Weighted Contrastive Learning for Anomaly-Aware Time-Series Forecasting

TL;DR

This paper introduces WECA, a weighted contrastive learning method that improves multivariate time-series forecasting accuracy during anomalies by aligning normal and anomalous data representations, ensuring robustness without sacrificing normal performance.

Contribution

The paper proposes WECA, a novel weighted contrastive loss that enhances anomaly-aware forecasting by preserving relevant information during distribution shifts.

Findings

WECA improves SMAPE by 6.1 percentage points on anomaly-affected data.

WECA maintains performance on normal data with negligible degradation.

The method demonstrates robustness in real-world ATM transaction datasets.

Abstract

Reliable forecasting of multivariate time series under anomalous conditions is crucial in applications such as ATM cash logistics, where sudden demand shifts can disrupt operations. Modern deep forecasters achieve high accuracy on normal data but often fail when distribution shifts occur. We propose Weighted Contrastive Adaptation (WECA), a Weighted contrastive objective that aligns normal and anomaly-augmented representations, preserving anomaly-relevant information while maintaining consistency under benign variations. Evaluations on a nationwide ATM transaction dataset with domain-informed anomaly injection show that WECA improves SMAPE on anomaly-affected data by 6.1 percentage points compared to a normally trained baseline, with negligible degradation on normal data. These results demonstrate that WECA enhances forecasting reliability under anomalies without sacrificing performance during regular operations.