Is a MASE calculation meaningful if the Test Set is all zeros?

Asked Mar 06 '18 at 20:36

Active Mar 07 '18 at 12:32

Viewed 83 times

I'm dealing with a dataset of intermittent demand (orders) for the inventory of over 6000 parts. I've extracted one of the parts just to show the type of output I'm getting for a MASE calculation. They're all not as simple as this, but I'm wondering if this is a meaningful calculation or should I just discount all zero MASE calculations?

library(forecast)
part<-c(1,2,0,0,0,0,0,0,0,0,2,0,0,0,0,0,0,0,0,0,0,0,0,0)
part<-ts(part,start=1990,end=2013)
part2<-window(part,start=1990,end=2001)
partfit<-naive(part2,h=12)
part3<-window(part,start=2002)
accuracy(partfit,part3)

with output

                  ME     RMSE       MAE  MPE MAPE MASE       ACF1 Theil's U
Training set -0.09090909 1.087115 0.6363636 -Inf  Inf    1 -0.4654289        NA
Test set      0.00000000 0.000000 0.0000000  NaN  NaN    0        NaN       NaN

edited Mar 07 '18 at 12:32

asked Mar 06 '18 at 20:36

Angus

You mention MAE in the title but not in the content... – Richard Hardy Mar 07 '18 at 07:28
Corrected, see title. This is a problem with intermittent data. – Angus Mar 07 '18 at 20:48

Is a MASE calculation meaningful if the Test Set is all zeros?

0 Answers0