Stationarity-Aware Retrieval-Augmented Time Series Forecasting

2026-06-02

AI Summary

Real-world time series often exhibit non-stationarity and abrupt regime shifts, which undermine the effectiveness of conventional similarity-based retrieval approaches for forecasting. To address this challenge, this work proposes SARAF, a novel framework that introduces stationarity awareness into retrieval-augmented prediction. SARAF first constructs a candidate segment pool based on temporal similarity, then adaptively modulates the strength of diversity selection according to the underlying data stationarity, and finally integrates retrieved segments via a stationarity-aware fusion strategy. By dynamically balancing relevance and diversity, the method achieves significantly superior performance over strong baselines across eight real-world datasets, with particularly notable gains in prediction accuracy and robustness under non-stationary conditions.

Abstract

Time series forecasting relies on historical patterns, but real-world series often exhibit non-stationarity and regime shifts that challenge fully parametric forecasters. Inspired by Retrieval-Augmented Generation (RAG), recent work augments forecasters by retrieving relevant historical segments and using them as external evidence at inference time. However, due to the intrinsic non-stationarity of real-world time series, a highly similar past segment does not necessarily imply a similar future, rendering similarity-only retrieval brittle and prone to redundancy. We propose Stationarity-Aware Retrieval-Augmented Time Series Forecasting (SARAF), a framework that adaptively balances relevance and diversity in retrieval. SARAF first forms a candidate pool via temporal similarity with time-aligned enhancement, then applies a diversity-aware selection strategy to cover heterogeneous historical regimes, with the diversification strength automatically modulated by dataset-level stationarity. Moreover, SARAF uses stationarity-aware aggregation to fuse the retrieved futures. Extensive experiments on eight real-world datasets show that SARAF achieves competitive forecasting performance and improves average accuracy and robustness over strong baselines, with particularly clear benefits under challenging non-stationary settings. Code: https://github.com/ShiqiaoZhou/SARAF.

Problem

• non-stationarity
• regime shifts
• time series forecasting
• retrieval-augmented generation
• historical similarity

Innovation

• Retrieval-Augmented Forecasting
• Non-stationary Time Series
• Diversity-aware Retrieval
• Stationarity-aware Aggregation
• Regime Shift