vita (version 1.0.0)

PIMP: PIMP-algorithm for the permutation variable importance measure

Description

PIMP implements the test approach of Altmann et al. (2010) for the permutation variable importance measure VarImp in a random forest for classification and regression.

Usage

"PIMP"(X, y, rForest, S = 100, parallel = FALSE, ncores=0, seed = 123, ...)
"print"(x, ...)

Arguments

X

a data frame or a matrix of predictors

y

a response vector. If a factor, classification is assumed, otherwise regression is assumed.

rForest

an object of class randomForest, importance must be set to True.

S

The number of permutations for the response vector y. Default is S=100.

parallel

Should the PIMP-algorithm run parallel? Default is parallel=FALSE and the number of cores is set to one. The parallelized version of the PIMP-algorithm are based on mclapply and so is not available on Windows.

ncores

The number of cores to use, i.e. at most how many child processes will be run simultaneously. Must be at least one, and parallelization requires at least two cores. If ncores=0, then the half of CPU cores on the current host are used.

seed

a single integer value to specify seeds. The "combined multiple-recursive generator" from L'Ecuyer (1999) is set as random number generator for the parallelized version of the PIMP-algorithm. Default is seed = 123.

...

optional parameters for randomForest

x

for the print method, an PIMP object

Value

VarImp

the original permutation variable importance measures of the random forest.

PerVarImp

a matrix, where each row is a vector containing the S permuted VarImp measures for each predictor variables.

type

one of regression, classification

Details

The PIMP-algorithm by Altmann et al. (2010) permutes $S$ times the response variable $y$. For each permutation of the response vector $y^{*s}$, a new forest is grown and the permutation variable importance measure ($VarImp^{*s}$) for all predictor variables $X$ is computed. The vector perVarImp of $S$ $VarImp$ measures for every predictor variables are used to approximate the null importance distributions (PimpTest).

References

Breiman L. (2001), Random Forests, Machine Learning 45(1),5-32,

Altmann A.,Tolosi L., Sander O. and Lengauer T. (2010),Permutation importance: a corrected feature importance measure, Bioinformatics Volume 26 (10), 1340-1347,

See Also

PimpTest, importance, randomForest, mclapply

Examples

###############################
#      Regression            #
##############################
##############################
## Simulating data
X = replicate(12,rnorm(100))
X = data.frame(X) #"X" can also be a matrix
y = with(X,2*X1 + 1*X2 + 2*X3 + 1*X4 - 2*X5 - 1*X6 - 1*X7 + 2*X8 )

##############################
## Regression with Random Forest:
library("randomForest")
reg.rf = randomForest(X,y,mtry = 3,ntree=500,importance=TRUE)
##############################
## PIMP-Permutation variable importance measure
# the parallelized version of the PIMP-algorithm
system.time(pimp.varImp.reg<-PIMP(X,y,reg.rf,S=10, parallel=TRUE, ncores=2))
# the non parallelized version of the PIMP-algorithm
system.time(pimp.varImp.reg<-PIMP(X,y,reg.rf,S=10, parallel=FALSE))

##############################
#      Classification        #
##############################
## Simulating data
X = replicate(12,rnorm(100))
X= data.frame( X) #"X" can also be a matrix
z  = with(X,2*X1 + 3*X2 + 2*X3 + 1*X4 -
            2*X5 - 2*X6 - 2*X7 + 1*X8 )
pr = 1/(1+exp(-z))         # pass through an inv-logit function
y = as.factor(rbinom(100,1,pr))

##############################
## Classification with Random Forest:
cl.rf = randomForest(X,y,mtry = 3,ntree = 500, importance = TRUE)
##############################
## PIMP-Permutation variable importance measure
# the parallelized version of the PIMP-algorithm
system.time(pimp.varImp.cl<-PIMP(X,y,cl.rf,S=10, parallel=TRUE, ncores=2))
# the non parallelized version of the PIMP-algorithm
system.time(pimp.varImp.cl<-PIMP(X,y,cl.rf,S=10, parallel=FALSE))