For the random forest method, what value do we use to rank
the instances? Using classification, we have votes. Using
regression, we have predicted values...
Thanks.