[R] stlf algorithm failure

Christian Scherer scherer-christian at web.de
Thu Jul 5 11:55:18 CEST 2012


   Dear R-helpers,
   currently I am writing a thesis using R in forecasting scenarios.
   No the time is shortening and I wanted to set up all test cases, including
   the STLF algorithm of the forecast package.
   I am using 15 minutes (96 a day) data of two weeks to forecast one week...
   The frequency used is 672 (so two weeks)
   This should be fine because it has to be at least 2 priods..
   Unfortunately the stlf algorithm claims that my time series is "not periodic
   or has less than two periods" . But both is definitely true... If I add one
   more value it works but this would falsen the forecast and lead to more
   problems since I use different granularities and had to add 8 more values in
   the worst case....
   I have this problem only with the stlf algorithm. Holt Winters and other
   ones work perfectly...
   Do you know any solution for this problem?
   A fast answer would be really great because I am very close to the release
   of my work for my university
   Kind regards and thanks in advance
   Christian


   Ihr WEB.DE Postfach immer dabei: die kostenlose WEB.DE Mail App für iPhone
   und Android.
   [1]https://produkte.web.de/freemail_mobile_startseite/

References

   1. https://produkte.web.de/freemail_mobile_startseite/


More information about the R-help mailing list