What is Bagging
An ensemble learning method where multiple models are trained on different subsamples of data and then their responses are combined.
Definition
Bagging helps reduce scatter and make the model more stable. Instead of one model, many similar models are trained on random samples with reversion. The answers are then averaged or chosen by majority vote. A well-known example of this approach is random forest.
Example
To classify applications, several decision trees are trained on different parts of the data, and the final answer is chosen by voting.
Why it matters
The term is important for understanding ensembles: sometimes a group of simple models works more reliably than one complex one.
How it works
The algorithm creates several training sets from the original data, trains individual models and aggregates their predictions.
Where it is used
- classification
- regression
- stable models on tabular data
Limitations
Bagging increases computational costs and does not always help if the underlying models are too simple or the data is poorly prepared.