Abstract
Earlier research has shown that non-overlapping temporal aggregation of auto-correlated demand
can improve the forecast accuracy of single exponential smoothing, especially for negative or low positive
autocorrelation parameter. In this paper, we analyse the impact of non-overlapping temporal aggregation when
an optimal forecasting method is used. We consider an AR(1) demand process and a minimum mean square
error (MMSE) forecasting method. The expressions of the mean square error (MSE) before and after the
aggregation of the demand are derived. The numerical results of the comparison of the MSEs show that by
using the optimal MMSE forecasting method, regardless of the aggregation level and the autocorrelation
parameter, the non-overlapping temporal aggregation approach is outperformed by the non-aggregation one.
Original language | English |
---|---|
Title of host publication | ILS 2016 - 6th International Conference on Information Systems, Logistics and Supply Chain |
Publication status | Published - 2016 |
Event | International Conference on Information Systems, Logistics and Supply Chain - Bordeaux, France Duration: 1 Jun 2016 → 4 Jun 2016 |
Conference
Conference | International Conference on Information Systems, Logistics and Supply Chain |
---|---|
Abbreviated title | ILS 2016 |
Country/Territory | France |
City | Bordeaux |
Period | 1/06/16 → 4/06/16 |
Bibliographical note
The full text is available from: ils2016conference.com/wp-content/uploads/2015/03/ILS2016_SB05_4.pdfKeywords
- Demand aggregation
- Forecast accuracy
- Mean square error