Weighted Contrastive Learning for Anomaly-Aware Time-Series Forecasting

2025-12-08

AI Summary

To address the insufficient prediction robustness of multivariate time series models under distributional shifts (e.g., abrupt changes in ATM cash demand), this paper proposes Weighted Contrastive Adaptation (WCA). WCA integrates domain-informed anomaly injection with weighted contrastive learning to jointly optimize representation alignment for both normal and anomalous data within deep time series models—preserving discriminative features of anomalies while maintaining consistency of regular dynamics. Experiments on real-world ATM transaction data demonstrate that WCA reduces SMAPE by 6.1 percentage points under anomaly conditions, with negligible degradation in standard forecasting performance. To the best of our knowledge, this work is the first to introduce weighted contrastive learning into multivariate time series forecasting under anomalous conditions, significantly enhancing model generalization under distributional shift and improving operational utility in real-world financial applications.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Cognitive Modeling & Cognitive Systems: Analogy

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

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.

Problem

• Improves time-series forecasting accuracy during anomalous events.
• Addresses distribution shifts that degrade deep learning model performance.
• Enhances reliability without compromising normal operation forecasting.

Innovation

• Weighted contrastive objective aligns normal and anomaly representations
• Preserves anomaly information while maintaining consistency
• Improves forecasting accuracy under anomalies without degrading normal performance