0

I have performed principal component analysis (PCA) in Matlab via

[U,S,V] = svd(X)

Now I have to show the principal directions of the considered principal components; either by finding a way to plot them or interpret them in terms of the features.

If I have chosen 6 of the 12 PCs, I am still not able to plot them. So how can I interpret them in terms of the features?

amoeba
  • 93,463
  • 28
  • 275
  • 317
Jamgreen
  • 363
  • 1
  • 3
  • 9
  • 1
    You will most likely have to plot each PC to an other one, such as (1,2), (1,3), (1,4) ... (2,3), (2,4) etc. How much inertia do those 6 PC gather? Ploting them is the easiest way, otherwise, you'd have to look at each coordinates for each point, and see where the maxima and the minima are (within the [-1, 1] range) – Amandine FAURILLOU Sep 29 '15 at 07:08
  • 2
    Try the functions `pca` & `biplot`. – Steve S Sep 29 '15 at 08:05
  • +1 to both comments above. Is there a particular reason you chose six instead of, say, three PCs? The "feature vectors" are the PCs. You might want to use your PC scores to cluster/visualise your sample more effectively. – usεr11852 Sep 29 '15 at 08:54
  • Also, please note that we have a lot of useful questions and answers about PCA here. Your questions suggest that you did not really explore the existing answers. In this case, looking at threads tagged with [pca] and [data-visualization] might be helpful for you here. For example, http://stats.stackexchange.com/questions/7860. – amoeba Sep 29 '15 at 08:59

0 Answers0