MAPE:
Mean Absolute percentage Error
Mean absolute percentage error (MAPE) is a very popular error measure which recommended by many textbooks, and it also is the main method used in M-competition. In most of the cases, MAPE are used with the other method together, such as MAE, RMSE. The formula for MAPE is:
However, there is serious problem when using the percentage error measures, Hyndman and Koehler (2005) pointed out that because the methods need to find out the relative percentage, pt, thus the percentage error will be undefined if any yt = 0 (which is frequently occur in intermittent demand), and when any yt close to zero, the result will be a very skewed distribution. Although in M3-competition, inordinately large (or infinite) MAPE results were avoided by excluding the zero-value data (Makridakis and Hibon, 2000), but this is an artificial solution which is hardly to apply in partial situations (Hyndman and Koehler,2005).
Copyright © 2015 | University of Warwick, Coventry | [email protected]
POWERED BY KASPAR CHEN