1.1 - 8.10 - ChangePointDetectionRT (ML Engine) - Teradata Vantage

Teradata Vantage™ - Machine Learning Engine Analytic Function Reference

Teradata Vantage
Release Number
October 2019
Content Type
Programming Reference
Publication ID
English (United States)

Real-time change-point detection typically uses the sliding window algorithm and assumes that the data follows a distribution, but has different parameters θ in different segments. In the two following hypotheses about the parameter,

H 0:θ = θ 0

H 1:θ = θ 1

the reference part has parameter θ 0 and the testing part in the sliding window has parameter θ 1, with the following notation:

Notation used by Machine Learning Engine function ChangePointDetectionRT

This is the decision rule for testing the sliding window:

If S j k < h 0, choose H 0.

If S j k h 0, choose H 1.

The ChangePointDetectionRT function detects change points in a stochastic process or time series, using real-time change-point detection, implemented with these algorithms:

  • Search algorithm: sliding window
  • Segmentation algorithm: normal distribution

Use this function when the input data cannot be stored in memory, or when the application requires a real-time response. If the input data can be stored in memory and the application does not require a real-time response, use the function ChangePointDetection (ML Engine).