Forecast2Anomaly (F2A): Adapting Multivariate Time Series Foundation Models for Anomaly Prediction

2025-11-05

AI Summary

To address the challenge of dynamically evolving anomalous patterns in multivariate time series and the poor generalizability of existing methods, this paper proposes a zero-shot anomaly forecasting framework. Methodologically, it (1) adapts a pre-trained time-series foundation model (TSFM) by introducing a joint prediction-anomaly loss function to jointly optimize normal forecasting accuracy and anomaly detection performance; and (2) incorporates a retrieval-augmented generation (RAG) module that enables adaptive distribution shift mitigation without parameter updates. Crucially, the framework requires neither fine-tuning nor labeled anomaly data, enabling cross-system and cross-scenario prior anomaly warnings. Evaluated on 16 real-world datasets, it consistently outperforms state-of-the-art approaches, demonstrating superior generalizability and computational efficiency.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Anomaly/Outlier Detection
• Multiagent Systems: Other Foundations of Multi Agent Systems

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Forecasting anomalies (anomaly prediction) in multivariate time series from different real-world, dynamic, and complex systems is vital for preempting critical failures, leading to a substantial minimization in operational costs and human labor. Yet, existing methods are limited to specific systems while failing to generalize to evolving anomaly patterns over time. In contrast, pretrained Time Series Foundation Models (TSFMs) have recently demonstrated strong generalization and zero-shot forecasting capabilities. However, their potential remains untapped for anomaly prediction, a task fundamentally different from forecasting normal behavior. Thus, we present Forecast2Anomaly (F2A), a novel framework that empowers TSFMs with anomaly prediction abilities through two key innovations. First, we propose a joint forecast-anomaly loss that fine-tunes TSFMs to accurately forecast future signals even at anomalous time points. Second, we introduce a Retrieval-Augmented Generation (RAG) module that retrieves historically relevant horizons and conditions predictions on them. This component dynamically adapts to distributional shifts at inference time, enabling F2A to track evolving anomalies without requiring model updates. By combining targeted fine-tuning with dynamic retrieval, F2A bridges the gap between robust TSFM zero-shot forecasting and zero-shot anomaly prediction. Extensive experiments across 16 diverse datasets and multiple TSFM backbones show that F2A consistently outperforms state-of-the-art methods, offering a scalable, zero-shot anomaly prediction solution for real-world applications.

Problem

• Adapting foundation models for anomaly prediction in multivariate time series
• Enabling zero-shot forecasting to handle evolving anomaly patterns over time
• Bridging the gap between robust forecasting and anomaly detection capabilities

Innovation

• Fine-tunes foundation models with joint forecast-anomaly loss
• Uses retrieval-augmented generation to adapt to distribution shifts
• Enables zero-shot anomaly prediction without model updates