Hello -

Has anyone ever played with ML models for forecasting?

Google has TimesFM -- https://github.com/google-research/timesfm -- which 
is a time series forecasting model that is designed to do predictions with 
far fewer inputs. (It's pre-trained on *lots* of time series data, so 
taking a smaller sample of data it can forecast from there.)

Potential use cases -- no idea how well it'd work:

   - Scoring an anomaly score for each transaction -- how much is this away 
   from the predicted normal of a transaction.
   - Forecasting transactions for the next three months, six months, or 
   further -- what is the trend line? This could be based on only the previous 
   few months or a longer term history.

Mark



-- 
You received this message because you are subscribed to the Google Groups 
"Beancount" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
To view this discussion visit 
https://groups.google.com/d/msgid/beancount/9000f5f3-865e-4cfe-b2df-ef8ef36cff61n%40googlegroups.com.

Reply via email to