20% of the data set reserved for testing

·
  • We’ll reserve 10%, or 15 points, of the data set, with 5 picked randomly from each of the three classes. We’ll refer to this as our test set.
  • Is that enough?
  • On larger data sets, the general guideline is to use 20% of the data set for testing. Since our data set here is quite small, we limit the test set to 10%.
  • Okay.

Link:: The Little Learner

Обратные ссылки