Big data and quantile risk control in digital asset markets

The paper develops a theoretical and empirical interpretation of the role of big data in digital asset risk management. The subject of the study is risk-control regimes based on quantile forecasts of Bitcoin and Solana returns as two assets with different volatility structures and different degrees of market maturity. The purpose is to show under which conditions a broader information set creates an economic effect rather than merely increasing computational complexity. The methodological design combines the economics of uncertainty, quantile regression, LightGBM quantile boosting, the continuous-time graph model DG-Neural ODE, and the indicators of pinball loss, interval coverage, Sharpe ratio, and maximum drawdown. Big data are interpreted as an integrated information system that joins long return series, heterogeneous market indicators, news-related signals, and inter-market dependencies. The results demonstrate that statistical forecast quality and economic usefulness are not identical: for Bitcoin, the best balance between return and risk is delivered by LightGBM, whereas for Solana an additional effect is generated by DG-Neural ODE because it captures regime shifts more effectively. The paper concludes that the core issue in the theory of risk control in digital asset markets is not model complexity itself, but the relation between interval width, selective market participation, and institutional rules of position limitation.