prophet_rolling_crossvalidation: Prophet Rolling Mean Out of Sample Error (Hyndman's...

In JedStephens/prophethyperbayes: Bayesian Hyperparameter Optimisation Framework for Prophet

Description

This function implements Hyndman's Evaluation on a rolling forecasting origin. (See https://robjhyndman.com/hyndsight/tscv/). The arguments of the function are used to configure it.

Usage

prophet_rolling_crossvalidation(prophet_configuration_object, data,
  cores_used, error_metric = c("ME", "RMSE", "MAE", "MPE", "MAPE",
  "MASE"), aggregating_metric = "MEAN", process_starting_row,
  per_validation_period, graph = FALSE)

Arguments

prophet_configuration_object	The result of your configuration of the prophet_configuration() function template. (Described on the GitHub in the README)
data	a data object containing all data (test and train). This must be a dataframe with at least the column ds (standing for date in POSXIT) and y which are the observed values
cores_used	The numebr of cores upon which the paralised optimisation process is allowed to occur. (Leave at least one core open.)
error_metric	your choice of: "ME", "RMSE", "MAE", "MPE", "MAPE", "MASE". Provide only one option as a character string.
aggregating_metric	either the mean (specify "MEAN") or a vector of weights (which will calculate a weighted mean).
process_starting_row	The row for which it is the first point in Hyndman's Evaluation on a rolling forecasting origin
per_validation_period	The number of periods per each of the forward Hyndman's Evaluation on a rolling forecasting origin.
graph	The graph argument when set to 'TRUE' provides a visualisation of the rolling forecasting method.

Value

A list with two metrics: Score (which is the value to be maximized) and Pred which is not currely being used.

JedStephens/prophethyperbayes documentation built on May 23, 2019, 5:01 a.m.