Understanding the logic behind anomaly detection

The algorithm behind the anomaly detection feature in Power BI has been developed by Microsoft and is designed for real-time time-series data in many different applications. The purpose of this section is not to go over all technical details but instead to keep it pragmatic. In this section, we will explore the key considerations of this algorithm that help us understand how to use the feature in Power BI appropriately. For more in-depth information, you can read the paper Time-Series Anomaly Detection Service at Microsoft by Ren et al., 2019.

The algorithms behind Microsoft's anomaly detection feature

First of all, the algorithm created by Microsoft is a combination of two methods and is referred to as SR-CNN. SR stands for Spectral Residual and CNN stands for Convolutional Neural Network. Both of these deep learning (DL) methods are most often used for analyzing images, and it's the combination of these two techniques...