The DOI matches its Crossref record; 124 of the 124 references it cites were found in a registry or an index; 4 of the 112 sources its claims cite were checked for support.
A random forest guided tour
Open accessAbstract
The random forest algorithm, proposed by L. Breiman in 2001, has been extremely successful as a general-purpose classification and regression method. The approach, which combines several randomized decision trees and aggregates their predictions by averaging, has shown excellent performance in settings where the number of variables is much larger than the number of observations. Moreover, it is versatile enough to be applied to large-scale problems, is easily adapted to various ad hoc learning tasks, and returns measures of variable importance. The present article reviews the most recent theoretical and methodological developments for random forests. Emphasis is placed on the mathematical forces driving the algorithm, with special attention given to the selection of parameters, the resampling mechanism, and variable importance measures. This review is intended to provide non-experts easy access to the main ideas.
Overview
CoChat Summary
## TL;DR
This paper is a survey of random forests that reviews the algorithm’s construction, theoretical analyses, variable importance measures, and several extensions. It emphasizes that random forests are highly successful in practice, but that much of their behavior still lacks complete mathematical explanation.
## Key findings
- Breiman’s original random forest builds M randomized trees, each from aₙ sampled observations, using mtry randomly chosen split directions at each node and stopping when a node has fewer than nodesize points; the default implementation cited is M = 500, mtry = ⌈p/3⌉ for regression, aₙ = n, and nodesize = 5 for regression.
- The paper states that the variance of the forest decreases as M grows, and that choosing a large number of trees does not lead to overfitting; it also notes that the error of the finite forest can be made arbitrarily close to that of the infinite forest.
- For simplified forest models, the review reports consistency results such as centered forests being consistent when k → ∞ and n / 2^k → ∞, and states that for sparse regression the convergence rate depends on the number of active variables |S| rather than the ambient dimension p.
- For variable importance, the paper describes two measures: MDI and MDA; it reports that correlated variables can harm MDA performance, and in a Gaussian setting the variable importance decreases as the inverse of the square of p when the number of correlated variables increases.
## Methods
- This is a narrative theoretical review, not an empirical study; it surveys the random forest literature and focuses on the mathematical structure of the method, parameter tuning, resampling, split selection, variable importance, and extensions.
- The paper’s “data” are the cited results from prior theoretical and methodological work on random forests, including Breiman’s original algorithm, simplified forest models, bootstrap/subsampling analyses, and variable-importance theory.
- The main analysis is synthesis and comparison of existing results, including descriptions of algorithmic steps, consistency and asymptotic normality results, kernel/nearest-neighbor connections, and known behavior of importance measures.
## Limitations
- The authors state that little has been proven mathematically for the original Breiman procedure, and that the role of the bootstrap in random forests is still poorly understood.
- They explicitly note that the consistency of the unpruned procedure comes at the price of a conjecture regarding CART behavior that is difficult to verify.
- Evident from the text: this is a review article, so it does not present a new dataset or new experimental evaluation of random forests itself.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
4 Valid · 4 Not valid · 83 Not checked · 47 Own results and interpretations
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“He also establishes that the infinitesimal jackknife (Efron, 1979) consistently estimates the forest variance.”
What the sources say
- Not checked yetcites Wager (2014)
- Supportedcites Efron (1979) source
The citation to Efron (1979) is attached to “infinitesimal jackknife,” identifying the method, not attributing Wager’s forest-variance result to Efron. Section 5 (p. 11) explicitly discusses Jaeckel’s infinitesimal jackknife. (LLM, conf=0.92)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
Every source the paper cites for this claim was checked and supports it.
In the paper
“Assuming some regularity conditions on the regression model, and considering a modification of Breiman's forests in which all directions are preselected for splitting, Scornet et al. (2015) prove that, with probability 1 -ξ , for all n large enough and all 1 ≤ q ≤ k ,”
What the sources say
- Supportedcites Scornet et al. (2015) source
Split into 2 atoms — 2 confirmed, 0 partially supported, 0 disputed, 0 misattributed, 0 unverifiable. Expand the claim to see per-atom verdicts.
