Early stopping of feature selection loop if certain criteria is met

Can anyone create a simple workflow that stops a feature selection loop early if say the accuracy goes over a certain percentage. I’m really struggling with this. Thanks

Hi,

yeah that’s a tricky one, as the “Feature Selection Loop Start” is a kind of recursive loop.

image

If you want to use random feature selection strategy you can create your own feature list and iterate through it. With the “variable condition loop end” it’s possible to stop the iteration.

I uploaded the Workflow here:

The Feature List creation within the meta node looks like this (I use Python, as this permutation of the column names is not that easy in KNIME):

(Apologies for having the Classic UI enabled. Looks a bit different in modern UI)

1 Like