I competed an H2O Random Forest Regression and the table on the left is the internal ranked importance measures. I also attempted a Permutation and Target Shuffling workflow with the results in the right table. (I found the permutation/target shuffling workflow in the KNIME forum and attempted a customization for my specific purposes.) (Permutation Feature Importance for Linear Regression )
I believe I must have done something wrong with the permutation workflow as the results appear to be nearly the exact opposite of the RFR results (tables above).
Would someone please look at my permutation target reshuffling workflow and tell me if you see where I may have screwed up.
Thank you
Feature Importance.knwf (194.8 KB)
Sold Prices by ListingID.xlsx (60.5 KB)
temp for shap testing.xlsx (1016.9 KB)