CoLDM: Componentwise Lead–Lag-Aware Dependence Modeling for Multivariate Time-Series Anomaly Detection in IoT
IEEE Internet of Things Journal · 01.06.2026 · 13 (11) · 24792-24805
DOI: 10.1109/jiot.2026.3674919 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Internet of Things Journal, erschienen am 01.06.2026. Verfasst von Yaqi Fan und 5 weiteren Autor:innen. Die Arbeit nennt 60 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Beijing Key Laboratory of Multimodal Data Intelligent Perception and Governance.
Abstract (Originalsprache)
In Internet of Things (IoT) and industrial IoT (IIoT) systems, interconnected sensors continuously generate nonstationary multivariate time series (MTS), where intervariable dependencies vary across operating regimes. Consequently, anomalies may manifest not only as pointwise deviations but also as abnormal changes in dependence patterns, especially lead–lag interactions induced by propagation delays and system responses. Reliably detecting such anomalies requires accurately modeling time-varying intervariable dependencies. However, existing methods mainly focus on instantaneous correlations or lag-agnostic dependence learning in the feature space, thereby overlooking lead–lag-aware dependence structures which are crucial for detecting anomalies in nonstationary environments. We propose CoLDM, an unsupervised reconstruction-based framework that learns lead–lag-aware intervariable dependence representations for MTS anomaly detection (AD). CoLDM first extracts statistical lead–lag cues and converts them into a soft structural prior, which is injected via structural preference modulation (SPM) to guide dependence learning. Moreover, CoLDM adopts a componentwise dual-branch architecture to capture complementary lead–lag dependence information, and introduces a component-adaptive latent fusion module that adaptively integrates the two representations across regimes, reducing redundant overlap and improving robustness. Experiments on real-world benchmarks demonstrate that CoLDM generalizes well and remains robust under nonstationary regimes.
Förderung
- National Natural Science Foundation of China
- Beijing Key Laboratory of Multimodal Data Intelligent Perception and Governance
Zitieren
APA
Fan, Y., Lin, W., Chen, K., Wang, X., Yang, J., & Qiao, Y. (2026). CoLDM: Componentwise Lead–Lag-Aware Dependence Modeling for Multivariate Time-Series Anomaly Detection in IoT. IEEE Internet of Things Journal, 13(11), 24792-24805. https://doi.org/10.1109/jiot.2026.3674919
Vancouver
Fan Y, Lin W, Chen K, Wang X, Yang J, Qiao Y. CoLDM: Componentwise Lead–Lag-Aware Dependence Modeling for Multivariate Time-Series Anomaly Detection in IoT. IEEE Internet of Things Journal. 2026;13(11):24792-24805. doi:10.1109/jiot.2026.3674919
BibTeX
@article{Fan2026,
author = {Fan, Yaqi and Lin, Wenhui and Chen, Kexin and Wang, Xiangchao and Yang, Jie and Qiao, Yuanyuan},
title = {CoLDM: Componentwise Lead–Lag-Aware Dependence Modeling for Multivariate Time-Series Anomaly Detection in IoT},
journal = {IEEE Internet of Things Journal},
year = {2026},
volume = {13},
number = {11},
pages = {24792--24805},
doi = {10.1109/jiot.2026.3674919}
}
EUVIMED