rf.modelSel: Random Forest Model Selection

Description

Implements Murphy et al., (2010) Random Forests model selection approach.

Usage

rf.modelSel(xdata, ydata, imp.scale = "mir", r = c(0.25, 0.5, 0.75), final.model = FALSE, seed = NULL, parsimony = NULL, ...)

Arguments

xdata

X Data for model

ydata

Y Data for model

imp.scale

Type of scaling for importance values (mir or se), default is mir

Vector of importance percentiles to test i.e., c(0.1, 0.2, 0.5, 0.7, 0.9)

final.model

Run final model with selected variables (TRUE/FALSE)

seed

Sets random seed in the R global environment. This is highly suggested.

parsimony

Threshold for competing model (0-1)

...

Additional arguments to pass to randomForest (e.g., ntree=1000, replace=TRUE, proximity=TRUE)

Value

A list class object with the following components: @return rf.final Final selected model, if final = TRUE(randomForest model object) @return sel.vars Final selected variables (vector) @return test Validation parameters used on model selection (data.frame) @return sel.importance Importance values for selected model (data.frame) @return importance Importance values for all models (data.frame) @return parameters Variables used in each tested model (list) @return s Type of scaling used for importance

References

Evans, J.S. and S.A. Cushman (2009) Gradient Modeling of Conifer Species Using Random Forest. Landscape Ecology 5:673-683.

Murphy M.A., J.S. Evans, and A.S. Storfer (2010) Quantify Bufo boreas connectivity in Yellowstone National Park with landscape genetics. Ecology 91:252-261

Evans J.S., M.A. Murphy, Z.A. Holden, S.A. Cushman (2011). Modeling species distribution and change using Random Forests CH.8 in Predictive Modeling in Landscape Ecology eds Drew, CA, Huettmann F, Wiersma Y. Springer

Examples

Run this code

# Classification on iris data
require(randomForest)
data(iris)
  iris$Species <- as.factor(iris$Species)
( rf.class <- rf.modelSel(iris[,1:4], iris[,"Species"], seed=1234, imp.scale="mir") )
( rf.class <- rf.modelSel(iris[,1:4], iris[,"Species"], seed=1234, imp.scale="mir", 
                          parsimony=0.03) )

   plot(rf.class)              # plot importance for selected variables
   plot(rf.class, imp = "all") # plot importance for all variables 

 vars <- rf.class$selvars
 ( rf.fit <- randomForest(x=iris[,vars], y=iris[,"Species"]) )

# Regression on airquality data
data(airquality)
  airquality <- na.omit(airquality)
( rf.regress <- rf.modelSel(airquality[,2:6], airquality[,1], imp.scale="se") )
( rf.regress <- rf.modelSel(airquality[,2:6], airquality[,1], imp.scale="se", parsimony=0.03) )

   plot(rf.regress)              # plot importance for selected variables
   plot(rf.regress, imp = "all") # plot importance for all variables 

# To use parameters from competing model
vars <- rf.regress$parameters[[3]]

# To use parameters from selected model
vars <- rf.regress$selvars 

( rf.fit <- randomForest(x=airquality[,vars], y=airquality[,1]) )