Scale a group time series by using a factor derived from the MASE error function

scale_by_mase() scales a group time series by using a factor derived from the MASE error function.

Usage

scale_by_mase(.tbl, .value, ...)

Arguments

.tbl: tibble; data with a value (class: numeric) column and group (class: character) column(s)
.value: numeric; unquoted name of the column that contains the numeric values
...: character; one or more unquoted grouping columns

Value

The original tibble with the .value column back-transformed to the orginal scale.

Details

Scaling a grouped time series can be helpful for global forecasting methods when using machine learning and deep learning algorithms. Scaling by MASE and using MASE as the error function is equivalent to to minimizing the MAE in the preprocessed time series.

For each series, a MASE scale factor is calculated using the denominator of the MASE scaled error equation. Then, the series is divided by this factor.

References

Pablo Montero-Manso, Rob J. Hyndman, Principles and algorithms for forecasting groups of time series: Locality and globality, International Journal of Forecasting, 2021 link

Examples


library(dplyr, warn.conflicts = FALSE)

group_ts_tbl <- tsbox::ts_tbl(fpp2::arrivals)

head(group_ts_tbl)
#> # A tibble: 6 × 3
#>   id    time       value
#>   <chr> <date>     <dbl>
#> 1 Japan 1981-01-01 14.8 
#> 2 Japan 1981-04-01  9.32
#> 3 Japan 1981-07-01 10.2 
#> 4 Japan 1981-10-01 19.5 
#> 5 Japan 1982-01-01 17.1 
#> 6 Japan 1982-04-01 10.6 

new_tbl <- scale_by_mase(.tbl = group_ts_tbl, .value = value, id)

head(new_tbl)
#> # A tibble: 6 × 3
#>   id    time       value
#>   <chr> <date>     <dbl>
#> 1 Japan 1981-01-01 0.722
#> 2 Japan 1981-04-01 0.456
#> 3 Japan 1981-07-01 0.497
#> 4 Japan 1981-10-01 0.954
#> 5 Japan 1982-01-01 0.837
#> 6 Japan 1982-04-01 0.519

attributes(new_tbl)$scale_factors
#> # A tibble: 4 × 2
#>   id    scale
#>   <chr> <dbl>
#> 1 Japan 20.4 
#> 2 NZ    33.0 
#> 3 UK    37.5 
#> 4 US     9.69