Robust Anomaly Detection for Multivariate Time Series through Stochastic Recurrent Neural Network
The DOI matches its Crossref record; 29 of the 30 references it cites were found in a registry or an index.
Abstract
Industry devices (i.e., entities) such as server machines, spacecrafts, engines, etc., are typically monitored with multivariate time series, whose anomaly detection is critical for an entity's service quality management. However, due to the complex temporal dependence and stochasticity of multivariate time series, their anomaly detection remains a big challenge. This paper proposes OmniAnomaly, a stochastic recurrent neural network for multivariate time series anomaly detection that works well robustly for various devices. Its core idea is to capture the normal patterns of multivariate time series by learning their robust representations with key techniques such as stochastic variable connection and planar normalizing flow, reconstruct input data by the representations, and use the reconstruction probabilities to determine anomalies. Moreover, for a detected entity anomaly, OmniAnomaly can provide interpretations based on the reconstruction probabilities of its constituent univariate time series. The evaluation experiments are conducted on two public datasets from aerospace and a new server machine dataset (collected and released by us) from an Internet company. OmniAnomaly achieves an overall F1-Score of 0.86 in three real-world datasets, signicantly outperforming the best performing baseline method by 0.09. The interpretation accuracy for OmniAnomaly is up to 0.89.
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“The representation learning method based on VAE can model the distribution of each time step to better capture the complex spatiotemporal dependencies and provide better interpretability in time series modeling tasks.”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [81]
In the citing paper
“Note that in time series anomaly detection task, the adjustment strategy is widely adopted: if a time point in a certain successive abnormal segment is detected, all anomalies in this abnormal segment are viewed to be correctly detected [81], [126], [163].”
From Self-Supervised Learning for Time Series Analysis: Taxonomy, Progress, and Prospects (2024) · cites this paper as [81]
In the citing paper
“Present a comprehensive evaluation of the TVGeAN model using four widely cited MTS benchmarks for supervised and unsupervised learning tasks: C-MPASS [35], SMD [36], MSL, and SMAP [37].”
From TVGeAN: Tensor Visibility Graph-Enhanced Attention Network for Versatile Multivariant Time Series Learning Tasks (2024) · cites this paper as [36]
Citations
Cites
Auto-Encoding Variational Bayes
Kingma, D. P., & Welling, M. (2013) · arXiv · doi:10.48550/arxiv.1312.6114 · Cited by 20,067 · cited 5×
“Stochastic Gradient Variational Bayes (SGVB) [7] is a variational inference algorithm often used in VAE to train the parameters φ and θ by maximizing the evidence of lower bound (ELBO), L(xt):”
BackgroundMethodInfluentialEmpirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling
Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1412.3555 · Cited by 11,008 · cited 5×
“Simple RNN could fail to learn the long-term dependence in a sequence [1].”
BackgroundMethodInfluentialMonte Carlo Statistical Methods
Kim, H. (2000) · Technometrics · doi:10.1080/00401706.2000.10485722 · Cited by 5,508
Monte Carlo Statistical Methods
Robert, C. P., & Casella, G. (2004) · Springer Texts in Statistics · doi:10.1007/978-1-4757-4145-2 · Cited by 4,271
Detecting Spacecraft Anomalies Using LSTMs and Nonparametric Dynamic Thresholding
Hundman, K., Constantinou, V., Laporte, C., Colwell, I., & Soderstrom, T. (2018) · Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3219819.3219845 · Cited by 1,771 · cited 9×
“Mars Science Laboratory rover [6] 27 55 power, computational”
BackgroundMethodInfluentialParallel Networks that Learn to Pronounce English Text
B, T. J. S. A. (1987) · Cited by 1,514
Variational Inference with Normalizing Flows
Rezende, D. J., & Mohamed, S. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1505.05770 · Cited by 1,253 · cited 6×
“K) are invertible mapping functions (f k (zk−1 t ) = zk−1 t + utanh(wTzk−1 t + b), u, w and b are the parameters) [18].”
