基于多尺度的多变量时间序列异常检测模型

doi:10.11772/j.issn.1001-9081.2025030302

摘要/Abstract

摘要：

多变量时间序列数据常表现出多尺度特征和复杂的相互依赖性，给异常检测带来了挑战。为了解决这一问题，提出一种基于多尺度的多变量时间序列异常检测模型M3AD（Multi-scale Mamba Multi-layer perceptron Anomaly Detection）。首先，采用多尺度特征提取方法，即通过在不同时间尺度上分割时间序列，捕捉短期和长期模式；其次，利用多层感知机（MLP）和卷积层进行特征学习，提取局部和高级特征表示；再次，引入选择性状态空间模型（SSM） Mamba，通过它高效的处理能力捕捉长序列中的关键信息；最后，通过基于KL （Kullback-Leibler）散度的损失函数和异常分数计算，实现对异常的敏感检测。为了验证模型的有效性，将M3AD与Anomaly Transformer和MEMTO等7种模型在4个公共数据集上对比。实验结果表明，M3AD在精确率、召回率和F1分数等关键指标上相较于对比方法展现出显著优势和领先性能，验证了M3AD在多变量时间序列异常检测任务中的有效性和优越性。

关键词: 异常检测, 多尺度特征提取, 时间序列, 状态空间模型, 无监督学习

Abstract:

Multivariate time series data often exhibit multi-scale characteristics and complex interdependencies， which makes challenges for anomaly detection. To address this issue， a multi-scale based multivariate time series anomaly detection model was developed， namely M3AD （Multi-scale Mamba Multi-layer perceptron Anomaly Detection）. Firstly， a multi-scale feature extraction method was employed， which means segmenting time series across different temporal scales to capture short- and long-term patterns. Secondly， a Multi-Layer Perceptron （MLP） and convolutional layers were utilized for feature learning for extracting local and high-level feature representations. Thirdly， a selective State Space Model （SSM） Mamba was introduced to capture key information in long sequences through its efficient processing capability. Finally， sensitive anomaly detection was achieved through a loss function based on KL （Kullback-Leibler） divergence and anomaly score calculation. To validate the model’s effectiveness， M3AD was compared with seven models， such as Anomaly Transformer and MEMTO， on four public datasets. Experimental results show that M3AD has significant advantages and leading performance compared to the other methods in key metrics such as precision， recall， and F1 score， verifying the effectiveness and superiority of M3AD in multivariate time series anomaly detection tasks.

Key words: anomaly detection, multi-scale feature extraction, time series, State Space Model (SSM), unsupervised learning

中图分类号:

TP391.4

尹春勇, 张不凡. 基于多尺度的多变量时间序列异常检测模型[J]. 计算机应用, 2026, 46(3): 790-797.

Chunyong YIN, Bufan ZHANG. Multi-scale based multivariate time series anomaly detection model[J]. Journal of Computer Applications, 2026, 46(3): 790-797.

