rfUtilities 2.1.4
R package for random forests model selection, class balance and validation
Random Forests Model Selection, inference, fit and performance evaluation
New release of of "rfUtilities" 2.1-3 includes new functions for combining random forests ensembles, bivariate partial dependence plots and unsupervised classification using random forests.
New release of of "rfUtilities" 2.1-2 includes new functions for calculating Log Loss performance evaluation a function implementing an Isotonic regression for calibration of the estimated posterior probabilities of a model. There is also a new function for deriving parameter effect size based on partial dependency (Cafri & Bailey, 2016). The statistics Mean Absolute Error (mae) and Mean Bias Error (mbe) were added to the rf.crossValidation function.
Available functions in rfUtilities are:
accuracy - A function, called by the rf.crossValidation function or independently, that provides validation statistics for
binomial or regression models
bivariate.partialDependence - Bivariate partial-dependency plot
logLoss - Calculates Logarithmic loss (logLoss)
multi.collinear - Multi-collinearity test with matrix permutation.
occurrence.threshold - A statistical sensitivity test for occurrence probability thresholds
probability.calibration - Isotonic probability calibration
rf.class.sensitivity - Random Forests class-level sensitivity analysis
rf.classBalance - Random Forests Class Balance (Zero Inflation Correction) Model
rf.combine - Combine Random Forests Ensembles
rf.crossValidation - Random Forests classification or regression cross-validation
rf.effectSize - Random Forests parameter effect size
rf.imp.freq - Random Forests variable selection frequency
rf.modelSel - Random Forests Model Selection
rf.partial.ci - Random Forests regression partial dependency plot with confidence intervals
rf.partial.prob - Random Forest probability scaled partial dependency plots
rf.regression.fit - Evaluates fit and overfit of random forests regression models
rf.significance - Significance test for classification or regression random forests models
rf.unsupervised - Unsupervised Random Forests
Bugs: Users are encouraged to report bugs here. Go to issues in the menu above, and press new issue to start a new bug report, documentation correction or feature request. You can direct questions to jeffrey_evans@tnc.org.
To install rfUtilities in R use install.packages() to download curent stable release from CRAN
or, for the development version, run the following (requires the remotes package):
remotes::install_github("jeffreyevans/rfUtilities")
Tutorial: See (http://evansmurphy.wixsite.com/evansspatial/random-forest-sdm).