randomUniformForest (version 1.0.6)

init_values: Training and validation samples from data

Description

Draw training and test samples from data. Samples can be accessed by subsctioting original data or by their own references.

Usage

init_values(X, Y = NULL, sample.size = 0.5, 
data.splitting = "ALL", 
unit.scaling = FALSE, 
scaling = FALSE, 
regression = FALSE)

Arguments

X

a matrix or dataframe to be splitted in training and validation sample

Y

a response vector for the observed data.

sample.size

size of the needed training sample in proportion of the nulber of observations in original data.

data.splitting

not currently used.

unit.scaling

if TRUE, scale all data in X between 0 and 1, if they are all positive, or between -1 and 1.

scaling

if TRUE, centers and scales data, so each variable willhave mean 0 abd variance 1.

regression

if TRUE and scaling = TRUE, Y will also be scaled.

Value

• a list with the following components :
• xtrain
  a matrix or data frame representing the training sample.
• ytrain
  a response vector representing the training responses according to the training sample.
• xtest
  a matrix or data frame representing the validation sample.
• ytest
  a response vector representing the validation responses according to the validation sample.
• train_idx
  subscripts of the training sample.
• test_idx
  subscripts of the validation sample.

Examples

data(iris)
Y <- iris$Species
X <- iris[,-which(colnames(iris) == "Species")]
trainingAndValidationsamples <- init_values(X, Y, sample.size = 0.5)

Xtrain = trainingAndValidationsamples$xtrain
Ytrain = trainingAndValidationsamples$ytrain
Xvalid = trainingAndValidationsamples$xtest
Yvalid = trainingAndValidationsamples$ytest