We've been trying several techniques training over multiple months, model tuning, ensembling methods, and extensive feature engineering. Interestingly, our best results so far come from training the model using all the available features and months, even though some months are imbalanced. However, we still can't seem to break the 0.92 barrier on the leaderboard.

For those who have reached 0.93+ LB scores, would you be open to sharing any insights or strategies that have helped you get there? Any tips would be truly appreciated!