.png){kind=icon}
.png){kind=icon}
Return
Reinforcement Learning-Driven Enterprise Financial Management and Spatiotemporal Convolution Optimization Model
.png){kind=icon}
PRE
AI
W
Z
DOI：10.1002/eng2.70730.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
To address sequential decision-making problems in corporate financial management involving multiple objectives, multiple constraints, and high uncertainty, this study proposes a corporate financial management framework based on the joint optimization of reinforcement learning and spatiotemporal convolution. The framework uses spatiotemporal convolution to capture cross-period dependencies and cross-entity transmission relationships in corporate financial data. It embeds structured state representations into the reinforcement learning policy. Additionally, funding cost, liquidity requirements, tail risk, and compliance thresholds are incorporated as endogenous constraints in the policy update process. This design enables end-to-end optimization and avoids the fragmented structure of the traditional predict-then-decide paradigm. Experimental results show that the proposed model consistently outperforms baseline methods on key indicators such as funding cost, liquidity buffer capacity, and operational efficiency. The model also demonstrates robust performance on risk indicators, including maximum drawdown, cash shortfall, and compliance trigger events. These results confirm the effectiveness and interpretability of end-to-end joint optimization in complex financial environments. The study further indicates that when decision policies capture both long-term temporal dependencies and transmission mechanisms within organizational networks, firms can achieve more balanced coordination among cost control, operational efficiency, financial robustness, and regulatory compliance. Integrating risk and compliance constraints directly into the decision process further improves the stability of financial strategies. Therefore, this study makes a meaningful contribution to research on intelligent financial decision-making and cross-departmental capital allocation in corporate finance.
Keywords:
enterprise financial management
reinforcement learning
risk constraints
spatiotemporal convolution
Journal
IF:
2
Papers:
362
Citations:
1.7K
Organization
No organization information available
Cited Papers