Single feature elimination using loop

Hi all,

 

I reviewed the Feature Elimination node in Knime, but my dataset contains over 2000 features and over 2000 instances, so such a node would be impossible for me to use.

 

However, I have managed to narrow my dataset down to about 50 or so "top features." Is there a way for me to design a loop that eliminates one feature at a time and then displays the accuracy/statistics for a classifier run on that dataset?

 

For example, if I have features labled "A", "B", "C",..."X", "Y", "Z", "Class," my code would first eliminate "A", from the dataset, run the classifier predicting "Class," generate accuracy rates, and then reinsert "A" and eliminate "B", run the classifier again predicting "Class", etc.

Hi Kashish,

 

so you basically want to run over your data and in each run you want to leave one feature out.

 

So, let's do it the opposite way :-) Please see the attached workflow. After the reference column filter you get in each run one feature left out to process further.

Iris