Anomaly detection model based on math expression could not exclude datetime range

Question

I created math expression based on FILL(LOG(RATE(METRIC)),0) math expression, and excluded time range which was wrong, but AI model could not learn and after that wrong period predicting wrong values. Please help me with this - why it is happening and how can I fix it ? For excluding datetime I'm using boto3 library and put_anomaly_detection endpoint. I'm getting already created models and putting that models by excluding datetime.

Answer

Anomaly Detection models are not supported for math expressions containing 'METRICS' expression. I have assumed you use it; based on the `METRIC` present in your expression `FILL(LOG(RATE(METRIC)),0)`. Please correct me if this is not the case.

Along with this, there's a bunch of other math expressions currently not supported for AD models. [Click here for the full list](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/using-metric-math.html#using-anomaly-detection-on-metric-math).

---
On a side note, you can consider allowing some time for the model to retrain, based on sudden spikes or changes. Models are retrained [continually](https://docs.aws.amazon.com/AmazonCloudWatch/latest/monitoring/CloudWatch_Anomaly_Detection.html#CloudWatch_Anomaly_Detection_Algorithm). If this wait does not bring any improvements, I would suggest to delete & recreate the model from scratch with appropriate expression & time ranges.

Anomaly detection model based on math expression could not exclude datetime range

Relevant content