0

Intuitively, I have a greater confidence in a forecast performed on data at an aggregate (e.g. total company) level than the sum of forecasts at made a detailed level (e.g. product).

However, when there is a requirement to produce both, what are the recommended approaches to bring one in line with the other?

I've just started to use the wonderful fable package (and other associated R packages).

My current thought is to do both, and then proportionally reduce the detailed forecast so that its sum is equal to the aggregate forecast.

kjetil b halvorsen
  • 63,378
  • 26
  • 142
  • 467
aja
  • 1
  • 1
  • See https://otexts.com/fpp2/hierarchical.html – Rob Hyndman Sep 14 '20 at 09:26
  • Thank you @rob-hyndman. Will have a read. – aja Sep 14 '20 at 10:42
  • Some similar posts: https://stats.stackexchange.com/questions/373876/combination-of-hierarchial-time-series-forecasts-with-different-methods-settin, https://stats.stackexchange.com/questions/155694/how-can-i-forecast-interrelated-hierarchies, https://stats.stackexchange.com/questions/122599/interpretation-of-demand-forecasting-hierarchy, https://stats.stackexchange.com/questions/31473/forecasting-hierarchical-time-series-r-package, https://stats.stackexchange.com/questions/234192/optimal-combination-approach-and-forecasted-proportions – kjetil b halvorsen Sep 14 '20 at 22:29
  • Many thanks for taking the time to share those links @kjetil-b-halvorsen. Will take a look – aja Sep 16 '20 at 05:49

0 Answers0