rf.cross.validation: rf.cross.validation

In shihuang047/crossRanger: Cross-Application of Ranger Models using microbiome data

rf.cross.validation

Description

It runs standard random forests with n-folds cross-validation error estimation for both classification and regression using rf.out.of.bag.

Usage

rf.cross.validation(
  x,
  y,
  nfolds = 3,
  ntree = 500,
  verbose = FALSE,
  sparse = FALSE,
  imp_pvalues = FALSE,
  ...
)

Arguments

x	Training data: data.matrix or data.frame.
y	A response vector. If a factor, classification is assumed, otherwise regression is assumed.
nfolds	The number of folds in the cross validation. If nfolds > length(y) or nfolds==-1, uses leave-one-out cross-validation. If nfolds was a factor, it means customized folds (e.g., leave-one-group-out cv) were set for CV.
ntree	The number of trees.
verbose	A boolean value indicates if showing computation status and estimated runtime.
sparse	A boolean value indicates if the input matrix transformed into sparse matrix for rf modeling.
imp_pvalues	If compute both importance score and pvalue for each feature.
...	Other parameters applicable to 'ranger'.

Value

Object of class rf.cross.validation with elements including a ranger object and mutiple metrics for model evaluations.

Author(s)

Shi Huang

See Also

ranger

Examples

x0 <- data.frame(t(rmultinom(16,160,c(.001,.5,.3,.3,.299))) + 0.65)
x <- data.frame(rbind(t(rmultinom(7, 75, c(.201,.5,.02,.18,.099))),
            t(rmultinom(8, 75, c(.201,.4,.12,.18,.099))),
            t(rmultinom(15, 75, c(.011,.3,.22,.18,.289))),
            t(rmultinom(15, 75, c(.091,.2,.32,.18,.209))),
            t(rmultinom(15, 75, c(.001,.1,.42,.18,.299)))))
s<-factor(c(rep("A", 15), rep("B", 15), rep("C", 15), rep("D", 15)))
y<-factor(rep(c("Y", "N"), 30))
y0<-factor(c(rep("A", 10), rep("B", 30), rep("C", 5), rep("D", 15)))
system.time(rf.cross.validation(x, y, imp_pvalues=FALSE))
system.time(rf.cross.validation(x, y, imp_pvalues=TRUE))
rf.cross.validation(x, y0, imp_pvalues=FALSE)
y_n<- 1:60
rf.cross.validation(x, y_n, nfolds=5, imp_pvalues=FALSE)
rf.cross.validation(x, y_n, nfolds=5, imp_pvalues=TRUE)
# when nfolds is a factor, it actually run a leave-one-group-out cv
rf.cross.validation(x, y, nfolds=s, imp_pvalues=TRUE)
rf.cross.validation(x, y_n, nfolds=s, imp_pvalues=FALSE)

shihuang047/crossRanger documentation built on Feb. 7, 2023, 10:03 p.m.