Abstract

The M4 forecast competition required forecasts of 100,000 time series at different frequencies. We provide a detailed description of the calibrated average of Rho and Delta (Card) forecasting method that we developed for this purpose. Delta estimates a dampened trend from the growth rates, while Rho estimates an adaptive but simple autoregressive model. Calibration estimates a more elaborate autoregressive model, treating the averaged forecasts from Rho and Delta as if they were observed. The proposed method is easy to understand, combining very fast execution with an excellent forecast performance.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call