ranking, forecast,
cumulative, accuracy and bands of
a backtest, total_forecast(), find_outliers(),
fourier_terms() and seasonal_dummies().
as.data.frame() on a backtest still gives a plain data
frame. The time column of forecasts and outliers is a date
for monthly and quarterly series.First release: R interface to the Rust crate foresight 0.7.3.
model_*()): benchmarks,
Theta, Holt-Winters, log-linear regression, seasonal ARIMA with
regressors and automatic orders, exponential smoothing with automatic
choice, a Prophet-style trend with events, TBATS, Croston with its
variants, forecasts of the seasonally adjusted series, ensembles and
Box-Cox scales.backtest(): rolling origin on several threads
(foresight_threads()), errors by horizon, the choice by
out-of-sample error and empirical intervals by horizon and for totals
(total_forecast()).decompose_stl(), decompose_mstl(),
fill_gaps(), find_outliers(),
clean_series(), tests of stationarity and seasonality,
accuracy measures.autoplot() and
theme_foresight().