foresightr
Time series forecasting that replays the past before trusting a model: a rolling-origin backtest, Tashman (2000) <doi:10.1016/S0169-2070(00)00065-0>, refits every candidate at each origin, the choice is made by out-of-sample error, and prediction intervals are quantiles of the errors actually observed, for each horizon and for totals of the next periods. Models include seasonal ARIMA by exact maximum likelihood with automatic orders, Hyndman and Khandakar (2008) <doi:10.18637/jss.v027.i03>; exponential smoothing in state space form, Hyndman, Koehler, Snyder and Grose (2002) <doi:10.1016/S0169-2070(01)00110-8>; the Theta method, Assimakopoulos and Nikolopoulos (2000) <doi:10.1016/S0169-2070(00)00066-2>; a trend with changepoints and events in the manner of Taylor and Letham (2018) <doi:10.1080/00031305.2017.1380080>; TBATS, De Livera, Hyndman and Snyder (2011) <doi:10.1198/jasa.2011.tm09771>; seasonal-trend decompositions by LOESS for one or several seasonal periods, Bandara, Hyndman and Bergmeir (2021) <doi:10.48550/arXiv.2107.13462>; the method of Croston (1972) <doi:10.1057/jors.1972.50> for intermittent demand; and ensembles. The computations are done by the 'Rust' crate 'foresight', bundled with the package.
- A state space framework for automatic forecasting using exponential smoothing methods
- Automatic Time Series Forecasting: TheforecastPackage forR
- Forecasting and Stock Control for Intermittent Demands
- Forecasting at Scale
- Forecasting time series with complex seasonal patterns using exponential smoothing
- Out-of-sample tests of forecasting accuracy: an analysis and review
- The theta model: a decomposition approach to forecasting
This page was built for software: foresightr
