|
|
Ensemble Methods
| Description: |
Basic idea:
- Build many models (e.g. trees)
- When making prediction, use a voting mechanism (e.g. majority)
These methods have had the most profound impact on prediction accuracy of any method in the last decade.
Ways to form ensembles:
- Bagging (and it’s variants)
- Boosting (and it’s variants)
|
| Publications: |
Schapire/90a: The Strength of Weak Learnability
|
|
|
