I'm a beginner in Time Series Forecasting and I have some basic questions. I hope someone would give me some advice. Thanks!

assuming that I have the daily sales data of 2018 to today (3/5/2020) and I want to predict the future (from today) one year daily sale. how to make the a one-year daily prediction?


  • $\begingroup$ What methods are you thinking about for this prediction ? From a practical point of view, chaining predictions in the way you describe will degrade your prediction accuracy really quick regardless of what you use $\endgroup$ – AnarKi Mar 5 at 23:40
  • $\begingroup$ I agree with you. what is a good method for this long term (365 days) prediction? would you please give me some advice? thanks. $\endgroup$ – user7264299 Mar 6 at 2:34
  • $\begingroup$ I suggest you edit your question. You didn't mention a thing about what you would like to use, granularity of data etc. your question is very vague... Anyway if you only have sales data i.e. 1 feature, then lookup ARIMA and the likes based on the data specifics (seasonality, etc.). Forget Machine learning for the moment (as tagged) since you wont be really taking advantage of that. $\endgroup$ – AnarKi Mar 7 at 11:05

If I was unfamiliar with time series I would run exponential smoothing models, there are many and I would run more than one against a hold out data set to see which works best. They are a lot simpler to do than ARIMA, have a good track record of predicting, and are robust to violations of assumptions. Nothing is perfect, they are a good starting point.

| cite | improve this answer | |

Your Answer

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

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