The Data driven structure learning algorithm and the LASSO-VAR model were used to construct a time-varying dynamic Bayesian complex network model comprising high-dimensional continuous variables, to identify and warning systemic risk. The findings indicate that the financial industry achieved an average ODC of over 58% in the early stages of the global financial crisis, demonstrating the strongest risk spillover effect. The process of systemic risk contagion is manifested as a linkage process of internal industry diffusion to inter-industry diffusion, and key risk nodes can be identified in the early stages. The real estate industry is a source of financial system risk is verified through sensitivity analysis to validate the dynamic Bayesian complex network model of high-dimensional continuous variables. The model can identify and warn institutional level system risks in the early stages of risk outbreaks, and it helps to manage and control system risks in the early stages of crises.