Unsupervised Change Point Detection for heterogeneous sensor signals

TL;DR

This paper reviews unsupervised change point detection methods for heterogeneous sensor signals, emphasizing their flexibility and evaluating their effectiveness across different data sources without requiring labeled training data.

Contribution

It provides a comprehensive examination and comparison of unsupervised change point detection algorithms tailored for diverse sensor data sources.

Findings

Algorithms vary in effectiveness depending on data heterogeneity

Unsupervised methods do not require labeled training data

Evaluation criteria help select suitable algorithms for specific applications

Abstract

Change point detection is a crucial aspect of analyzing time series data, as the presence of a change point indicates an abrupt and significant change in the process generating the data. While many algorithms for the problem of change point detection have been developed over time, it can be challenging to select the appropriate algorithm for a specific problem. The choice of the algorithm heavily depends on the nature of the problem and the underlying data source. In this paper, we will exclusively examine unsupervised techniques due to their flexibility in the application to various data sources without the requirement for abundant annotated training data and the re-calibration of the model. The examined methods will be introduced and evaluated based on several criteria to compare the algorithms.