Mission control
Generate data, build diverse trees, and reach the target score.
⏱ 00:00
0 of 15 trees builtTarget: 82% test score
Interactive forest canvas
Click any point to inspect every tree prediction.Tree vote comparison
Sort columns by clicking their headings.A random forest becomes stronger when accurate trees also make different mistakes.
| Tree | Bootstrap records | Features used | Depth | Prediction | Confidence | Correct | OOB score | Vote weight |
|---|
Learning summary
Bootstrap samplingEach tree receives a different resampled training set. Duplicate and out-of-bag records create useful variation.
Random feature subsetsEach split considers only some features. This reduces tree correlation and prevents one feature dominating every tree.
Voting and averagingClassification uses combined votes. Regression averages tree predictions, reducing unstable errors from individual trees.