// interactive walkthrough
How a Random Forest trains — tree by tree
Bagging + feature subsampling + averaging — the three tricks that turn high-variance decision trees into a stable ensemble. Watch each tree get a different bootstrap sample, grow on a different feature subset, and contribute one vote to the final prediction. OOB error tracks held-out accuracy without a separate validation split.
settings changed — forest has been reset to apply.
01
Task & impurity criterion
// per-tree split rule
task
criterion
02
Hyperparameters
// drag to tune · changes reset the forest
bootstrap
03
Training controls
// build the forest tree by tree
data & ensemble prediction
selected tree
click ▶ Next tree to start growing the forest.
internal
leaf
04
Bag & feature subset for the selected tree
// what randomness this tree seesBuild a tree to see its bootstrap sample and feature subset here.
05
Forest gallery
// click a tree to inspect it
OOB error vs n_trees
OOB error
train error
ensemble vs actuals
06
Feature importance
// total weighted impurity decrease attributed to each featureNo trees yet.
07
Forest history
// one row per treeNo trees yet.