You do not have permission to delete messages in this group
Copy link
Report message
Show original message
Either email addresses are anonymous for this group or you need the view member email addresses permission to view the original message
to Beancount
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.