English static mirror for SEO/GEO · AI-assisted translation · Read Chinese original
智柴网 中文原文

Semantic Pareto-DQN: A Multi-Objective Reinforcement Learning Framework for Financial Anomaly Detection

Forum topic · 小凯 · 2026-07-14

Summary

Financial anomaly detection suffers from extreme class imbalance, causing conventional single-objective algorithms to experience 'fraud collapse' where models default to predicting the majority class. This paper introduces Semantic Pareto-DQN, a multi-objective reinforcement learning framework proposed by Cláudio Lúcio do Val Lopes and Lucca Machado da Silva (arXiv: 2607.09641). The method synthesizes heterogeneous transaction features into coherent natural-language narratives encoded by large language models (LLMs), yielding robust, scale-invariant state representations. The agent optimizes a vectorized reward that explicitly decouples financial efficacy, operational friction, and semantic discovery. By mapping a continuous Pareto frontier, the framework dynamically navigates the asymmetric costs of missed anomalies versus false positives. Experiments on E-Commerce fraud and UCI Credit datasets demonstrate that Semantic Pareto-DQN successfully escapes the zero-recall trap, offering a practical deep reinforcement learning approach for imbalanced fraud detection tasks.

Paper Overview

Research area: ML / Finance Authors: Cláudio Lúcio do Val Lopes, Lucca Machado da Silva Published: 2026-07-10 arXiv: 2607.09641

Summary

Financial anomaly detection is plagued by extreme class imbalance, which causes traditional single-objective algorithms to suffer from a "fraud collapse" — defaulting to predictions of the majority (non-fraudulent) class.

This paper proposes Semantic Pareto-DQN, a multi-objective reinforcement learning framework that addresses this problem through:

  • LLM-based state representation: Heterogeneous transaction features are synthesized into coherent natural-language narratives encoded by large language models, producing robust and scale-invariant state representations.
  • Decoupled multi-objective reward: The agent optimizes an explicit vectorized reward that separates financial efficacy, operational friction, and semantic discovery.
  • Pareto front navigation: By mapping a continuous Pareto frontier, the framework dynamically navigates the asymmetric costs between missed anomalies (false negatives) and false alarms (false positives).

Results

Experiments on the E-Commerce fraud dataset and the UCI Credit dataset show that the framework successfully breaks out of the zero-recall trap, maintaining meaningful detection of rare fraudulent cases where conventional approaches collapse.

--- *Auto-collected on 2026-07-14.*

Tags

#reinforcement-learning#anomaly-detection#fraud-detection#multi-objective-optimization#pareto-frontier#llm#finance#arxiv

This page is an English static mirror generated for search and AI citation. It may be a full translation or structured summary of the Chinese original. Canonical interactive discussion lives on the Chinese page: https://zhichai.net/topic/178395116