Systemic Risk in Chinese Commercial Banking Systems Based on Complex Networks and Deep Learning

Given the pronounced nonlinearity and time-varying characteristics of risk contagion in modern interbank markets, traditional econometric methods based on linear assumptions and static frameworks are insufficient to effectively capture the tail risk spillover effects under extreme market conditions. Therefore, this article constructs a systematic risk analysis framework integrating deep learning algorithms and complex network theory, builds a directed weighted network of tail risk in the banking system, and analyzes the topological structure of risk transmission by combining Granger causality tests and the Infomap community detection algorithm. Finally, this article introduces time-varying network analysis and the PELT changepoint detection method to endogenously identify structural abrupt changes in risk evolution. The LSTM-CoVaR estimates are validated against linear quantile-regression, copula, and unconditional CAViaR benchmarks through conditional-coverage backtests and an out-of-sample evaluation. The study finds that the Chinese banking system has formed a significant core-periphery asymmetric two-layer structure, with the risk transmission mechanism evolving from early single-center diffusion to a core spillover-driven and regional multi-layered resonance model. Systemically important banks are the main positive sources of risk spillover, and their fluctuations are highly synchronized with policy cycles. Local and rural commercial banks, as net receivers, amplify local shocks in the chain structure of core, regional, and periphery. The surge in network connectivity after 2020 has propelled the system into a highly coupled, globally resonant phase, enhancing overall resilience while significantly increasing the fragility of local structures. These findings suggest that macroprudential monitoring may benefit from extending its focus beyond single hubs to the multi-layered nested structure of risk transmission, where regional disruptions can propagate into broader financial instability.