My understanding is, the result should be:
1. rank auc for each dataset
2. average all the ranks to arrive at final rank.
So the rank should be ordered small -> large, right? Current ranking on the results page seems reversed.
(Not that I'm arguing for the current position, my submission is the baseline file =] )Posted by: xiayunsun @ Aug. 10, 2018, 7:45 a.m.
Thanks for noticing this, we are working on this =)
Interesting!!!!I am the best, but!!! Why auc is so low............and why someone's auc is negetive?Posted by: Lean-Y @ Aug. 10, 2018, 12:25 p.m.
We are using gini index as evaluation measure, check the scoring program!
BestPosted by: hugo.jair @ Aug. 10, 2018, 1:24 p.m.
The ranking is now in the correct order, apologies for the issue
Thanks a lot.Posted by: xiayunsun @ Aug. 10, 2018, 2:43 p.m.
Hi Hugo -- I think "Duration" column is sorted in the wrong order too? A shorter duration should imply a better model?Posted by: xiayunsun @ Aug. 14, 2018, 2:28 p.m.
You are right, thank you very much!, duration is now shown in the right order. Please note that duration is only taken into account for tie-breaking
HJPosted by: hugo.jair @ Aug. 14, 2018, 2:35 p.m.
Yeah I understand. Thanks for the fast response!Posted by: xiayunsun @ Aug. 14, 2018, 2:45 p.m.