The Machine Learning “Advent Calendar” Day 19: Bagging in Excel

Ensemble learning is a meta-modeling approach that combines multiple base models to improve performance. Bagging, short for Bootstrap Aggregating, is a key ensemble technique that involves creating numerous training datasets through bootstrapping, which is sampling with replacement. A base model is trained on each of these new datasets, and their final predictions are aggregated, typically by averaging. This method is demonstrated with linear regression to explain the estimation of confidence intervals and with decision trees to show how combining weak learners can result in a stronger, smoother model.