You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Using ChatGPT or another method, create an online method for detecting notable outliers or potential data errors.
Prior art
See the "Find interesting views" tab of the ETL Insighter, which truncates an indicator to the last 20 years and feeds the data plus a prompt to ChatGPT 4o.
Challenges
The 128k context window is not enough to put the entire data series in
You could pass the data file for use with code interpreter (likely worse results)
You could truncate the data (e.g. last 20 years)
You could split the data into multiple chunks, and do detection on each chunk
You could thumbnail subsets of the data, and do detection based on thumbnails
The text was updated successfully, but these errors were encountered:
Overview
Using ChatGPT or another method, create an online method for detecting notable outliers or potential data errors.
Prior art
See the "Find interesting views" tab of the ETL Insighter, which truncates an indicator to the last 20 years and feeds the data plus a prompt to ChatGPT 4o.
Challenges
The text was updated successfully, but these errors were encountered: