Fig. 4.

Schematic representation of random forest algorithm. The three major steps in the random forest algorithm are bootstrapping, bagging, and aggregation. During bootstrapping, the training dataset is resampled into several small datasets, which are then bagged for the decision tree. The size of the bagged dataset remains the same but bootstrapped decision trees are different from each other. All decision trees make predictions on test data, and in the aggregation step, all predictions are combined for the final prediction. For a classification problem, the final prediction is made by major voting, but for a regression problem, the final prediction uses the mean or median value.