BackgroundMethodInfluentialA Multimodal Anomaly Detector for Robot-Assisted Feeding Using an LSTM-Based Variational Autoencoder
Park, D., Hoshi, Y., & Kemp, C. C. (2018) · IEEE Robotics and Automation Letters · doi:10.1109/lra.2018.2801475 · Cited by 1,115
A survey on unsupervised outlier detection in high‐dimensional numerical data
Zimek, A., Schubert, E., & Kriegel, H. (2012) · Statistical Analysis and Data Mining The ASA Data Science Journal · doi:10.1002/sam.11161 · Cited by 864 · cited 1×
“However, despite the rich literature in multivariate time series anomaly detection in different areas [4, 6, 11, 16, 17, 20, 26, 27], previous studies either take deterministic approaches [4, 6, 11] to model time series, or take stochastic approach but ignore the temporal dependence of observations in the time series…”
BackgroundLSTM-based Encoder-Decoder for Multi-sensor Anomaly Detection
Malhotra, P., Ramakrishnan, A., Anand, G., Vig, L., Agarwal, P., & Shroff, G. (2016) · arXiv · doi:10.48550/arxiv.1607.00148 · Cited by 819 · cited 9×
“To demonstrate the effectiveness of OmniAnomaly, we first compare it with four state-of-the-art unsupervised approaches for multivariate time series anomaly detection: LSTM with nonparametric dynamic thresholding (LSTM-NDT for short) [6], EncDec-AD [11], DAGMM [27], and LSTM-VAE [16], which have been described in Sect…”
BackgroundMethodInfluentialUnsupervised Anomaly Detection via Variational Auto-Encoder for Seasonal KPIs in Web Applications
Xu, H., Feng, Y., Chen, J., Wang, Z., Qiao, H., Chen, W., Zhao, N., Li, Z., Bu, J., Li, Z., et al. (2018) · Proceedings of the 2018 World Wide Web Conference on World Wide Web - WWW '18 · doi:10.1145/3178876.3185996 · Cited by 808 · cited 10×
“[24] shows that univariate time series in an online shopping website can present complex temporal relationships.”
BackgroundMethodInfluentialSelf-citationSmoothness Priors Analysis of Time Series
Kitagawa, G., & Gersch, W. (1996) · Lecture Notes in Statistics · doi:10.1007/978-1-4612-0761-0 · Cited by 646
A Recurrent Latent Variable Model for Sequential Data
Chung, J., Kastner, K., Dinh, L., Goel, K., Courville, A. C., & Bengio, Y. (2015) · arXiv (Cornell University) · doi:10.48550/arxiv.1506.02216 · Cited by 635 · cited 1×
“As suggested by [2], for sequential data modeling, in order to well represent the input data, it is beneficial to include information coming from zt−1 as part of zt.”
MethodAnomaly Detection in Streams with Extreme Value Theory
Siffer, A., Fouque, P.-A., Termier, A., & Largouet, C. (2017) · Proceedings of the 23rd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/3097983.3098144 · Cited by 553 · cited 17×
“Peaks-Over-Threshold (POT) [22] is the second theorem in EVT.”
BackgroundMethodInfluentialGeneric and Scalable Framework for Automated Time-series Anomaly Detection
Laptev, N., Amizadeh, S., & Flint, I. (2015) · Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2783258.2788611 · Cited by 425 · cited 2×
“Industry devices, such as server machines [9, 14], spacecrafts [6], robot-assisted systems [16, 17], engines [11], are typically monitored with multiple time series metrics (also called telemetry data, sensor data, etc.”
BackgroundFast Memory-efficient Anomaly Detection in Streaming Heterogeneous Graphs
Manzoor, E., Milajerdi, S. M., & Akoglu, L. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939783 · Cited by 291 · cited 1×
“1 INTRODUCTION Anomaly detection has been an active research topic in SIGKDD community with applications in graph [3, 12], log messages [15, 23], time series [6, 9, 14, 22], etc.”
BackgroundOn top-k recommendation using social networks
Yang, X., Steck, H., Guo, Y., & Liu, Y. (2012) · Proceedings of the sixth ACM conference on Recommender systems · doi:10.1145/2365952.2365969 · Cited by 162 · cited 1×
“Since there exist no established metrics to evaluate the interpretability of entity anomalies, motivated by the idea of HitRate@K for recommender systems [25], we define a new metric HitRate@P%= Hit@ ⌊P%×|GTt | ⌋ |GTt | , where|GTt | is the length of GTt and P can be 100 or 150.”
