Top Page | Upper Page | Contents | About This Site | JAPANESE

Ensemble Learning

Almost all mthods of Multi-Variable Analysis look for the models that match all data. This approach is good because it is simple. But it is bad that it has problems oversights of important feature of the data and Over Fitting .

Ensemble Learning uses the approach that is divide the data into some groups at first. Total output is made from partial outputs.

This is similar to Stratified Sampling . I often use Ensemble Learning as "automatic stratified sampling".

Random Forest uses Ensemble Learning for Decision Tree.

NEXT Batch learning and Online learning