Explainable anomaly detection on high-dimensional time series data

Rad, Bijan; Song, Fei; Jacob, Vincent; Diao, Yanlei

doi:10.1145/3465480.3468292

Cited by 6 publications

(3 citation statements)

References 27 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…There are several previous approaches to anomaly detection as listed and extracted on "Result and Discussions". However, one crucial difference between some of those approaches and the case we are interested in is that there is no further explanation of the difference between anomaly and extreme observations [38], [39].…”

Section: ) Evaluationmentioning

confidence: 99%

A Systematic Review of Anomaly Detection within High Dimensional and Multivariate Data

Suboh

Aziz

Shaharudin

et al. 2023

JOIV : Int. J. Inform. Visualization

View full text Add to dashboard Cite

In data analysis, recognizing unusual patterns (outliers’ analysis or anomaly detection) plays a crucial role in identifying critical events. Because of its widespread use in many applications, it remains an important and extensive research brand in data mining. As a result, numerous techniques for finding anomalies have been developed, and more are still being worked on. Researchers can gain vital knowledge by identifying anomalies, which helps them make better meaningful data analyses. However, anomaly detection is even more challenging when the datasets are high-dimensional and multivariate. In the literature, anomaly detection has received much attention but not as much as anomaly detection, specifically in high dimensional and multivariate conditions. This paper systematically reviews the existing related techniques and presents extensive coverage of challenges and perspectives of anomaly detection within high-dimensional and multivariate data. At the same time, it provides a clear insight into the techniques developed for anomaly detection problems. This paper aims to help select the best technique that suits its rightful purpose. It has been found that PCA, DOBIN, Stray algorithm, and DAE-KNN have a high learning rate compared to Random projection, ROBEM, and OCP methods. Overall, most methods have shown an excellent ability to tackle the curse of dimensionality and multivariate features to perform anomaly detection. Moreover, a comparison of each algorithm for anomaly detection is also provided to produce a better algorithm. Finally, it would be a line of future studies to extend by comparing the methods on other domain-specific datasets and offering a comprehensive anomaly interpretation in describing the truth of anomalies.

show abstract

Section: ) Evaluationmentioning

confidence: 99%

A Systematic Review of Anomaly Detection within High Dimensional and Multivariate Data

Suboh

Aziz

Shaharudin

et al. 2023

JOIV : Int. J. Inform. Visualization

View full text Add to dashboard Cite

show abstract

“…There are also DNN-specific post-hoc methods such as DeepLift [20] and IntegratedGradients [21], which leverage the gradients of DNN outputs with respect to the input features. While these posthoc models were originally developed for supervised DNNs, recent works have extended them to the AE and its probabilistic variant, the variational autoencoder (VAE) [22,23,24,25,26,27].…”

Section: Explainable Deep Learningmentioning

confidence: 99%

Coalitional Bayesian autoencoders: Towards explainable unsupervised deep learning with applications to condition monitoring under covariate shift

Yong

Brintrup

2022

Applied Soft Computing

View full text Add to dashboard Cite

“…Explainability for time series can be done by using dimension reduction (see e.g. [RSJD21]) or deep learning methods (see e.g. [AGP + 20]).…”

mentioning

confidence: 99%

Explainable multi-class anomaly detection on functional data

Cura¹,

Firdova²,

Labart³

et al. 2022

Preprint

View full text Add to dashboard Cite

In this paper we describe an approach for anomaly detection and its explainability in multivariate functional data. The anomaly detection procedure consists of transforming the series into a vector of features and using an Isolation forest algorithm. The explainable procedure is based on the computation of the SHAP coefficients and on the use of a supervised decision tree. We apply it on simulated data to measure the performance of our method and on real data coming from industry.

show abstract

Explainable anomaly detection on high-dimensional time series data

Cited by 6 publications

References 27 publications

A Systematic Review of Anomaly Detection within High Dimensional and Multivariate Data

A Systematic Review of Anomaly Detection within High Dimensional and Multivariate Data

Coalitional Bayesian autoencoders: Towards explainable unsupervised deep learning with applications to condition monitoring under covariate shift

Explainable multi-class anomaly detection on functional data

Contact Info

Product

Resources

About