Forecast combinations using multiple starting points

Disney, Stephen Michael ORCID: https://orcid.org/0000-0003-2505-9271 and Petropoulos, Fotios 2015. Forecast combinations using multiple starting points. Presented at: Logistics and Operations Management Section Annual Conference, Cardiff, 9 January 2015.

Full text not available from this repository.

Abstract

Motivated by a real forecasting problem we investigate the issue of when to start forecasting. We show that different starting points yield different forecasts with different levels of accuracy. We propose a new forecasting procedure that sets the final forecast to the average of these future forecasts. Using the M3 competition data we show set that the Average Relative Mean Absolute Error of the out-of-sample forecast can be reduced by up to 7% when compared to a single forecast based on all available in-sample data. Practical, context specific insights are discussed and a future research agenda is highlighted.

Item Type:	Conference or Workshop Item (Lecture)
Date Type:	Completion
Status:	Unpublished
Schools:	Centre for Advanced Manufacturing Systems At Cardiff (CAMSAC) Business (Including Economics)
Subjects:	H Social Sciences > HD Industries. Land use. Labor H Social Sciences > HD Industries. Land use. Labor > HD28 Management. Industrial Management
Last Modified:	27 Oct 2022 10:28
URI:	https://orca.cardiff.ac.uk/id/eprint/70566

Actions (repository staff only)

Edit Item