I'm trying to include some of the dimension reduction technique, which already exist into a workflow framework. As far as I understood PCA, the PCA-compute & -apply nodes create vectors for all number-variables. After that I well replace all my number-variables by those vectors and can continue using other model learners and predictors.
But this means: I will have continue to working on a "unreadable" data set. Would it be possible, to somehow reduce the original data set (dimensions) on basis of the vectors calculated by PCA? (e.g. like the Correlation Filter?)
Thanks you in advance