rfPred: Variable selection in Random Forest

In fifer: A Biostatisticians Toolbox for Various Activities, Including Plotting, Data Cleanup, and Data Analysis

Description

Variable selection for prediction purposes using Random Forest. See rfThresh for complete documentation.

Usage

rfPred(object, importance = "permutation", nfor.pred = 25, nmj = 1,
  outfile = "rfPred.file", named.file = "rfPredResults", ...)

Arguments

object	an object returned from rfInterp
importance	what importance measure should be used? Either "permutation" or "gini."
nfor.pred	number of forests to grow
nmj	a contant used for setting the threshold for variable selection. Higher values indicate a less stringent threshold.
outfile	The file location where the rfPred object should be stored. Defaults to storing it in rfPred.file in the default directory.
named.file	What should the rfPred object be named when saved? Defaults to "rfPredResults".
...	other arguments passed to cforest or randomForest

Value

varselect.pred	The variables selected for Prediction (sorted)
err.interp	The error at each stage of the stepwise variable inclusion.
mean.jump	The threshold for variable inclusion.
stepwise.error	The OOB error rate at each iteration of the stepwise procedure.
num.varselect.pred	The number of variables selected for prediction.
comput.time	Computation time of the procedure.
model	The final model, either a randomForest or cforest object.

Author(s)

Robin Genuer, Jean-Michel Poggi and Christine Tuleau-Malot, with modifications by Dustin Fife

See Also

rfInterp, rfThresh

Examples

## Not run: data(iris); 
data = iris; 
formula = as.formula("Species ~ Sepal.Length + Sepal.Width + Petal.Length + Petal.Width")
thresh = rfThresh(formula, data=iris, nruns=2, importance="permutation"); 
interp = rfInterp(thresh, importance="permutation");
predic = rfPred(interp, importance="gini")
predic
## End(Not run)

fifer documentation built on May 30, 2017, 7:40 a.m.