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Scornet et al. (2015) proved a consistency result in the context of additive regression models for the pruned version of Breiman's forest.”
What the sources say
- Partly supportedcites Scornet et al. (2015) sourceExperimental — accuracy not yet measured
Our copy confirms the core attribution: the Abstract and Section 3 (assumption H1) state an L2 consistency result for Breiman's (2001) random forest in the context of additive regression models. However, the paper never describes its result as holding for a 'pruned version': Section 2/Section 3 set out two regimes — Theorem 1, where trees are 'not fully developed' because growth stops at t_n leaves (t_n < a_n), and Theorem 2, where trees are 'fully grown' (t_n = a_n, Breiman's default) and consistency comes from the subsampling rate — and the paper presents both as results for Breiman's original procedure (bootstrapping replaced by subsampling), not for a pruned variant. The 'pruned version' qualifier is therefore not supported by our copy. (LLM, conf=0.72)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
Every source the paper cites for this claim was checked and supports it.
In the paper
“It is shown in Scornet et al. (2015) that random forests asymptotically perform, with high probability, splits along the S informative variables (in the sense of Section 3.1).”
What the sources say
- Supportedcites Scornet et al. (2015) source
Section 3 (Main results) of our copy states verbatim that 'Proposition 1 below shows that random forests nicely adapt to the sparsity setting by asymptotically performing, with high probability, splits along the S informative variables,' and Proposition 1 formalizes this (with probability 1−ξ, for n large enough, the first k cut directions j_{q,n}(X) lie in {1,...,S}). The same section defines the S informative variables as the first S components of the additive model Y = Σ_{j≤S} m_j(X^{(j)}) + ε under the sparsity constraint S < p, matching the claim's qualifier. The claim omits the proposition's conditions (m_try = p, (H1), and no m_j constant on an interval), but that level of specificity is not carried by the citing sentence. (LLM, conf=0.92)
boostedstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Rests on an assumption not shown here.
- Supported
In the paper
“The resulting Dynamic Random Forest (DRF) shows significant improvement in terms of accuracy on 20 real-based data sets compared to the standard, static, algorithm.”
What the sources say
- Not checked yetcites Bernard et al. (2012)
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
51 more claims
Create a free account to see all 138 claims and to verify the 108 references still open.
- Probably not a substantial claim. 63
- It may state two separate claims as one. 40
- The quote only partly states this claim. 1
Citations
Cites
Breiman, L. (2001) · Machine Learning · doi:10.1023/a:1010933404324 · Cited by 133,538
The Elements of Statistical Learning
Hastie, T., Tibshirani, R., & Friedman, J. (2009) · Springer Series in Statistics · doi:10.1007/978-0-387-84858-7 · Cited by 25,970
The Elements of Statistical Learning
Hastie, T., Friedman, J., & Tibshirani, R. (2001) · Springer Series in Statistics · doi:10.1007/978-0-387-21606-5 · Cited by 24,162
Classification and Regression Trees.
Gordon, A. D., Breiman, L., Friedman, J. H., Olshen, R. A., & Stone, C. J. (1984) · Biometrics · doi:10.2307/2530946 · Cited by 23,522
Classification and Regression Trees.
