Hello Superset Community,

I would like to start the discussion for SIP-213, which proposes the
addition of visual anomaly detection for timeseries charts

Please discuss and subscribe on Github:
https://github.com/apache/superset/issues/40520

A prototype implementation for the above is available for reference at the
following PR:
https://github.com/apache/superset/pull/40954

The proposal is for the addition of statistical anomaly detection (Z-score,
MAD, and Prophet-based) as a post-processing step for timeseries ECharts,
visually highlighting anomalies as red scatter points. This feature mirrors
the architecture of the existing Prophet forecasting feature to provide
management-level dashboards with fast, low-effort visual feedback.

I have kept the pull request in draft status and will update the
implementation based on the feedback during the SIP discussion.

Thank you for your time and feedback.

Best Regards,
Ayush Raj
(DeepFriedYeti)

Reply via email to