I have very small time series data(24 points) for sales for different-different regions. I need to build Range Forecast (Confidence/Credible Intervals) for sales around it for every region. I tried FB's Prophet, XGB, ARIMA, SARIMAX but my ranges are very wide (more than 100% of my historical sales) as my data for some region is non-normal and for some other region either its bi-modal or multi-modal.

Can someone suggest some other approach where I can have narrow range with high level of confidence? In other words, I am looking for the narrowest interval method which has a fixed level of confidence.

  • 1
    $\begingroup$ What is very small? 5 observations? 10 observations? If so, what you are seeing might just be the bitter reality. You have little data, so your model parameters cannot be estimated with high precision. Accounting for that makes the prediction intervals quite wide. $\endgroup$ Apr 4, 2022 at 16:57
  • $\begingroup$ I have 24 data points to be exact @RichardHardy $\endgroup$
    – Amit Kumar
    Apr 5, 2022 at 2:53


Your Answer

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