You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Describe the feature:
Being able to set cross condtion from other detectors or a specific field in the datafeed.
Describe a specific use case for the feature:
We are trying to detect anomalies from a wide population of pumps (influencers). One specific job is to monitor the motor temperature (detector) and when it's increasing or decrasing (compared to the average of the history ) we may have an issue. That's true when the the pump is running stable at the nominal speed. It's not true anymore when we have to stop and restart the pump, motor temperature is going to drop and speed is equal to zero. We don't want this transitional behavior to impact the anomaly detection (it's not an anomaly it's just a start and stop).
Something like: do not consider the motor temperature if the speed is unstable over the last 15min would do the job. It's not possible to filter the datafeed because each pump has its own nominal speed.
@elastic/machine-learning
The text was updated successfully, but these errors were encountered:
Hello,
Here is an other idea related to ML custom rules:
Consider anomaly if the average value of the detector over the last the bucket is 10% higher (or lower) compared to the average. We have some issue where the parameter is super stable and a slight variation creates an important severity.
Describe the feature:
Being able to set cross condtion from other detectors or a specific field in the datafeed.
Describe a specific use case for the feature:
We are trying to detect anomalies from a wide population of pumps (influencers). One specific job is to monitor the motor temperature (detector) and when it's increasing or decrasing (compared to the average of the history ) we may have an issue. That's true when the the pump is running stable at the nominal speed. It's not true anymore when we have to stop and restart the pump, motor temperature is going to drop and speed is equal to zero. We don't want this transitional behavior to impact the anomaly detection (it's not an anomaly it's just a start and stop).
Something like: do not consider the motor temperature if the speed is unstable over the last 15min would do the job. It's not possible to filter the datafeed because each pump has its own nominal speed.
@elastic/machine-learning
The text was updated successfully, but these errors were encountered: