Gets the best model from a modeltime table

Source: R/modeltime_multibestmodel.R

this feature allows you to select the best model for each series, based on a specific evaluation metric.

modeltime_multibestmodel(
  .table,
  .metric = NULL,
  .minimize = TRUE,
  .forecast = TRUE
)

Arguments

.table

'table_time**' tibble generated with the modeltime_multifit() function.
.metric

evaluation metric, from modeltime_accuracy() of 'modeltime' package: 'mae', 'mape','mase','smape','rmse','rsq'.
.minimize

boolean (default = TRUE), TRUE if the error metric should be minimized, FALSE in order to maximize it.
.forecast

boolean (default = TRUE), If it is TRUE, it indicates that the modeltime_multi forecast() function has already been applied to the object that enters the ".table" parameter. This is evaluated by the existence of the column "nested_forecast".

Value

table_time tibble filtered by the best model.

Details

take the object 'table_time' from the output of the function modeltime_multifit(), and selects the best model based on the selected metric.

Examples

# Data
data_serie <- sknifedatar::table_time
                                      
# best_model_emae
sknifedatar::modeltime_multibestmodel(.table = data_serie$table_time,
                                      .metric = "rmse",
                                      .minimize = TRUE,
                                      .forecast = FALSE)

#> # A tibble: 2 x 7
#>   sector  nested_column   m_ets m_nnetar nested_model   calibration   best_model
#>   <chr>   <list>          <lis> <list>   <list>         <list>        <list>    
#> 1 Comerc… <tibble [49 × … <fit… <fit[+]> <model_time [… <model_time … <int [1]> 
#> 2 Ensena… <tibble [49 × … <fit… <fit[+]> <model_time [… <model_time … <int [1]>