Forecasting - Combination of methods

dplyr
tidyverse

#1

Hi there,
I have several forecasts and I am determined to combine them according to several weights.
Holt Winders = [rows = Calenderweeks, column =products]
Smoothing = [samething]c
and so on
As anyone suggestions on how I can manage that the best way?

Regarding that question: Can I combine columns simply by calling the columns iteratively of the different forecasting matrices?

Best regards

Anna Urban


#2

Could you please include a minimal example of what you're working with. It's much easier to troubleshoot code with code:

Ideally you could do this using reprex (short for minimal reproducible example). It will help us help you if we can be sure we're all working with/looking at the same stuff.

If you've never heard of a reprex before, you might want to start by reading the tidyverse.org help page. The reprex dos and don'ts are also useful.

Also see the community FAQ on reprex:


#3

Could this be a good usecase for broom?

Here's a nice review of R packages for forecast combinations. It was written in 2016 and the space is moving quickly


I 2nd the reprex request.


#4

In addition to the blog post that @EconomiCurtis shared, Hyndman also has an excellent book on forecasting that is freely available online and includes R code - it's one of the most practical and useful textbooks I've ever read. (The second edition is 2018 and the online version was last updated earlier this month). Here is his section on combining forecasts: https://otexts.org/fpp2/combinations.html

You may find the whole book useful though, depending on your project and timeline.

Update: To save everyone from needing to skim the link, the gist of the advice is that weighted averages don't seem to perform any better than unweighted
averages:

While there has been considerable research on using weighted averages, or some other more complicated combination approach, using a simple average has proven hard to beat.