RED-F: Reconstruction-Elimination based Dual-stream Contrastive Forecasting for Multivariate Time Series Anomaly Prediction

TL;DR

Red-F, a Reconstruction-Elimination based Dual-stream Contrastive Forecasting framework, comprising the Reconstruction-Elimination Model (REM) and the Dual-stream Contrastive Forecasting Model (DFM), employs a contrastive forecast that transforms the difficult task of absolute signal detection into a simpler, more robust task of relative trajectory comparison by computing the divergence between these two predictive streams.

Abstract

The proactive prediction of anomalies (AP) in multivariate time series (MTS) is a critical challenge to ensure system dependability. The difficulty lies in identifying subtle anomaly precursors concealed within normal signals. However, existing unsupervised methods, trained exclusively on normal data, demonstrate a fundamental propensity to reconstruct normal patterns. Consequently, when confronted with weak precursors, their predictions are dominated by the normal pattern, submerging the very signal required for prediction. To contend with the limitation, we propose RED-F, a Reconstruction-Elimination based Dual-stream Contrastive Forecasting framework, comprising the Reconstruction-Elimination Model (REM) and the Dual-stream Contrastive Forecasting Model (DFM). The REM utilizes a hybrid time-frequency mechanism to mitigate the precursor, generating a purified, normal-pattern baseline. The DFM then receives this purified baseline and the original sequence which retains the precursor as parallel inputs. At the core of our framework, RED-F employs a contrastive forecast that transforms the difficult task of absolute signal detection into a simpler, more robust task of relative trajectory comparison by computing the divergence between these two predictive streams. This contrastive mechanism serves to amplify the faint precursor signal. Furthermore, the DFM is trained with a novel Multi-Series Prediction (MSP) objective, which leverages distant future context to enhance its predictive sensitivity. Extensive experiments on six real-world datasets demonstrate the superior capability of RED-F in anomaly prediction tasks.

Figures (7)

• Figure 1: Two anomaly precursor patterns exhibited in the SMD dataset. The blue curve represents normal time series values, and the red background indicates the actual anomaly segment. The red dashed box highlights the anomaly precursor: before the anomaly occurs, the time series data already exhibits discernible fluctuations that differ from the historical normal pattern. Fig. \ref{['fig:1a']} illustrates a point anomaly precursor, and Fig. \ref{['fig:1b']} illustrates a series anomaly precursor.
• Figure 2: The architecture of RED-F. It consists of two components: (1) Reconstruction-Elimination Model(REM): Eliminates the subtle fluctuations in the anomaly precursor to obtain its normal pattern; (2) Dual-Stream Contrastive Forecasting Model(DFM): Predicts future anomalies based on the predictive comparison between the anomaly precursor and the normal pattern.
• Figure 3: The architecture of DFM, illustrating how the main model is augmented by auxiliary MSP modules.
• Figure 4: Sensitivity analysis of RED-F's hyperparameters, with input length $L$=196 and prediction horizon $H=$32.
• Figure 5: Efficacy of REM on handling anomaly precursors in the SMD dataset. Blue curves represent the original input, while orange curves are the REM reconstruction. The top row shows a normal window, the bottom row shows a precursor window, where REM effectively suppresses the anomalous fluctuations.