Abstract
As regional financial risks continue to evolve and increasingly intertwine, traditional regulatory frameworks face challenges in timely and accurate risk identification. To address this, the study proposes an optimized regulatory approach that integrates social network analysis (SNA) with a random forest (RF) model. This approach enhances regional financial risk monitoring and early warning capabilities by leveraging network structure indicators and machine learning predictions. Financial data from a certain province between 2016 and 2022 are used to construct an undirected weighted network. Empirical results show that network density increases while node degree centrality decreases, suggesting more balanced risk propagation. However, high-risk areas exhibit significant nonlinear amplification effects. The RF model incorporating SNA features performs exceptionally well in six-level risk classification. It achieves F1 scores between 0.941 and 0.985 and an area under the curve (AUC) of 0.975, significantly outperforming baseline models such as support vector machines and logistic regression. Feature importance analysis reveals that weighted node strength and network density contribute approximately 60%, highlighting the central role of network topology in risk prediction. Based on these findings, the study proposes a differentiated regulatory strategy guided by centrality and clustering indicators. This strategy aims to promote smarter and more precise regional financial supervision.