Hello, I'm trying to better understand how the expected range in Elementary is calculated. Given the configuration below, I noticed that the graph shows expected ranges for a duration of around 7 months, with each data point displaying its expected range. From the documentation, I understand that the expected values themselves are not stored, but rather generated in run time based on: • The training data is saved in the

METRICS_ANOMALY_SCORE

table. • The metric values for all time buckets are stored in the

DATA_MONITORING_METRICS

table. Could someone explain how Elementary is able to show expected ranges spanning nearly 7 months by using only 14 days of training data? Configuration:

days_back: 30
min_training_set_size: 14
      - elementary.volume_anomalies:
          seasonality: 'day_of_week'

Hey @Pavan Kumar Kota! When setting the seasonality, Elementary automatically multiplies the training period by 7 to ensure there are enough data points to compare to. You can adjust the training period to have it smaller - 4 days would make it 4 weeks.

Thanks Noya! Can you point me to table in which this training data is stored. In METRICS_ANOMALY_SCORE, I see only 14 days of training data! That's why I was asking - how Elementary is able to show expected ranges spanning nearly 7 months by using only 14 days of training data?