Figure 4.
A comprehensive analysis of the TreeBagger training. (a) The OOB error while applying different numbers of grown trees. (b) The importance estimates of each predictor, σ indicates the standard deviation of the profile, and pn indicates the n elements of the profile. The confusion matrices of the train and test sets are demonstrated in (c) and (d), respectively. The order of the predicted class in (c) and (d) is (1) Decline, (2) Flat, (3) Ascend, and (4) Irregular.