Lab – creating an anomaly detection system
Azure Stream Analytics (ASA) simplifies the process of creating and training custom ML models by incorporating built-in anomaly detection powered by ML. It offers the convenience of performing anomaly detection through straightforward function calls. Two novel unsupervised ML functions have been introduced by Microsoft to identify two prevalent types of anomalies: transient and enduring. These are common anomalies, so you don’t have to create your own detection algorithm but can use the ones provided by Microsoft.
The AnomalyDetection_SpikeAndDip function is designed to pinpoint transient or short-lived anomalies, such as spikes or dips, leveraging the widely recognized kernel density estimation (KDE) algorithm.
On the other hand, the AnomalyDetection_ChangePoint function is employed to identify persistent or long-lasting anomalies, such as bi-level shifts, gradual increases, and gradual decreases. It relies on the established...
