Bootstrapping method to introduce democracy in decision making.
General steps:
- Generate multiple trees by bootstrapping the samples.
- choosing times with replacement from all available training cases.
- For each node of the tree, randomly choose variables on which t base that decision at that node.
- You only consider these nodes to do the splitting, and ignore the rest of the nodes.
- The final classifier outputs the class that is most agreed by the individual trees → voting!