43 Data Slicing

When using the createDataParition() function, the first argument you pass is what you want the function to “Split” on. Here we will set y to be the spam$type factor. The second argument tells the function what percentage of the data you want in the training set, in the code below it has been set to a \(75\% ~:~25\%\) split.

# Load the package and the data
library(caret); library(kernlab); data(spam)

set.seed(123)

# Use data partition functions to create the training and testing sets
inTrain <- createDataPartition(y = spam$type,
                               p = 0.75,
                               list=FALSE)
training <- spam[inTrain, ]
testing <- spam[-inTrain, ]

# Check that it works
dim(training)

## [1] 3451   58

dim(testing)

## [1] 1150   58

43.1 SPAM Example: K-Fold

Important functions here:

• createFolds()

set.seed(123)

# Create the folds returning the training data using the `createFolds` function
folds <- createFolds(y = spam$type,
                     k = 10,
                     list = TRUE,
                     returnTrain = TRUE)

# Check how many data points are in each fold for the training data to make sure they're the same length
sapply(folds, length)

## Fold01 Fold02 Fold03 Fold04 Fold05 Fold06 Fold07 Fold08 Fold09 Fold10 
##   4141   4140   4140   4142   4142   4141   4141   4141   4140   4141

folds[[1]][1:10]

##  [1]  1  2  3  4  5  6  7  8 10 11

# Create the folds returning the testing data using the `createFolds` function
folds <- createFolds(y = spam$type,
                     k = 10,
                     list = TRUE,
                     returnTrain = FALSE)

# Check how many data points are in each fold for the testing data to make sure they're the same length
sapply(folds, length)

## Fold01 Fold02 Fold03 Fold04 Fold05 Fold06 Fold07 Fold08 Fold09 Fold10 
##    460    460    460    459    460    461    459    461    461    460

folds[[1]][1:10]

##  [1]   1  16  28  35  40  68  78  87 111 114

43.2 Spam Example: Resampling

If you want to resample or bootstrap instead of doing k-folds cross validation, you can use the createResample() function. The argument ‘times’ tells the function how many times you want to resample the data and the ‘list’ argument lets the function know what structure you would like the data to be outputted as.

set.seed(123)

# Create resample using 'createResample()' function
folds <- createResample(y = spam$type, times = 10,
                        list = TRUE)

sapply(folds, length)

## Resample01 Resample02 Resample03 Resample04 Resample05 Resample06 Resample07 
##       4601       4601       4601       4601       4601       4601       4601 
## Resample08 Resample09 Resample10 
##       4601       4601       4601

# As we're resampling here, you can get the same output back multiple times, for example, we have 1 and 5 return twice in the first fold.
folds[[1]][1:10]

##  [1]  1  1  2  5  5  9 13 16 17 18

43.3 SPAM Example: Time Slices

If we want time slices for any particular reason, we can use the createTimeSlices() function to do that. The ‘initial Window’ argument lets the function know how long the slices will be (how many samples from our arbitrary time vector below) and the ‘horizon’ argument lets the function know how many samples we would like to predict using our window specified beforehand.

set.seed(123)

# Create our arbitrary time vector
time <- 1:1000

# Create the folds uing our 'createTimeSlices' function.
folds <- createTimeSlices(y = time, initialWindow = 20,
                          horizon = 10)

#What does the 'folds' object look like now?
names(folds)

## [1] "train" "test"

# Check how many samples are in both our window and our horizon
folds$train[[1]]

##  [1]  1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20

folds$test[[1]]

##  [1] 21 22 23 24 25 26 27 28 29 30