Comparative analysis of cryptocurrency series prediction models

this article presents a comparative analysis of modern approaches to forecasting the dynamics and volatility of cryptocurrency series. It considers classical statistical models (ARIMA, GARCH), seasonal forecasting methods (Prophet), and neural network algorithms (LSTM, GRU), as well as their hybrid architectures. Particular attention is paid to the applicability of models in conditions of limited samples and high information sensitivity of the cryptocurrency market. It is shown that ARIMA retains its significance as an interpretable benchmark for short-term analysis, GARCH remains a key tool for volatility assessment, while neural network and hybrid approaches demonstrate advantages with large data sets but are limited in terms of interpretability. The work contributes to the development of a methodology for forecasting abnormal price movements in decentralized markets, justifying the need to integrate statistical and information-theoretical methods. Prospects for further research are related to the use of entropy analysis and the development of hybrid models to improve the accuracy and stability of forecasts in conditions of high uncertainty. The work was carried out within the framework of the project «Development of a methodology for the formation of an instrumental base for analysis and modeling of spatial socio-economic development of systems in the context of digitalization based on internal reserves» (FSEG- 2023-0008).

cryptocurrency market, volatility forecasting, ARIMA, GARCH, Prophet, LSTM, hybrid models, anomalous price movements, risk management