forecast.run_cross_validation()

Run rolling forecast origin cross-validation.

Usage

Source

forecast.run_cross_validation(
    forecast_function,
    historic,
    params,
    horizon,
    step,
    min_train=365 * 2,
    cores=1,
    error_horizons=(7, 14, 21, 28, 35, 42)
)

This can take some time because it fits a separate forecasting model for every combination of fold, metric, and area. Progress is displayed with tqdm, including the number of completed forecasts. The function also prints the number of worker processes used and the total run time.

Parameters

forecast_function: prophet | arima

Forecasting function to run.

historic: pd.DataFrame

Historic data used to create rolling training and test samples.

params: ProphetParams | ARIMAParams

Parameters for the selected forecasting model.

horizon: int

Number of daily observations in each test set.

step: int

How many days to move by before creating a new sample. Warning: using a step of 365 will produce test samples all at approximately the same time of year.

min_train: int = 365 * 2

Minimum number of days to include in training sample. By default, set to 2 years as that allows detection of yearly seasonality.

cores: int = 1

Number of CPU cores to use. Set to 1 for sequential processing or -1 to use all available cores.

error_horizons=(7, 14, 21, 28, 35, 42)

Horizons to calculate error at.

Returns

forecasts: pd.DataFrame

Forecasts contains results for every fold, metric, and area.