rfMCCV: Monte Carlo cross-validation of Random Forest models

In RFmarkerDetector: Multivariate Analysis of Metabolomics Data using Random Forests

Description

This function allows to perform a Monte Carlo cross-validation of a Random Forest

Usage

rfMCCV(data, nsplits, test_prop, opt_params)

Arguments

data	a n x p dataframe used to build the models. The first two columns must represent respectively the sample names and the class labels related to each sample
nsplits	the number of random splittings of the original dataset into training and test data sets
test_prop	the percentage (expressed as a real number) of the observations of the original dataset to be included in each test set
opt_params	a list of optional parameters characterizing both the model to be validated and the input dataset. It may include parameters like the number of trees (ntree), the mtry, or the eventual reference class label (ref_level) of the dataset

Value

a list of three elements:

• a n x p dataframe representing the predictions during the cross-validation process. Its number of lines is equal to the number of observations included in each test set and the number of columns is equal to the number of test sets (defined by the nsplits input parameter).

• a n x p dataframe representing the labels associated to the samples of each test set. Its number of lines n is equal to the number observations in each test set while p is the number of test sets defined by the nsplits input parameter

• a list of p random forest models tested in the cross-validation process

@examples data(cachexiaData) params <- list(ntrees = 500, ref_level = levels(cachexiaData[,2])[1] ) mccv_obj <- rfMCCV(cachexiaData, nsplits = 5, test_prop = 1/3, opt_params = params)

Author(s)

Piergiorgio Palla

RFmarkerDetector documentation built on May 2, 2019, 3:42 p.m.