Random forest
Bosques aleatorios (random forest)
- Date
- 2001 · confidence high
- Place
- University of California, Berkeley, California, United States
- Credited to
- Leo Breiman
- Field
- Knowledge & symbols
What was discovered
Breiman combined bagging (bootstrap sampling) with random feature selection at each node to build ensembles of decision trees that are weakly correlated with each other, whose combined vote greatly improves accuracy and reduces overfitting compared with a single tree.
Tin Kam Ho had created the first 'random decision forest' algorithm in 1995 using the random subspace method; Breiman formalised the modern version combining bagging and random feature selection in 2001, and it is the one that gave the technique its popular name.
What it led to
- One of the most widely used machine learning algorithms for classification and regression
- Variable-importance methods for black-box models
Built on
- Bootstrap aggregating1994-1996
- Decision trees
- Bagging (bootstrap aggregating), by Breiman himself
- Tin Kam Ho's random subspace method (1995)