caret (version 6.0-24)

createDataPartition: Data Splitting functions

Description

A series of test/training partitions are created using createDataPartition while createResample creates one or more bootstrap samples. createFolds splits the data into k groups while createTimeSlices creates cross-validation sample information to be used with time series data.

Usage

createDataPartition(y, 
                    times = 1,
                    p = 0.5,
                    list = TRUE,
                    groups = min(5, length(y)))
createResample(y, times = 10, list = TRUE)
createFolds(y, k = 10, list = TRUE, returnTrain = FALSE)
createMultiFolds(y, k = 10, times = 5)
createTimeSlices(y, initialWindow, horizon = 1, fixedWindow = TRUE)

Arguments

y

a vector of outcomes. For createTimeSlices, these should be in chronological order.

times

the number of partitions to create

p

the percentage of data that goes to training

list

logical - should the results be in a list (TRUE) or a matrix with the number of rows equal to floor(p * length(y)) and times columns.

groups

for numeric y, the number of breaks in the quantiles (see below)

k

an integer for the number of folds.

returnTrain

a logical. When true, the values returned are the sample positions corresponding to the data used during training. This argument only works in conjunction with list = TRUE

initialWindow

The initial number of consecutive values in each training set sample

horizon

The number of consecutive values in test set sample

fixedWindow

A logical: if FALSE, the training set always start at the first sample.

Value

• A list or matrix of row position integers corresponding to the training data

Details

For bootstrap samples, simple random sampling is used.

For other data splitting, the random sampling is done within the levels of y when y is a factor in an attempt to balance the class distributions within the splits.

For numeric y, the sample is split into groups sections based on percentiles and sampling is done within these subgroups. For createDataPartition, the number of percentiles is set via the groups argument. For createFolds and createMultiFolds, the number of groups is set dynamically based on the sample size and k. For smaller samples sizes, these two functions may not do stratified splitting and, at most, will split the data into quartiles.

Also, for createDataPartition, very small class sizes (<= 3)="" the="" classes="" may="" not="" show="" up="" in="" both="" training="" and="" test="" data<="" p="">

For multiple k-fold cross-validation, completely independent folds are created. The names of the list objects will denote the fold membership using the pattern "Foldi.Repj" meaning the ith section (of k) of the jth cross-validation set (of times). Note that this function calls createFolds with list = TRUE and returnTrain = TRUE.

Hyndman and Athanasopoulos (2013)) discuss rolling forecasting origin< techniques that move the training and test sets in time. createTimeSlices can create the indices for this type of splitting.

References

http://caret.r-forge.r-project.org/splitting.html

Hyndman and Athanasopoulos (2013), Forecasting: principles and practice. https://www.otexts.org/fpp

Examples

data(oil)
createDataPartition(oilType, 2)

x <- rgamma(50, 3, .5)
inA <- createDataPartition(x, list = FALSE)

plot(density(x[inA]))
rug(x[inA])

points(density(x[-inA]), type = "l", col = 4)
rug(x[-inA], col = 4)

createResample(oilType, 2)

createFolds(oilType, 10)
createFolds(oilType, 5, FALSE)

createFolds(rnorm(21))

createTimeSlices(1:9, 5, 1, fixedWindow = FALSE)
createTimeSlices(1:9, 5, 1, fixedWindow = TRUE)
createTimeSlices(1:9, 5, 3, fixedWindow = TRUE)
createTimeSlices(1:9, 5, 3, fixedWindow = FALSE)