I have 4 predictive features and 4800 observations. I did PCA fit like this:

pca = PCA(n_components=X_train_scaled.shape[1],whiten=True)

and I got

enter image description here

I have a few question regarding this:

Does this mean that the first feature, which is age in this case, is the most important? I know that this is the variance of principle components, but why are there 4 PCs just like the features? what can I infer from this plot about my predictive features?

In other words, how can this help me in the down-stream analysis in regards to choosing maybe the best features from the 4?

  • $\begingroup$ Yes, Thanks! but I still have this one question: so the PCs are the new features, instead of the original 4. Each PC is a linear combination of features, right? so how can I find which features are in PC1, and PC2 and so on... ? in other words, why is the variance in PC1 for example, higher than the variance of PC2 ? is PC1 a result of different linear combination? If so, which features participated in PC1? $\endgroup$
    – CORy
    Jan 2 at 21:14
  • 1
    $\begingroup$ You just need to read more about PCA. All of that (& more!) is covered in various places on the site. Click on the pca tag, sort by votes, & start reading some of the most influential threads. You might also try the wikipedia page for PCA, & Python's help pages for your functions. $\endgroup$ Jan 2 at 21:17