Forecasting Bitcoin Price Volatility: A Comparative Analysis of Fuzzy Time Series, ARIMA, and GARCH Models for Short-Term and Long-Term Predictions

Abstract

Bitcoin’s pronounced price volatility continues to challenge the reliability of forecasting models, posing significant risks for traders, investors, and financial analysts. This study conducts a comprehensive comparative evaluation of several fuzzy time series models against traditional time series models, including ARIMA and GARCH. Using a decade-long dataset of daily Bitcoin closing prices, the study assesses both short-term and long-term predictive performance across multiple error-based and accuracy-based metrics. The findings reveal a clear horizon-dependent pattern: FTS models, particularly those incorporating Markov transitions, excel in short-term forecasting by capturing nonlinear and rapidly shifting market behavior, while ARIMA and GARCH models demonstrate superior long-term performance due to their ability to model broader trends and volatility structures. The study concludes that no single model is universally optimal; instead, aligning the forecasting method with the intended horizon and Bitcoin’s market dynamics is essential for improving decision-making in volatile financial environments.