图/表 8

参考文献 35

[1]	HILAL W， GADSDEN S A， YAWNEY J. Financial fraud： a review of anomaly detection techniques and recent advances［J］. Expert Systems with Applications， 2022， 193： No.116429.
[2]	SUN Y， CHEN T， NGUYEN Q V H， et al. TinyAD： memory-efficient anomaly detection for time-series data in industrial IoT ［J］. IEEE Transactions on Industrial Informatics， 2024， 20（1）： 824-834.
[3]	TORABI H， MIRTAHERI S L， GRECO S. Practical autoencoder based anomaly detection by using vector reconstruction error［J］. Cybersecurity， 2023， 6： No.1.
[4]	PARK D， HOSHI Y， KEMP C C. A multimodal anomaly detector for robot-assisted feeding using an LSTM-based variational autoencoder ［J］. IEEE Robotics and Automation Letters， 2018， 3（3）： 1544-1551.
[5]	DU B， SUN X， YE J， et al. GAN-based anomaly detection for multivariate time series using polluted training set［J］. IEEE Transactions on Knowledge and Data Engineering， 2023， 35（12）： 12208-12219.
[6]	LI D， CHEN D， JIN B， et al. MAD-GAN： multivariate anomaly detection for time series data with generative adversarial networks［C］// Proceedings of the 2019 International Conference on Artificial Neural Networks， LNCS 11730. Cham： Springer， 2019： 703-716.
[7]	GU Y， JAZIZADEH F. Time series anomaly detection using generative adversarial network discriminators and density estimation for infrastructure systems［J］. Automation in Construction， 2024， 165： No.105500.
[8]	GU A， DAO T. Mamba： linear-time sequence modeling with selective state spaces ［EB/OL］. ［2024-12-25］. .
[9]	LIU Y， TIAN Y， ZHAO Y， et al. VMamba： visual state space model［C］// Proceedings of the 38th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2024： 103031-103063.
[10]	AHAMED M A， CHENG Q. TimeMachine： a time series is worth 4 mambas for long-term forecasting ［C］// Proceedings of the 27th European Conference on Artificial Intelligence. Amsterdam： IOS Press， 2024： 1688-1695.
[11]	LENZ B， LIEBER O， ARAZI A， et al. Jamba： hybrid Transformer-Mamba language models ［EB/OL］. ［2024-12-13］..
[12]	LI K， CHEN G， YANG R， et al. SPMamba： state-space model is all you need in speech separation ［EB/OL］. ［2024-12-15］..
[13]	SHUMWAY R H， STOFFER D S. ARIMA models［M］// 4th ed. Time series analysis and its applications： with R examples， STS. Cham： Springer， 2017： 75-163.
[14]	KAYASTHA N， THOMAS V， GALBRAITH J， et al. Monitoring wetland change using inter-annual landsat time-series data［J］. Wetlands， 2012， 32： 1149-1162.
[15]	MA J， PERKINS S. Time-series novelty detection using one-class support vector machines［C］// Proceedings of the 2003 International Joint Conference on Neural Networks. Piscataway： IEEE， 2003： 1741-1745.
[16]	QIN Y， LOU Y. Hydrological time series anomaly pattern detection based on isolation forest［C］// Proceedings of the IEEE 3rd Information Technology， Networking， Electronic and Automation Control Conference. Piscataway： IEEE， 2019： 1706-1710.
[17]	YIN C， ZHANG S， WANG J， et al. Anomaly detection based on convolutional recurrent autoencoder for IoT time series ［J］. IEEE Transactions on Systems， Man， and Cybernetics： Systems， 2022， 52（1）： 112-122.
[18]	LIN S， CLARK R， BIRKE R， et al. Anomaly detection for time series using VAE-LSTM hybrid model ［C］// Proceedings of the 2020 IEEE International Conference on Acoustics， Speech and Signal Processing. Piscataway： IEEE， 2020： 4322-4326.
[19]	WANG Y， DU X， LU Z， et al. Improved LSTM-based time-series anomaly detection in rail transit operation environments［J］. IEEE Transactions on Industrial Informatics， 2022， 18（12）： 9027-9036.
[20]	XU J， WU H， WANG J， et al. Anomaly Transformer： time series anomaly detection with association discrepancy ［EB/OL］. ［2024-12-11］..
[21]	YANG Y， ZHANG C， ZHOU T， et al. DCdetector： dual attention contrastive representation learning for time series anomaly detection［C］// Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2023： 3033-3045.
[22]	DENG A， HOOI B. Graph neural network-based anomaly detection in multivariate time series［C］// Proceedings of the 35th AAAI Conference on Artificial Intelligence. Palo Alto： AAAI Press， 2021： 4027-4035.
[23]	段雪源，付钰，王坤. 基于VAE-WGAN的多维时间序列异常检测方法［J］. 通信学报， 2022， 43（3）： 1-13.
	DUAN X Y， FU Y， WANG K. Multi-dimensional time series anomaly detection method based on VAE-WGAN［J］. Journal on Communications， 2022， 43（3）： 1-13.
[24]	PHAM T A， LEE J H， PARK C S. MST-VAE： multi-scale temporal variational autoencoder for anomaly detection in multivariate time series［J］. Applied Sciences， 2022， 12（19）： No.10078.
[25]	DARBAN Z Z， WEBB G I， PAN S， et al. CARLA： self-supervised contrastive representation learning for time series anomaly detection［J］. Pattern Recognition， 2025， 157： No.110874.
[26]	KIM J， KANG H， KANG P. Time-series anomaly detection with stacked Transformer representations and 1D convolutional network［J］. Engineering Applications of Artificial Intelligence， 2023， 120： No.105964.
[27]	WANG Z， PEI C， MA M， et al. Revisiting VAE for unsupervised time series anomaly detection： a frequency perspective［C］// Proceedings of the ACM Web Conference 2024. New York： ACM， 2024： 3096-3105.
[28]	GU A， GOEL K， RÉ C. Efficiently modeling long sequences with structured state spaces ［EB/OL］. ［2024-11-25］..
[29]	HUNDMAN K， CONSTANTINOU V， LAPORTE C， et al. Detecting spacecraft anomalies using LSTMs and nonparametric dynamic thresholding ［C］// Proceedings of the 24th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2018： 387-395.
[30]	ABDULAAL A， LIU Z， LANCEWICKI T. Practical approach to asynchronous multivariate time series anomaly detection and localization［C］// Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2021： 2485-2494.
[31]	LI Z， ZHAO Y， HAN J， et al. Multivariate time series anomaly detection and interpretation using hierarchical inter-metric and temporal embedding ［C］// Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2021： 3220-3230.
[32]	SONG J， KIM K， OH J， et al. MEMTO： memory-guided transformer for multivariate time series anomaly detection［C］// Proceedings of the 37th International Conference on Neural Information Processing Systems. Red Hook： Curran Associates Inc.， 2023： 57947-57963.
[33]	SU Y， ZHAO Y， NIU C， et al. Robust anomaly detection for multivariate time series through stochastic recurrent neural network［C］// Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. New York： ACM， 2019： 2828-2837.
[34]	ZHOU B， LIU S， HOOI B， et al. BeatGAN： anomalous rhythm detection using adversarially generated time series［C］// Proceedings of the 28th International Joint Conference on Artificial Intelligence. California： IJCAI.org， 2019： 4433-4439.
[35]	XIAO C， GOU Z， TAI W， et al. Imputation-based time-series anomaly detection with conditional weight-incremental diffusion models ［C］// Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining. New York： ACM， 2023： 2742-2751.

