s_Ranger: Random Forest Classification and Regression (C, R)

In egenn/rtemis: Machine Learning and Visualization

Random Forest Classification and Regression (C, R)

Description

Train a Random Forest for regression or classification using ranger

Usage

s_Ranger(
  x,
  y = NULL,
  x.test = NULL,
  y.test = NULL,
  x.name = NULL,
  y.name = NULL,
  n.trees = 1000,
  weights = NULL,
  ifw = TRUE,
  ifw.type = 2,
  ifw.case.weights = TRUE,
  ifw.class.weights = FALSE,
  upsample = FALSE,
  downsample = FALSE,
  resample.seed = NULL,
  autotune = FALSE,
  classwt = NULL,
  n.trees.try = 500,
  stepFactor = 2,
  mtry = NULL,
  mtryStart = NULL,
  inbag.resample = NULL,
  stratify.on.y = FALSE,
  grid.resample.params = setup.resample("kfold", 5),
  gridsearch.type = c("exhaustive", "randomized"),
  gridsearch.randomized.p = 0.1,
  metric = NULL,
  maximize = NULL,
  probability = NULL,
  importance = "impurity",
  local.importance = FALSE,
  replace = TRUE,
  min.node.size = NULL,
  splitrule = NULL,
  strata = NULL,
  sampsize = if (replace) nrow(x) else ceiling(0.632 * nrow(x)),
  tune.do.trace = FALSE,
  imetrics = FALSE,
  n.cores = rtCores,
  print.tune.plot = FALSE,
  print.plot = FALSE,
  plot.fitted = NULL,
  plot.predicted = NULL,
  plot.theme = rtTheme,
  question = NULL,
  grid.verbose = verbose,
  verbose = TRUE,
  outdir = NULL,
  save.mod = ifelse(!is.null(outdir), TRUE, FALSE),
  ...
)

Arguments

x	Numeric vector or matrix / data frame of features i.e. independent variables
y	Numeric vector of outcome, i.e. dependent variable
x.test	Numeric vector or matrix / data frame of testing set features Columns must correspond to columns in x
y.test	Numeric vector of testing set outcome
x.name	Character: Name for feature set
y.name	Character: Name for outcome
n.trees	Integer: Number of trees to grow. Default = 1000
weights	Numeric vector: Weights for cases. For classification, weights takes precedence over ifw, therefore set weights = NULL if using ifw. Note: If weight are provided, ifw is not used. Leave NULL if setting ifw = TRUE.
ifw	Logical: If TRUE, apply inverse frequency weighting (for Classification only). Note: If weights are provided, ifw is not used.
ifw.type	Integer 0, 1, 2 1: class.weights as in 0, divided by min(class.weights) 2: class.weights as in 0, divided by max(class.weights)
ifw.case.weights	Logical: If TRUE, define ranger's case.weights using IPW. Default = TRUE Note: Cannot use case.weights together with stratify.on.y or inbag.resample
ifw.class.weights	Logical: If TRUE, define ranger's class.weights using IPW. Default = FALSE
upsample	Logical: If TRUE, upsample training set cases not belonging in majority outcome group
downsample	Logical: If TRUE, downsample majority class to match size of minority class
resample.seed	Integer: If provided, will be used to set the seed during upsampling. Default = NULL (random seed)
autotune	Logical: If TRUE, use randomForest::tuneRF to determine mtry
classwt	Vector, Float: Priors of the classes for randomForest::tuneRF if autotune = TRUE. For classification only; need not add up to 1
n.trees.try	Integer: Number of trees to train for tuning, if autotune = TRUE
stepFactor	Float: If autotune = TRUE, at each tuning iteration, mtry is multiplied or divided by this value. Default = 1.5
mtry	[gS] Integer: Number of features sampled randomly at each split. Defaults to square root of n of features for classification, and a third of n of features for regression.
mtryStart	Integer: If autotune = TRUE, start at this value for mtry
inbag.resample	List, length n.tree: Output of setup.resample to define resamples used for each tree. Default = NULL
stratify.on.y	Logical: If TRUE, overrides inbag.resample to use stratified bootstraps for each tree. This can help improve test set performance in imbalanced datasets. Default = FALSE. Note: Cannot be used with ifw.case.weights
grid.resample.params	List: Output of setup.resample defining grid search parameters.
gridsearch.type	Character: Type of grid search to perform: "exhaustive" or "randomized".
gridsearch.randomized.p	Float (0, 1): If gridsearch.type = "randomized", randomly test this proportion of combinations.
metric	Character: Metric to minimize, or maximize if maximize = TRUE during grid search. Default = NULL, which results in "Balanced Accuracy" for Classification, "MSE" for Regression, and "Coherence" for Survival Analysis.
maximize	Logical: If TRUE, metric will be maximized if grid search is run.
probability	Logical: If TRUE, grow a probability forest. See ranger::ranger. Default = FALSE
importance	Character: "none", "impurity", "impurity_corrected", or "permutation" Default = "impurity"
local.importance	Logical: If TRUE, return local importance values. Only applicable if importance is set to "permutation".
replace	Logical: If TRUE, sample cases with replacement during training.
min.node.size	[gS] Integer: Minimum node size
splitrule	Character: For classification: "gini" (Default) or "extratrees"; For regression: "variance" (Default), "extratrees" or "maxstat". For survival "logrank" (Default), "extratrees", "C" or "maxstat".
strata	Vector, Factor: Will be used for stratified sampling
sampsize	Integer: Size of sample to draw. In Classification, if strata is defined, this can be a vector of the same length, in which case, corresponding values determine how many cases are drawn from the strata.
tune.do.trace	Same as do.trace but for tuning, when autotune = TRUE
imetrics	Logical: If TRUE, calculate interpretability metrics (N of trees and N of nodes) and save under the extra field of rtMod
n.cores	Integer: Number of cores to use.
print.tune.plot	Logical: passed to randomForest::tuneRF.
print.plot	Logical: if TRUE, produce plot using mplot3 Takes precedence over plot.fitted and plot.predicted.
plot.fitted	Logical: if TRUE, plot True (y) vs Fitted
plot.predicted	Logical: if TRUE, plot True (y.test) vs Predicted. Requires x.test and y.test
plot.theme	Character: "zero", "dark", "box", "darkbox"
question	Character: the question you are attempting to answer with this model, in plain language.
grid.verbose	Logical: Passed to gridSearchLearn
verbose	Logical: If TRUE, print summary to screen.
outdir	String, Optional: Path to directory to save output
save.mod	Logical: If TRUE, save all output to an RDS file in outdir save.mod is TRUE by default if an outdir is defined. If set to TRUE, and no outdir is defined, outdir defaults to paste0("./s.", mod.name)
...	Additional arguments to be passed to ranger::ranger

