Remove correlated features from training set and apply the same features to test set

Andy3

Hello all,

I just wondering how you achieve to remove pairwise correlated features from your training set (using the Remove Correlated Attributes operator) and apply the same features to your test set? If I should compare this operation to something I think about the "Apply feature set" (as exists for the features selection operator) or somewhat OHE and the Preprocessing model output. See screenshot below of the process. I have normally these two training and test preprocessing operations in two different processes.

Thanks for your help.

MartinLiebig

Hi@Andy3,

you usually don't need to do it. Keep in mind that Apply Model is ignoring additional attributes.

Best,

Martin

MartinLiebig

Hi@Andy3,

if you need to do it, you can use Data to Weights for it. Attached is an example.

BR,

Martin

<宏/ >

<参数键= value =“process_duration_for_mail30"/>






























<参数键=“k”值="10"/>

Andy3

Yeah, fair enough though I was hoping the where an operator(s) like this so I could have consistency through the various data sets (and in my mind). I leave it there.

Thanks for the help.