Time-point encoding improves anomaly detection in multivariate time series
TENET: Time-point Encoding Network for Multivariate Time Series Anomaly Detection
Machine LearningArtificial Intelligence
Summary
Detecting unusual patterns in complex time-based data can be tricky because small changes might look like errors when they are not. The authors present TENET, a new approach that looks at the current moment in context with its surroundings to better spot real problems. Instead of trying to recreate or predict the whole data segment, TENET creates a summary focused on the current point and scores how normal it looks. Tests show TENET works better than other methods on standard datasets.
What this means in practice
- •For industrial monitoring teams: Monitor complex sensor data streams to better identify abnormal system behavior without false alarms from local variations.
- •For network operations centers: Improve detection of unusual network activity by modeling time series data with current-point focused latent representations.
Authors
Dongchan Cho, Juwon Hwang, Keumyeong Kang, Jiho Han, Namsoon Jung
Abstract
In unsupervised multivariate time series anomaly detection (MTSAD), local variation can inflate reconstruction and forecasting errors, making observation-space discrepancy an unreliable indicator of abnormal system behavior. Since latent-space scoring can mitigate this limitation, we reframe one-class MTSAD as current-point system-state compatibility estimation and propose TENET, the Time-point Encoding Network for Multivariate Time Series Anomaly Detection. We define the current-point system state as a latent representation of the target time point, constructed from variable-level temporal context in the input window. TENET constructs this state through a temporal layer that retrieves window context relative to the current point, encodes time with current-point-centered embeddings, and adaptively mixes the retrieved context with the current representation. With this design, TENET treats the input window not as the object to reconstruct, forecast, summarize, or score, but as context for constructing a latent representation evaluated with a fixed standard-normal negative log-likelihood score. Across benchmark datasets, TENET achieves state-of-the-art performance. Additional experiments further show that TENET remains effective under alternative one-class objectives and outperforms substitutions based on generic sequence encoders, suggesting that its advantage comes from decision-aligned representation construction rather than from a specialized scoring rule. These results indicate that one-class MTSAD can be strong when the latent representation is designed around the current time point.