feature importance for prediction

hi dear knime form user

ı have a questions. ı want to prediction my dataset. One or some of the columns causes the prediction algorithm to memorize. For this reason, I need to check my columns’ Feature Importance values for MLR

how can I do that

Hi @BerkayAkar -

I’m afraid I don’t fully understand your question. Are you asking about the mlr package available in R? Or about other options for evaluating feature importance in KNIME generally?

hi @ScottF

ı want to make a sales price prediction from some costumer or item features. the machine cares more about some columns when estimating. so I want to look at the feature importance values and ı want to remove some columbs because the values it cares about are wrong values.

Fair enough - but is this about the mlr package specifically? Or are you content to use other algorithms native to KNIME? In the latter case I can try to help, depending on which algorithm makes sense for you.

(If you must use mlr, detailed help on that is beyond the scope of what we can assist with on this forum.)

Since it sounds like you have a use case already, maybe you could post your workflow in progress?

If it is similar to python you could call feature importance directly on certain ml predictors (e.g Trees, Forests)
But that’s not KNIME specific and rather relates to the package

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