Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

FinArena: A Human-Agent Collaboration Framework for Financial Market Analysis and Forecasting

Congluo Xu, Zhaobin Liu, Ziyang Li

TL;DR

FinArena presents a mixture-of-experts–inspired human–agent framework for financial analysis that fuses multimodal data (time-series, news, and financial statements) through specialized LLM agents and a universal expert, guided by investors' risk preferences. It introduces an uncertainty-driven adaptive Retrieval-Augmented Generation module for news data and an iterative reasoning process for financial statements, integrated by a Report Agent for human-in-the-loop decisions. Empirical results show improved stock-movement prediction and promising trading simulations in the US market, with some limitations in the A-share market due to information-disclosure and data-quality issues. The work demonstrates the practical potential of personalized, multimodal AI-assisted investing for retail users and provides a publicly available small-scale dataset to foster further research.

Abstract

To improve stock trend predictions and support personalized investment decisions, this paper proposes FinArena, a novel Human-Agent collaboration framework. Inspired by the mixture of experts (MoE) approach, FinArena combines multimodal financial data analysis with user interaction. The human module features an interactive interface that captures individual risk preferences, allowing personalized investment strategies. The machine module utilizes a Large Language Model-based (LLM-based) multi-agent system to integrate diverse data sources, such as stock prices, news articles, and financial statements. To address hallucinations in LLMs, FinArena employs the adaptive Retrieval-Augmented Generative (RAG) method for processing unstructured news data. Finally, a universal expert agent makes investment decisions based on the features extracted from multimodal data and investors' individual risk preferences. Extensive experiments show that FinArena surpasses both traditional and state-of-the-art benchmarks in stock trend prediction and yields promising results in trading simulations across various risk profiles. These findings highlight FinArena's potential to enhance investment outcomes by aligning strategic insights with personalized risk considerations.

FinArena: A Human-Agent Collaboration Framework for Financial Market Analysis and Forecasting

TL;DR

FinArena presents a mixture-of-experts–inspired human–agent framework for financial analysis that fuses multimodal data (time-series, news, and financial statements) through specialized LLM agents and a universal expert, guided by investors' risk preferences. It introduces an uncertainty-driven adaptive Retrieval-Augmented Generation module for news data and an iterative reasoning process for financial statements, integrated by a Report Agent for human-in-the-loop decisions. Empirical results show improved stock-movement prediction and promising trading simulations in the US market, with some limitations in the A-share market due to information-disclosure and data-quality issues. The work demonstrates the practical potential of personalized, multimodal AI-assisted investing for retail users and provides a publicly available small-scale dataset to foster further research.

Abstract

To improve stock trend predictions and support personalized investment decisions, this paper proposes FinArena, a novel Human-Agent collaboration framework. Inspired by the mixture of experts (MoE) approach, FinArena combines multimodal financial data analysis with user interaction. The human module features an interactive interface that captures individual risk preferences, allowing personalized investment strategies. The machine module utilizes a Large Language Model-based (LLM-based) multi-agent system to integrate diverse data sources, such as stock prices, news articles, and financial statements. To address hallucinations in LLMs, FinArena employs the adaptive Retrieval-Augmented Generative (RAG) method for processing unstructured news data. Finally, a universal expert agent makes investment decisions based on the features extracted from multimodal data and investors' individual risk preferences. Extensive experiments show that FinArena surpasses both traditional and state-of-the-art benchmarks in stock trend prediction and yields promising results in trading simulations across various risk profiles. These findings highlight FinArena's potential to enhance investment outcomes by aligning strategic insights with personalized risk considerations.

Paper Structure

This paper contains 26 sections, 15 equations, 4 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Financial Modeling based on Historical Stock Price
  4. LLM Agents for Financial Market Analysis
  5. Discussion on the Existing Mutil-Agent LLM Systems
  6. Framework
  7. Framework Overview
  8. Generative Time Series Agent
  9. Uncertainty-Driven Adaptive News Agent
  10. Iterative Reasoning Agent for Financial Statement Analysis
  11. Information Aggregation System and Human-Agent Collaboration
  12. Dataset
  13. Stock Price Data
  14. Financial News
  15. Financial Statements
  16. ...and 11 more sections

Figures (4)

  • Figure 1: The Framework of FinArena
  • Figure 2: The Comparison of Different RAG Application Strategy
  • Figure 3: The Results of Ablation Experiment. Using RAG or not are marked with orange and gray respectively, and the dotted line is used to connect the two and display the difference between them. The triangle's dots represent the average performance of all companies.
  • Figure 4: An example of the process of pre-processing news data using LLM, with the red box displaying the "bias information" present in the news data and the blue box indicating the modification commands given by LLM.