Anomaly Detection Principles and Algorithms by Kishan G. Mehrotra Chilukuri K. Mohan & HuaMing Huang

Author:Kishan G. Mehrotra, Chilukuri K. Mohan & HuaMing Huang
Language: eng
Format: epub
Publisher: Springer International Publishing, Cham

6.1 Distance from the Rest of the Data

The simplest anomaly detection algorithms are based on the assumptions about the data distribution, e.g., that data is one-dimensional and normally distributed with a mean of and standard deviation of σ. A large distance from the center of the distribution implies that the probability of observing such a data point is very small. Since there is only a low probability of observing such a point (drawn from that distribution), a data point at a large distance from the center is considered to be an anomaly. More specifically, in such simple cases, anomaly detection algorithms may rely on the fact that, as z increases, the number of data points found at a distance of zσ away from decreases rapidly. For instance, only about 0.1% of the data points exceed , and this can be used to justify the following well-known criterion for detecting an outlier: If a data point is z (typically z = 3) or more standard deviations away from the arithmetic mean, then it is an outlier.

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.