Bitcoin Volatility Forecasting Using Statistical Analysis and AI Models - A Comparative Study

This work offers a comparative analysis of the effectiveness of two analytical cultures: statistical analysis and data modeling (DMC), and algorithmic AI analysis (AMC, artificial network LSTM) to forecast volatility of Bitcoin obtain by using Exponentially Weighted Moving Averages (EWMA). The analysis was based on data between March 10, 2016 and April 23, 2024, divided into two sets: a training set (1906 data points) and a test set (35 data points). Models’ results of test sets’ data have been compared using MSE (Mean Square Error) and MAE (Mean Absolute Error). The study has shown that on a shorter time horizon a statistical analysis resulted in a better fit than AI methods (LSTM and LSTM+x). Conversely, LSTM+x was better at a longer time period when data displays higher volatility and higher peaks. In conclusion, both methods are applicable in predicting volatility of cryptocurrencies.