0

I would like to build a do my performance evaluation on top 3 predictions (by that, I mean, top N classes that are the most probable to assign an item, within a cross-validation).

Of course, I can keep the cross-validation result, go through each and look into the probabilities of assignment, sort them and pick the top 3 ones, and then build my confusion matrix. But I wonder if the functionality already exist?

Areza
  • 5,623
  • 7
  • 48
  • 79
  • The hit ratio table has this information. – TomKraljevic Apr 18 '18 at 11:09
  • I read about it http://h2o-release.s3.amazonaws.com/h2o/rel-lambert/5/docs-website/userguide/scorehitratio.html - but to me, didn't really sound like what I am looking for. from hit ratio you can't go to a confusion matrix, neither calculate precision-recall. unless, I misunderstood. – Areza Apr 19 '18 at 20:00

0 Answers0