Forecasting Bitcoin return using GARCH-LSTM

The high volatility of Bitcoin returns has become a significant concern for investors making investment decisions. Therefore, Bitcoin return forecasting is essential to maximize profit potential and minimize the risk of loss for Bitcoin investors. This paper proposes developing a hybrid model of GARCH combined with the machine learning technique known as LSTM. In hybrid modelling, the GARCH model's estimated volatility functions as an input for the LSTM model to capture the volatility trends in Bitcoin returns. The LSTM model component of the hybrid model can capture the long-term dependence of sequential Bitcoin returns. To validate the accuracy of the proposed model, we use the formulas for HMAE and HMSE. These formulas are commonly used to measure the accuracy of financial forecasting models, especially when dealing with heteroskedasticity in financial data. The results show that the GARCH-LSTM hybrid model can predict future Bitcoin returns effectively. This model is better than the GARCH model in terms of accuracy and provides stable and consistent predictions.