Neural network forecasting of systemic financial risk based on blockchain market signals

Анотація

The modern financial system operates amid digitalization and the growing integration of blockchain infrastructure into global market processes. The expansion of digital assets and the increasing speed of financial transactions generate new sources of systemic instability that are not fully captured by traditional macro-financial indicators. This necessitates the application of analytical tools that account for nonlinearity, temporal dynamics, and the networked nature of financial processes. The study aims to substantiate the feasibility of using structured blockchain signals to model the dynamics of systemic financial risk and to test a neural network approach to its forecasting. Methods. The methodological framework integrates economic-statistical analysis of blockchain metrics with deep learning techniques. It is planned to systematize blockchain signals by functional group based on their economic content, and to develop a set of relevant predictors for further algorithmic analysis. Modeling is based on a recurrent neural network that accounts for the temporal structure of the data and lag dependencies. The model design focuses on developing an integrated indicator of systemic financial risk that accounts for nonlinear changes. Results. The findings reveal a statistically significant relationship between classified blockchain signals and periods of increased market turbulence. The proposed neural network model enables the timely identification of systemic financial risk escalation and reduces the forecasting lag compared to conventional approaches. Empirical validation confirms the model’s ability to capture nonlinear dependencies and the cumulative effects of market signals. Conclusions. The study substantiates the relevance of blockchain signals as an informational basis for forecasting systemic financial risk. The proposed approach expands the analytical toolkit for financial stability monitoring and provides a foundation for developing adaptive early warning systems. The results may be applied in risk management practice, macro-financial analysis, and regulatory supervision.