Cryptocurrency Market Forecasting Based On Garch-Lstm Neural Networks: A Case Study Of Bitcoin And Ethereum

This study investigates the effectiveness of a hybrid forecasting model that combines Generalized Autoregressive Conditional Heteroskedasticity (GARCH) with Long Short-Term Memory (LSTM) neural networks, specifically applied to the cryptocurrency market, focusing on Bitcoin and Ethereum. The inherent volatility of cryptocurrencies presents substantial challenges for accurate price prediction, necessitating advanced methodologies that can adapt to fluctuating market conditions. We first utilize GARCH models to analyze and capture the time-varying volatility in the returns of Bitcoin and Ethereum, enabling a comprehensive understanding of the underlying market dynamics. Following this, we implement LSTM networks to exploit their capability to model complex, non-linear relationships in sequential data, enhancing the predictive power of the model. The performance of the GARCH-LSTM framework is rigorously evaluated using historical price data for Bitcoin and Ethereum, employing key metrics such as Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) to assess forecasting accuracy. The results demonstrate that the hybrid approach significantly outperforms traditional forecasting methods, providing more reliable predictions and insights into market trends. This study contributes to the growing body of literature on cryptocurrency forecasting by illustrating the potential of combining econometric techniques with advanced machine learning methods, offering valuable implications for traders and investors in the cryptocurrency ecosystem. However, the experimental results revealed that the LSTM model outperformed the other eight methods in terms of forecasting performance measures, the RMSPE validation is 0.112561, and the RMSE validation is 0.011456.