Retraining an Anomaly Alert
When an anomaly alert has reached the Buffering or Monitoring status, a retraining service can be launched to refresh the existing anomaly model. During retraining, the alert is returned to the Calibrating status where new data is collected so that a new model can be built and a new validation set generated. The ability to retrain an anomaly alert is useful when changes occur in the operating environment of the device you are monitoring.
There are two ways to launch the retraining service in ThingWorx Composer:
• Execute the RetrainAlert service from the Generic list on the Services page of the remote Thing.
• Click the retrain icon 
in the Status column on the Alerts page of the remote Thing. (This icon is only available when the alert is in Monitoring status.)
in the Status column on the Alerts page of the remote Thing. (This icon is only available when the alert is in Monitoring status.)