Hi Dojo
Is there an example of how to configure anomaly detection alarming? I.e. how do we know the format of these alarms generated, also how we configure the severities of the alarms and, if already possible, thresholds on deviation from usual ranges?
The use case if for anomaly detection based on dynamic time rages as described on this use case:
Whereby, if possible, we could also define which timespan length we need to look at to analyse the changing point.
Hi Elvio,
Configuration of Anomaly detection is not possible for the moment. The severity of the alarm is calculated by Analytics considering multiple factors such as the alarmtemplate itself, how much the anomaly deviates from normal behavior, the other changepoints in the past and the direction of the anomaly.
Thanks for this feedback and the need for more configuration is something we often hear. We are taking this in consideration. Would you be interested to be a stakeholder and join a review meeting if we would take up this task?
The format of the suggestion alarm is as follows:
The alarm is on the parameter the anomaly was detected and has a source “Suggestion Engine”. In the value there is a brief description about the anomaly that has been found. In this case there was a level shift detected.
In the use case there might be a triggering of a trend change anomaly or a variance anomaly. We should see the real data to be sure whether this is the case or not. But I think the solution suggestion by Bert is the best one.
Hi Willem
Thanks for the reply! Yes, I am available to participate.
Regarding the thresholds configuration, we could also consider in the form of configuring the level of confidence (%) required to get the alarm suggestion, hence manage the amount of alarms and evaluate the effectiveness of the configuration