Cryptocurrency volatility prediction based on price, return and volatility cross-correlation using LSTM

Abstract

The rise of cryptocurrencies as a major economic factor has drawn interest from both individual investors and regulators. This has led researchers to explore new ways to predict cryptocurrency volatility. This research proposes a new approach for more accurately predicting the conditional variance of cryptocurrencies while previous research did not investigate cross-correlation among cryptocurrency mean and volatility landscapes. This research precisely uses a method to select the most relevant features based on the cross-correlation between price, returns, and volatility. Experiments on eight cryptocurrencies that have the largest market capacity from 2018 to 2024 show that this approach is effective. By considering stronger cross-correlation when choosing the data used for volatility prediction using LSTM, the research substantially reduces the error of prediction. The findings of this research can be used to predict the volatility of other cryptocurrencies and the stock markets.