Benchmarking Inductive Biases for Multivariate Time-Series Anomaly Detection with a Robust Multi-View Channel-Graph Detector

2026-05-27

AI Summary

This study addresses the lack of a unified benchmark for multivariate time series anomaly detection, which hinders systematic evaluation of methods in terms of effectiveness, robustness, and generalization. The authors establish a comprehensive experimental framework evaluating ten representative detectors across five datasets and propose an adaptive multi-view detection method that uniquely integrates a NOTEARS-constrained directed channel graph, block-wise attention mechanisms, and temporal correlation views. Their contributions include the first benchmark that jointly considers effectiveness, efficiency, robustness, and cross-dataset generalization; the identification of three empirical insights, including the absence of a universally dominant method; and a novel approach that achieves a macro-averaged VUS-ROC of 0.675—outperforming the next-best method by 5.1 percentage points, consistently ranking among the top three across all datasets, and demonstrating superior robustness under noise, channel dropout, and time-shift perturbations.

Abstract

We present a unified experiment, analysis, and benchmark study of multivariate time-series (MTS) anomaly detection. Ten family-representative detectors -- spanning statistical, reconstruction, association, frequency, and generic-transformer families -- are evaluated on five datasets (SMD, MSL, SMAP, PSM, and MSDS) under effectiveness, efficiency, robustness, and cross-dataset generalisation. All methods share the same windowing, scoring, hardware, and metric protocols. Effectiveness, ablation, and robustness use three random seeds; cross-dataset transfer uses seed~0 because each extra seed requires $250$ source-target evaluations. The benchmark yields three method-independent findings: no single-bias baseline dominates; absolute perturbation VUS-ROC is more informative than retention ratios; and MSDS behaves as an event-dense deployment workload rather than a sparse point-anomaly benchmark. Under this protocol we also introduce \ours{}, an adaptive detector family combining a NOTEARS-constrained directed channel-graph view with optional patch-attention and temporal-association views. \ours{} achieves the best macro-average VUS-ROC ($0.675$, $+5.1$~pt over the second-best LSTM-AE), ranks first overall, and reaches the top-3 on all five datasets. Its wins on MSL and MSDS are narrow, while its average and robustness gains are larger: under the same three-seed robustness protocol for every method, it obtains the strongest absolute VUS-ROC across noise, channel dropout, and time-shift perturbations. We release the MSDS preprocessing protocol, configurations, scripts, and seed-level metric dumps.

Problem

• multivariate time-series
• anomaly detection
• inductive biases
• benchmarking
• robustness

Innovation

• channel-graph
• NOTEARS
• multi-view anomaly detection
• robustness benchmarking
• multivariate time-series