Ryzin, J. V., Breiman, L., Friedman, J. H., Olshen, R. A., & Stone, C. J. (1986) · Journal of the American Statistical Association · doi:10.2307/2288003 · Cited by 20,825
Classification and Regression by randomForest
Liaw, A. I., & Wiener, M. C. (2007) · Cited by 18,594
MapReduce: simplified data processing on large clusters
Dean, J., & Ghemawat, S. (2008) · Communications of the ACM · doi:10.1145/1327452.1327492 · Cited by 18,476
Bootstrap methods: another look at the jackknife
Efron, B. (1979) · The Annals of Statistics · doi:10.1214/aos/1176344552 · Cited by 17,602
Breiman, L. (1996) · Machine Learning · doi:10.1007/bf00058655 · Cited by 16,325
Gene selection for cancer classification using support vector machines
Guyon, I., Weston, J., Barnhill, S., & Vapnik, V. (2002) · Machine Learning · doi:10.1023/a:1012487302797 · Cited by 9,954
Research works printed but not linked: 3
- Cutler A, Zhao G (2001) PERT—perfect random tree ensembles. Comput Sci Stat 33:490–497
- Howard J, Bowles M (2012) The two most important algorithms in predictive modeling today. In: Strata Conference: Santa Clara. http://strataconf.com/strata2012/public/schedule/detail/22658
- Wager S, Hastie T, Efron B (2014) Confidence intervals for random forests: the jackknife and the infinitesimal jackknife. J Mach Learn Res 15:1625–1651
Cited by
Athey, S., Tibshirani, J., & Wager, S. (2019) · The Annals of Statistics · doi:10.1214/18-aos1709 · Cited by 1,870
Hyperparameters and tuning strategies for random forest
Probst, P., Wright, M. N., & Boulesteix, A. (2019) · WIREs Data Mining and Knowledge Discovery · doi:10.1002/widm.1301 · Cited by 1,640
A systematic literature review of machine learning methods applied to predictive maintenance
Carvalho, T. P., Soares, F. A. A. M. N., Vita, R., Francisco, R. D. P., Basto, J. P., & Alcalá, S. G. S. (2019) · Computers & Industrial Engineering · doi:10.1016/j.cie.2019.106024 · Cited by 1,576
Hengl, T., Nussbaum, M., Wright, M. N., Heuvelink, G. B. M., & Gräler, B. (2018) · PeerJ · doi:10.7717/peerj.5518 · Cited by 1,094
Global soil pollution by toxic metals threatens agriculture and human health
Hou, D., Jia, X., Wang, L., McGrath, S. P., Zhu, Y.-G., Hu, Q., Zhao, F.-J., Bank, M. S., O’Connor, D., & Nriagu, J. (2025) · Science · doi:10.1126/science.adr5214 · Cited by 1,016
Metadata
- Authors
- Biau, G., & Scornet, E.
- Affiliations
- Centre National de la Recherche Scientifique (FR)
- ORCID
- Scornet, Erwan
- Venue
- Test
- Published
- 2016-04-19
- Volume
- 25
- Issue
- 2
- Pages
- 197-227
- ISSN
- 1133-0686 (ISSN-L)
- 1863-8260
- OpenAlex
- W2216946510
- Cited by
- 4,149
- References
- 124
- Citations by year
- 2026: 570 · 2025: 934 · 2024: 829 · 2023: 701 · 2022: 463 · 2021: 317 · 2020: 170 · 2019: 100 · 2018: 43 · 2017: 15 · 2016: 6
- FWCI
- 55.39
- Type
- journal article
- Language
- en
- Open access
- green
- arxiv.org
- Licence
- www.springer.com/tdm (tdm)
- Locations
- Test (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- HAL (submittedVersion, other-oa)
- HAL (submittedVersion)
- RePEc: Research Papers in Economics (submittedVersion)
- arXiv (Cornell University)
- Topics
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Inference
- Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning and Data Classification
- Physical Sciences › Mathematics › Statistics and Probability › Statistical Methods and Applications
- Keywords
- random forest, ensemble learning, decision tree, variable importance measures, resampling, parameter selection, classification, regression analysis, high-dimensional data
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
How to cite
Cite CoChat for the checks on this page, and the paper for its own findings.
- This page
- CoChat. (2026). Citation check: A random forest guided tour. https://app.cochat.ai/papers/10.1007/s11749-016-0481-7
- The paper
- Biau, G., & Scornet, E. (2016). A random forest guided tour. Test. https://doi.org/10.1007/s11749-016-0481-7
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.