Multivariate Anomaly Detection - Training the Model

What is multivariate anomaly detection?

Multivariate anomaly detection is the advanced form of anomaly detection where the machine learning algorithm needs to learn multiple features that can cause an anomaly in the system. It means that there is a dependency on more than one feature which can cause an anomaly. All the features are associated with a particular timestamp. For example, to determine the anomalies in an environment of a room, multiple features could be involved like temperature, humidity, oxygen level, and so on.

Implementation

We need to follow the steps mentioned in the below illustration to implement multivariate anomaly detection with Azure anomaly detector service.

Let’s make sure we understand the implementation steps displayed in the illustration above:

• Step 1: We need to prepare the dataset. Now, in this lesson, we’ll consider a dataset containing the temperature and the humidity values associated with the timestamp. Then, we need to create the ZIP file and upload it to Azure Blob Storage and generate a SAS URL for the ZIP file.
• Step 2: We’ll use the SAS URL as our data source and submit a request to train a multivariate anomaly detection model to Azure anomaly detector service.
• Step 3: We’ll check the status of whether the model is finished training and ready to be consumed.
• Step 4: We’ll use the trained model to detect the anomalies in

...