Fine Grained control on Anomaly Detection for different series_type

[Fletchersan]

Status Quo of current architecture:

Currently the AnomalyDetectionController class has a single detect function which runs anomaly detection for all 3 series types:

• overall
• sudim
• data_quality

A more fine grained control would help in multiple scenarios down the line, including in cases where we would run anomaly detection for the overall kpi or a combination of overall and data_quality etc. giving the user a choice for what to run.

This would also help in handling scheduler load, by staggering anomaly detection for different series_type to coincide with live data updation.

Proposed Solution Implementation:

Creation of another field in the anomaly_params dictionary which will contain a series of flags signalling which series_type to run anomaly detection for.