Wang Sujuan, Mansoor Musadaq
School of Accounting, Zhengzhou University of Economics and Business, Zhengzhou, Henan, China.
School of Computing Sciences, Pak-Austria Fachhochschule Institute of Applied Sciences and Technology, Haripur, Pakistan.
PeerJ Comput Sci. 2025 Apr 7;11:e2812. doi: 10.7717/peerj-cs.2812. eCollection 2025.
This study explores the complexities of enterprise financial management by optimizing financial models with a particular focus on enhancing risk prediction performance. A multi-objective mathematical model is first developed to establish key optimization goals, including cost reduction, improved capital utilization, and increased economic benefits. This model systematically defines decision variables and optimization objectives, providing a comprehensive framework for enterprise financial management. To improve predictive accuracy, the study integrates genetic algorithms with back-propagation (BP) neural networks, leveraging genetic algorithms to optimize the neural network's parameters and structure. Additionally, a hierarchical reinforcement learning model based on fuzzy reasoning (HRL-FR) is proposed to enhance decision-making capabilities. This model employs hierarchical decision-making and policy optimization, incorporating fuzzy reasoning to address uncertainties in complex and dynamic financial environments. Experimental validation using the Compustat dataset confirms the effectiveness of the proposed model. Key financial variables, including the working capital asset ratio and debt-to-equity ratio, are identified as significant influencers of prediction accuracy, reinforcing the model's robustness. The genetic algorithm's search and optimization process identifies parameter combinations that maximize neural network performance, further improving predictive capabilities. Comprehensive evaluations conducted on the Center for Research in Security Prices (CRSP) and Compustat datasets for 2022 confirm the HRL-FR model's superior ability to predict and analyze enterprise financial management information accurately. The model demonstrates higher profitability, enhanced efficiency, and predictive curves that closely align with optimal financial models. These findings highlight the HRL-FR model's potential as a powerful tool for enterprise financial management optimization, offering valuable insights for risk mitigation and strategic decision-making.
本研究通过优化财务模型来探索企业财务管理的复杂性，特别关注提高风险预测性能。首先开发了一个多目标数学模型来确立关键优化目标，包括成本降低、资本利用改善和经济效益提高。该模型系统地定义了决策变量和优化目标，为企业财务管理提供了一个全面的框架。为提高预测准确性，该研究将遗传算法与反向传播（BP）神经网络相结合，利用遗传算法优化神经网络的参数和结构。此外，还提出了一种基于模糊推理的分层强化学习模型（HRL-FR）来增强决策能力。该模型采用分层决策和策略优化，纳入模糊推理以应对复杂动态财务环境中的不确定性。使用标准普尔COMPUSTAT数据库进行的实验验证证实了所提出模型的有效性。关键财务变量，包括营运资本资产比率和资产负债率，被确定为预测准确性的重要影响因素，增强了模型的稳健性。遗传算法的搜索和优化过程确定了使神经网络性能最大化的参数组合，进一步提高了预测能力。对2022年证券价格研究中心（CRSP）和标准普尔COMPUSTAT数据库进行的综合评估证实了HRL-FR模型在准确预测和分析企业财务管理信息方面的卓越能力。该模型显示出更高的盈利能力、更高的效率以及与最优财务模型紧密匹配的预测曲线。这些发现凸显了HRL-FR模型作为企业财务管理优化有力工具的潜力，为风险缓解和战略决策提供了有价值的见解。
