Learning approaches for anomaly detection

In this section, we'll look at the most popular and straightforward methods we can use for anomaly detection.

Detecting anomalies with statistical tests

Statistical tests are usually used to catch extreme values for individual features. The general name for this type of test is extreme-value analysis. An example of such a test is the use of the Z-score measure:

Here, is a sample from the dataset, µ is the mean of all samples from the dataset, and is the standard deviation of samples in the dataset. Z-values are placed in the interval [-1, 1] and the smallest values that are close to zero are the most normal or expected ones. The following graph shows which values from...