0

I am new in the field of time series data analysis. I am trying to forecast Production data by using SARIMA models. There is a clear seasonal pattern in the data. Unfortunately, I only have 41 observations in total that I can use to build the models. From theory, I know the test dataset should be at least 20-30%, but as my dataset is too small, I am doubting that if only 30-32 observations are enough to find the best models from a GRID SEARCH? As in this case we are not considering the most recent observations.

So, what should be the train:test ratio in my case?

Sharif Al-Mahmud
  • 1
  • 2
  • I'm not voting to close yet but your question doesn't seem to be in line of what's expected of questions on StackOverflow, you are a bit off topic https://stackoverflow.com/help/on-topic - can you make your question more specific or provide some background that makes this more specific about the tool or a programming problem? Currently, your question seems to be more theoretical and about data analysis. – Grismar Dec 25 '21 at 23:20
  • I would use one period, two at most. why? because is difficult to generalize with such small records, so what you can do is precisely use moving averages knowing that this methods (of the moving average family) do well on short term predictions. – Javier Espinoza Dec 25 '21 at 23:22
  • 1
    I’m voting to close this question because this doesn't seem to be about programming as defined in the [help/on-topic]. This seems to be more appropriate on [Cross Validated](https://stats.stackexchange.com/help/on-topic), which is the Stack Exchange site for "*for statisticians, data miners, and anyone else doing data analysis or interested in it as a discipline*". Check the link on what questions are on-topic and their [How to Ask](https://stats.stackexchange.com/help/how-to-ask) page. – Gino Mempin Dec 26 '21 at 00:04
  • Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. – Community Jan 03 '22 at 12:37

0 Answers0