Details

You should cconsider, or try, setting mtry to NCOL(x), especially for small number of features. By default mtry is set to NCOL(x) for NCOL(x) <= 20. For imbalanced datasets, setting stratify.on.y = TRUE should improve performance. If autotune = TRUE, randomForest::tuneRF will be run to determine best mtry value. [gS]: indicated parameter will be tuned by grid search if more than one value is passed

See Tech Report comparing balanced (ifw.case.weights = TRUE) and weighted (ifw.class.weights = TRUE) Random Forests.

Value

rtMod object

Author(s)

E.D. Gennatas

See Also

train_cv for external cross-validation

Other Supervised Learning: s_AdaBoost(), s_AddTree(), s_BART(), s_BRUTO(), s_BayesGLM(), s_C50(), s_CART(), s_CTree(), s_EVTree(), s_GAM(), s_GBM(), s_GLM(), s_GLMNET(), s_GLMTree(), s_GLS(), s_H2ODL(), s_H2OGBM(), s_H2ORF(), s_HAL(), s_KNN(), s_LDA(), s_LM(), s_LMTree(), s_LightCART(), s_LightGBM(), s_MARS(), s_MLRF(), s_NBayes(), s_NLA(), s_NLS(), s_NW(), s_PPR(), s_PolyMARS(), s_QDA(), s_QRNN(), s_RF(), s_RFSRC(), s_SDA(), s_SGD(), s_SPLS(), s_SVM(), s_TFN(), s_XGBoost(), s_XRF()

Other Tree-based methods: s_AdaBoost(), s_AddTree(), s_BART(), s_C50(), s_CART(), s_CTree(), s_EVTree(), s_GBM(), s_GLMTree(), s_H2OGBM(), s_H2ORF(), s_LMTree(), s_LightCART(), s_LightGBM(), s_MLRF(), s_RF(), s_RFSRC(), s_XGBoost(), s_XRF()

Other Ensembles: s_AdaBoost(), s_GBM(), s_RF()

egenn/rtemis documentation built on Nov. 22, 2024, 4:12 a.m.