prediction profiling / optimization in knime?


I was wondering if there are tools or workflows for exploring and optimizing target prediction for different values of features after building a model. I’ve used other tools that can:

1- Interactively change the values of the features to see the effect on the prediction
2- Run an optimization of target prediction given the features and maybe constraints

If this is not clear, I’m pretty much looking for the capabilities shown in the short video here:

This is probably one of the last capabilities I’m trying to move over to knime from other tools

Maybe if I rephrase my question a little differently? After building a model, what do you do if you want to find the best/range of features that maximize the target for example?

Hi @tnad -

I think the closest thing we have to this interactive type of functionality is probably the Binary Classification Inspector node, but that’s not quite what you’re looking for.

I can submit your feedback as a feature request for future development though :slight_smile:


From what I read there is no equivalent to the JMP profiler in KNIME but you might take a look at these examples and the included links about model interpretability.


@ScottF Thanks. It doesn’t even have to be interactive (although it’s a nice way to present to non-computational people). An ability to run an optimization on the target given the model and descriptors with constraints would be great at giving insight into how features can manipulated to get a desired target.

@mlauber71 I will look at model interpretability workflows. Thank you.

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