<![CDATA[Technological advancements have increased interest in cryptocurrency investments, particularly Bitcoin and Ethereum, despite the high price volatility that remains a major challenge for investors. This study aims to predict cryptocurrency price fluctuations using Long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU) models, both of which are well-regarded for time series data analysis. Model performance was evaluated using parameters such as learning rate, timestamps, batch size, number of epochs, and Early Stopping callbacks. The evaluation metrics included Mean Absolute Percentage Error (MAPE), Root Mean Square Error (RMSE), and the coefficient of determination (R²). The results indicate that the GRU model outperforms LSTM in predicting cryptocurrency prices. For Bitcoin data, GRU achieved a MAPE of 0.38%, RMSE of 343.02, and R² of 0.9988, surpassing LSTM, which recorded a MAPE of 0.41%, RMSE of 356.01, and R² of 0.9987. Similarly, for Ethereum data, GRU achieved a MAPE of 0.45%, RMSE of 20.89, and R² of 0.9983, outperforming LSTM with a MAPE of 0.49%, RMSE of 22.29, and R² of 0.9980. These findings demonstrate that GRU is more accurate and efficient in modeling cryptocurrency price patterns, offering strategic opportunities for investors to make more informed decisions in navigating the complexities of the cryptocurrency market.]]>
Copyrights © 2025
