QuiZSF: An efficient data-model interaction framework for zero-shot time-series forecasting

2025-08-09

AI Summary

Zero-shot forecasting (ZSF) for time series suffers from poor performance under data-scarce conditions, and existing time-series pre-trained models (TSPMs) struggle to dynamically integrate external knowledge. Method: We propose QuiZSF—a unified framework featuring (i) ChronoRAG Base, a hierarchical tree-structured knowledge base enabling scalable, domain-aware retrieval; (ii) a multi-granularity sequence interaction learner and a dual-branch collaborative alignment mechanism, jointly supporting both non-LLM and LLM-based TSPMs; and (iii) retrieval-augmented generation (RAG) for dynamic knowledge injection. Contribution/Results: Extensive experiments across diverse zero-shot forecasting settings demonstrate that QuiZSF achieves state-of-the-art (SOTA) performance on 75% of tasks with non-LLM TSPMs and 87.5% with LLM-based TSPMs, while maintaining high memory efficiency and inference speed.

Technology Category

• Machine Learning: Time-Series/Data Streams
• Data Mining & Knowledge Management: Intelligent Query Processing
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Time series forecasting has become increasingly important to empower diverse applications with streaming data. Zero-shot time-series forecasting (ZSF), particularly valuable in data-scarce scenarios, such as domain transfer or forecasting under extreme conditions, is difficult for traditional models to deal with. While time series pre-trained models (TSPMs) have demonstrated strong performance in ZSF, they often lack mechanisms to dynamically incorporate external knowledge. Fortunately, emerging retrieval-augmented generation (RAG) offers a promising path for injecting such knowledge on demand, yet they are rarely integrated with TSPMs. To leverage the strengths of both worlds, we introduce RAG into TSPMs to enhance zero-shot time series forecasting. In this paper, we propose QuiZSF (Quick Zero-Shot Time Series Forecaster), a lightweight and modular framework that couples efficient retrieval with representation learning and model adaptation for ZSF. Specifically, we construct a hierarchical tree-structured ChronoRAG Base (CRB) for scalable time-series storage and domain-aware retrieval, introduce a Multi-grained Series Interaction Learner (MSIL) to extract fine- and coarse-grained relational features, and develop a dual-branch Model Cooperation Coherer (MCC) that aligns retrieved knowledge with two kinds of TSPMs: Non-LLM based and LLM based. Compared with contemporary baselines, QuiZSF, with Non-LLM based and LLM based TSPMs as base model, respectively, ranks Top1 in 75% and 87.5% of prediction settings, while maintaining high efficiency in memory and inference time.

Problem

• Enhancing zero-shot forecasting with dynamic external knowledge integration
• Overcoming data scarcity in domain transfer and extreme conditions
• Bridging retrieval-augmented generation with time series pre-trained models

Innovation

• Hierarchical tree-structured storage for scalable retrieval
• Multi-grained interaction learner extracting relational features
• Dual-branch model aligning knowledge with TSPMs