Azure Anomaly Detector: Strange Things Are Afoot
Microsoft is not the first company to scan data sets for irregularities (Oracle and Google come to mind), but its Anomaly Detector uses the power of machine learning to make intelligent observations. This is useful for organizations that deal with large amounts of data in time series.
Microsoft’s goal is to make it easy for developers to integrate machine learning functionality into their applications without significant experience in the space. Just as AWS seeks to “make machine learning boring and totally vanilla,” Microsoft’s aim is to democratize access to a powerful tool and leverage its considerable cloud infrastructure to that end.
Anomaly Detector is currently in preview as one of Microsoft’s Cognitive Services. Microsoft bills users per thousand transactions (beyond an initial trial), and promises 99.9% uptime. Anomaly Detector is an API, a stateless service that does not store user data, and is compatible with C#, Java, and Python.
Our Take
Microsoft promises simple, seamless algorithmic analysis of large data sets as part of the democratization of machine learning. Anyone familiar with the Azure console should have little difficulty using the service. Where pattern recognition is important, event management for purposes of fraud detection, anti-money laundering, and proactive problem management as part of ITSM, Anomaly Detector is another arrow in the Microsoft Admin’s quiver.