Change detection

From Wikipedia, the free encyclopedia

In statistical analysis, change detection tries to identify changes in the probability distribution of a stochastic process^[1]. More generally it also includes the detection of anomalous behavior. Change detection test are often used in manufacturing (quality control), intrusion detection, spam filtering, and medical diagnostics.

Using the sequential analysis ("online") approach, any change test must make a trade-off between these common metrics:

• False alarm rate
• Misdetection rate
• Detection delay

Easily-detectable changes may be diagnosed using SPRT or CUSUM. More subtle changes—sometimes intentionally so, to evade detection—may be diagnosed using commensurately more sophisticated algorithms^[2].

Offline algorithms typically employ clustering based on maximum likelihood estimation.

There are also heuristic algorithms to detect change, specific to the problem being studied.

[edit] See also

• Receiver operating characteristic
• Recall rate
• Hypothesis testing
• Detection theory
• Quick detection

[edit] Notes and references

1. ^ A specific application may be concerned with changes in the mean, variance, correlation, or spectral density
2. ^ Tarem Ahmed, Mark Coates, and Anukool Lakhina (May, 2007). "Multivariate Online Anomaly Detection Using Kernel Recursive Least Squares". INFOCOM '07: 625-633. doi:10.1109/INFCOM.2007.79. 

• Michèle Basseville and Igor V. Nikiforov (April 1993). Detection of Abrupt Changes: Theory and Application. Prentice-Hall, Englewood Cliffs, N.J.. ISBN 0-13-126780-9.