Simple Training/Test Set Splitting for Time SeriesSource:
time_series_split creates resample splits using
returns only a single split. This is useful when creating a single
time_series_split( data, date_var = NULL, initial = 5, assess = 1, skip = 1, lag = 0, cumulative = FALSE, slice = 1, point_forecast = FALSE, ... )
A data frame.
A date or date-time variable.
The number of samples used for analysis/modeling in the initial resample.
The number of samples used for each assessment resample.
A integer indicating how many (if any) additional resamples to skip to thin the total amount of data points in the analysis resample. See the example below.
A value to include an lag between the assessment and analysis set. This is useful if lagged predictors will be used during training and testing.
A logical. Should the analysis resample grow beyond the size specified by
initialat each resample?.
Returns a single slice from time_series_cv
Whether or not to have the testing set be a single point forecast or to be a forecast horizon. The default is to be a forecast horizon. Default:
These dots are for future extensions and must be empty.
rsplit object that can be used with the
functions to extract the data in each split.
lag variables can be specified as:
initial = 24
initial = "2 years", if the
date_var(date or datetime)
Initial (Training Set) and Assess (Testing Set)
The main options,
assess, control the number of
data points from the original data that are in the analysis (training set)
and the assessment (testing set), respectively.
skip enables the function to not use every data point in the resamples.
skip = 1, the resampling data sets will increment by one position.
Example: Suppose that the rows of a data set are consecutive days. Using
skip = 7
will make the analysis data set operate on weeks instead of days. The
assessment set size is not affected by this option.
The Lag parameter creates an overlap between the Testing set. This is needed when lagged predictors are used.
Cumulative vs Sliding Window
cumulative = TRUE, the
initial parameter is ignored and the
analysis (training) set will grow as
resampling continues while the assessment (testing) set size will always remain
cumulative = FALSE, the
initial parameter fixes the analysis (training)
set and resampling occurs over a fixed window.
This controls which slice is returned. If
slice = 1, only the most recent
slice will be returned.
library(dplyr) # DATA ---- m750 <- m4_monthly %>% dplyr::filter(id == "M750") # Get the most recent 3 years as testing, and previous 10 years as training m750 %>% time_series_split(initial = "10 years", assess = "3 years") #> Using date_var: date #> <Analysis/Assess/Total> #> <120/36/306> # Skip the most recent 3 years m750 %>% time_series_split( initial = "10 years", assess = "3 years", skip = "3 years", slice = 2 # <- Returns 2nd slice, 3-years back ) #> Using date_var: date #> <Analysis/Assess/Total> #> <120/36/306> # Add 1 year lag for testing overlap m750 %>% time_series_split( initial = "10 years", assess = "3 years", skip = "3 years", slice = 2, lag = "1 year" # <- Overlaps training/testing by 1 year ) #> Using date_var: date #> <Analysis/Assess/Total> #> <120/48/306>