Create model weights based on ranking and apply to forecast data

This function creates weights based on a ranked dataframe of models. It takes in either a path containing csv files, a list of csv files, or a dataframe with a column named "model_name" containing the respective model names and applies the weights to each model's predictions. The output is a new csv file with the weighted ensemble forecast and a dataframe of the calculated weights.

Usage

propose_weights(
  forecast_data,
  ingest = c("dataframe", "path", "list"),
  rank_df,
  rankCol = NULL,
  drop = NULL,
  team = "TeamName",
  mod_name = "my_ensemble"
)

Arguments

forecast_data: A dataframe, path to directory with csv files, or list of csv files containing the forecasts to be weighted.
ingest: A character vector that determines whether the forecast_data parameter represents a dataframe ("dataframe"), a directory path as a string (e.g. "./forecast_data") of csv files containing forecast data ("path"), or a list of strings each representing the path to a csv file ("list").
rank_df: A dataframe containing columns for each model's name and its rank.
rankCol: The name of the column in the rank_df dataframe that contains the ranking information.
drop: An integer indicating how many models at the bottom of the ranking should be excluded from the analysis.
team: A string representing the name of the team generating the forecast data.
mod_name: A string representing the name of the weighted ensemble model.

Value

Returns a dataframe containing the calculated weights, which is ordered by rank, and writes a csv file of the weighted ensemble forecast to a specified directory.

Examples

props_weights <- propose_weights(forecast_data = "path/to/forecast_data", ingest = "path", rank_df, rankCol = "rank", drop = 2, team = "my_team", mod_name = "weighted_ensemble")
#> Error in nrow(rank_df): object 'rank_df' not found