Introduction
A random forest is an ensemble of decision trees trained with bagging — plus one extra twist that makes the trees even more different from one another. It is one of the most useful algorithms in all of machine learning:
- works well out of the box, with little tuning;
- handles classification and regression;
- copes with many features, mixed feature types and outliers;
- needs no feature scaling;
- tells you which features matter most.
If someone asks "what should I try first on a table of data?", a random forest (or gradient boosting, topic 3.5) is a very good answer.