[Paper Review] Enhancing Anomaly Detection in Financial Markets with an LLM-based Multi-Agent Framework

Taejin Park|arXiv (Cornell University)|Mar 28, 2024

Complex Systems and Time Series Analysis6 citations

TL;DR

The paper proposes an LLM-based multi-agent framework to validate and interpret financial market anomalies, demonstrated on the S&P 500 data to reduce manual verification.

ABSTRACT

This paper introduces a Large Language Model (LLM)-based multi-agent framework designed to enhance anomaly detection within financial market data, tackling the longstanding challenge of manually verifying system-generated anomaly alerts. The framework harnesses a collaborative network of AI agents, each specialised in distinct functions including data conversion, expert analysis via web research, institutional knowledge utilization or cross-checking and report consolidation and management roles. By coordinating these agents towards a common objective, the framework provides a comprehensive and automated approach for validating and interpreting financial data anomalies. I analyse the S&P 500 index to demonstrate the framework's proficiency in enhancing the efficiency, accuracy and reduction of human intervention in financial market monitoring. The integration of AI's autonomous functionalities with established analytical methods not only underscores the framework's effectiveness in anomaly detection but also signals its broader applicability in supporting financial market monitoring.

Motivation & Objective

• Motivate and address the manual verification bottleneck in anomaly alerts from market data.
• Propose a collaborative multi-agent AI system that validates, explains, and consolidates anomaly findings.
• Demonstrate the framework with a S&P 500 time series to show improvements in efficiency and accuracy.
• Highlight how AI autonomy can complement traditional analysis methods in market monitoring.

Proposed method

• Convert tabular anomaly data into LLM-friendly formats with metadata to enable context-rich processing.
• Deploy specialized data expert agents (web research, institutional knowledge, cross-checking) to validate anomalies via diverse sources.
• Consolidate expert analyses into a summary report through a dedicated reporting agent.
• Facilitate management discussion among domain agents to derive strategic recommendations.
• Present a human-facing briefing with final decisions made by a human analyst.
• Use a multi-agent workflow to automate anomaly validation from detection through reporting.

Experimental results

Research questions

• RQ1 How can an LLM-based multi-agent system improve validation and interpretation of detected anomalies in financial market data?
• RQ2 To what extent can autonomous agents reduce human intervention while preserving analytical rigor in anomaly workflows?
• RQ3 How does the framework perform when applied to real-world financial series (e.g., S&P 500) with historical anomalies?

Key findings

• The framework enhances efficiency, accuracy, and reduces human intervention in anomaly validation.
• Applied to the S&P 500, the system identifies notable outliers and confirms context with historical events.
• Web research and institutional knowledge agents provide corroborating evidence for anomalies.
• Cross-checking reveals that some previously labeled null values are actually data changes, improving data integrity.
• Management and summary agents produce a consolidated assessment ready for human decision-making.
• The approach demonstrates potential for broader applicability in financial market monitoring.