A generalized matrix profile framework with support for contextual series analysis

Paepe, Dieter De; Hautte, Sander Vanden; Steenwinckel, Bram; Turck, Filip De; Ongenae, Femke; Janssens, Olivier; Hoecke, Sofie Van

doi:10.1016/j.engappai.2020.103487

Cited by 15 publications

(12 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Note that existing literature argues that ground truth in some time series are not reliable (De Paepe et al. 2020 ). We have annotated such time series in the experiments, but report results on all of them for completeness.…”

Section: Experimental Evaluationmentioning

confidence: 99%

See 1 more Smart Citation

AURORA: A Unified fRamework fOR Anomaly detection on multivariate time series

Zhang

McNeil

et al. 2021

Data Min Knowl Disc

View full text Add to dashboard Cite

The ability to accurately and consistently discover anomalies in time series is important in many applications. Fields such as finance (fraud detection), information security (intrusion detection), healthcare, and others all benefit from anomaly detection. Intuitively, anomalies in time series are time points or sequences of time points that deviate from normal behavior characterized by periodic oscillations and long-term trends. For example, the typical activity on e-commerce websites exhibits weekly periodicity and grows steadily before holidays. Similarly, domestic usage of electricity exhibits daily and weekly oscillations combined with long-term season-dependent trends. How can we accurately detect anomalies in such domains while simultaneously learning a model for normal behavior? We propose a robust offline unsupervised framework for anomaly detection in seasonal multivariate time series, called AURORA. A key innovation in our framework is a general background behavior model that unifies periodicity and long-term trends. To this end, we leverage a Ramanujan periodic dictionary and a spline-based dictionary to capture both seasonal and trend patterns. We conduct experiments on both synthetic and real-world datasets and demonstrate the effectiveness of our method. AURORA has significant advantages over existing models for anomaly detection, including high accuracy (AUC of up to 0.98), interpretability of recovered normal behavior ( accuracy in period detection), and the ability to detect both point and contextual anomalies. In addition, AURORA is orders of magnitude faster than baselines.

show abstract

Section: Experimental Evaluationmentioning

confidence: 99%

“… Indices marked with asterisk (*) have been reported to have questionable anomaly labels in recent work (De Paepe et al. 2020 ), but we include them for completeness …”

Section: Experimental Evaluationmentioning

confidence: 99%

AURORA: A Unified fRamework fOR Anomaly detection on multivariate time series

Zhang

McNeil

et al. 2021

Data Min Knowl Disc

View full text Add to dashboard Cite

show abstract

“…Zhu et al have suggested a way to calculate the Matrix Profile when the data contains missing values, using knowledge about the range of the data [25]. Lastly, we presented the Contextual Matrix Profile [5] as a generalization of the Matrix Profile that is capable of tracking multiple matches over configurable time spans.…”

Section: Related Workmentioning

confidence: 99%

“…A distance measure that performs a non-linear transformation along the time axis and can ignore the prefix or suffix of sequences being matched, based on Dynamic Time Warping, has been suggested by Furtado Silva et al [6]. Recently, we suggested the Series Distance Matrix framework [5] as a way to easily combine different distance measures with the techniques processing these distances in a plug-and-play way.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Implications of Z-Normalization in the Matrix Profile

Paepe

Avendano

Hoecke

2020

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Companies are increasingly measuring their products and services, resulting in a rising amount of available time series data, making techniques to extract usable information needed. One state-of-theart technique for time series is the Matrix Profile, which has been used for various applications including motif/discord discovery, visualizations and semantic segmentation. Internally, the Matrix Profile utilizes the z-normalized Euclidean distance to compare the shape of subsequences between two series. However, when comparing subsequences that are relatively flat and contain noise, the resulting distance is high despite the visual similarity of these subsequences. This property violates some of the assumptions made by Matrix Profile based techniques, resulting in worse performance when series contain flat and noisy subsequences. By studying the properties of the z-normalized Euclidean distance, we derived a method to eliminate this effect requiring only an estimate of the standard deviation of the noise. In this paper we describe various practical properties of the z-normalized Euclidean distance and show how these can be used to correct the performance of Matrix Profile related techniques. We demonstrate our techniques using anomaly detection using a Yahoo! Webscope anomaly dataset, semantic segmentation on the PAMAP2 activity dataset and for data visualization on a UCI activity dataset, all containing real-world data, and obtain overall better results after applying our technique. Our technique is a straightforward extension of the distance calculation in the Matrix Profile and will benefit any derived technique dealing with time series containing flat and noisy subsequences.

show abstract

Anomaly Detection in Time Series

Borges

Akbarinia

Masséglia

2021

Transactions on Large-Scale Data- And Knowledge-Centered Systems L

View full text Add to dashboard Cite

Data mining has become an important task for researchers in the past few years, including detecting anomalies that may represent events of interest. The problem of anomaly detection refers to discovering the patterns that do not conform to expected behavior. This paper analyzes recent studies on the detection of anomalies in time series. The goal is to provide an introduction to anomaly detection and a survey of recent research and challenges. The article is divided into three main parts. First, the main concepts are presented. Then, the anomaly detection task is defined. Afterward, the main approaches and strategies to solve the problem are presented.

show abstract

A generalized matrix profile framework with support for contextual series analysis

Cited by 15 publications

References 17 publications

AURORA: A Unified fRamework fOR Anomaly detection on multivariate time series

AURORA: A Unified fRamework fOR Anomaly detection on multivariate time series

Implications of Z-Normalization in the Matrix Profile

Anomaly Detection in Time Series

Contact Info

Product

Resources

About