Hi all,
Please discuss & subscribe here:
https://github.com/apache/superset/issues/40520
----
This feature adds statistical anomaly detection as a post-processing step
for all timeseries Echarts. It follows a very similar architecture as the
already existing Prophet based forecasting method creating a visual
hgihlight for the same, which is very helpful for management level
dashboards where visual feedback takes highest priority for a low
time-to-insight
Key points:
- 3 detection methods: Z-score, MAD, and Prophet-based
- Red scatter points as highlight
- No new dependency
I have a working implementation ready and will open a draft PR soon
Looking forward to feedback and discussions

Thanks and Regards,
Ayush Raj

Reply via email to