Concept - Random Forest
A Concept - Bagging method for decision trees that also randomly selects features (s.350).
- Pipeline (s.353, Towards Data Science): initial dataset -> L bootstrap samples + selected features -> deep trees fitted on each bootstrap sample, considering only the selected features -> random forest = a kind of average of the trees.
- Reduces variance (s.348); a strong learner example (s.343).
- Related variant: ExtraTrees (randomized top-down split) (s.350). Slide accuracy: RandomForest Gini 0.75107, Entropy 0.75222 (s.352).
- In stacking demo (s.360) its decision regions are blocky/axis-aligned.
Related
Concept - Decision Tree | Concept - Bootstrapping | Concept - Majority Voting | Concept - Ensemble Learning
Source chunk: 03 Text Classification - Chunk 12 (s.331-360)