In cross validation should feature selection be applied only on the training data or also on the test set?
Asked
Active
Viewed 74 times
1
gung - Reinstate Monica
- 132,789
- 81
- 357
- 650
-
[This thread](http://stats.stackexchange.com/q/64147/28500) goes into extensive detail. Note that your setting aside of a separate training set might give poor performance unless you have thousands of samples; see the answer from Frank Harrell in that thread. – EdM Jan 04 '17 at 16:21