Closed simecek closed 2 years ago
Currently we report an accuracy and F1. Accuracy is problematic for unbalanced datasets. F1 seems to have issues for unbalanced datasets as well (https://en.wikipedia.org/wiki/F-score) and seems to depends on the ordering of the label (https://stats.stackexchange.com/questions/76776/is-the-f-1-score-symmetric).
Currently we report an accuracy and F1. Accuracy is problematic for unbalanced datasets. F1 seems to have issues for unbalanced datasets as well (https://en.wikipedia.org/wiki/F-score) and seems to depends on the ordering of the label (https://stats.stackexchange.com/questions/76776/is-the-f-1-score-symmetric).