I get a 0.46 in my local validation, but in the leaderboard i get 0.2,
I need your opinion about this situation and if you have any idea about the correlation between local validation and leaderboard.
thank you all
Yes, it can be that bad. Keep working on your model. The deadline is near.
Feature engineering, cross-validation might help. Don't stack models, I don't think it helps.
thank you brother
One of the best strategies is to select the submission with your best leaderboard score, and the best score on your local validation.
Yes, it can be that bad. Keep working on your model. The deadline is near.
Feature engineering, cross-validation might help. Don't stack models, I don't think it helps.
thank you brother
One of the best strategies is to select the submission with your best leaderboard score, and the best score on your local validation.