BackgroundSequential Neural Models with Stochastic Layers
Fraccaro, M., Sønderby, S. K., Paquet, U., & Winther, O. (2016) · Research at the University of Copenhagen (University of Copenhagen) · Cited by 157
Anomaly Detection Using Program Control Flow Graph Mining From Execution Logs
Nandi, A., Mandal, A., Atreja, S., Dasgupta, G. B., & Bhattacharya, S. (2016) · Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining · doi:10.1145/2939672.2939712 · Cited by 154 · cited 1×
“1 INTRODUCTION Anomaly detection has been an active research topic in SIGKDD community with applications in graph [3, 12], log messages [15, 23], time series [6, 9, 14, 22], etc.”
BackgroundSpotLight: Detecting Anomalies in Streaming Graphs
Eswaran, D., Faloutsos, C., Guha, S., & Mishra, N. (2018) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3219819.3220040 · Cited by 145
Filonov, P., Lavrentyev, A., & Vorontsov, A. (2016) · arXiv · doi:10.48550/arxiv.1612.06676 · Cited by 140 · cited 1×
“However, despite the rich literature in multivariate time series anomaly detection in different areas [4, 6, 11, 16, 17, 20, 26, 27], previous studies either take deterministic approaches [4, 6, 11] to model time series, or take stochastic approach but ignore the temporal dependence of observations in the time series…”
BackgroundCorrelating events with time series for incident diagnosis
Chen, L., Lou, J., Lin, Q., Fu, Q., Ding, R. F., Zhang, D., & Wang, Z. (2014) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/2623330.2623374 · Cited by 118 · cited 1×
“1 Problem Statement A time series contains successive observations which are usually collected at equal-space timestamps [10].”
BackgroundFailure detection in assembly: Force signature analysis
Rodriguez, A., Bourne, D., Mason, M. T., Rossano, G. F., & Wang, J. (2010) · IEEE International Conference on Automation Science and Engineering (CASE) · doi:10.1109/coase.2010.5584452 · Cited by 93 · cited 2×
“However, despite the rich literature in multivariate time series anomaly detection in different areas [4, 6, 11, 16, 17, 20, 26, 27], previous studies either take deterministic approaches [4, 6, 11] to model time series, or take stochastic approach but ignore the temporal dependence of observations in the time series…”
BackgroundMethodA multimodal execution monitor with anomaly classification for robot-assisted feeding
Park, D., Kim, H., Hoshi, Y., Erickson, Z., Kapusta, A. S., & Kemp, C. C. (2017) · IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) · doi:10.1109/iros.2017.8206437 · Cited by 61 · cited 6×
“Industry devices, such as server machines [9, 14], spacecrafts [6], robot-assisted systems [16, 17], engines [11], are typically monitored with multiple time series metrics (also called telemetry data, sensor data, etc.”
BackgroundMethodInfluentialLearning a Hierarchical Monitoring System for Detecting and Diagnosing Service Issues
Nair, V. B., Raul, A., Khanduja, S., Bahirwani, V., Shao, Q., Sellamanickam, S., Keerthi, S., Herbert, S., & Dhulipalla, S. (2015) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/2783258.2788624 · Cited by 61 · cited 2×
“Industry devices, such as server machines [9, 14], spacecrafts [6], robot-assisted systems [16, 17], engines [11], are typically monitored with multiple time series metrics (also called telemetry data, sensor data, etc.”
BackgroundEnhanced Telemetry Monitoring with Novelty Detection
Heras, J. A. M., & Donati, A. (2014) · AI Magazine · doi:10.1609/aimag.v35i4.2553 · Cited by 40 · cited 1×
“Supervised learning methods [17, 20] need labeled data for model training and can only identify anomaly occurrences for known anomaly types [13].”
MethodStructural Event Detection from Log Messages
Wu, F. Q., Anchuri, P., & Li, Z. (2017) · ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) · doi:10.1145/3097983.3098124 · Cited by 26 · cited 1×
“1 INTRODUCTION Anomaly detection has been an active research topic in SIGKDD community with applications in graph [3, 12], log messages [15, 23], time series [6, 9, 14, 22], etc.”
BackgroundLinear Gaussian State Space Modeling
Kitagawa, G., & Gersch, W. (1996) · Lecture notes in statistics · doi:10.1007/978-1-4612-0761-0_5 · Cited by 13 · cited 9×
“Inspired by [5], to explicitly model temporal dependence among stochastic variables in the latent space, we propose the stochastic variable connection technique: Linear Gaussian State Space Model [8] connection between stochastic variables, and the concatenation of stochastic variable and GRU latent variable.”
