Forecasting the volume of digital financial assets in Russia using instrumental methods

The article addresses methodological approaches to forecasting the volume of digital financial assets (DFAs) in the Russian Federation, focusing on the application of instrumental time series models under conditions of high volatility and institutional uncertainty. The empirical analysis relies on monthly data on the Russian DFA market for the period 2022-2025, collected from official publications of the Bank of Russia, sectoral reviews, and specialized statistical platforms. Descriptive statistics revealed a high variability of the series, right-skewed distribution, bimodal character, and multiple regime shifts. Structural breaks were identified in June 2023 and August 2024, corresponding to regulatory interventions and technological advancements. A log-transformation and differencing procedures were applied to stabilize variance and achieve stationarity, while step and ramp intervention variables were introduced to capture the long-term effects of structural changes. The study compares the predictive performance of three forecasting techniques: ARIMA, SARIMA with intervention variables, and Prophet. The highest accuracy was achieved by the Prophet model, which proved most adaptive to nonlinear patterns and seasonal structures. Prophet effectively captured the August 2024 surge, modeled annual cycles, and forecasted a peak in August 2025 followed by a sharp correction, reflecting cyclical investment behavior. Accuracy metrics confirmed its superiority: Prophet recorded the lowest error levels, residual diagnostics further demonstrated proximity to white noise, absence of autocorrelation, and only moderate deviations in extreme values, underscoring the model's resilience in volatile environments. The comparative evaluation of forecasting results highlights the critical importance of model selection in volatile financial markets. While ARIMA proved insufficient for regime-shifting data, SARIMA's explicit inclusion of structural interventions improved performance, and Prophet's nonlinear adaptability yielded the most reliable forecasts. The findings confirm that DFA markets exhibit phase-dependent and discontinuous behavior that cannot be effectively captured by traditional linear tools alone. This underscores the need for flexible, adaptive, and ensemble approaches that integrate regime shifts, seasonal cycles, and exogenous shocks into forecasting strategies. The study's contribution lies in providing empirical evidence on the predictive capabilities of alternative time series models for the DFA market in Russia and demonstrating the role of adaptive algorithms in capturing structural and cyclical transformations. The results have direct implications for regulators, market participants, and researchers seeking to design evidence-based policies, manage investment risks, and develop strategic planning frameworks in the digital economy. Future research directions include hybrid forecasting architectures, integration of machine learning techniques, and scenario-based simulations to further enhance predictive reliability in the context of rapidly evolving digital financial ecosystems. The authors declare no conflicts of interests.

Keywords

Authors

References