TradExpert: Revolutionizing Trading with Mixture of Expert LLMs

2024-10-16

arXiv.org

AI Summary

This paper addresses the challenge of integrating heterogeneous multi-source financial data—news, market quotes, Alpha factors, and fundamentals—in quantitative trading. We propose the first Mixture-of-Experts (MoE) Large Language Model (LLM) framework tailored for quant finance: four domain-specific expert LLMs process respective data modalities in parallel, while a generalist expert LLM unifies cross-modal representations for joint decision-making, supporting both return prediction and stock ranking tasks. Key contributions include: (1) the first MoE-LLM architecture designed specifically for finance; (2) a switchable dual-task mechanism enabling flexible inference modes; and (3) an open-source, large-scale financial evaluation benchmark. Leveraging multi-source prompt engineering, domain-adaptive fine-tuning, and unified structured/unstructured modeling, our framework achieves statistically significant improvements over monolithic LLMs and state-of-the-art baselines across multiple live-trading benchmarks, demonstrating both the efficacy of cross-modal financial insight fusion and strong real-world deployability.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision

Application Category

• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

The integration of Artificial Intelligence (AI) in the financial domain has opened new avenues for quantitative trading, particularly through the use of Large Language Models (LLMs). However, the challenge of effectively synthesizing insights from diverse data sources and integrating both structured and unstructured data persists. This paper presents TradeExpert, a novel framework that employs a mix of experts (MoE) approach, using four specialized LLMs, each analyzing distinct sources of financial data, including news articles, market data, alpha factors, and fundamental data. The insights of these expert LLMs are further synthesized by a General Expert LLM to make a final prediction or decision. With specific prompts, TradeExpert can be switched between the prediction mode and the ranking mode for stock movement prediction and quantitative stock trading, respectively. In addition to existing benchmarks, we also release a large-scale financial dataset to comprehensively evaluate TradeExpert's effectiveness. Our experimental results demonstrate TradeExpert's superior performance across all trading scenarios.

Problem

• Synthesizing insights from diverse financial data sources
• Integrating structured and unstructured financial data effectively
• Improving stock movement prediction and quantitative trading performance

Innovation

• Mixture of Expert LLMs for diverse data analysis
• General Expert LLM synthesizes insights for decisions
• Switchable prediction and ranking modes for trading