Scholarlink

<p>Automation and intelligent optimization of financial processes are fundamentally transforming the financial industry by enhancing operational accuracy, efficiency, and strategic decision-making. Deep learning models play a pivotal role in&#xa0;automating complex, high-stakes tasks such as invoice processing, payment scheduling, and financial forecasting. However, existing rule-based and static machine learning methods often struggle with dynamic adaptability, real-time optimization, and complex financial scenarios, leading to delayed payments, missed discounts, and suboptimal cash flow management. To overcome these limitations, this paper proposes a novel framework:&#xa0;Deep Reinforcement Learning for Financial Process Optimization (DRL-FPO). The framework trains an intelligent agent that dynamically learns optimal payment policies through continuous interaction with the evolving financial environment.&#xa0;The agent maximizes cumulative rewards by balancing early payment discounts, penalty avoidance, and liquidity preservation. Applied to automated invoice processing and payment scheduling integrated with enterprise resource planning (ERP) systems and real-time data feeds,&#xa0;DRL-FPO demonstrates significant quantitative improvements: a 17.5% increase in payment accuracy, a 22.3% reduction in late payment penalties, and a 34.8% decrease in processing latency compared to traditional methods like CTPN, MHEW, and Q-VMD, which lack dynamic adaptability.&#xa0;These results highlight the framework’s potential to enhance cash flow management, reduce operational costs, and optimize complex financial workflows in volatile corporate environments, thereby delivering substantial business value.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Automation and intelligent optimization of financial processes using deep reinforcement learning and ERP integration

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月24日

查看全文
  • Xinfeng Li,
  • Yameng Bai

摘要

Automation and intelligent optimization of financial processes are fundamentally transforming the financial industry by enhancing operational accuracy, efficiency, and strategic decision-making. Deep learning models play a pivotal role in automating complex, high-stakes tasks such as invoice processing, payment scheduling, and financial forecasting. However, existing rule-based and static machine learning methods often struggle with dynamic adaptability, real-time optimization, and complex financial scenarios, leading to delayed payments, missed discounts, and suboptimal cash flow management. To overcome these limitations, this paper proposes a novel framework: Deep Reinforcement Learning for Financial Process Optimization (DRL-FPO). The framework trains an intelligent agent that dynamically learns optimal payment policies through continuous interaction with the evolving financial environment. The agent maximizes cumulative rewards by balancing early payment discounts, penalty avoidance, and liquidity preservation. Applied to automated invoice processing and payment scheduling integrated with enterprise resource planning (ERP) systems and real-time data feeds, DRL-FPO demonstrates significant quantitative improvements: a 17.5% increase in payment accuracy, a 22.3% reduction in late payment penalties, and a 34.8% decrease in processing latency compared to traditional methods like CTPN, MHEW, and Q-VMD, which lack dynamic adaptability. These results highlight the framework’s potential to enhance cash flow management, reduce operational costs, and optimize complex financial workflows in volatile corporate environments, thereby delivering substantial business value.