Labels Matter More Than Models: Quantifying the Benefit of Supervised Time Series Anomaly Detection

TL;DR

The paper tackles time-series anomaly detection under limited labeling, arguing that model complexity is less critical than leveraging anomaly labels. It introduces STAND, a simple supervised baseline, and conducts the first dedicated benchmark comparing supervised versus unsupervised TSAD methods. Across five datasets, results show that even minimal supervision yields substantial gains over state-of-the-art unsupervised approaches, with STAND delivering better prediction consistency and anomaly localization. The work advocates a data-centric shift in TSAD research and provides open-source code to support broader evaluation and deployment.

Abstract

Time series anomaly detection (TSAD) is a critical data mining task often constrained by label scarcity. Consequently, current research predominantly focuses on Unsupervised Time-series Anomaly Detection (UTAD), relying on complex architectures to model normal data distributions. However, this approach often overlooks the significant performance gains available from limited anomaly labels achievable in practical scenarios. This paper challenges the premise that architectural complexity is the optimal path for TSAD. We conduct the first methodical comparison between supervised and unsupervised paradigms and introduce STAND, a streamlined supervised baseline. Extensive experiments on five public datasets demonstrate that: (1) Labels matter more than models: under a limited labeling budget, simple supervised models significantly outperform complex state-of-the-art unsupervised methods; (2) Supervision yields higher returns: the performance gain from minimal supervision far exceeds that from architectural innovations; and (3) Practicality: STAND exhibits superior prediction consistency and anomaly localization compared to unsupervised counterparts. These findings advocate for a data-centric shift in TSAD research, emphasizing label utilization over purely algorithmic complexity. The code is publicly available at https://github.com/EmorZz1G/STAND.

Figures (14)

• Figure 1: Classification of time series anomaly detection approaches.
• Figure 2: Multi-metric performance comparison across different categories of time series anomaly detection methods.
• Figure 3: The overall architecture of the STAND framework. An input time series is first projected into latent space by the feature embedding. The resulting latent sequence is then fed into the temporal encoding module, which captures bidirectional temporal context. Finally, an anomaly scoring module outputs anomaly scores for each time step.
• Figure 4: Sensitivity analysis of model performance with respect to three key hyperparameters on five datasets. (a)-(b) show the impact of model dimension ($d_{\text{model}}$) on CCE and VUS-PR, respectively; (c)-(d) illustrate the effect of the number of TEM layers; (e)-(f) depict the influence of window size.
• Figure 5: Sensitivity analysis of model performance with respect to three key hyperparameters on four subsets of the PSM dataset. (a)-(b) show the impact of model dimension ($d_{\text{model}}$) on CCE and VUS-PR, respectively; (c)-(d) illustrate the effect of the number of TEM layers; (e)-(f) depict the influence of window size.

Theorems & Definitions (3)

• Definition 1: Time Series Anomaly Detection
• Definition 2: Unsupervised Time-series Anomaly Detection (UTAD)
• Definition 3: Supervised Time-series Anomaly Detection (STAD)