Classification and Regression by randomForest

By Andy Liaw, Matthew C. Wiener. 2007

19,523 citations recorded by the source corpus.

Subfields: Evaluation & Benchmarks, Algorithms & Complexity

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Evaluation & Benchmarks

In Algorithms & Complexity

Citations in the corpus

Explanations of model predictions with live and breakDown packages (2018, methodology): “Find more examples for classification and regression models created with caret, mlr, randomForest (Liaw and Wiener, 2002) and other frameworks in package vignettes6.”
Iterative random forests to discover predictive and stable high-order interactions (2017, methodology): “The tuning parameters of Random Forests were set to default levels for the R randomForest package (Liaw and Wiener, 2002), M = 100 random intersection trees of depth 5 were grown with nchild = 2, and B = 20 bootstrap replicates were taken to determine the stability scores of recovered interactions.”
Forest Floor Visualizations of Random Forests (2016, methodology): “For an original RF implementation [10], predictions of terminal nodes of classification trees are reduced to a single majority vote.”
Interpreting tree ensembles with inTrees (2014, methodology): “In the inTrees R implementation, the tree format is defined as the same as the “randomForest”[11] and “RRF”[5] R packages shown in Table 1.”
Interpreting random forest classification models using a feature contribution method (2013, methodology): “4 The distribution Ŷi is calculated by the function predict in the R package randomForest [13] when the type of prediction is set to prob.”
Peeking Inside the Black Box: Visualizing Statistical Learning With Plots of Individual Conditional Expectation (2013, methodology): “We use Random Forests ( RF ) implemented in R (Liaw and Wiener, 2002) to ﬁt ˆ f .”
Interpretable classification models for recidivism prediction (2015, methodology): “Random Forests RF randomForest (Liaw and Wiener 2002) 19 values of W + Support Vector Machines (Radial Basis Kernel) SVM RBF e1071 (Meyer et al.”
Relational autoencoder for feature extraction (2017, methodology): “Although feature selection methods, such as Subset Selection [7] and Random Forest [8], are effective in ﬁltering out unnecessary features, if all input features contribute to ﬁnal results to some extent, feature selection is more likely to give rise to information loss than feature extraction…”
Learning Optimal Classification Trees: Strong Max-Flow Formulations (2020, background): “For example, they are the building blocks for random forests, one of the most popular and stable machine learning techniques available, see e.g., Liaw and Wiener (2002).”
The age of secrecy and unfairness in recidivism prediction (2018, methodology): “To then investigate whether the COMPAS general score depends in a more complex way on the Criminal Involvement Subscale items listed in the Appendix in Table 12, we ran several machine learning algorithms (random forests (31), boosted decision trees (32), and support vector machines with a radial ba”
Estimation and Inference of Heterogeneous Treatment Effects using Random Forests (2015, neutral): “…Gaussian: Moreover, the asymptotic variance σ n can be consistently estimated using the inﬁnitesimal jackknife (7) : b (cid:14) We note that Theorem 1 also holds for binary classiﬁ-cation forests with leaf size k = 1, as is default in the R -package randomForest [Liaw and Wiener, 2002].”
Uniform Convergence of Random Forests via Adaptive Concentration (2015, methodology): “Unlike competing non-parametric techniques such as kernel methods or neural networks, random forests require very little tuning; experience has shown that we can often obtain good predictive models out-of-the-box with standard software like randomForest for R [27].”
Optimal randomized classification trees (2021, methodology): “The CART classifier is built with the rpart R package [51] and RF with the randomForest R package [52] both with the default tuning parameters, while the ORCT classifier outlined below is built with depth D = 1.”
Gene selection and classification of microarray data using random forest (2006, neutral): “The most important parameter to choose is mtry, the number of input variables tried at each split, but it has been reported that the default value is often a good choice [24].”