I would like to know which method is used by AWS Forecast to generate lower bound and upper bound time series forecasts at a given quantile?

More generally, what is the method employed to make quantile forecasts? I would be glad if you can share some related papers, articles, ...

Source: https://docs.aws.amazon.com/forecast/latest/dg/metrics.html

  • 1
    $\begingroup$ See d1.awsstatic.com/whitepapers/… $\endgroup$ Jan 12, 2021 at 7:15
  • $\begingroup$ Thank you for the link. If I have well understood, instead of minimizing the RMSE (for example) when making point forecast, they minimize the Quantile Loss to generate quantile forecasts. So does it mean that it is "as simple as" changing the loss function? $\endgroup$
    – yoyoog
    Jan 12, 2021 at 8:53

1 Answer 1


Yes to me that's all it is, changing the loss function.
This is the case for the linear regression VS quantile regression.

Your coefficients won't have the same meaning, from Wikipedia :
"Whereas the method of least squares estimates the conditional mean of the response variable across values of the predictor variables, quantile regression estimates the conditional median (or other quantiles) of the response variable."


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.