MethodInfluential
Research works printed but not linked: 3
- Fraccaro Marco, 2016, Ulrich Paquet, and Ole Winther.
- Diederik, 2014, Proceedings of the Second International Conference on Learning Representations (ICLR
- Zong Bo, 2018, International Conference on Learning Representations.
Cited by
A Review on Outlier/Anomaly Detection in Time Series Data
Blázquez-García, A., Conde, A., Mori, U., & Lozano, J. A. (2022) · ACM Computing Surveys · doi:10.1145/3444690 · Cited by 1,102
“an outlier. Given that the results directly vary depending on that selection and few techniques give an automatic way to determine the threshold [Hundman et al. 2018;Lu et al. 2018;Siffer et al. 2017;Su et al. 2019], an interesting future line of research would be to deepen on the dynamic and adaptive selection of thr…”
BackgroundTranAD: deep transformer networks for anomaly detection in multivariate time series data
Tuli, S., Casale, G., & Jennings, N. R. (2022) · Proceedings of the VLDB Endowment · doi:10.14778/3514061.3514067 · Cited by 1,063
“The OmniAnomaly, CAE-M and MSCRED models use sequential observations as input, allowing these methods to retain the temporal information.”
BackgroundMethodInfluentialUSAD: UnSupervised Anomaly Detection on Multivariate Time Series
Audibert, J., Michiardi, P., Guyard, F., Marti, S., & Zuluaga, M. A. (2020) · Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining · doi:10.1145/3394486.3403392 · Cited by 1,094
“On top, the results obtained with SWaT and WADI datasets are presented, whereas the bottom part of the table reports obtained results from the benchmark proposed by [17], using three remaining datasets.”
BackgroundMethodResultInfluentialMultivariate Time-Series Anomaly Detection via Graph Attention Network
Zhao, H., Wang, Y., Duan, J., Huang, C., Cao, D., Tong, Y., Xu, B., Bai, J., Tong, J., & Zhang, Q. (2020) · 2020 IEEE International Conference on Data Mining (ICDM) · doi:10.1109/icdm50108.2020.00093 · Cited by 743
“Following the evaluation strategy in [4], we treat the whole segment as correct if any observation in this segment is detected as anomaly correctly.”
BackgroundMethodInfluentialDeep Learning for Time Series Anomaly Detection: A Survey
Zamanzadeh Darban, Z., Webb, G. I., Pan, S., Aggarwal, C., & Salehi, M. (2025) · ACM Computing Surveys · doi:10.1145/3691338 · Cited by 613
“2) LSTM-VAE [143] T Un P OmniAnomaly [163] T Un W ✓ Point + Subseq ✓ STORN [159] ST Un P Point GGM-VAE [76] T Un W Subseq SISVAE [114] T Un W Point VAE-GAN [138] T Semi W Point VELC [186] T Un TopoMAD [81] ST Un W Subseq ✓ PAD [34] T Un W Subseq InterFusion [119] ST Un W ✓ Subseq MT-RVAE* [177] ST Un W RDSMM [115] T U…”
Influential
Metadata
- Authors
- Su, Y., Zhao, Y., Niu, C., Liu, R., Sun, W., & Pei, D.
- Affiliations
- Tsinghua University (CN)
- Stevens Institute of Technology (US)
- Beijing University of Posts and Telecommunications (CN)
- Venue
- Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining
- Publisher
- ACM
- Published
- 2019-07-25
- Pages
- 2828-2837
- OpenAlex
- W2950361482
- Cited by
- 1,804
- References
- 30
- Citations by year
- 2026: 369 · 2025: 458 · 2024: 388 · 2023: 283 · 2022: 172 · 2021: 99 · 2020: 33 · 2019: 2
- FWCI
- 29.27
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD) (publishedVersion)
- Funders
- Okawa Research Grant
- Beijing National Research Center for Information Science and Technology (BNRist)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Anomaly Detection Techniques and Applications
- Physical Sciences › Computer Science › Computer Networks and Communications › Software System Performance and Reliability
- Physical Sciences › Engineering › Control and Systems Engineering › Fault Detection and Control Systems
- Keywords
- multivariate time series anomaly detection, stochastic recurrent neural networks, reconstruction-based anomaly detection, interpretability
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.