数据集	样本数		异常比/%	特征数
数据集	训练集	测试集	异常比/%	特征数
MSL	46 653	73 729	10.5	55
PSM	105 984	26 497	27.8	25
SMAP	108 146	27 037	12.8	25
SWaT	495 000	449 919	12.1	51

数据集	样本数		异常比/%	特征数
数据集	训练集	测试集	异常比/%	特征数
MSL	46 653	73 729	10.5	55
PSM	105 984	26 497	27.8	25
SMAP	108 146	27 037	12.8	25
SWaT	495 000	449 919	12.1	51

参数	值	参数	值
初始学习率	0.000 1	编码器层数	6
批处理大小	64	时间窗口大小	105
异常比率	0.01	补丁大小	（3，5，7）

参数	值	参数	值
初始学习率	0.000 1	编码器层数	6
批处理大小	64	时间窗口大小	105
异常比率	0.01	补丁大小	（3，5，7）

模型	MSL			SMAP			SWaT			PSM
模型	P	R	F1	P	R	F1	P	R	F1	P	R	F1
BeatGAN^［34］	89.75	85.42	87.53	92.38	55.85	69.61	64.01	87.46	73.92	90.30	93.84	92.04
LSTM-VAE^［4］	85.49	79.94	82.62	92.20	67.75	78.10	76.00	89.50	82.20	73.62	89.92	80.96
OmniAnomaly^［33］	89.02	86.37	87.67	92.49	81.99	86.92	81.42	84.30	82.83	88.39	74.46	80.83
AnomalyTrans^［20］	92.09	95.15	93.59	94.13	99.40	96.69	91.55	96.73	94.07	96.91	98.90	97.89
InterFusion^［31］	81.28	92.70	86.62	89.77	88.52	89.14	80.59	85.58	83.01	83.61	83.45	83.52
MEMTO^［32］	92.07	96.76	94.36	93.76	99.63	96.61	94.18	97.54	95.83	97.46	99.23	98.34
DiffAD^［35］	92.23	95.34	93.75	96.39	97.02	96.70	94.08	96.78	95.41	96.87	98.79	97.82
M3AD	92.10	97.73	94.83	93.92	99.70	96.72	94.33	99.87	97.02	97.56	98.85	98.20