Retrieval-augmented Large Language Models for Financial Time Series Forecasting

2025-02-09

AI Summary

Existing financial time-series retrieval methods struggle to precisely identify key factors influencing stock price movements, thereby limiting forecasting accuracy. To address this, we propose the first Retrieval-Augmented Generation (RAG) framework tailored for financial time-series forecasting, which jointly models textual semantics and numerical dynamics. Our approach introduces FinSeer—a domain-specific retriever for financial data—alongside an LLM-feedback-driven candidate sequence filtering mechanism and a training objective that maximizes historical significance-aware sequence similarity. Furthermore, we release a high-quality, human-annotated dataset integrating financial indicators and stock prices. Evaluated on the BIGDATA22 benchmark using a fine-tuned 1B-parameter StockLLM, our method achieves an 8% absolute accuracy improvement over strong baselines and random retrieval. It significantly enhances identification of critical temporal patterns while robustly suppressing noise interference.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Machine Learning: Time-Series/Data Streams
• Natural Language Processing: Generation

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
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Stock movement prediction, a fundamental task in financial time-series forecasting, requires identifying and retrieving critical influencing factors from vast amounts of time-series data. However, existing text-trained or numeric similarity-based retrieval methods fall short in handling complex financial analysis. To address this, we propose the first retrieval-augmented generation (RAG) framework for financial time-series forecasting, featuring three key innovations: a fine-tuned 1B parameter large language model (StockLLM) as the backbone, a novel candidate selection method leveraging LLM feedback, and a training objective that maximizes similarity between queries and historically significant sequences. This enables our retriever, FinSeer, to uncover meaningful patterns while minimizing noise in complex financial data. We also construct new datasets integrating financial indicators and historical stock prices to train FinSeer and ensure robust evaluation. Experimental results demonstrate that our RAG framework outperforms bare StockLLM and random retrieval, highlighting its effectiveness, while FinSeer surpasses existing retrieval methods, achieving an 8% higher accuracy on BIGDATA22 and retrieving more impactful sequences. This work underscores the importance of tailored retrieval models in financial forecasting and provides a novel framework for future research.

Problem

• Enhance stock movement prediction accuracy
• Retrieve critical factors from financial data
• Develop retrieval-augmented framework for forecasting

Innovation

• Fine-tuned 1B parameter StockLLM
• LLM feedback-based candidate selection
• Maximized query-sequence similarity training