decompose_ts_object_for_ML: Decompose time series object

In ing-bank/tsforecast: Time Series Forecasting Pipeline

Description

decompose_ts_object_for_ML Add columns to a time series object that contains features of the time variable, including a periods column. This is in order to prepare the dataset for models that are not suited to handle time series objects like CART trees or RandomForest forests.

Usage

decompose_ts_object_for_ML(
  ts_object,
  filter_stationary = T,
  filter_date_features = F,
  add_xreg_deltas = T
)

Arguments

ts_object	A time series object, with the column of interest. It can also contain external regressor columns.
filter_stationary	Boolean, indicating whether to filter out stationary features of the decomposed time series. For instance, if data contains monthly data, then the "minute" feature will be the same across all periods. Setting filter_stationary to TRUE will eliminate all columns which have the same value in all rows.
filter_date_features	Boolean, indicating whether to filter out any column that contains a date feature (TRUE) or not (FALSE). This can be used for forecast methods which are not able to handle date objects as features.
add_xreg_deltas	Boolean, indicating whether to add as additional column(s) the deltas between consecutive rows for the external regressor column(s) (TRUE) or not (FALSE).

Value

A tibble that contains the original variable(s) along with date featurs as additional columns

Examples

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() %>%
  decompose_ts_object_for_ML()

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