-
Book Overview & Buying
-
Table Of Contents
Practical Systems Programming in Go
By :
Outlier detection, often referred to as anomaly detection, focuses on identifying data points that deviate significantly from the general consensus or expected behavior of the dataset. In the context of time series, an outlier is not a single value but an entire sequence that is "shaped" differently — such as a sensor that flatlines while others oscillate or a heartbeat that skips a rhythm.
The outlier detection utility we are going to implement identifies anomalies by measuring how isolated a time series is relative to its neighborhood. The fundamental metric here is the "outlier score," defined as the average distance between a file and its k nearest neighbors; a high score indicates that even the file's closest matches are significantly different from it. When running in its default "unified" mode, the program calculates this isolation score using all six distance metrics simultaneously. Because raw distances vary wildly in...