Multilabel classification under unbalanced class distributions

[smth]

See if this comment might help. You can use some kind of weighted sampling to rebalance the class distributions when sampling from the dataset: How to Prevent Overfitting