Forecasting autoregressive demands with temporal aggregation
Abstract
This paper examines the role of temporal aggregation in forecasting autoregressive demand processes. Motivated by the widespread presence of autocorrelation in demand data and the common use of aggregation in practice, the study analyzes how different aggregation levels affect forecast accuracy. The authors focus on autoregressive demand structures and investigate both theoretical properties and empirical performance. The results show that temporal aggregation can substantially improve forecasting accuracy under certain conditions, but that the benefits depend on the interaction between aggregation level, demand dynamics, and forecasting method. These findings provide guidance for practitioners on when and how temporal aggregation should be applied in operational forecasting contexts.