I can found two. Feature ranking is computed to

  • reduce the dimensionality, when the number of features is prohibitively large for a predictive model to be applied. In this case, we actually compute feature ranking to perform feature selection.
  • be able to focus in later experiments (which may be performed in some biological laboratory and/or may be very expensive/time consuming) only on some candidate features and save some money/time.

Are there any other motives?

  • $\begingroup$ I asked a very similar question here: stats.stackexchange.com/questions/202277/… $\endgroup$ Mar 14 '17 at 23:27
  • $\begingroup$ @MatthewDrury Yes, you did indeed. However, after reading the answers, my question "Are there any other motives" remains unaswered ... $\endgroup$
    – Antoine
    Mar 15 '17 at 13:30
  • $\begingroup$ Agreed, I wasn't claiming it a duplicate, just thought it would be useful to you. $\endgroup$ Mar 15 '17 at 13:39
  • $\begingroup$ @MatthewDrury Of course. Thank you for pointing it out, it was usefull, since i) now it seems that the answer to my question is probably not, and ii) it seems that there really is no strict definition of the relevance. (That would be my next question, but first we need to know, why does one want a feature ranking). $\endgroup$
    – Antoine
    Mar 15 '17 at 13:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.