cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Showing results for 
Search instead for 
Did you mean: 

How to calculate the Outbound Anomaly Rate

LR_9796586
10-Marble

How to calculate the Outbound Anomaly Rate


I’m using the below signal to train the Anomaly Detection, the first 6 cycles (before de red line) corresponds to normal operation, the cycles after that are obtained when a failure in the system had happened.

LR_9796586_0-1626193396925.png

I’m using the first 5 cycles for training, as recommended in  the documentation, and monitoring the signal after that. The bellow image depicts the current results:

LR_9796586_1-1626193417623.png

 

It is detected as an anomaly, the first cycle after training (cycle 6) which corresponds to normal behavior.   One of the parameters I think could be causing this is the Outbound Anomaly Rate, until now I have set this parameter as a function of the signals period, first as ⅙  and then as a complete period, but got the same result. 

 

Is there a reference to set this parameter? Does it depend on the shape or length of the signal?

 

 

1 ACCEPTED SOLUTION

Accepted Solutions

Hello,

 

In the Help Center here is an overview of the parameters available in Anomaly Detection.  The outbound Anomaly rate is the rate at which anomalies will be reported. Data is buffered for this amount of time (for example, 3 seconds). When averaged, if there is an anomalous value for more than 50% of that time, an anomaly is reported.

 

I would recommend setting this value to several seconds or perhaps longer depending on the incoming data frequency.  Also another parameter you might want to look into is the certainty parameter.  More information can be found here but setting the certainty to a higher value, like 99.99% should help to reduce false positives.

 

Warm Regards,

 

John

View solution in original post

2 REPLIES 2

Hello,

 

In the Help Center here is an overview of the parameters available in Anomaly Detection.  The outbound Anomaly rate is the rate at which anomalies will be reported. Data is buffered for this amount of time (for example, 3 seconds). When averaged, if there is an anomalous value for more than 50% of that time, an anomaly is reported.

 

I would recommend setting this value to several seconds or perhaps longer depending on the incoming data frequency.  Also another parameter you might want to look into is the certainty parameter.  More information can be found here but setting the certainty to a higher value, like 99.99% should help to reduce false positives.

 

Warm Regards,

 

John

View solution in original post

slangley
23-Emerald I
(To:LR_9796586)

Hi @LR_9796586

 

If the previous response answered your question, please mark it as the Accepted Solution for the benefit of others with the same question.

 

Regards.

 

--Sharon

Announcements