KNIME and H2O

KNIME Analytics Platform
1

Dear KNIMErs,

I've recently come across H2O, an exciting Java-based ML library under Apache licence which integratess rather nicely with R and Spark. Of course both R and Spark give me options of using it from KNIME, but one would think that there shoud be an easier way of exploiting a Java-based package from KNIME, shouldn't there? Just thinking out loud here... :-)

Cheers,
E

2

Yes, heard a lot of good things about it. The usual problem is that someone needs to drive the integration. At this point they don't have the visibility like R and/or Spark where there was enough community (or business) interest in having an integration. Of course you can always approach us and sponsor the integration (hint hint), or better yet, approach H20  and have them partner with us.

3

Bernd,

Let's see what we can do... can you reach out to me on business e-mail and schedule a call please?

Thanks
E

4

Sounds very interesting!

5

Hi Guys

Any news about this?

Tks

6

Hi Fabio,

sure, look out for the new KNIME release scheduled next week :-)

3.4.0

Cheers, Iris 

7

Cool!

Tks Iris

8

Hi,

I just installed and tried the new H2O integration for Knime. And I think I already love it, but...

...there seems to be a funny bug (or I do not understand the concept).

The H2O Predictor (Classification) can not be attached to the H2O Random Forest Learner (Classification) and tells me "Please use H2O Predictor (Regression) in order to use regression models for prediction."

And yes, the H2O Predictor (Regression) can be connected to the H2O RF Leaner (Classification) and actually does its "Regression-Job". Vice versa I do not tested. 

Any recommendations?

 

9

Hi Thomas,

thanks your post. This is really a weird one and of course shouldn't happen. As I can't reproduce locally: Any chance that you can mock a little workflow reproducing the problem and attach it to this thread (e.g. using the Data Generator) or send it to me via E-Mail (christian.dietz@knime.com)?

Thank you again and sorry for the trouble,

Christian

10

I want to implement H2O random forest for regression, is it necessary to work with hadoop or not
my source file is xls or arff how to introduce the data source and thanks; please provide me the workflow

 

thank you thomas

11

Hi Thomas,

can you start a new thread for your question? I will take care that you get a timely answer.

You can find here example workflow how the H2O Integration in KNIME is used https://www.knime.com/nodeguide/scripting/h2o-machine-learning and here https://www.youtube.com/watch?v=cex1xzq5OC8 starting around minute 11 you find a video introducing it. 

 

Best regards, Iris