Research map / Learning theory and algorithms / Prediction and decision losses
To Tune or Not to Tune the Number of Trees in Random Forest
Philipp Probst, Anne-Laure Boulesteix
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| increase number of trees from 11 to 2000 | random forest · OpenML binary classification (149 tasks) | OOB error rate: -0.0324 | Improved |
| increase number of trees from 11 to 2000 | random forest · OpenML binary classification (149 tasks) | OOB Brier score: -0.0683 | Improved |
| increase number of trees from 11 to 2000 | random forest · OpenML binary classification (149 tasks) | OOB AUC: 0.0553 | Improved |
| increase number of trees from 11 to 2000 | random forest · OpenML multiclass classification (44 tasks) | OOB error rate: -0.0753 | Improved |
| increase number of trees from 11 to 2000 | random forest · OpenML regression (113 tasks) | OOB R2: 0.1249 | Improved |
| increase number of trees beyond the minimum (error-rate curve) | random forest · OpenML binary classification (149 tasks) | OOB error rate: 15 of 149 curves rise again | Mixed |
| increase number of trees (Brier score and log loss curves) | random forest · OpenML binary classification (149 tasks) | Brier score and log loss | Improved |
| increase number of trees (median-based regression measures) | random forest · OpenML regression (113 tasks) | median squared and absolute error | Mixed |
| increase number of trees (mean squared error) | random forest · OpenML regression (113 tasks) | mean squared error | Improved |
| increase number of trees from 250 to 500 | random forest · OpenML datasets (small) | OOB performance | No clear effect |
| increase number of trees to convergence on non-monotonous error curves | random forest · OpenML datasets with non-monotonous error rate curve | OOB error rate: less than 0.01 worse than minimum | Hurt |
Closest papers by meaning
- When do random forests fail? (NeurIPS 2018 · similarity 0.88)
- Towards Convergence Rate Analysis of Random Forests for Classification (NeurIPS 2020 · similarity 0.87)
- Randomization as Regularization: A Degrees of Freedom Explanation for Random Forest Success (JMLR 2020 · similarity 0.86)
- On Computing Optimal Tree Ensembles (ICML 2023 · similarity 0.86)
- Random Forests, Decision Trees, and Categorical Predictors: The "Absent Levels" Problem (JMLR 2018 · similarity 0.86)
- Minimax Rates and Spectral Distillation for Tree Ensembles (NeurIPS 2026 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
