Mixed pooling of seasonality for time series forecasting: An application to pallet transport data

Multiple seasonal patterns, which often interact with each other, play a key role in time series forecasting, especially for business where effects are dramatic. Previous approaches including Fourier decomposition, exponential smoothing, and autoregressive integrated moving average (SARIMA) models do not reflect the distinct characteristics of period patterns. We propose mixed hierarchical seasonality (MHS) model. Intermediate parameters first estimated, mixture intermediate is taken. This results model that automatically learns relative importance addresses interactions between them. The implemented Stan, probabilistic language, was compared three existing on real-world dataset pallet transport from logistic network. Our new achieved considerable improvements terms out sample prediction error predictive density to complete pooling, SARIMA