Uses an existing decision tree to compute class labels for input vectors.
Learns a random forest for classification.