Hands-On Recommendation Systems with Python

Bagging and random forests

Bagging is short for bootstrap aggregating. Like most other ensemble methods, it averages over a large number of base classification models and averages their results to deliver its final prediction.

These are the steps involved in building a bagging model:

1. A certain percentage of the data points are sampled (say 10%). The Sampling is done with replacement. In other words, a particular data point can appear in multiple iterations.
2. A baseline classification model (typically a decision tree) is trained on this sampled data.
3. This process is repeated until n number of models are trained. The final prediction delivered by the bagging model is the average of all the predictions of all the base models.

An improvement ...