i am trying to build a general forecast at work, but i am in trouble about the frequency of my series. Can a subject reason guide me through the frequency selection? I have daily data, so i could choose frequency=7 for weekly seasonality. But the data is about mobile traffic, and the cycles long 30 days. When i compare both frequencys (7,30), the models with 30 frequency have lower information criteria and it looks like have a better fit. Can the business nature of the data and the information criteria results give me support to select frequency=30?

  • 1
    $\begingroup$ You should probably account for both frequencies. See the multiple-seasonalities tag info for some ideas. Regression of Fourier terms corresponding to all the different frequencies with ARMA errors could also be considered. $\endgroup$ – Richard Hardy Mar 31 '20 at 19:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.