Tools for Data Splitting.

{splitTools}

Overview

{splitTools} is a toolkit for fast data splitting. It does not have any dependencies.

Its two main functions partition() and create_folds() support

• data partitioning (e.g. into training, validation and test),
• creating (in- or out-of-sample) folds for cross-validation (CV),
• creating repeated folds for CV,
• stratified splitting,
• grouped splitting as well as
• blocked splitting (if the sequential order of the data should be retained).

The function create_timefolds() does time-series splitting where the out-of-sample data follows the (extending or moving) in-sample data.

The result of create_folds() can be directly passed to the folds argument in CV functions of XGBoost or LightGBM. Since these functions expect out-of-sample indices, set the option invert = TRUE.

Installation

# From CRAN
install.packages("splitTools")

# Development version
devtools::install_github("mayer79/splitTools")

Usage

library(splitTools)

p <- c(train = 0.5, valid = 0.25, test = 0.25)

# Train/valid/test indices for iris data stratified by Species
str(inds <- partition(iris$Species, p, seed = 1))

# List of 3
#  $ train: int [1:73] 1 3 5 7 8 10 12 13 14 15 ...
#  $ valid: int [1:38] 4 9 19 21 27 28 29 30 32 35 ...
#  $ test : int [1:39] 2 6 11 16 18 22 26 37 38 40 ...

# Same, but different output interface
head(inds <- partition(iris$Species, p, split_into_list = FALSE, seed = 1))

# [1] train test  train valid train test 
# Levels: train valid test

# In-sample indices for 5-fold CV (stratified by Species)
str(inds <- create_folds(iris$Species, k = 5, seed = 1))

# List of 5
#  $ Fold1: int [1:120] 2 4 5 6 7 8 9 10 11 15 ...
#  $ Fold2: int [1:120] 1 2 3 4 5 6 9 10 11 12 ...
#  $ Fold3: int [1:120] 1 2 3 4 6 7 8 9 11 12 ...
#  $ Fold4: int [1:120] 1 3 5 6 7 8 10 11 12 13 ...
#  $ Fold5: int [1:120] 1 2 3 4 5 7 8 9 10 12 ...

# In-sample indices for 3 times repeated 5-fold CV (stratified by Species)
str(inds <- create_folds(iris$Species, k = 5, m_rep = 3, seed = 1))

# List of 15
#  $ Fold1.Rep1: int [1:120] 2 4 5 6 7 8 9 10 11 15 ...
#  $ Fold2.Rep1: int [1:120] 1 2 3 4 5 6 9 10 11 12 ...
#  $ Fold3.Rep1: int [1:120] 1 2 3 4 6 7 8 9 11 12 ...
#  $ Fold4.Rep1: int [1:120] 1 3 5 6 7 8 10 11 12 13 ...
#  $ Fold5.Rep1: int [1:120] 1 2 3 4 5 7 8 9 10 12 ...
#  $ Fold1.Rep2: int [1:120] 1 2 3 4 5 6 8 9 11 12 ...
#  $ Fold2.Rep2: int [1:120] 1 3 6 7 8 9 10 12 13 14 ...
# [...]

# Indices for time-series splitting
str(inds <- create_timefolds(1:100, k = 5))

# List of 5
# $ Fold1:List of 2
#  ..$ insample : int [1:17] 1 2 3 4 5 6 7 8 9 10 ...
#  ..$ outsample: int [1:17] 18 19 20 21 22 23 24 25 26 27 ...
# $ Fold2:List of 2
#  ..$ insample : int [1:34] 1 2 3 4 5 6 7 8 9 10 ...
#  ..$ outsample: int [1:17] 35 36 37 38 39 40 41 42 43 44 ...
# $ Fold3:List of 2
# [...]

For more details, check out the vignette.