Highly skewed dataset for Random forrest

[lion137]

I’m having problems with applying rf model to some tabular data, this is histogram of target variable:

Tried to combine over with under sampling, but that lead to the lower accuracy.
Any ideas how to improve accuracy?

[vbakshi]

Not sure if it’s standard practice—but what if you tried applying log to the target variable and then using that during training?

[lion137]

Thanks, will try and let know!