PMML to cell error

I used 10 fold cross-validation to partition the data. When I tried to compare the model performance, I used PMML to cell which is connected to learner node, however I encountered an error “cannot continue the loop as the workflow was executed while loop being partially executed”. If I use partitioning node then it is fine. My question is: can you use the PMML to cell node if you do cross-validation?


Hi there!

Which node reports error? How did you mean to compare models? Can you maybe post a screenshot of a workflow for better understanding of connections?

Here is a workflow example where different models are trained and evaluated with cross validation and the best is chosen. Hope it will help.*CB0u_eLmzlghiZI2



Thank you. That workflow is helpful. I was trying to replicate this workflow by using cross validation rather than partitioning node. I think that would not be feasible.

I do have a question on the credit scoring workflow you posted. When it writes out the best model, it should be the model that was obtained using cross-validation right? So if I want to deploy the best model, I need to run it on the entire dataset and then get the model that I can use for possible new data. Is that correct?

Hi @Learn2019!

Glad you find it helpful. On KNIME Hub you can search for nodes, workflows and in future much more :slight_smile:

Haven’t checked in details but don’t see why wouldn’t it be feasible to use cross validation on workflow from picture.

No. Cross-validation is used for validating model and that is why in Credit scoring workflow you use Scorer node after cross-validation. Bottom branch is for model you will deploy - if it is the best at the end of course.

Here is another simpler workflow for cross-validation and corresponding topic from forum that can help.

If you will have any more questions don’t hesitate :wink:


1 Like

Thank you very much!