Hi Daniel, and welcome to the community!
The only reasonable way to do this is with script processes that either run inside Airtable using the Script Block or Custom Script Apps (React Javascript environment). While both of these approaches make forecasting possible, I tend to use a more distant approach with Google Apps Script, NodeJS, or even Julia if I need a compiled app in an embedded device.
But I did create an example using script blocks and it worked pretty well as a demonstration of pure machine-learning in simple javascript. I must warn you though - the code is wonky because it is based on some very old data so it simulates years ahead and it can be a little wonky to read.
Not only transfer to the next month, but embracing all earlier data/forecasts to derive a newly updated forecast based on the most recent information, right? This is the machine learning component of the forecasting methodology and you can see how I did that in the example.
