Retrieval-augmented Large Language Models for Financial Time Series Forecasting
Mengxi Xiao, Zihao Jiang, Lingfei Qian, Zhengyu Chen, Yueru He, Yijing Xu, Yuecheng Jiang, Dong Li, Ruey-Ling Weng, Min Peng, Jimin Huang, Sophia Ananiadou, Qianqian Xie
2025-02-12
Summary
This paper talks about a new way to predict stock prices using advanced AI technology. The researchers created a system called FinSeer that combines a powerful language model with a smart way of finding relevant historical data to make better predictions about stock movements.
What's the problem?
Predicting stock prices is really hard because there's so much information to consider, and traditional methods aren't great at handling all this complex financial data. Current systems either focus too much on text or on numbers, but they struggle to put it all together effectively.
What's the solution?
The researchers developed FinSeer, which uses three main tricks to improve predictions. First, they used a super-smart AI that understands language really well. Second, they created a clever way to pick out the most important historical information. Third, they trained the system to find patterns that are most similar to what's happening right now in the stock market. They also made new datasets with both financial numbers and historical stock prices to train and test their system.
Why it matters?
This matters because better stock predictions can help investors make smarter decisions and potentially earn more money. It's not just about stocks though - this kind of technology could be used to predict other financial trends too. By showing that their system works better than existing methods, the researchers have opened up new possibilities for using AI in finance, which could lead to more accurate forecasts and better financial planning in the future.
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.