Bagging

An Ensemble Method.

Analogy

Diagnosis based on multiple doctors majority vote.

In each iteration some samples are drawn with replacement from the dataset (Bootstap), on which a classifier model is trained.

Each classifier will predict for a given sample and the so called bagged classifier will count the votes and assign the class with the most votes to the sample.

This method can also be applied to Regression problems by averaging the predictions from each model.

Advantages

• often better than single models derived from the whole dataset
• more robust against noise
• improved Accuracy in prediction