Intelligent Construction of a Supply Chain Finance Decision Support System and Financial Benefit Analysis Based on Deep Reinforcement Learning and Particle Swarm Optimization Algorithm

Abstract

With the continuous development of global supply chain finance, leveraging advanced artificial intelligence technologies to enhance the intelligence of decision support systems has become a key focus in both academia and industry. This paper aims to construct a supply chain finance decision support system based on deep reinforcement learning and the particle swarm optimization (PSO) algorithm. By effectively capturing the dynamic characteristics of supply chain finance data, optimizing model parameters, and improving decision-making accuracy and response speed, this study further evaluates the system's impact on enhancing corporate financial benefits. Starting with the theoretical foundation of supply chain finance and decision support systems, this paper analyzes the prevalent challenges in the field, such as inaccurate decision-making, slow response times, and insufficient model robustness. Next, it provides a detailed discussion on the application of artificial intelligence in financial decision-making, outlining the fundamental principles and core algorithms of deep reinforcement learning (e.g., DQN, DDPG) and the advantages and implementation mechanisms of PSO in parameter optimization. The paper also systematically reviews the integrated application of these two approaches. The proposed system architecture consists of four main components: data collection and preprocessing, decision-making modules, model optimization, and a feedback mechanism. The decision-making module utilizes deep reinforcement learning to construct a dynamic decision model based on state-action-reward principles, enabling real-time learning of key supply chain nodes and precise identification of risks and opportunities in complex financial environments. PSO is embedded in the model optimization process to perform global search and adaptive tuning of deep reinforcement learning hyperparameters, ensuring optimal model performance across various data scenarios. To validate the effectiveness of the proposed approach, experiments were conducted using real-world and simulated supply chain finance data. Key evaluation metrics included decision accuracy, response time, risk prediction accuracy, and corporate financial performance indicators (such as cost reduction rate, profit growth rate, and liquidity improvement). The results were compared with those of traditional decision support methods. Experimental results demonstrate that the proposed decision support system, integrating deep reinforcement learning and PSO, exhibits significant advantages in capturing dynamic supply chain finance data, optimizing decision-making strategies, and mitigating risks. The system effectively enhances corporate financial performance and operational efficiency. Additionally, through deployment and feedback analysis in real-world applications, this study explores areas for improvement in data quality, real-time response, and model generalization capabilities. Future research directions, such as multi-algorithm collaboration and cross-domain data integration, are also proposed. In summary, this study validates the effective application of deep reinforcement learning and PSO in supply chain finance decision support systems through system construction and empirical analysis. The research highlights the potential impact of intelligent decision-making on corporate financial performance and provides valuable insights and guidance for future exploration in this field.