PCA: Which principal components capture the main feature?

KNIME Hub
#1

Hey all,

i have read in the book (Data Mining and Knowledge Discovery for Process Monitoring and Control [Wang 1999-09-15]) that the first principal components can capture the main feature.
But not in every cases.
Is it possible to present the result as in this excerpt, to be sure which components have the biggerst part?

image
image530×567 67.3 KB

Thank you for your answer.

Have a good christmas time.

Johannes

#2

Hi @Koalabar -

If you take the second output port of the PCA Compute node (the spectral decomposition) and transpose it, you get something pretty similar to the table you’ve posted above.

#3

Hey Scott,

thank you for your answer.
Is it as well possible to get a percent value ?
And gives any opportunities to do this with the simple PCA Node?

Thank you Scott

#4

I think at the moment, the answer is no to both questions.

If you want to replicate the table above exactly, you could do it with an R Snippet and a few lines of code.