Assume you have a training dataset consisting of 1 million observations. Suppose fitting the model with p features on this data takes 10^-5p seconds. Suppose you want to choose the value of p by searching over 20 possible features. If you only want to spend about 20 minutes to decide p, what value of k should you use for a k-fold cross-validation (assume that the computation time grows linearly with respect to k)?

A

6

B

12

C

15

D

20