I've tried 3 options so far:

  1. Linear regressions: fast + 50% accurancy.
  2. FB Prophet: fast + 100% accurancy.
  3. SARIMAX (ARIMA): slow + 100% accurancy.

I can't use FB prophet on my project and SARIMAX is crazy slow. What are the other options to get a reasonable (not 50%) accurancy?

  • $\begingroup$ TBATS package in R - it's like the adult version of FB Prophet $\endgroup$
    – Skander H.
    May 25, 2018 at 19:33
  • $\begingroup$ 100% accuracy seems a bit high. $\endgroup$
    – zbicyclist
    May 25, 2018 at 21:20

1 Answer 1


seems like u work on seasonal data with high seasonal index, I got stuck at the same point now ,try out auto_arima by pyramid (I dont know if you have already tried it)to double the speed of SARIMAX in the process of best parameter search.

Else you could also try out PyFlux, TBATS and some other reliable techinques that have been experimented in terms of computation and time complexity. https://www.youtube.com/watch?v=pl6u8PC_1Ns

Bit lengthy but could help you in finding a better match.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.