An Introspective Comparison of Random Forest-Based Classifiers for the Analysis of Cluster-Correlated Data by Way of RF++
Figure 5
Each tree is grown on a different subject-level bootstrap set of samples (left) producing a forest (middle). New subject samples are piped down the forest and each tree casts a vote for each sample. A subject classification is computed as the class with the maximum number of votes across all samples for that subject among all trees (right). Proportions of votes for each class are also produced.