Is there a way to retrieve the groups of features that, jointly, show a high loading in each LV. I'm aware that I can retrieve this by digging into
x_loadings_, but given that covariance within and between groups is maximized, I'm wondering if there is a way to retrieve the combinations of features rather than individual features?
TL;DR: How to get strongly correlated features? Can we define these by their fitting coefficients?
This is how I'm performing the PLS-DA
from sklearn.cross_decomposition import PLSRegression plsr=PLSRegression(n_components=2, scale=True) plsr_fit=plsr.fit(input_data, y) #x scores plsr_fit.x_scores_ #x loadings plrs_fit.x_loadings_ #x weights plrs_fit.x_weights_