Why did you use random forest and not a simpler classifier for one of the classification problems?
Sigiloso
I have found random forest to have a balanced bias/variance trade-off, and is is also more interpretable, as compared to logistic/naivebayes (didn't seem convinced).