I keep getting very low f1 score even after oversampling the dataset. How did you guys combat it?
haaaa, how low is your score?
There is this tutorial video: https://www.youtube.com/watch?v=yMK2wmMBjxc&t=1608s shared in this discussion https://zindi.africa/competitions/autoinland-vehicle-insurance-claim-challenge/discussions/5851 that can serve as a reference start point.
i get a high f1 score but the score on the leaderboard is too low
This could be a sign of overfitting in your model or overlap between your training and validation data set locally. You might want to review these two aspects.