Why does the Time Series Anomaly Detection module uses Martingale? RRS feed

  • Question

  • Hello, 

    for my research I wanted to find out the best way of finding anomalies in time series with machine learning methods using Microsoft Azure Machine Learning Studio. Therefore I read a lot about "Anomaly Detection", "Anomaly Detection and Machine Learning", "Time Series", "Time Series and Anomaly Detection" and so on. 

    In nearly none of the papers I have read about a connection between "Anomaly Detection" or "Time series analysis" and Martingales, except for the Azure ML Studio module for time series anomaly detection. So now my question: 

    Why does Microsoft Azure ML Studio uses the Martingale framework for anomaly detection? And how does this exactly work?

    I also read the white paper given on the modul reference page, but I don't get the transfer from martingales (originally beloning to the class of betting strategies) to anomaly detection in time series. 

    I hope somebody can help me! :-) 


    Friday, November 1, 2019 1:16 PM

All replies

  • Hi Sarah,

    I have moved this thread to the right place for help. Thanks.



    Friday, November 1, 2019 11:01 PM