Red-RF: Reduced Random Forest for Big Data Using Priority Voting & Dynamic Data Reduction

Random Forests have been used as effective ensemble models for classification. We present in this paper a new type of Random Forests (RFs) called Red(uced) RF that adopts a new dynamic data reduction principle and a new voting mechanism called Priority Vote Weighting (PV) which improve accuracy, execution time and AUC values compared to Breiman´s RF. Red-RF also shows that the strength of a random forest can increase without noticeably increasing correlation between the trees. We then compare performance of Red-RF and Breiman´s RF in 8 experiments that involve classification problems with datasets of different sizes. Finally, we conduct 2 additional experiments that involve considerably big datasets with one million points in each.