Is there a way to adjust anomaly detection "treshold" to raise an alarm and which severity it is? Example optical level change could trigger an alarm like this "Level decreased by 0.12 dBm (from -0.12 dBm to -0.24 dBm)". On that environment this isn't a significant change and I'd like to teach that to anomaly detection.
Dear Jarkko,
this is an excellent question and one that we've recently been discussing internally. Unfortunately, I have to say that it is not yet possible at the moment, but we are exploring multiple options to make this possible.
A first track that we would like to follow is to improve our AI so that it can learn how large a level increase/decrease should be in order to justify the triggering of an alarm. This avoids that we trigger alarms for very tiny level increases/decreases. The learning could be done through user feedback (like a thumbs up / thumbs down).
The second approach is to implement a way for users to configure this minimal increase/decrease. If the user doesn't configure anything, the AI could still propose a value and use that instead.
Both tracks are placed on the roadmap, but they are not officially approved yet to start working on. If this feature is important to you, I would propose to boost its position on the roadmap through submitting a feature request here: https://community.dataminer.services/feature-suggestions
Thanks for making us aware that this use case would be interesting for you,
Dennis