Discussing the article: "Neural Networks in Trading: Piecewise Linear Representation of Time Series"

New comment
 

Check out the new article: Neural Networks in Trading: Piecewise Linear Representation of Time Series.

This article is somewhat different from my earlier publications. In this article, we will talk about an alternative representation of time series. Piecewise linear representation of time series is a method of approximating a time series using linear functions over small intervals.

Time series anomaly detection is a major subfield of time series data mining. Its purpose is to identify unexpected behavior throughout the data set. Since anomalies are often caused by different mechanisms, there are no specific criteria for their detection. In practice, data that exhibits expected behavior tends to attract more attention, while anomalous data is often perceived as noise which is usually ignored or eliminated. However, anomalies can contain useful information and thus detection of such anomalies can be important. Accurate anomaly detection can help mitigate unnecessary adverse impacts in various fields such as environment, industry, finance and others.

Anomalies in time series can be divided into the following three categories:

  1. Point anomalies: a data point is considered to be anomalous relative to other data points. These anomalies are often caused by measurement errors, sensor failures, data entry errors, or other exceptional events;
  2. Contextual anomalies: a data point is considered anomalous in a certain context, but not otherwise;
  3. Collective anomalies: a subsequence of a time series that exhibits anomalous behavior. This is quite a difficult task because such anomalies cannot be considered anomalous when analyzed individually. Instead, it is the collective behavior of the group that is anomalous.

Collective anomalies can provide valuable information about the system or process being analyzed, as they may indicate a group-level problem that needs to be addressed. Thus, detecting collective anomalies can be an important task in many fields such as cybersecurity, finance, and healthcare. The authors of the BPLR method focused in their work on identifying collective anomalies.

Author: Dmitriy Gizlyk

New comment