get the
algorithm for the trees and final prediction for this model printed out in R
that I can directly implement in SQL?  Or is the only way to do this is to use a
wrapper and call R through SQL and predict the value?

Thanks
Suresh

	[[alternative HTML version deleted]]

Nallan Suresh wrote:
> Hi - I am new to RandomForest and want to be able to predict for a set of
values outside of R.  So for example, if I create the model in R and want to get
the prediction for an observation in a database, is there a way I can get the
algorithm for the trees and final prediction for this model printed out in R
that I can directly implement in SQL?  Or is the only way to do this is to use a
wrapper and call R through SQL and predict the value?

I think getting rules out of R to be interpreted by another software is 
at least not easy. My way would be to SQL through R in order to import 
data to R.

Uwe Ligges


> Thanks
> Suresh
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.

To get the prediction in the database see
help("predict.randomForest").
This will allow you to score the existing data: you then export the 
scores to the database using "RODBC" or similar package.  This is what
most people mean when the say they want to get the model into the database.

If you really want the model out so you can execute the model in the 
database (on new data not previously known to R) then either

1. Use the "pmml" package wich does have support for
"randomForest" and
consult your database for how to execute PMML models (also see 
http://www.dmg.org/ for more on PMML), or

[1.b. If your database is not supporting PMML directly, consider 
something like Zementis ADAPA http://www.zementis.com/ or another 
similar product, or]

2. Upgrade (!) to PostgreSQL, install the PL/R extension 
(http://www.joeconway.com/plr/), and run R directly in your database as 
a stored procedure or trigger language.  (Neat.)

Allan

Nallan Suresh wrote:
> Hi - I am new to RandomForest and want to be able to predict for a set of
values outside of R.  So for example, if I create the model in R and want to get
the prediction for an observation in a database, is there a way I can get the
algorithm for the trees and final prediction for this model printed out in R
that I can directly implement in SQL?  Or is the only way to do this is to use a
wrapper and call R through SQL and predict the value?
>
> Thanks
> Suresh
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>