Concept - Majority Voting

Way of combining ensemble outputs (s.352):

• Equal: same weight for all models.
• Weighted: the best model gets more weight in a vote.

Slide example (public accuracy score): ExtraTrees Gini (Best) 0.75571 is the best single model; Voting Ensemble (Democracy) 0.75337 is worse; Voting Ensemble (3*Best vs. Rest) 0.75667 is best overall. Other rows: GradientBoostingMachine 0.65057, RandomForest Gini 0.75107, RandomForest Entropy 0.75222, ExtraTrees Entropy 0.75524.

Related

Concept - Bagging | Concept - Ensemble Learning | Concept - Stacking (learned combiner instead of a fixed voting rule)

Source chunk: 03 Text Classification - Chunk 12 (s.331-360)