Using the Result feedback button, you can mark results as true or false positives. This feedback can significantly improve the accuracy of detection.

Some results trigger an automated workflow, and all are manually reveiewd by the Elementary team.

Just so you know - Our machine learning models thrive on your feedback! We’re always hustling to make them even better, and your feedback play a huge role in helping us achieve that. So keep those comments coming!

​
False positive feedback

To get context on your false positive result feedback and trigger a response, we ask you to select a reason:

  • Insignificant change - The anomaly is not drastic enough for me to care about it. Usually the action item is to relax anomaly detection sensitivity.
  • Expected outlier - This isn’t an anomaly and should be within the expected range. The action item will be to re-train the model, sometimes with a wider training set.
  • Business anomaly - This is an anomaly, but one we expected to happen due to intentional change or business event. The action item will be to exclude the anomaly from the training set.
  • Not an interesting table - I don’t want to monitor this table. The action item is to delete the monitor.
  • Other - Non of the other reasons are a fit. Please add a comment to describe the use case.