Error Measurement
It is always a question that which method is the best in forecasting, unfortunately, there is no a universally agreed answer for this, a more appropriate saying is that it depends. Although lots of methods have been compared on real and simulated data in M-, M2-, M3- competitions (Makridakis et al. 1982, 1993; Makridakis and Hibon 2000), there are still hefty different situations and forecast environments in real life that might perform differently while forecasting.
Many examples can be found in real life such as in continuous demand forecasting usually is using SES, but it is undeniable that in many cases WMA can work much better than SES. In Intermittent demand forecasting, CR is still the most popular method, but the different versions of CR perform much better in statistical forecasting software such as SAP (Teunter, Syntetos, and Babai 2011).
Therefore, the essential is to determine the performance of each method, which the determination should not only contain subjective judgements, but also present a well quantitative result, and the techniques to approving the quantitative result is call Error Measurement. Three error measurement techniques would be introduced in here, and in the year 2 syllabus, MAE(MAD) and TS methods are mentioned.
Copyright © 2015 | University of Warwick, Coventry | [email protected]
POWERED BY KASPAR CHEN