Configure the threshold
Configure the adaptive alert's threshold.
Note: It’s recommended to wait at least a week to gather data about real-life traffic and then create the adaptive alert with appropriate settings.
Select the Predictive Model:
- Daily - if your everyday data are comparable with each other.
- Weekly - if your data are more likely to be similar on each particular day of the week.
Set model sensitivity:
- None - if you don’t want the alert to monitor the upper and lower prediction's values.
- Forgiving - if you want the alert to be triggered only after more deviations and more anomalies occur in a longer window of time. Can result in false negatives.
- Sensitive - if you want the alert to be triggered when fewer anomalies are detected. Can result in false positives.
- Custom - if you want to set your own model properties.
Set alert firing properties:
- Number of Anomalies required to trigger the alert - Anomaly is the number of polling cycles with atypical traffic required to trigger the alert. The most forgiving is 32; the most sensitive is 1.
- Number of Deviations required to trigger the alert - Deviation is the number of units of traffic measurements for which the detected traffic must differ from the predicted one to trigger the alert. The most forgiving is 8; the most sensitive is 1.
- Number of Polling cycles (time frames of five minutes) in which the anomalies and deviations must take place.
What you should see
- The creation of the alert for a new CP code may take some time due to processing of necessary data.
- You may see some alerts in the Alert History tab. It shows all anomalous periods from the enrolled data.