ML train/test with predetermined splits

Hi Everyone,

I was looking at the x-partition and x-agg nodes for machine learning. They look really slick to neatly test the performance of the model over a number of iterations.

I have another model which is outside of KNIME and I want to train/test and compare that model with the same train/validate/test sets that were used in that model. Is there a way to use x-part and x-agg or another strategy to do this?

I have 10 sets of train/val/test that I would like to use.

Thanks in advance!
Jason

If you have the training and validation sets already as separate files, then I imagine you can just import them with the relevant reader node and pipe them directly to the model builder nodes.

Did you have something else in mind?

1 Like

@elsamuel Thanks for the reply! I don’t have a lot of experience yet in this area so I thought there was something about the X-nodes which are required to do the model scoring/ROC curve analysis afterwards. If I can pull off the same thing with feeding in the 10 sets and using regular loops that would be great. Can you point me to any examples? If not I can hack away at it :slight_smile:

Thanks,
Jason

Hello @j_ochoada,

you can always go to KNIME Hub and look for examples. Here are search results if you look for cross validation:

Br,
Ivan

This topic was automatically closed 182 days after the last reply. New replies are no longer allowed.