split_ts_object: Split ts object into two parts

In ing-bank/tsforecast: Time Series Forecasting Pipeline

Description

split_ts_object is a function to split a time series object into two parts based on a specified split date. The two parts can then be used separately, e.g. one part for training a time series model and the other part for validation of the model.

Usage

split_ts_object(
  ts_object,
  ts_split_date,
  output = c("both", "train", "valid"),
  max_length = Inf
)

Arguments

ts_object	A time series object to be split into two.
ts_split_date	A date object or period value (in yyyymm numeric format) indicating the date at which to split the time series object into two.
output	What kind of output is wanted from the split, either 'train' (which is the time range before and on the split date), 'valid' (which is the time range after the split date) or 'both'.
max_length	A positive integer value indicating the maximum length the resulting ts object should be. This is in case the ts object requires trimming, e.g. to limit the historic data used for training.

Value

Depending on the value for the 'output' parameter:

• 'train' - ts object containing the time range before and on the split date

• 'valid' - ts object containing the time range after the split date

• 'both' - named list containing both of the above objects

Examples

ts_object <- tstools::initialize_ts_forecast_data(
      data = dummy_gasprice,
      date_col = "year_month",
      col_of_interest = "gasprice",
      group_cols = c("state", "oil_company"),
      xreg_cols = c("spotprice", "gemprice")
   ) %>%
   dplyr::filter(grouping == "state = New York   &   oil_company = CompanyA") %>%
   tstools::transform_data_to_ts_object(seasonal_periods = 3)
ts_train_data <- split_ts_object(
   ts_object = ts_object,
   ts_split_date = 200101,
   output = "train",
   max_length = Inf
)

ing-bank/tsforecast documentation built on Sept. 18, 2020, 9:40 a.m.